HEAL Network

Abstract

Atrial fibrillation (AF) is the most common arrhythmia worldwide, and patient perceptions significantly influence shared treatment decisions. Artificial intelligence-driven analysis of social media may offer valuable insights into contemporary public attitudes toward AF outside clinical settings.This qualitative study used large language modeling and advanced artificial intelligence topic modeling techniques to analyze public perceptions of AF from Reddit discussions between April 2006 and November 2023.We curated 86 323 AF-related conversations (18 754 posts, 67 569 comments) across 38 183 unique users by searching terms related to AF. Our topic modeling identified 65 distinct discussion topics organized into 9 thematic groups, with topics including personal experiences with treatments (eg, ablation, rate versus rhythm control), roles of health care providers and community support, AF triggers (diet, illicit substances, supplements, stress, caffeine), and anecdotes highlighting the difficulties of living with AF. Discussions commonly reflected 3 main themes: (1) advantages and limitations of wearable devices for AF monitoring, (2) hesitancy and misconceptions about AF treatment, and (3) patient-centered challenges following an AF diagnosis.The artificial intelligence-enabled analysis underscored substantial public discourse around patient experiences with AF detection and management. Leveraging social media data to understand patient perspectives on cardiovascular health may inform patient-centered resources and future research directions to better support patients living with AF.

View details for DOI 10.1161/JAHA.125.043999

View details for PubMedID 41404746

Abstract

Statin adherence impacts cardiovascular outcomes, yet disparities persist. Understanding the sociodemographic factors and barriers is crucial for targeted interventions.To investigate the relationship between sociodemographic factors and statin adherence across racial and ethnic groups.This retrospective study examined sociodemographic factors, prescription records, clinical factors, and responses from the Demographic, Drug Exposure, Healthcare Utilization Survey (HUS) in the All of Us (AoU) cohort. Multivariable logistic regression models were used to assess the impact of sociodemographic factors on adherence stratified by race.Adult participants with statin prescription records. Subgroup analyses included those who responded to the HUS.Statin prescription.We calculated percent days covered (PDC) as the proportion of days within a year in which a person prescribed a statin filled a prescription. Adequate adherence was defined as PDC ≥ 80%.Among the 17,029 adults with a statin prescription, the mean PDC was 57%, and 66% had PDC ≤ 80%. In multivariable analyses stratified by race and ethnicity, distinct barriers to adherence emerged. Among the non-Hispanic White participants, barriers to consistent healthcare [odds ratio (OR) = 0.60, 95% CI (0.42-0.87)] and lack of provider identity concordance [OR = 0.83, 95% CI (0.72-0.97)] were associated with lower adherence. In the non-Hispanic Black participants, Medicare [OR = 0.54, 95% CI (0.32-0.90)] and Veterans Affairs insurance [OR = 0.44, 95% CI (0.20-0.96)], as well as financial barriers [OR = 0.69, 95% CI (0.51-0.96)], reduced adherence. Among the Hispanic participants, provider-related anxiety [OR = 0.13, 95% CI (0.02-0.87)], immigrant status [OR = 0.25, 95% CI (0.08-0.72)], and Medicaid coverage [OR = 0.11, 95% CI (0.03-0.45)] predicted lower adherence.Addressing cardiovascular disease disparities requires recognizing unique sociodemographic barriers to statin adherence within racial and ethnic groups. Our findings highlight the need for tailored strategies considering the diverse barriers each group faces. Targeted interventions can bridge adherence gaps and improve cardiovascular outcomes across populations. This approach recognizes that although race and ethnicity may correlate with specific barriers, the underlying social determinants of health often play the key role in statin adherence.

View details for DOI 10.3389/fcvm.2025.1541082

View details for PubMedID 41458993

View details for PubMedCentralID PMC12740900

Abstract

Acute care use (ACU) represents a major economic burden in oncology, which can ideally be prevented. Existing models effectively predict such events.We aimed to quantify the cost savings achieved by implementing a model to predict ACU in oncology patients undergoing systemic therapy.This retrospective cohort study analyzed patients with cancer at an academic medical center from 2010 to 2022. We included patients who received systemic therapy and identified ACU events occurring after treatment initiation, excluding those with known death dates within the study period. Data on ACU-related expenses were gathered from Medicare claims and mapped to service codes in electronic health records, yielding average daily costs for each patient over 180 days following the start of therapy. The exposure was an ACU event.The main outcome was the average daily cost per patient at the end of the first 180 days of systemic therapy. We observed that expense accumulation flattened earlier and more rapidly among non-ACU patients. This study included 20,556 patients, of whom 3820 (18.58%) experienced at least 1 ACU. The average daily cost per patient for those with and without ACU was US

View details for DOI 10.1016/j.jid.2025.10.616

View details for PubMedID 41308732

View details for DOI 10.1161/circ.152.suppl_3.4370304

View details for Web of Science ID 001613838800020

View details for DOI 10.1161/circ.152.suppl_3.4366575

View details for Web of Science ID 001618614500044

Abstract

Dynamic predictive modeling using electronic health record data has gained significant attention in recent years. The reliability and trustworthiness of such models depend heavily on the quality of the underlying data, which is, in part, determined by the stages preceding the model development: data extraction from electronic health record systems and data preparation. In this paper, we identified over 40 challenges encountered during these stages and provided actionable recommendations for addressing them. These challenges are organized into 4 categories: cohort definition, outcome definition, feature engineering, and data cleaning. This comprehensive list serves as a practical guide for data extraction engineers and researchers, promoting best practices and improving the quality and real-world applicability of dynamic prediction models in clinical settings.

View details for DOI 10.2196/73987

View details for PubMedID 41105947

View details for DOI 10.1038/s41591-025-04008-8

View details for PubMedID 41102561

View details for PubMedCentralID 11756841

Abstract

Patient-reported outcomes (PROs) are vital in assessing disease activity and treatment outcomes in inflammatory bowel disease (IBD). However, manual extraction of these PROs from the free-text of clinical notes is burdensome. We aimed to improve data curation from free-text information in the electronic health record, making it more available for research and quality improvement. This study aimed to compare traditional natural language processing (tNLP) and large language models (LLMs) in extracting 3 IBD PROs (abdominal pain, diarrhea, fecal blood) from clinical notes across 2 institutions.Clinic notes were annotated for each PRO using preset protocols. Models were developed and internally tested at the University of California, San Francisco, and then externally validated at Stanford University. We compared tNLP and LLM-based models on accuracy, sensitivity, specificity, positive, and negative predictive value. In addition, we conducted fairness and error assessments.Interrater reliability between annotators was >90%. On the University of California, San Francisco test set (n = 50), the top-performing tNLP models showcased accuracies of 92% (abdominal pain), 82% (diarrhea) and 80% (fecal blood), comparable to GPT-4, which was 96%, 88%, and 90% accurate, respectively. On external validation at Stanford (n = 250), tNLP models failed to generalize (61%-62% accuracy) while GPT-4 maintained accuracies >90%. Pathways Language Model-2 and Generative Pre-trained Transformer-4 showed similar performance. No biases were detected based on demographics or diagnosis.LLMs are accurate and generalizable methods for extracting PROs. They maintain excellent accuracy across institutions, despite heterogeneity in note templates and authors. Widespread adoption of such tools has the potential to enhance IBD research and patient care.

View details for DOI 10.1016/j.gastha.2024.10.003

View details for PubMedID 39877865

View details for PubMedCentralID PMC11772946

Abstract

Objectives: The growing use of machine learning (ML) in healthcare raises concerns about how data biases affect real-world model performance. While existing frameworks evaluate algorithmic fairness, they often overlook the impact of bias on generalizability and clinical utility, which are critical for safe deployment. Building on prior methods, this study extends bias analysis to include clinical utility, addressing a key gap between fairness evaluation and decision-making.Materials and Methods: We applied a 3-phase evaluation to a previously developed model predicting prolonged opioid use (POU), validated on Veterans Health Administration (VHA) data. The analysis included internal and external validation, model retraining on VHA data, and subgroup evaluation across demographic, vulnerable, risk, and comorbidity groups. We assessed performance using area under the receiver operating characteristic curve (AUROC), calibration, and decision curve analysis, incorporating standardized net-benefits to evaluate clinical utility alongside fairness and generalizability.Results: The internal cohort (N=41929) had a 14.7% POU prevalence, compared to 34.3% in the external VHA cohort (N=397150). The model's AUROC decreased from 0.74 in the internal test cohort to 0.70 in the full external cohort. Subgroup-level performance averaged 0.69 (SD=0.01), showing minimal deviation from the external cohort overall. Retraining on VHA data improved AUROCs to 0.82. Clinical utility analysis showed systematic shifts in net-benefit across threshold probabilities.Discussion: While the POU model showed generalizability and fairness internally, external validation and retraining revealed performance and utility shifts across subgroups.Conclusion: Population-specific biases affect clinical utility-an often-overlooked dimension in fairness evaluation-a key need to ensure equitable benefits across diverse patient groups.

View details for DOI 10.1093/jamiaopen/ooaf115

View details for PubMedID 41036091

Abstract

OBJECTIVE: Building upon our previous work on predicting treatment retention in medications for opioid use disorder, we aimed to improve 6-month retention prediction in buprenorphine-naloxone (BUP-NAL) therapy by incorporating features derived from large language models (LLMs) applied to unstructured clinical notes.MATERIALS AND METHODS: We used de-identified electronic health record (EHR) data from Stanford Health Care (STARR) for model development and internal validation, and the NeuroBlu behavioral health database for external validation. Structured features were supplemented with 13 clinical and psychosocial features extracted from free-text notes using the CLinical Entity Augmented Retrieval pipeline, which combines named entity recognition with LLM-based classification to provide contextual interpretation. We trained classification (Logistic Regression, Random Forest, XGBoost) and survival models (CoxPH, Random Survival Forest, Survival XGBoost), evaluated using Receiver Operating Characteristic-Area Under the Curve (ROC-AUC) and C-index.RESULTS: XGBoost achieved the highest classification performance (ROC-AUC=0.65). Incorporating LLM-derived features improved model performance across all architectures, with the largest gains observed in simpler models such as Logistic Regression. In time-to-event analysis, Random Survival Forest and Survival XGBoost reached the highest C-index (0.65). SHapley Additive exPlanations analysis identified LLM-extracted features like Chronic Pain, Liver Disease, and Major Depression as key predictors. We also developed an interactive web tool for real-time clinical use.DISCUSSION: Features extracted using NLP and LLM-assisted methods improved model accuracy and interpretability, revealing valuable psychosocial risks not captured in structured EHRs.CONCLUSION: Combining structured EHR data with LLM-extracted features moderately improves BUP-NAL retention prediction, enabling personalized risk stratification and advancing AI-driven care for substance use disorders.

View details for DOI 10.1093/jamia/ocaf157

View details for PubMedID 40977375

Abstract

In 2023, the American Heart Association PREVENT (Predicting Risk of Cardiovascular Disease Events) equations were introduced as a tool to improve cardiovascular disease (CVD) risk prediction. This study tests their performance in a diverse socioeconomic cohort.We analyzed All of Us participants aged 30 to 79 years without baseline CVD who had required PREVENT input data over a 5.4-year follow-up. Discrimination was assessed using Harrell's C-statistic, with calibration by comparing predicted and observed 5-year CVD rates across 10-year risk deciles. Mean data are ±SD.We examined 9010 individuals (mean age, 63.0±11.0 years; 45.5% male). Racial and ethnic composition was 61.7% non-Hispanic White, 17.2% non-Hispanic Black, 4.5% multiracial/other, 1.3% non-Hispanic Asian, and 11.2% Hispanic or Latino. The "other" race/ethnic category reflects participants who self-identified as "other" in response to the, "Which category describes you?" item in the Basics survey. Over a mean follow-up of 3.6±1.8 years, 9.0% experienced a cardiovascular event. The mean 10-year predicted risks were 0.23±0.17 for total CVD, 0.13±0.10 for atherosclerotic CVD (ASCVD), and 0.19±0.17 for heart failure. The predicted-to-observed rate ratios were 5.3 for CVD and 3.3 for ASCVD. The C statistic for the overall sample was 0.732 (95% CI, 0.718-0.752) for CVD, 0.716 (95% CI, 0.698-0.741) for ASCVD, and 0.777 (95% CI, 0.757-0.800) for heart failure.The PREVENT equations showed strong discrimination across all strata in this national cohort. Overprediction of CVD events likely reflects baseline differences in comorbidity burden between the PREVENT development cohort and this All of Us cohort, particularly due to the exclusion of individuals missing estimated glomerular filtration rate, a variable not routinely collected and likely missing, not at random. Strong discrimination supports potential clinical utility, though further work is needed to improve calibration in this population.

View details for DOI 10.1161/JAHA.125.041549

View details for PubMedID 40932135

Abstract

Consumer wearable technologies have wide applications, including some that have US Food and Drug Administration clearance for health-related notifications. While wearable technologies may have premarket testing, validation, and safety evaluation as part of a regulatory authorization process, information on their postmarket use remains limited. The Stanford Center for Digital Health organized 2 pan-stakeholder think tank meetings to develop an organizing concept for empirical research on the postmarket evaluation of consumer-facing wearables.The postmarket evaluation of consumer wearables involves broad consideration of an individual consumer's journey from acquisition, intended and unintended use of the wearable, and access to health care resources on receipt of a notification. For individuals who do access the health care system, a wearable's downstream effects can be studied through appropriate clinical evaluation, delivery of guideline-directed treatments, shared decision-making in areas of clinical equipoise, and analysis of clinical end points and patient harms. Effective postmarket research draws from denominators appropriate to the clinical question, with clearly defined parameters for success and failure. Generalizability related to data completeness and reliability should also be considered. As patients increasingly integrate wearables into their health monitoring, cross-platform data sharing with a focus on privacy and data quality can drive patient-centered innovation and identify opportunities to bridge gaps in medical care.The think tank identified priorities in postmarket research, comprising the journey from consumer to patient and accounting for patient, clinician, health care delivery system, and societal impacts of consumer wearables. Overall, this approach serves not only to organize the study of consumer wearables but also to act as a guidepost for using real-world data in postmarket research.

View details for DOI 10.1001/jamacardio.2025.3006

View details for PubMedID 40928810

Abstract

Statin adherence impacts cardiovascular outcomes, yet disparities persist. Understanding sociodemographic factors and barriers is crucial for targeted interventions.To investigate the relationship between sociodemographic factors and statin adherence across racial and ethnic groups.This retrospective study examined sociodemographic factors, prescription records, clinical factors, and responses from the Demographic, Drug Exposure, Healthcare Utilization Survey (HUS) in the All of Us (AoU) cohort. Multivariable logistic regression models assessed the impact of sociodemographic factors on adherence stratified by race.Adult participants with statin prescription records. Subgroup analyses included those who responded to the HUS.Statin prescription.Percent days covered (PDC), calculated as the proportion of days within a year in which a person prescribed a statin filled a prescription. Adequate adherence was defined as PDC ≥ 80%.Of the 17,029 participants with a statin prescription, the mean statin PDC was 57%, with 66% reporting a PDC ≤ 80%. Racial and ethnic differences in adherence were observed, with Non-Hispanic White (NHW) participants having a median PDC of 74% (IQR [0.25,0.98]), Non-Hispanic Black (NHB) 49% (IQR [0.25,0.98]), and Hispanic participants 25% (IQR [0.08,0.49]). NHW participants faced employment barriers (OR 0.63, 95% CI [0.46, 0.86]) and provider inaccessibility (OR 0.56, 95% CI [0.40, 0.76]) as significant factors for lower adherence. NHB participants experienced patient anxiety (OR 0.53, 95% CI [0.30, 0.90]) and financial barriers (OR 0.65, 95% CI [0.50, 0.85]), while Hispanic participants showed patient anxiety (OR 0.14, 95% CI [0.02, 0.60]) and immigrant status (OR 0.36, 95% CI [0.17, 0.76]) as significant factors for lower adherence.To address cardiovascular disease disparities, it is crucial to recognize unique sociodemographic barriers to statin adherence within racial and ethnic groups. Our findings highlight the need for tailored strategies considering the diverse barriers each group faces. Targeted interventions can bridge adherence gaps and improve cardiovascular outcomes across populations. This approach recognizes that while race and ethnicity may correlate with specific barriers, it is the underlying SDoH that often play a pivotal role in statin adherence.

View details for DOI 10.1101/2025.08.26.25334490

View details for PubMedID 40909854

View details for PubMedCentralID PMC12407661

Abstract

Cluster analysis, a machine learning-based and data-driven technique for identifying groups in data, has demonstrated its potential in a wide range of contexts. However, critical appraisal and reproducibility are often limited by insufficient reporting, ultimately hampering the interpretation and trust of key stakeholders. The present paper describes the protocol that will guide the development of a reporting guideline and checklist for studies incorporating cluster analyses-Transparent Reporting of Cluster Analyses.Following the recommended steps for developing reporting guidelines outlined by the Enhancing the QUAlity and Transparency Of health Research Network, the work will be divided into six stages. Stage 1: literature review to guide development of initial checklist. Stage 2: drafting of the initial checklist. Stage 3: internal revision of checklist. Stage 4: Delphi study in a global sample of researchers from varying fields (n=≈) to derive consensus regarding items in the checklist and piloting of the checklist. Stage 5: consensus meeting to consolidate checklist. Stage 6: production of statement paper and explanation and elaboration paper. Stage 7: dissemination via journals, conferences, social media and a dedicated web platform.Due to local regulations, the planned study is exempt from the requirement of ethical review. The findings will be disseminated through peer-reviewed publications. The checklist with explanations will also be made available freely on a dedicated web platform (troca-statement.org) and in a repository.

View details for DOI 10.1136/bmjopen-2025-099609

View details for PubMedID 40840990

View details for DOI 10.1016/j.xops.2025.100717

View details for Web of Science ID 001460350300001

Abstract

Existing ophthalmology studies for clinical phenotypes identification in real-world datasets (RWD) rely exclusively on structured data elements (SDE). We evaluated the performance, generalizability, and fairness of multimodal ensemble models that integrate real-world SDE and free-text data compared to SDE-only models to identify patients with glaucoma.This is a retrospective cross-sectional study involving 2 health systems- University of Michigan (UoM) and Stanford University (SU). It involves 1728 patients visiting eye clinics during 2012-2021. Free-text embeddings extracted using BioClinicalBERT were combined with SDE. EditedNearestNeighbor (ENN) undersampling and Borderline-Synthetic Minority Over-sampling Technique (bSMOTE) addressed class imbalance. Lasso Regression (LR), Random Forest (RF), Support Vector Classifier (SVC) models were trained on UoM imbalanced (imb) and resampled data along with bagging ensemble method. Models were externally validated with SU data. Fairness was assessed using equalized odds difference (EOD) and Target Probability Difference (TPD).Among 900 and 828 patients from UoM and SU, 10% and 23% respectively had glaucoma as confirmed by ophthalmologists. At UoM, multimodal LRimb (F1 = 76.60 [61.90-88.89]; AUROC = 95.41 [87.01-99.63]) outperformed unimodal RFimb (F1 = 69.77 [52.94-83.64]; AUROC = 97.72 [95.95-99.18]) and ICD-coding method (F1 = 53.01 [39.51-65.43]; AUROC = 90.10 [84.59-93.93]). Bagging (BM = LRENN + LRbSMOTE) improved performance achieving an F1 of 83.02 [70.59-92.86] and AUROC of 97.59 [92.98-99.88]. During external validation BM achieved the highest F1 (68.47 [62.61-73.75]), outperforming unimodal (F1 = 51.26 [43.80-58.13]) and multimodal LRimb (F1 = 62.46 [55.95-68.24]). BM EOD revealed lower disparities for sex (<0.1), race (<0.5) and ethnicity (<0.5), and had least uncertainty using TDP analysis as compared to traditional models.Multimodal ensemble models integrating structured and unstructured EHR data outperformed traditional SDE models achieving fair predictions across demographic sub-groups. Among ensemble methods, bagging demonstrated better generalizability than stacking, particularly when training data is limited.This approach can enhance phenotype discovery to enable future research studies using RWD, leading to better patient management and clinical outcomes.

View details for DOI 10.1093/jamiaopen/ooaf080

View details for PubMedID 40799932

View details for PubMedCentralID PMC12342940

Abstract

BACKGROUND: Real-world medical environments such as oncology are highly dynamic due to rapid changes in medical practice, technologies, and patient characteristics. This variability, if not addressed, can result in data shifts with potentially poor model performance. Presently, there are few easy-to-implement, model-agnostic diagnostic frameworks to vet machine learning models for future applicability and temporal consistency.METHODS: We extracted clinical data from EHR for a cohort of over 24,000 patients who received antineoplastic therapy within a distinct year. The label of this study are acute careutilization (ACU) events, i.e., emergency department visits and hospitalizations, within 180 days of treatment initiation. Our cross-sectional data spans treatment initiation points from 2010-2022. We implemented three models within our validation framework: Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest(RF), and Extreme Gradient Boosting (XGBoost).RESULTS: Here, we introduce a model-agnostic diagnostic framework to validate clinical machine learning models on time-stamped data, consisting of four stages. First, the framework evaluates performance by partitioning data from multiple years into training and validation cohorts. Second, it characterizes the temporal evolution of patient outcomes and characteristics. Third, model longevity and trade-offs between data quantity and recency are explored. Finally, feature importance and data valuation algorithms are applied for feature reduction and data quality assessment. When applied to predicting ACU in cancer patients, the framework highlights fluctuations in features, labels, and data values over time.CONCLUSIONS: The work in this study emphasizes the importance of data timeliness and relevance. The results on ACU in cancer patients show moderate signs of drift and corroborate the relevance of temporal considerations when validating machine learning models for deployment at the point of care.

View details for DOI 10.1038/s43856-025-00965-w

View details for PubMedID 40596645

Abstract

The notion of medical digital twins is gaining popularity both within the scientific community and among the general public; however, much of the recent enthusiasm has occurred in the absence of a consensus on their fundamental make-up. Digital twins originate in the field of engineering, in which a constantly updating virtual copy enables analysis, simulation, and prediction of a real-world object or process. In this Health Policy paper, we evaluate this concept in the context of medicine and outline five key components of the medical digital twin: the patient, data connection, patient-in-silico, interface, and twin synchronisation. We consider how various enabling technologies in multimodal data, artificial intelligence, and mechanistic modelling will pave the way for clinical adoption and provide examples pertaining to oncology and diabetes. We highlight the role of data fusion and the potential of merging artificial intelligence and mechanistic modelling to address the limitations of either the AI or the mechanistic modelling approach used independently. In particular, we highlight how the digital twin concept can support the performance of large language models applied in medicine and its potential to address health-care challenges. We believe that this Health Policy paper will help to guide scientists, clinicians, and policy makers in creating medical digital twins in the future and translating this promising new paradigm from theory into clinical practice.

View details for DOI 10.1016/j.landig.2025.02.004

View details for PubMedID 40518342

Abstract

In the ongoing opioid epidemic, the mortality risk of opioid initiation in patients with dementia or mild cognitive impairment (MCI) remains understudied despite their vulnerability. This study evaluates mortality risks associated with opioid exposure in patients diagnosed with dementia or MCI by comparing outcomes between the initiation and continuation groups.We conducted a retrospective cohort study using data from a Northern California academic healthcare system (Stanford Health Care Alliance; 2015/01/01-2024/07/31), including 27,757 patients aged 50-100 with dementia or MCI. Of these, 14,105 received opioids after diagnosis and were classified as initiation (opioid-naïve; n=9443) or continuation (non-naïve; n=4662) groups. Cox regression assessed 14-day mortality risk. Aalen's additive model examined time-varying impact up to 180 days. Potential causes of death were extracted from clinical notes using GPT-3.5-Turbo. We also analyzed an independent community healthcare system cohort (Providence Health & Service; n=208,306) from western US states (2015/01/01-2023/05/31) as a replication cohort.In the primary cohort, 4.1% (572/14,105) of patients died within 14 days of opioid exposure. The initiation group had a significantly higher 14-day mortality risk than the continuation group (adjusted hazard ratio (aHR), 2.00 (1.59-2.52); P<0.0001). The replication cohort had a 14-day mortality rate of 6.2% (7022/113,343) with a smaller difference between the initiation (n=77,168) and continuation (n=36,175) groups (aHR 1.22 (1.16-1.30); P<0.0001). In both cohorts, elevated risk stabilized after day 30. In the primary cohort, respiratory conditions (62% vs. 48%, P<0.1), particularly pneumonia (38% vs. 19%, P<0.05), were more prevalent among the initiation group who died early.Starting opioids in patients with dementia or MCI is associated with elevated short-term mortality risks, with the initiation group having twice the 14-day mortality risk in academic settings and a smaller but significant increase in community healthcare systems. The first 30 days after initiation represent a critical risk window, likely due to a lack of tolerance to opioid adverse effects. These findings underscore the need for cautious initiation, tailored follow-up protocols accounting for healthcare setting characteristics, and close monitoring during the first month in this vulnerable population.

View details for DOI 10.1186/s12916-025-04172-1

View details for PubMedID 40484923

View details for PubMedCentralID 4229610

View details for Web of Science ID 001560037700016

Abstract

Preoperative opioid misuse is associated with worse postoperative outcomes. This prospective longitudinal cohort study evaluated the association between preoperative opioid misuse and prolonged pain and opioid use after elective spine surgery; and examined postoperative trajectories of patient-reported outcomes over one year.Fifty-two patients undergoing elective spine surgery completed presurgical and weekly postoperative longitudinal assessments of pain and opioid use and monthly assessments of depression, anxiety, sleep disturbance, and physical function. Cox regression analyzed the effect of preoperative opioid misuse on time to pain and opioid cessation while linear mixed-effects models examined longitudinal changes in postoperative outcomes.Adjusting for age, sex, operative region, number of spinal levels, and any preoperative opioid use, preoperative opioid misuse (COMM-Positive) was associated with a delayed return to baseline opioid dose (hazard ratio [HR], 0.35; 95% confidence interval [CI], 0.14-0.88; p=0.02) and delayed opioid cessation (HR, 0.25; 95% CI, 0.09-0.59; p=0.008). All patients experienced comparable reductions in current and average pain intensity, and pain interference over time. COMM-Positive patients reported a normalization of postoperative anxiety and depression 1 month after surgery with a rebound at 3 months while patients without preoperative opioid misuse remained stable over time.Preoperative opioid misuse is a significant risk factor for delayed opioid cessation even after adjusting for preoperative opioid use, and is associated with a transient normalization of anxiety and depressive symptoms with a rebound 3 months following spine surgery. Targeted screening and risk reduction strategies are needed for patients reporting preoperative opioid misuse before spine surgery.

View details for DOI 10.14245/ns.2550394.197

View details for PubMedID 40625009

View details for DOI 10.1200/JCO.2025.43.16_suppl.1572

View details for Web of Science ID 001509282500001

Abstract

OBJECTIVES: This scoping review assesses machine learning (ML) tools that predicted falls, relying on information in health records without using any sensor data. The aim was to assess the available evidence on innovative techniques to improve fall prevention management.METHODS: Studies were included if they focused on predicting fall risk with machine learning in elderly populations and were written in English. There were 13 different extracted variables, including population characteristics (community dwelling, inpatients, age range, main pathology, ethnicity/race). Furthermore, the number of variables used in the final models, as well as their type, was extracted.RESULTS: A total of 6331 studies were retrieved, and 19 articles met criteria for data extraction. Metric performances reported by authors were commonly high in terms of accuracy (e.g., greater than 0.70). The most represented features included cardiovascular status and mobility assessments. Common gaps identified included a lack of transparent reporting and insufficient fairness assessments.CONCLUSIONS: This review provides evidence that falls can be predicted using ML without using sensors if the amount of data and its quality is adequate. However, further studies are needed to validate these models in diverse groups and populations.

View details for DOI 10.1186/s41512-025-00190-y

View details for PubMedID 40325490

Abstract

The Global Initiative on Artificial Intelligence for Health (GI-AI4H), established by the World Health Organization, serves to harmonize governance standards for artificial intelligence (AI). The GI-AI4H spearheads novel on-the-ground efforts, especially in low- and middle-income countries, to advance ethical, regulatory, implementation, and operational dimensions of global governance for health AI. The GI-AI4H's efforts across the United Nations drives safe, ethical, equitable, and sustainable health AI use for the global community.

View details for DOI 10.1038/s41746-025-01618-x

View details for PubMedID 40269244

View details for PubMedCentralID PMC12019307

Abstract

Artificial intelligence (AI) tools to support clinicians in providing patient-centered care can contribute to patient empowerment and care efficiency. We aimed to draft potential AI tools for tailored patient support corresponding to patients' needs and assess clinicians' perceptions about the usefulness of those AI tools. To define patients' issues, we analyzed 528,199 patient messages of 11,123 patients with diabetes by harnessing natural language processing and AI. Applying multiple prompt-engineering techniques, we drafted a series of AI tools, and five endocrinologists evaluated them for perceived usefulness and risk. Patient education and administrative support for timely and streamlined interaction were perceived as highly useful, yet deeper integration of AI tools into patient data was perceived as risky. This study proposes assorted AI applications as clinical assistance tailored to patients' needs substantiated by clinicians' evaluations. Findings could offer essential ramifications for developing potential AI tools for precision patient care for diabetes and beyond.

View details for DOI 10.1038/s41746-025-01604-3

View details for PubMedID 40240489

View details for PubMedCentralID 5069713

Abstract

Rates of oral anticoagulation (OAC) nonprescription in atrial fibrillation approach 50%. Understanding reasons for OAC nonprescription may reduce gaps in guideline-recommended care. We aimed to identify reasons for OAC nonprescription from clinical notes using large language models.We identified all patients and associated clinical notes in our health care system with a clinician-billed visit for atrial fibrillation without another indication for OAC and stratified them on the basis of active OAC prescriptions. Three annotators labeled reasons for OAC nonprescription in clinical notes on 10% of all patients ("annotation set"). We engineered prompts for a generative large language model (Generative Pre-trained Transformer 4) and trained a discriminative large language model (ClinicalBERT) to identify reasons for OAC nonprescription and selected the best-performing model to predict reasons for the remaining 90% of patients ("inference set").A total of 35 737 patients were identified, of which 7712 (21.6%) did not have active OAC prescriptions. A total of 910 notes across 771 patients were annotated. Generative Pre-trained Transformer 4 outperformed ClinicalBERT (macro-F1 score across all reasons of 0.79, compared with 0.69 for ClinicalBERT). Using Generative Pre-trained Transformer 4 on the inference set, 61.1% of notes had documented reasons for OAC nonprescription, most commonly the alternative use of an antiplatelet agent (23.3%), therapeutic inertia (21.0%), and low burden of atrial fibrillation (17.1%).This is the first study using large language models to extract documented reasons for OAC nonprescription from clinical notes in patients with atrial fibrillation and reveals guideline-discordant practices and actionable insights for the development of health system interventions to reduce OAC nonprescription.

View details for DOI 10.1161/JAHA.124.040419

View details for PubMedID 40145287

Abstract

Artificial Intelligence (AI)-based prediction models, including risk scoring systems and decision support systems, are increasingly adopted in healthcare. Addressing AI fairness is essential to fighting health disparities and achieving equitable performance and patient outcomes. Numerous and conflicting definitions of fairness complicate this effort. This paper aims to structure the transition of AI fairness from theory to practical application with appropriate fairness metrics. For 27 definitions of fairness identified in the recent literature, we assess the relation with the model's intended use, type of decision influenced and ethical principles of distributive justice. We advocate that due to limitations in some notions of fairness, clinical utility, performance-based metrics (area under the receiver operating characteristic curve), calibration, and statistical parity are the most relevant group-based metrics for medical applications. Through two use cases, we demonstrate that different metrics may be applicable depending on the intended use and ethical framework. Our approach provides a foundation for AI developers and assessors by assessing model fairness and the impact of bias mitigation strategies, hence promoting more equitable AI-based implementations.

View details for DOI 10.1101/2025.03.24.25324500

View details for PubMedID 40196288

Abstract

With rapidly evolving artificial intelligence solutions, healthcare organizations need an implementation roadmap. A "clinical trials" informed approach can promote safe and impactful implementation of artificial intelligence. This framework includes four phases: (1) Safety; (2) Efficacy; (3) Effectiveness and comparison to an existing standard; and (4) Monitoring. Combined with inter-institutional collaboration and national funding support, this approach will advance safe, usable, effective, and equitable deployments of artificial intelligence in healthcare.

View details for DOI 10.1038/s41746-025-01506-4

View details for PubMedID 39962232

View details for PubMedCentralID 7518439

Abstract

Pressure injuries (PIs) pose a substantial healthcare burden and incur significant costs worldwide. Several risk prediction tools to allow timely implementation of preventive measures and a subsequent reduction in healthcare system burden are available and in use. The ability of risk prediction tools to correctly identify those at high risk of PI (prognostic accuracy) and to have a clinically significant impact on patient management and outcomes (effectiveness) is not clear. We aimed to evaluate the prognostic accuracy and clinical effectiveness of risk prediction tools for PI and to identify gaps in the literature.The umbrella review was conducted according to Cochrane guidance. Systematic reviews (SRs) evaluating the accuracy or clinical effectiveness of adult PI risk prediction tools in any clinical settings were eligible. Studies on paediatric tools, sensor-only tools, or staging/diagnosis of existing PIs were excluded. MEDLINE, Embase, CINAHL, and EPISTEMONIKOS were searched (inception to June 2024) to identify relevant SRs, as well as Google Scholar (2013 to 2024) and reference lists. Methodological quality was assessed using adapted AMSTAR-2 criteria. Results were described narratively. We identified 26 SRs meeting all eligibility criteria with 19 SRs assessing prognostic accuracy and 11 assessing clinical effectiveness of risk prediction tools for PI (4 SRs assessed both aspects). The 19 SRs of prognostic accuracy evaluated 70 tools (39 scales and 31 machine learning (ML) models), with the Braden, Norton, Waterlow, Cubbin-Jackson scales (and modifications thereof) the most evaluated tools. Meta-analyses from a focused set of included SRs showed that the scales had sensitivities and specificities ranging from 53% to 97% and 46% to 84%, respectively. Only 2/19 (11%) SRs performed appropriate statistical synthesis and quality assessment. Two SRs assessing machine learning-based algorithms reported high prognostic accuracy estimates, but some of which were sourced from the same data within which the models were developed, leading to potentially overoptimistic results. Two randomised trials assessing the effect of PI risk assessment tools (within the full test-intervention-outcome pathway) on the incidence of PIs were identified from the 11 SRs of clinical effectiveness; both were included in a Cochrane SR and assessed as high risk of bias. Both trials found no evidence of an effect on PI incidence. Limitations included the use of the AMSTAR-2 criteria, which may have overly focused on reporting quality rather than methodological quality, compounded by the poor reporting quality of included SRs and that SRs were not excluded based on low AMSTAR-2 ratings (in order to provide a comprehensive overview). Additionally, diagnostic test accuracy principles, rather than prognostic modelling approaches were heavily relied upon, which do not account for the temporal nature of prediction.Available systematic reviews suggest a lack of high-quality evidence for the accuracy of risk prediction tools for PI and limited reliable evidence for their use leading to a reduction in incidence of PI. Further research is needed to establish the clinical effectiveness of appropriately developed and validated risk prediction tools for PI.

View details for DOI 10.1371/journal.pmed.1004518

View details for PubMedID 39913541

Abstract

Population ageing and the rise in chronic diseases place continual stress on healthcare systems. Scarce resources often impede equitable access to healthcare, particularly in rural areas, resulting in prolonged waiting times and heightened risks of morbidity and mortality. Telemedicine has emerged as a promising solution, offering remote and equitable care that could potentially bridge access gaps and enhance health outcomes. This systematic review aims to quantitatively examine the impact of telemedicine implementation on waiting times, defined as the time passed from the booking of a visit for an outpatient to the administration of the service.A systematic review was conducted using studies on telemedicine interventions that specifically addressed waiting times. Bias assessment was performed with three tools: ROBINS-I ("Risk of Bias In Non-Randomized Studies of Interventions"), AXIS ("Appraisal tool for Cross-Sectional Studies") and RoB-2 ("Risk of Bias-2"). A weighted mean approach was used to synthesise results, with medians synthesised using a median approach.Articles in English were retrieved from the PubMed and Scopus databases.Studies were excluded if they did not specifically address waiting times related to telemedicine interventions. Only studies that considered waiting times defined as the time passed from the booking of a visit for an outpatient to the administration of the service and any telemedicine intervention were included.A total of 53 records were included, encompassing 270 388 patients in both the experimental and control groups. The weighted mean reduction in waiting times was calculated, and bias was assessed. No record was evaluated to be at high risk of bias, with 69.8% of studies evaluated at low risk and 26.4% at moderate risk (3.8% were surveys). Results were synthesised using a weighted mean approach for studies reporting means, and a median approach for studies reporting medians.Overall, a weighted mean reduction of 25.4 days in waiting times was observed. Focusing on clinical specialties (n=114 042), the weighted mean reduction amounted to 34.7 days, while in surgical patients (n=156 346), telemedicine was associated with a weighted mean of 17.3 days saved.The implementation of telemedicine solutions may significantly improve waiting times, potentially leading to more efficient and equitable healthcare systems.CRD42023490822.

View details for DOI 10.1136/bmjopen-2024-088153

View details for PubMedID 39884707

View details for PubMedCentralID PMC11784372

Abstract

For studies using real-world data, accurately identifying patients with phenotypes of interest is challenging. To identify cohorts of interest, most studies exclusively use the International Classification of Diseases (ICD) billing codes, which can be limiting. We developed a method to accurately identify the presence or absence of 3 common ocular diseases (diabetic retinopathy [DR], age-related macular degeneration [AMD], and glaucoma) using electronic health record (EHR) data.Database study.Three thousand nine hundred fourteen eyes from 1957 patients at 2 Sight OUtcomes Research CollaborativE (SOURCE) Ophthalmology Data Repository sites.We developed enhanced phenotype identification (EPI) algorithms that search EHR fields, including eye examination findings, orders, charges, medication prescriptions, and surgery data for evidence that a patient has glaucoma, DR, or AMD. We trained our EPI models using gold standard assessments of the EHR by ophthalmologists for the presence/absence of these conditions, compared the performance of our EPI models to models developed using ICD codes alone, and validated the performance of model using data from another SOURCE site.Area under the receiver operating curve (AUC), area under the precision-recall curve (AUPRC), and model calibration.The AUCs of our EPI models were better than ICD-only models for glaucoma (0.97 vs. 0.90), DR (0.997 vs. 0.98), and AMD (0.99 vs. 0.95). The AUPRCs of our EPI models were also much better than ICD-only models for glaucoma (0.79 vs. 0.32), DR (0.96 vs. 0.84), and AMD (0.74 vs. 0.55). When testing on patients from a second SOURCE site, the AUC and AUPRC for glaucoma (0.93, 0.74), DR (0.98, 0.77), and AMD (0.96, 0.64) were slightly worse than the primary site but still quite high. However, for all 3 conditions, model calibration was worse at the second site.Leveraging machine learning, we developed EPI models to accurately identify most patients with glaucoma, DR, and AMD in real-world datasets. The EPI models significantly outperform ICD-only models in identifying patients confirmed to have these conditions. These findings underscore the potential of using comprehensive EHR data combined with advanced machine learning techniques to improve the accuracy of patient phenotype identification, leading to better patient management and clinical outcomes.Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

View details for DOI 10.1016/j.xops.2025.100717

View details for PubMedID 40212931

View details for PubMedCentralID PMC11985028

Abstract

BACKGROUND: Generative AI, particularly large language models (LLMs), holds great potential for improving patient care and operational efficiency in healthcare. However, the use of LLMs is complicated by regulatory concerns around data security and patient privacy. This study aimed to develop and evaluate a secure infrastructure that allows researchers to safely leverage LLMs in healthcare while ensuring HIPAA compliance and promoting equitable AI.MATERIALS AND METHODS: We implemented a private Azure OpenAI Studio deployment with secure API-enabled endpoints for researchers. Two use cases were explored, detecting falls from electronic health records (EHR) notes and evaluating bias in mental health prediction using fairness-aware prompts.RESULTS: The framework provided secure, HIPAA-compliant API access to LLMs, allowing researchers to handle sensitive data safely. Both use cases highlighted the secure infrastructure's capacity to protect sensitive patient data while supporting innovation.DISCUSSION AND CONCLUSION: This centralized platform presents a scalable, secure, and HIPAA-compliant solution for healthcare institutions aiming to integrate LLMs into clinical research.

View details for DOI 10.1093/jamia/ocaf005

View details for PubMedID 39836496

Abstract

BACKGROUND: Pressure injuries (PIs) place a substantial burden on healthcare systems worldwide. Risk stratification of those who are at risk of developing PIs allows preventive interventions to be focused on patients who are at the highest risk. The considerable number of risk assessment scales and prediction models available underscores the need for a thorough evaluation of their development, validation, and clinical utility. Our objectives were to identify and describe available risk prediction tools for PI occurrence, their content and the development and validation methods used.METHODS: The umbrella review was conducted according to Cochrane guidance. MEDLINE, Embase, CINAHL, EPISTEMONIKOS, Google Scholar, and reference lists were searched to identify relevant systematic reviews. The risk of bias was assessed using adapted AMSTAR-2 criteria. Results were described narratively. All included reviews contributed to building a comprehensive list of risk prediction tools.RESULTS: We identified 32 eligible systematic reviews only seven of which described the development and validation of risk prediction tools for PI. Nineteen reviews assessed the prognostic accuracy of the tools and 11 assessed clinical effectiveness. Of the seven reviews reporting model development and validation, six included only machine learning models. Two reviews included external validations of models, although only one review reported any details on external validation methods or results. This was also the only review to report measures of both discrimination and calibration. Five reviews presented measures of discrimination, such as the area under the curve (AUC), sensitivities, specificities, F1 scores, and G-means. For the four reviews that assessed the risk of bias assessment using the PROBAST tool, all models but one were found to be at high or unclear risk of bias.CONCLUSIONS: Available tools do not meet current standards for the development or reporting of risk prediction models. The majority of tools have not been externally validated. Standardised and rigorous approaches to risk prediction model development and validation are needed.TRIAL REGISTRATION: The protocol was registered on the Open Science Framework ( https://osf.io/tepyk ).

View details for DOI 10.1186/s41512-024-00182-4

View details for PubMedID 39806510

View details for DOI 10.1016/j.gastha.2024.10.003

View details for Web of Science ID 001398433800001

Abstract

American Indian and Alaska Native (AI/AN) communities are at a critical juncture in health research, where combining participatory methods with advancements in artificial intelligence and machine learning (AI/ML) can promote equity. Community-based participatory research methods which emerged to help Alaska Native communities navigate the complicated legacy of historical research abuses provide a framework to allow emerging AI/ML technologies to align with their unique world views, community strengths, and healthcare goals. A consortium of researchers (including Alaska Native Tribal Health Consortium, the Center for Alaska Native Health Research at University of Alaska, Fairbanks, Stanford University, Southcentral Foundation, and Maniilaq Association) is using community-engaged AI/ML methods to address air medical ambulance (medevac) utilization in rural communities within the Alaska Tribal Health System (ATHS). This mixed-methods convergent triangulation study uses qualitative and quantitative analyses to develop AI/ML models tailored to community needs, provider concerns, and cultural contexts. Early successes have led to a second funded project to expand community perspectives, pilot models, and address issues of governance and ethics. Using the Ethical, Legal, and Social Implications of Research framework to address implementation of AI/ML in AI/AN communities, this second grant expands community engagement, technical capacity, and creates a body within the ATHS able to provide recommendations about AI/ML security, privacy, governance and policy. These two projects have the potential to provide equitable AI/ML implementation in Alaska Native healthcare and provide a roadmap for researchers and policy makers looking to effect similar change in other AI/AN and marginalized communities.

View details for DOI 10.3389/frai.2025.1568886

View details for PubMedID 40260415

View details for DOI 10.1136/emermed-2024-214444

View details for PubMedID 39658217

Abstract

Informal caregivers of people with Alzheimer's disease and related dementias (ADRD) are at risk of poor mental health. This study aimed to investigate the feasibility and validity of studying caregivers' mental stressors using online caregiving forum data (March 2018-February 2022) and natural language processing and machine learning (NLP/ML). NLP/ML topic modeling generated eight prominent topics, which we compared with qualitatively defined themes and the existing caregiving framework to assess validity. Among a total of 60,182 posts, 5848 were mental distress-related; for the ADRD patients (symptoms, medication, relocation, care duty share, diagnosis, conversation strategy) and the caregivers (caregiving burden and support). While we observed novel topics from NLP/ML-defined topics, mostly those were aligned with the existing framework. For feasibility assessment, qualitative title screening was done. The findings shed new light on the potential of NLP/ML text analysis of the online forum for informal caregivers to prepare tailored support for this vulnerable population.

View details for DOI 10.1038/s44184-024-00100-y

View details for PubMedID 39537826

View details for PubMedCentralID 9047171

View details for DOI 10.1136/bmjhci-2024-101101

View details for PubMedID 39496359

Abstract

The increasing interest in leveraging generative AI models in healthcare necessitates secure infrastructure at academic medical centers. Without an all-encompassing secure system, researchers may create their own insecure microprocesses, risking the exposure of protected health information (PHI) to the public internet or its inadvertent incorporation into AI model training. To address these challenges, our institution implemented a secure pathway to the Azure OpenAI Service using our own private OpenAI instance which we fully control to facilitate high-throughput, secure LLM queries. This pathway ensures data privacy while allowing researchers to harness the capabilities of LLMs for diverse healthcare applications. Our approach supports compliant, efficient, and innovative AI research in healthcare. This paper discusses the implementation, advantages, and use cases of this secure infrastructure, underscoring the critical need for centralized, secure AI solutions in academic medical environments.

View details for DOI 10.21203/rs.3.rs-5095287/v1

View details for PubMedID 39399679

View details for DOI 10.1038/s41588-024-01903-7

View details for PubMedID 39251787

View details for PubMedCentralID 6785182

Abstract

Patient-reported outcomes (PROs) are vital in assessing disease activity and treatment outcomes in inflammatory bowel disease (IBD). However, manual extraction of these PROs from the free-text of clinical notes is burdensome. We aimed to improve data curation from free-text information in the electronic health record, making it more available for research and quality improvement. This study aimed to compare traditional natural language processing (tNLP) and large language models (LLMs) in extracting three IBD PROs (abdominal pain, diarrhea, fecal blood) from clinical notes across two institutions.Clinic notes were annotated for each PRO using preset protocols. Models were developed and internally tested at the University of California San Francisco (UCSF), and then externally validated at Stanford University. We compared tNLP and LLM-based models on accuracy, sensitivity, specificity, positive and negative predictive value. Additionally, we conducted fairness and error assessments.Inter-rater reliability between annotators was >90%. On the UCSF test set (n=50), the top-performing tNLP models showcased accuracies of 92% (abdominal pain), 82% (diarrhea) and 80% (fecal blood), comparable to GPT-4, which was 96%, 88%, and 90% accurate, respectively. On external validation at Stanford (n=250), tNLP models failed to generalize (61-62% accuracy) while GPT-4 maintained accuracies >90%. PaLM-2 and GPT-4 showed similar performance. No biases were detected based on demographics or diagnosis.LLMs are accurate and generalizable methods for extracting PROs. They maintain excellent accuracy across institutions, despite heterogeneity in note templates and authors. Widespread adoption of such tools has the potential to enhance IBD research and patient care.

View details for DOI 10.1101/2024.09.05.24313139

View details for PubMedID 39281744

View details for PubMedCentralID PMC11398594

Abstract

This study aims to explore and develop tools for early identification of depression concerns among cancer patients by leveraging the novel data source of messages sent through a secure patient portal.We developed classifiers based on logistic regression (LR), support vector machines (SVMs), and 2 Bidirectional Encoder Representations from Transformers (BERT) models (original and Reddit-pretrained) on 6600 patient messages from a cancer center (2009-2022), annotated by a panel of healthcare professionals. Performance was compared using AUROC scores, and model fairness and explainability were examined. We also examined correlations between model predictions and depression diagnosis and treatment.BERT and RedditBERT attained AUROC scores of 0.88 and 0.86, respectively, compared to 0.79 for LR and 0.83 for SVM. BERT showed bigger differences in performance across sex, race, and ethnicity than RedditBERT. Patients who sent messages classified as concerning had a higher chance of receiving a depression diagnosis, a prescription for antidepressants, or a referral to the psycho-oncologist. Explanations from BERT and RedditBERT differed, with no clear preference from annotators.We show the potential of BERT and RedditBERT in identifying depression concerns in messages from cancer patients. Performance disparities across demographic groups highlight the need for careful consideration of potential biases. Further research is needed to address biases, evaluate real-world impacts, and ensure responsible integration into clinical settings.This work represents a significant methodological advancement in the early identification of depression concerns among cancer patients. Our work contributes to a route to reduce clinical burden while enhancing overall patient care, leveraging BERT-based models.

View details for DOI 10.1093/jamia/ocae188

View details for PubMedID 39018490

Abstract

The prevalence of obesity has been increasing worldwide, with substantial implications for public health. Obesity is independently associated with cardiovascular morbidity and mortality and is estimated to cost the health system over

Abstract

Postsurgical falls have significant patient and societal implications but remain challenging to identify and track. Detecting postsurgical falls is crucial to improve patient care for older adults and reduce healthcare costs. Large language models (LLMs) offer a promising solution for reliable and automated fall detection using unstructured data in clinical notes. We tested several LLM prompting approaches to postsurgical fall detection in two different healthcare systems with three open-source LLMs. The Mixtral-8*7B zero-shot had the best performance at Stanford Health Care (PPV = 0.81, recall = 0.67) and the Veterans Health Administration (PPV = 0.93, recall = 0.94). These results demonstrate that LLMs can detect falls with little to no guidance and lay groundwork for applications of LLMs in fall prediction and prevention across many different settings.

View details for DOI 10.1101/2024.06.25.24309480

View details for PubMedID 38978655

Abstract

Spinal CSF leaks lead to spontaneous intracranial hypotension (SIH). While International Classification of Headache Disorders, Third Edition (ICHD-3) criteria necessitate imaging confirmation or low opening pressure (OP) for SIH diagnosis, their sensitivity may be limited. We offered epidural blood patches (EBPs) to patients with symptoms suggestive of SIH, with and without a documented low OP or confirmed leak on imaging. This study evaluates the efficacy of this strategy.We conducted a prospective cohort study with a nested case-control design including all patients who presented to a tertiary headache clinic with clinical symptoms of SIH who completed study measures both before and after receiving an EBP between August 2016 and November 2018.The mean duration of symptoms was 8.7 ± 8.1 years. Of 85 patients assessed, 69 did not meet ICHD-3 criteria for SIH. At an average of 521 days after the initial EBP, this ICHD-3-negative subgroup experienced significant improvements in Patient-Reported Outcomes Measurement Information System (PROMIS) Global Physical Health score of +3.3 (95% CI 1.5-5.1), PROMIS Global Mental Health score of +1.8 (95% CI 0.0-3.5), Headache Impact Test (HIT)-6 head pain score of -3.8 (95% CI -5.7 to -1.8), Neck Disability Index of -4.8 (95% CI -9.0 to -0.6) and PROMIS Fatigue of -2.3 (95% CI -4.1 to -0.6). Fifty-four percent of ICHD-3-negative patients achieved clinically meaningful improvements in PROMIS Global Physical Health and 45% in HIT-6 scores. Pain relief following lying flat prior to treatment was strongly associated with sustained clinically meaningful improvement in global physical health at an average of 521 days (odds ratio 1.39, 95% CI 1.1-1.79; p < 0.003). ICHD-3-positive patients showed high rates of response and previously unreported, treatable levels of fatigue and cognitive deficits.Patients who did not conform to the ICHD-3 criteria for SIH showed moderate rates of sustained, clinically meaningful improvements in global physical health, global mental health, neck pain, fatigue, and head pain after EBP therapy. Pre-treatment improvement in head pain when flat was associated with later, sustained improvement after EBP therapy among patients who did not meet the ICHD-3 criteria.This study provides Class IV evidence that epidural blood patch is an effective treatment of suspected CSF leak not conforming to ICHD-3 criteria for SIH.

View details for DOI 10.1212/WNL.0000000000209449

View details for PubMedID 38820488

Abstract

Opioid use disorder (OUD) and opioid dependence lead to significant morbidity and mortality, yet treatment retention, crucial for the effectiveness of medications like buprenorphine-naloxone, remains unpredictable. Our objective was to determine the predictability of 6-month retention in buprenorphine-naloxone treatment using electronic health record (EHR) data from diverse clinical settings and to identify key predictors.This retrospective observational study developed and validated machine learning-based clinical risk prediction models using EHR data.Data were sourced from Stanford University's healthcare system and Holmusk's NeuroBlu database, reflecting a wide range of healthcare settings. The study analyzed 1800 Stanford and 7957 NeuroBlu treatment encounters from 2008 to 2023 and from 2003 to 2023, respectively.Predict continuous prescription of buprenorphine-naloxone for at least 6 months, without a gap of more than 30 days. The performance of machine learning prediction models was assessed by area under receiver operating characteristic (ROC-AUC) analysis as well as precision, recall and calibration. To further validate our approach's clinical applicability, we conducted two secondary analyses: a time-to-event analysis on a single site to estimate the duration of buprenorphine-naloxone treatment continuity evaluated by the C-index and a comparative evaluation against predictions made by three human clinical experts.Attrition rates at 6 months were 58% (NeuroBlu) and 61% (Stanford). Prediction models trained and internally validated on NeuroBlu data achieved ROC-AUCs up to 75.8 (95% confidence interval [CI] = 73.6-78.0). Addiction medicine specialists' predictions show a ROC-AUC of 67.8 (95% CI = 50.4-85.2). Time-to-event analysis on Stanford data indicated a median treatment retention time of 65 days, with random survival forest model achieving an average C-index of 65.9. The top predictor of treatment retention identified included the diagnosis of opioid dependence.US patients with opioid use disorder or opioid dependence treated with buprenorphine-naloxone prescriptions appear to have a high (∼60%) treatment attrition by 6 months. Machine learning models trained on diverse electronic health record datasets appear to be able to predict treatment continuity with accuracy comparable to that of clinical experts.

View details for DOI 10.1111/add.16587

View details for PubMedID 38923168

Abstract

BACKGROUND: Timely heart failure (HF) diagnosis can lead to earlier intervention and reduced morbidity. Among historically marginalized patients, new-onset HF diagnosis is more likely to occur in acute care settings (emergency department or inpatient hospitalization) than outpatient settings. Whether inequity within outpatient clinician practices affects diagnosis settings is unknown.METHODS: We determined the setting of incident HF diagnosis among Medicare fee-for-service beneficiaries between 2013 and 2017. We identified sociodemographic and medical characteristics associated with HF diagnosis in the acute care setting. Within each outpatient clinician practice, we compared acute care diagnosis rates across sociodemographic characteristics: female versus male sex, non-Hispanic White versus other racial and ethnic groups, and dual Medicare-Medicaid eligible (a surrogate for low income) versus nondual-eligible patients. Based on within-practice differences in acute diagnosis rates, we stratified clinician practices by equity (high, intermediate, and low) and compared clinician practice characteristics.RESULTS: Among 315 439 Medicare patients with incident HF, 173 121 (54.9%) were first diagnosed in acute care settings. Higher adjusted acute care diagnosis rates were associated with female sex (6.4% [95% CI, 6.1%-6.8%]), American Indian (3.6% [95% CI, 1.1%-6.1%]) race, and dual eligibility (4.1% [95% CI, 3.7%-4.5%]). These differences persisted within clinician practices. With clinician practice adjustment, dual-eligible patients had a 4.9% (95% CI, 4.5%-5.4%) greater acute care diagnosis rate than nondual-eligible patients. Clinician practices with greater equity across dual eligibility also had greater equity across sex and race and ethnicity and were more likely to be composed of predominantly primary care clinicians.CONCLUSIONS: Differences in HF diagnosis rates in the acute care setting between and within clinician practices highlight an opportunity to improve equity in diagnosing historically marginalized patients.

View details for DOI 10.1161/CIRCHEARTFAILURE.123.010718

View details for PubMedID 38847082

View details for Web of Science ID 001313316209005

View details for Web of Science ID 001277006602343

View details for Web of Science ID 001313316201085

Abstract

Coronary artery calcium (CAC) is a powerful tool to refine atherosclerotic cardiovascular disease (ASCVD) risk assessment. Despite its growing interest, contemporary public attitudes around CAC are not well-described in literature and have important implications for shared decision-making around cardiovascular prevention. We used an artificial intelligence (AI) pipeline consisting of a semi-supervised natural language processing model and unsupervised machine learning techniques to analyze 5,606 CAC-related discussions on Reddit. A total of 91 discussion topics were identified and were classified into 14 overarching thematic groups. These included the strong impact of CAC on therapeutic decision-making, ongoing non-evidence-based use of CAC testing, and the patient perceived downsides of CAC testing (e.g., radiation risk). Sentiment analysis also revealed that most discussions had a neutral (49.5%) or negative (48.4%) sentiment. The results of this study demonstrate the potential of an AI-based approach to analyze large, publicly available social media data to generate insights into public perceptions about CAC, which may help guide strategies to improve shared decision-making around ASCVD management and public health interventions.

View details for DOI 10.1038/s41746-024-01077-w

View details for PubMedID 38555387

View details for PubMedCentralID PMC10981728

Abstract

Predictive models show promise in healthcare, but their successful deployment is challenging due to limited generalizability. Current external validation often focuses on model performance with restricted feature use from the original training data, lacking insights into their suitability at external sites. Our study introduces an innovative methodology for evaluating features during both the development phase and the validation, focusing on creating and validating predictive models for post-surgery patient outcomes with improved generalizability.Electronic health records (EHRs) from 4 countries (United States, United Kingdom, Finland, and Korea) were mapped to the OMOP Common Data Model (CDM), 2008-2019. Machine learning (ML) models were developed to predict post-surgery prolonged opioid use (POU) risks using data collected 6 months before surgery. Both local and cross-site feature selection methods were applied in the development and external validation datasets. Models were developed using Observational Health Data Sciences and Informatics (OHDSI) tools and validated on separate patient cohorts.Model development included 41 929 patients, 14.6% with POU. The external validation included 31 932 (UK), 23 100 (US), 7295 (Korea), and 3934 (Finland) patients with POU of 44.2%, 22.0%, 15.8%, and 21.8%, respectively. The top-performing model, Lasso logistic regression, achieved an area under the receiver operating characteristic curve (AUROC) of 0.75 during local validation and 0.69 (SD = 0.02) (averaged) in external validation. Models trained with cross-site feature selection significantly outperformed those using only features from the development site through external validation (P < .05).Using EHRs across four countries mapped to the OMOP CDM, we developed generalizable predictive models for POU. Our approach demonstrates the significant impact of cross-site feature selection in improving model performance, underscoring the importance of incorporating diverse feature sets from various clinical settings to enhance the generalizability and utility of predictive healthcare models.

View details for DOI 10.1093/jamia/ocae028

View details for PubMedID 38412331

Abstract

To measure pediatrician adherence to evidence-based guidelines in the treatment of young children with attention-deficit/hyperactivity disorder (ADHD) in a diverse healthcare system using natural language processing (NLP) techniques.We extracted structured and free-text data from electronic health records (EHRs) of all office visits (2015-2019) of children aged 4-6 years in a community-based primary healthcare network in California, who had ≥1 visits with an ICD-10 diagnosis of ADHD. Two pediatricians annotated clinical notes of the first ADHD visit for 423 patients. Inter-annotator agreement (IAA) was assessed for the recommendation for the first-line behavioral treatment (F-measure = 0.89). Four pre-trained language models, including BioClinical Bidirectional Encoder Representations from Transformers (BioClinicalBERT), were used to identify behavioral treatment recommendations using a 70/30 train/test split. For temporal validation, we deployed BioClinicalBERT on 1,020 unannotated notes from other ADHD visits and well-care visits; all positively classified notes (n = 53) and 5% of negatively classified notes (n = 50) were manually reviewed.Of 423 patients, 313 (74%) were male; 298 (70%) were privately insured; 138 (33%) were White; 61 (14%) were Hispanic. The BioClinicalBERT model trained on the first ADHD visits achieved F1 = 0.76, precision = 0.81, recall = 0.72, and AUC = 0.81 [0.72-0.89]. Temporal validation achieved F1 = 0.77, precision = 0.68, and recall = 0.88. Fairness analysis revealed low model performance in publicly insured patients (F1 = 0.53).Deploying pre-trained language models on a variable set of clinical notes accurately captured pediatrician adherence to guidelines in the treatment of children with ADHD. Validating this approach in other patient populations is needed to achieve equitable measurement of quality of care at scale and improve clinical care for mental health conditions.

View details for DOI 10.1093/jamia/ocae001

View details for PubMedID 38244997

Abstract

BACKGROUND: Patients with cancer starting systemic treatment programs, such as chemotherapy, often develop depression. A prediction model may assist physicians and health care workers in the early identification of these vulnerable patients.OBJECTIVE: This study aimed to develop a prediction model for depression risk within the first month of cancer treatment.METHODS: We included 16,159 patients diagnosed with cancer starting chemo- or radiotherapy treatment between 2008 and 2021. Machine learning models (eg, least absolute shrinkage and selection operator [LASSO] logistic regression) and natural language processing models (Bidirectional Encoder Representations from Transformers [BERT]) were used to develop multimodal prediction models using both electronic health record data and unstructured text (patient emails and clinician notes). Model performance was assessed in an independent test set (n=5387, 33%) using area under the receiver operating characteristic curve (AUROC), calibration curves, and decision curve analysis to assess initial clinical impact use.RESULTS: Among 16,159 patients, 437 (2.7%) received a depression diagnosis within the first month of treatment. The LASSO logistic regression models based on the structured data (AUROC 0.74, 95% CI 0.71-0.78) and structured data with email classification scores (AUROC 0.74, 95% CI 0.71-0.78) had the best discriminative performance. The BERT models based on clinician notes and structured data with email classification scores had AUROCs around 0.71. The logistic regression model based on email classification scores alone performed poorly (AUROC 0.54, 95% CI 0.52-0.56), and the model based solely on clinician notes had the worst performance (AUROC 0.50, 95% CI 0.49-0.52). Calibration was good for the logistic regression models, whereas the BERT models produced overly extreme risk estimates even after recalibration. There was a small range of decision thresholds for which the best-performing model showed promising clinical effectiveness use. The risks were underestimated for female and Black patients.CONCLUSIONS: The results demonstrated the potential and limitations of machine learning and multimodal models for predicting depression risk in patients with cancer. Future research is needed to further validate these models, refine the outcome label and predictors related to mental health, and address biases across subgroups.

View details for DOI 10.2196/51925

View details for PubMedID 38236635

Abstract

Assessment in medical education has evolved through a sequence of eras each centering on distinct views and values. These eras include measurement (e.g., knowledge exams, objective structured clinical examinations), then judgments (e.g., workplace-based assessments, entrustable professional activities), and most recently systems or programmatic assessment, where over time multiple types and sources of data are collected and combined by competency committees to ensure individual learners are ready to progress to the next stage in their training. Significantly less attention has been paid to the social context of assessment, which has led to an overall erosion of trust in assessment by a variety of stakeholders including learners and frontline assessors. To meaningfully move forward, the authors assert that the reestablishment of trust should be foundational to the next era of assessment. In our actions and interventions, it is imperative that medical education leaders address and build trust in assessment at a systems level. To that end, the authors first review tenets on the social contextualization of assessment and its linkage to trust and discuss consequences should the current state of low trust continue. The authors then posit that trusting and trustworthy relationships can exist at individual as well as organizational and systems levels. Finally, the authors propose a framework to build trust at multiple levels in a future assessment system; one that invites and supports professional and human growth and has the potential to position assessment as a fundamental component of renegotiating the social contract between medical education and the health of the public.

View details for DOI 10.5334/pme.1110

View details for PubMedID 38274558

View details for Web of Science ID 001483857800034

Abstract

The lack of data quality frameworks to guide the development of artificial intelligence (AI)-ready data sets limits their usefulness for machine learning (ML) research in health care and hinders the diagnostic excellence of developed clinical AI applications for patient care.To discern what constitutes high-quality and useful data sets for health and biomedical ML research purposes according to subject matter experts.This qualitative study interviewed data set experts, particularly those who are creators and ML researchers. Semistructured interviews were conducted in English and remotely through a secure video conferencing platform between August 23, 2022, and January 5, 2023. A total of 93 experts were invited to participate. Twenty experts were enrolled and interviewed. Using purposive sampling, experts were affiliated with a diverse representation of 16 health data sets/databases across organizational sectors. Content analysis was used to evaluate survey information and thematic analysis was used to analyze interview data.Data set experts' perceptions on what makes data sets AI ready.Participants included 20 data set experts (11 [55%] men; mean [SD] age, 42 [11] years), of whom all were health data set creators, and 18 of the 20 were also ML researchers. Themes (3 main and 11 subthemes) were identified and integrated into an AI-readiness framework to show their association within the health data ecosystem. Participants partially determined the AI readiness of data sets using priority appraisal elements of accuracy, completeness, consistency, and fitness. Ethical acquisition and societal impact emerged as appraisal considerations in that participant samples have not been described to date in prior data quality frameworks. Factors that drive creation of high-quality health data sets and mitigate risks associated with data reuse in ML research were also relevant to AI readiness. The state of data availability, data quality standards, documentation, team science, and incentivization were associated with elements of AI readiness and the overall perception of data set usefulness.In this qualitative study of data set experts, participants contributed to the development of a grounded framework for AI data set quality. Data set AI readiness required the concerted appraisal of many elements and the balancing of transparency and ethical reflection against pragmatic constraints. The movement toward more reliable, relevant, and ethical AI and ML applications for patient care will inevitably require strategic updates to data set creation practices.

View details for DOI 10.1001/jamanetworkopen.2023.45892

View details for PubMedID 38039004

Abstract

Limited sharing of data sets that accurately represent disease and patient diversity limits the generalizability of artificial intelligence (AI) algorithms in health care.To explore the factors associated with organizational motivation to share health data for AI development.This qualitative study investigated organizational readiness for sharing health data across the academic, governmental, nonprofit, and private sectors. Using a multiple case studies approach, 27 semistructured interviews were conducted with leaders in data-sharing roles from August 29, 2022, to January 9, 2023. The interviews were conducted in the English language using a video conferencing platform. Using a purposive and nonprobabilistic sampling strategy, 78 individuals across 52 unique organizations were identified. Of these, 35 participants were enrolled. Participant recruitment concluded after 27 interviews, as theoretical saturation was reached and no additional themes emerged.Concepts defining organizational readiness for data sharing and the association between data-sharing factors and organizational behavior were mapped through iterative qualitative analysis to establish a framework defining organizational readiness for sharing clinical data for AI development.Interviews included 27 leaders from 18 organizations (academia: 10, government: 7, nonprofit: 8, and private: 2). Organizational readiness for data sharing centered around 2 main constructs: motivation and capabilities. Motivation related to the alignment of an organization's values with data-sharing priorities and was associated with its engagement in data-sharing efforts. However, organizational motivation could be modulated by extrinsic incentives for financial or reputational gains. Organizational capabilities comprised infrastructure, people, expertise, and access to data. Cross-sector collaboration was a key strategy to mitigate barriers to access health data.This qualitative study identified sector-specific factors that may affect the data-sharing behaviors of health organizations. External incentives may bolster cross-sector collaborations by helping overcome barriers to accessing health data for AI development. The findings suggest that tailored incentives may boost organizational motivation and facilitate sustainable flow of health data for AI development.

View details for DOI 10.1001/jamanetworkopen.2023.48422

View details for PubMedID 38113040

Abstract

Seasonal influenza is a leading cause of death in the U.S., causing significant morbidity, mortality, and economic burden. Despite the proven efficacy of vaccinations, rates remain notably low, especially among Medicaid enrollees. Leveraging Medicaid claims data, this study characterizes influenza vaccination rates among Medicaid enrollees and aims to elucidate factors influencing vaccine uptake, providing insights that might also be applicable to other vaccine-preventable diseases, including COVID-19. This study used Medicaid claims data from nine U.S. states (2016-2021], encompassing three types of claims: fee-for-service, major Medicaid managed care plan, and combined. We included Medicaid enrollees who had an in-person healthcare encounter during an influenza season in this period, excluding those under 6 months of age, over 65 years, or having telehealth-only encounters. Vaccination was the primary outcome, with secondary outcomes involving in-person healthcare encounters. Chi-square tests, multivariable logistic regression, and Fisher's exact test were utilized for statistical analysis. A total of 20,868,910 enrollees with at least one healthcare encounter in at least one influenza season were included in the study population between 2016 and 2021. Overall, 15% (N = 3,050,471) of enrollees received an influenza vaccine between 2016 and 2021. During peri-COVID periods, there was an increase in vaccination rates among enrollees compared to pre-COVID periods, from 14% to 16%. Children had the highest influenza vaccination rates among all age groups at 29%, whereas only 17% were of 5-17 years, and 10% were of the 18-64 years were vaccinated. We observed differences in the likelihood of receiving the influenza vaccine among enrollees based on their health conditions and medical encounters. In a study of Medicaid enrollees across nine states, 15% received an influenza vaccine from July 2016 to June 2021. Vaccination rates rose annually, peaking during peri-COVID seasons. The highest uptake was among children (6 months-4 years), and the lowest was in adults (18-64 years). Female gender, urban residency, and Medicaid-managed care affiliation positively influenced uptake. However, mental health and substance abuse disorders decreased the likelihood. This study, reliant on Medicaid claims data, underscores the need for outreach services.

View details for DOI 10.3390/vaccines11111712

View details for PubMedID 38006044

Abstract

As the use of artificial intelligence has spread rapidly throughout the US health care system, concerns have been raised about racial and ethnic biases built into the algorithms that often guide clinical decision making. Race-based medicine, which relies on algorithms that use race as a proxy for biological differences, has led to treatment patterns that are inappropriate, unjust, and harmful to minoritized racial and ethnic groups. These patterns have contributed to persistent disparities in health and health care. To reduce these disparities, we recommend a race-aware approach to clinical decision support that considers social and environmental factors such as structural racism and social determinants of health. Recent policy changes in medical specialty societies and innovations in algorithm development represent progress on the path to dismantling race-based medicine. Success will require continued commitment and sustained efforts among stakeholders in the health care, research, and technology sectors. Increasing the diversity of clinical trial populations, broadening the focus of precision medicine, improving education about the complex factors shaping health outcomes, and developing new guidelines and policies to enable culturally responsive care are important next steps.

View details for DOI 10.1377/hlthaff.2023.00545

View details for PubMedID 37782875

Abstract

Opioids are commonly prescribed for postoperative pain, but may lead to prolonged use and addiction. Diabetes impairs nerve function, complicates pain management, and makes opioid prescribing particularly challenging.This retrospective observational study included a cohort of postoperative patients from a multisite academic health system to assess the relationship between diabetes, pain, and prolonged opioid use (POU), 2008-2019. POU was defined as a new opioid prescription 3-6 months after discharge. The odds that a patient had POU was assessed using multivariate logistic regression controlling for patient factors (e.g., demographic and clinical factors, as well as prior pain and opiate use).A total of 43,654 patients were included, 12.4% with diabetes. Patients with diabetes had higher preoperative pain scores (2.1 vs 1.9, p<0.001) and lower opioid naïve rates (58.7% vs 68.6%, p<0.001). Following surgery, patients with diabetes had higher rates of POU (17.7% vs 12.7%, p<0.001) despite receiving similar opioid prescriptions at discharge. Patients with Type I diabetes were more likely to have POU compared to other patients (Odds Ratio [OR]: 2.22; 95% Confidence Interval [CI]:1.69-2.90 and OR:1.44, CI: 1.33-1.56, respectively).In conclusion, surgical patients with diabetes are at increased risk for POU even after controlling for likely covariates, yet they receive similar postoperative opiate therapy. The results suggest a more tailored approach to diabetic postoperative pain management is warranted.

View details for DOI 10.1371/journal.pone.0287697

View details for PubMedID 37616195

View details for PubMedCentralID PMC10449216

Abstract

BACKGROUND: Treatment of surgical pain is a common reason for opioid prescriptions. Being able to predict which patients are at risk for opioid abuse, dependence, and overdose (opioid-related adverse outcomes [OR-AE]) could help physicians make safer prescription decisions. We aimed to develop a machine-learning algorithm to predict the risk of OR-AE following surgery using Medicaid data with external validation across states.METHODS: Five machine learning models were developed and validated across seven US states (90-10 data split). The model output was the risk of OR-AE 6-months following surgery. The models were evaluated using standard metrics and area under the receiver operating characteristic curve (AUC) was used for model comparison. We assessed calibration for the top performing model and generated bootstrap estimations for standard deviations. Decision curves were generated for the top-performing model and logistic regression.RESULTS: We evaluated 96,974 surgical patients aged 15 and 64. During the 6-month period following surgery, 10,464 (10.8%) patients had an OR-AE. Outcome rates were significantly higher for patients with depression (17.5%), diabetes (13.1%) or obesity (11.1%). The random forest model achieved the best predictive performance (AUC: 0.877; F1-score: 0.57; recall: 0.69; precision:0.48). An opioid disorder diagnosis prior to surgery was the most important feature for the model, which was well calibrated and had good discrimination.CONCLUSIONS: A machine learning models to predict risk of OR-AE following surgery performed well in external validation. This work could be used to assist pain management following surgery for Medicaid beneficiaries and supports a precision medicine approach to opioid prescribing.

View details for DOI 10.1371/journal.pcbi.1011376

View details for PubMedID 37578969

Abstract

The evolution of pre- versus postoperative risk factors remains unknown in the development of persistent postoperative pain and opioid use. We identified preoperative versus comprehensive perioperative models of delayed pain and opioid cessation after total joint arthroplasty including time-varying postoperative changes in emotional distress. We hypothesized that time-varying longitudinal measures of postoperative psychological distress, as well as pre- and postoperative use of opioids would be the most significant risk factors for both outcomes.A prospective cohort of 188 patients undergoing total hip or knee arthroplasty at Stanford Hospital completed baseline pain, opioid use, and emotional distress assessments. After surgery, a modified Brief Pain Inventory was assessed daily for 3 months, weekly thereafter up to 6 months, and monthly thereafter up to 1 year. Emotional distress and pain catastrophizing were assessed weekly to 6 months, then monthly thereafter. Stepwise multivariate time-varying Cox regression modeled preoperative variables alone, followed by all perioperative variables (before and after surgery) with time to postoperative opioid and pain cessation.The median time to opioid and pain cessation was 54 and 152 days, respectively. Preoperative total daily oral morphine equivalent use (hazard ratio-HR 0.97; 95% confidence interval-CI 0.96-0.98) was significantly associated with delayed postoperative opioid cessation in the perioperative model. In contrast, time-varying postoperative factors: elevated PROMIS (Patient-Reported Outcomes Measurement Information System) depression scores (HR 0.92; 95% CI 0.87-0.98), and higher Pain Catastrophizing Scale scores (HR 0.85; 95% CI 0.75-0.97) were independently associated with delayed postoperative pain resolution in the perioperative model.These findings highlight preoperative opioid use as a key determinant of delayed postoperative opioid cessation, while postoperative elevations in depressive symptoms and pain catastrophizing are associated with persistent pain after total joint arthroplasty providing the rationale for continued risk stratification before and after surgery to identify patients at highest risk for these distinct outcomes. Interventions targeting these perioperative risk factors may prevent prolonged postoperative pain and opioid use.

View details for DOI 10.1007/s40122-023-00543-9

View details for PubMedID 37556071

View details for PubMedCentralID 7317603

Abstract

Assessments of health-related quality of life (HRQOL) are conducted by health systems to improve patient-centered care. Studies have shown that the COVID-19 pandemic poses unique stressors for patients with cancer. This study investigates change in self-reported global health scores in patients with cancer before and during the COVID-19 pandemic. In this single-institution retrospective cohort study, patients who completed the Patient-Reported Outcomes Measurement Information System (PROMIS) at a comprehensive cancer center before and during the COVID-19 pandemic were identified. Surveys were analyzed to assess change in the global mental health (GMH) and global physical health (GPH) scores at different time periods (pre-COVID: 3/1/5/2019-3/15/2020, surge1: 6/17/2020-9/7/2020, valley1: 9/8/2020-11/16/2020, surge2: 11/17/2020-3/2/2021, and valley2: 3/3/2021-6/15/2021). A total of 25,192 surveys among 7209 patients were included in the study. Mean GMH score for patients before the COVID-19 pandemic (50.57) was similar to those during various periods during the pandemic: surge1 (48.82), valley1 (48.93), surge2 (48.68), valley2 (49.19). Mean GPH score was significantly higher pre-COVID (42.46) than during surge1 (36.88), valley1 (36.90), surge2 (37.33) and valley2 (37.14). During the pandemic, mean GMH (49.00) and GPH (37.37) scores obtained through in-person were similar to mean GMH (48.53) and GPH (36.94) scores obtained through telehealth. At this comprehensive cancer center, patients with cancer reported stable mental health and deteriorating physical health during the COVID-19 pandemic as indicated by the PROMIS survey. Modality of the survey (in-person versus telehealth) did not affect scores.

View details for DOI 10.1038/s41598-023-36025-3

View details for PubMedID 37311790

View details for PubMedCentralID 7450263

Abstract

BACKGROUND: Machine learning (ML) predictions are becoming increasingly integrated into medical practice. One commonly used method, ℓ1-penalised logistic regression (LASSO), can estimate patient risk for disease outcomes but is limited by only providing point estimates. Instead, Bayesian logistic LASSO regression (BLLR) models provide distributions for risk predictions, giving clinicians a better understanding of predictive uncertainty, but they are not commonly implemented.METHODS: This study evaluates the predictive performance of different BLLRs compared to standard logistic LASSO regression, using real-world, high-dimensional, structured electronic health record (EHR) data from cancer patients initiating chemotherapy at a comprehensive cancer centre. Multiple BLLR models were compared against a LASSO model using an 80-20 random split using 10-fold cross-validation to predict the risk of acute care utilization (ACU) after starting chemotherapy.FINDINGS: This study included 8439 patients. The LASSO model predicted ACU with an area under the receiver operating characteristic curve (AUROC) of 0.806 (95% CI: 0.775-0.834). BLLR with a Horseshoe+prior and a posterior approximated by Metropolis-Hastings sampling showed similar performance: 0.807 (95% CI: 0.780-0.834) and offers the advantage of uncertainty estimation for each prediction. In addition, BLLR could identify predictions too uncertain to be automatically classified. BLLR uncertainties were stratified by different patient subgroups, demonstrating that predictive uncertainties significantly differ across race, cancer type, and stage.INTERPRETATION: BLLRs are a promising yet underutilised tool that increases explainability by providing risk estimates while offering a similar level of performance to standard LASSO-based models. Additionally, these models can identify patient subgroups with higher uncertainty, which can augment clinical decision-making.FUNDING: This work was supported in part by the National Library Of Medicine of the National Institutes of Health under Award Number R01LM013362. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

View details for DOI 10.1016/j.ebiom.2023.104632

View details for PubMedID 37269570

Abstract

Background: Statins are the cornerstone of treatment of patients with atherosclerotic cardiovascular disease (ASCVD). Despite this, multiple studies have shown that women with ASCVD are less likely to be prescribed statins than men. The objective of this study was to use Natural Language Processing (NLP) to elucidate factors contributing to this disparity.Methods: Our cohort included adult patients with two or more encounters between 2014 and 2021 with an ASCVD diagnosis within a multisite electronic health record (EHR) in Northern California. After reviewing structured EHR prescription data, we used a benchmark deep learning NLP approach, Clinical Bidirectional Encoder Representations from Transformers (BERT), to identify and interpret discussions of statin prescriptions documented in clinical notes. Clinical BERT was evaluated against expert clinician review in 20% test sets.Results: There were 88,913 patients with ASCVD (mean age 67.8±13.1 years) and 35,901 (40.4%) were women. Women with ASCVD were less likely to be prescribed statins compared with men (56.6%vs 67.6%, p <0.001), and, when prescribed, less likely to be prescribed guideline-directed high-intensity dosing (41.4%vs 49.8%, p <0.001). These disparities were more pronounced among younger patients, patients with private insurance, and those for whom English is their preferred language. Among those not prescribed statins, women were less likely than men to have statins mentioned in their clinical notes (16.9%vs 19.1%, p <0.001). Women were less likely than men to have statin use reported in clinical notes despite absence of recorded prescription (32.8%vs 42.6%, p <0.001). Women were slightly more likely than men to have statin intolerance documented in structured data or clinical notes (6.0%vs 5.3%, p=0.003).Conclusions: Women with ASCVD were less likely to be prescribed guideline-directed statins compared with men. NLP identified additional sex-based statin disparities and reasons for statin non-prescription in clinical notes of patients with ASCVD.

View details for DOI 10.1016/j.ajpc.2023.100496

View details for PubMedID 37128554

Abstract

When patients with cancer develop depression, it is often left untreated. We developed a prediction model for depression risk within the first month after starting cancer treatment using machine learning and Natural Language Processing (NLP) models. The LASSO logistic regression model based on structured data performed well, whereas the NLP model based on only clinician notes did poorly. After further validation, prediction models for depression risk could lead to earlier identification and treatment of vulnerable patients, ultimately improving cancer care and treatment adherence.

View details for DOI 10.3233/SHTI230274

View details for PubMedID 37203503

View details for DOI 10.1038/s41746-023-00832-9

View details for PubMedID 37149704

View details for DOI 10.1038/s42256-023-00633-5

View details for Web of Science ID 000963904900003

Abstract

Despite compelling evidence that statins are safe, are generally well tolerated, and reduce cardiovascular events, statins are underused even in patients with the highest risk. Social media may provide contemporary insights into public perceptions about statins.To characterize and classify public perceptions about statins that were gleaned from more than a decade of statin-related discussions on Reddit, a widely used social media platform.This qualitative study analyzed all statin-related discussions on the social media platform that were dated between January 1, 2009, and July 12, 2022. Statin- and cholesterol-focused communities, were identified to create a list of statin-related discussions. An artificial intelligence (AI) pipeline was developed to cluster these discussions into specific topics and overarching thematic groups. The pipeline consisted of a semisupervised natural language processing model (BERT [Bidirectional Encoder Representations from Transformers]), a dimensionality reduction technique, and a clustering algorithm. The sentiment for each discussion was labeled as positive, neutral, or negative using a pretrained BERT model.Statin-related posts and comments containing the terms statin and cholesterol.Statin-related topics and thematic groups.A total of 10 233 unique statin-related discussions (961 posts and 9272 comments) from 5188 unique authors were identified. The number of statin-related discussions increased by a mean (SD) of 32.9% (41.1%) per year. A total of 100 discussion topics were identified and were classified into 6 overarching thematic groups: (1) ketogenic diets, diabetes, supplements, and statins; (2) statin adverse effects; (3) statin hesitancy; (4) clinical trial appraisals; (5) pharmaceutical industry bias and statins; and (6) red yeast rice and statins. The sentiment analysis revealed that most discussions had a neutral (66.6%) or negative (30.8%) sentiment.Results of this study demonstrated the potential of an AI approach to analyze large, contemporary, publicly available social media data and generate insights into public perceptions about statins. This information may help guide strategies for addressing barriers to statin use and adherence.

View details for DOI 10.1001/jamanetworkopen.2023.9747

View details for PubMedID 37093597

Abstract

Technological advances now make it possible to study a patient from multiple angles with high-dimensional, high-throughput multi-scale biomedical data. In oncology, massive amounts of data are being generated ranging from molecular, histopathology, radiology to clinical records. The introduction of deep learning has significantly advanced the analysis of biomedical data. However, most approaches focus on single data modalities leading to slow progress in methods to integrate complementary data types. Development of effective multimodal fusion approaches is becoming increasingly important as a single modality might not be consistent and sufficient to capture the heterogeneity of complex diseases to tailor medical care and improve personalised medicine. Many initiatives now focus on integrating these disparate modalities to unravel the biological processes involved in multifactorial diseases such as cancer. However, many obstacles remain, including lack of usable data as well as methods for clinical validation and interpretation. Here, we cover these current challenges and reflect on opportunities through deep learning to tackle data sparsity and scarcity, multimodal interpretability, and standardisation of datasets.

View details for DOI 10.1038/s42256-023-00633-5

View details for PubMedID 37693852

View details for PubMedCentralID PMC10484010

Abstract

Background Statins are guideline-recommended medications that reduce cardiovascular events in patients with diabetes. Yet, statin use is concerningly low in this high-risk population. Identifying reasons for statin nonuse, which are typically described in unstructured electronic health record data, can inform targeted system interventions to improve statin use. We aimed to leverage a deep learning approach to identify reasons for statin nonuse in patients with diabetes. Methods and Results Adults with diabetes and no statin prescriptions were identified from a multiethnic, multisite Northern California electronic health record cohort from 2014 to 2020. We used a benchmark deep learning natural language processing approach (Clinical Bidirectional Encoder Representations from Transformers) to identify statin nonuse and reasons for statin nonuse from unstructured electronic health record data. Performance was evaluated against expert clinician review from manual annotation of clinical notes and compared with other natural language processing approaches. Of 33 461 patients with diabetes (mean age 59±15 years, 49% women, 36% White patients, 24% Asian patients, and 15% Hispanic patients), 47% (15 580) had no statin prescriptions. From unstructured data, Clinical Bidirectional Encoder Representations from Transformers accurately identified statin nonuse (area under receiver operating characteristic curve [AUC] 0.99 [0.98-1.0]) and key patient (eg, side effects/contraindications), clinician (eg, guideline-discordant practice), and system reasons (eg, clinical inertia) for statin nonuse (AUC 0.90 [0.86-0.93]) and outperformed other natural language processing approaches. Reasons for nonuse varied by clinical and demographic characteristics, including race and ethnicity. Conclusions A deep learning algorithm identified statin nonuse and actionable reasons for statin nonuse in patients with diabetes. Findings may enable targeted interventions to improve guideline-directed statin use and be scaled to other evidence-based therapies.

View details for DOI 10.1161/JAHA.122.028120

View details for PubMedID 36974740

View details for Web of Science ID 000990866101649

View details for DOI 10.1007/s40620-023-01609-9

View details for PubMedID 36877370

Abstract

Undetected infection and delayed isolation of infected individuals are key factors driving the monkeypox virus (now termed mpox virus or MPXV) outbreak. To enable earlier detection of MPXV infection, we developed an image-based deep convolutional neural network (named MPXV-CNN) for the identification of the characteristic skin lesions caused by MPXV. We assembled a dataset of 139,198 skin lesion images, split into training/validation and testing cohorts, comprising non-MPXV images (n=138,522) from eight dermatological repositories and MPXV images (n=676) from the scientific literature, news articles, social media and a prospective cohort of the Stanford University Medical Center (n=63 images from 12 patients, all male). In the validation and testing cohorts, the sensitivity of the MPXV-CNN was 0.83 and 0.91, the specificity was 0.965 and 0.898 and the area under the curve was 0.967 and 0.966, respectively. In the prospective cohort, the sensitivity was 0.89. The classification performance of the MPXV-CNN was robust across various skin tones and body regions. To facilitate the usage of the algorithm, we developed a web-based app by which the MPXV-CNN can be accessed for patient guidance. The capability of the MPXV-CNN for identifying MPXV lesions has the potential to aid in MPXV outbreak mitigation.

View details for DOI 10.1038/s41591-023-02225-7

View details for PubMedID 36864252

Abstract

OBJECTIVES: In this systematic review we aimed at assessing how artificial intelligence (AI), including machine learning (ML) techniques have been deployed to predict, diagnose, and treat chronic kidney disease (CKD). We systematically reviewed the available evidence on theseinnovative techniques to improve CKD diagnosis and patient management.METHODS: We included English language studies retrieved from PubMed.The review istherefore to be classified as a "rapid review", since it includes one database only, and has language restrictions; the novelty and importanceof the issue make missing relevant papers unlikely. We extracted 16 variables, including: main aim, studied population, data source, sample size, problem type (regression, classification), predictors used, and performance metrics.We followedthe Preferred Reporting Items for Systematic Reviews (PRISMA) approach; all main steps were done in duplicate.The review was registered on PROSPERO.RESULTS: From a total of 648 studies initially retrieved, 68 articles met the inclusioncriteria. Models, as reported by authors, performed well, but the reported metrics were not homogeneous across articles and therefore direct comparison was not feasible. The most common aim was prediction of prognosis, followed by diagnosis of CKD. Algorithm generalizability, and testing on diverse populations was rarely taken into account. Furthermore, the clinical evaluation and validation of the models/algorithms was perused; only a fraction of the included studies, 6 out of 68, were performed in a clinical context.CONCLUSIONS: Machine learning is apromising toolfor the prediction of risk, diagnosis, and therapy management for CKDpatients. Nonetheless, future work is needed to address the interpretability, generalizability, and fairness of the models to ensure the safe application of such technologies in routine clinical practice.

View details for DOI 10.1007/s40620-023-01573-4

View details for PubMedID 36786976

Abstract

Medications such as buprenorphine-naloxone are among the most effective treatments for opioid use disorder, but limited retention in treatment limits long-term outcomes. In this study, we assess the feasibility of a machine learning model to predict retention vs. attrition in medication for opioid use disorder (MOUD) treatment using electronic medical record data including concepts extracted from clinical notes. A logistic regression classifier was trained on 374 MOUD treatments with 68% resulting in potential attrition. On a held-out test set of 157 events, the full model achieved an area under the receiver operating characteristic curve (AUROC) of 0.77 (95% CI: 0.64-0.90) and AUROC of 0.74 (95% CI: 0.62-0.87) with a limited model using only structured EMR data. Risk prediction for opioid MOUD retention vs. attrition is feasible given electronic medical record data, even without necessarily incorporating concepts extracted from clinical notes.

View details for PubMedID 38222349

View details for PubMedCentralID PMC10785878

Abstract

Ensemble learning is a powerful technique for improving the accuracy and reliability of prediction models, especially in scenarios where individual models may not perform well. However, combining models with varying accuracies may not always improve the final prediction results, as models with lower accuracies may obscure the results of models with higher accuracies. This paper addresses this issue and answers the question of when an ensemble approach outperforms individual models for prediction. As a result, we propose an ensemble model for predicting patients at risk of postoperative prolonged opioid. The model incorporates two machine learning models that are trained using different covariates, resulting in high precision and recall. Our study, which employs five different machine learning algorithms, shows that the proposed approach significantly improves the final prediction results in terms of AUROC and AUPRC.

View details for PubMedID 38222353

View details for PubMedCentralID PMC10785929

Abstract

Clinical notes are an essential component of a health record. This paper evaluates how natural language processing (NLP) can be used to identify the risk of acute care use (ACU) in oncology patients, once chemotherapy starts. Risk prediction using structured health data (SHD) is now standard, but predictions using free-text formats are complex. This paper explores the use of free-text notes for the prediction of ACU in leu of SHD. Deep Learning models were compared to manually engineered language features. Results show that SHD models minimally outperform NLP models; an ℓ1-penalised logistic regression with SHD achieved a C-statistic of 0.748 (95%-CI: 0.735, 0.762), while the same model with language features achieved 0.730 (95%-CI: 0.717, 0.745) and a transformer-based model achieved 0.702 (95%-CI: 0.688, 0.717). This paper shows how language models can be used in clinical applications and underlines how risk bias is different for diverse patient groups, even using only free-text data.

View details for PubMedID 37350895

View details for Web of Science ID 001243755000061

Abstract

The coronavirus disease 2019 public health emergency (PHE) caused extensive job loss and loss of employer-sponsored insurance. State Medicaid programs experienced a related increase in enrollment during the PHE. However, the composition of enrollment and enrollee changes during the pandemic is unknown. This study examined changes in Medicaid enrollment and population characteristics during the PHE. A retrospective study documenting changes in Medicaid new enrollment and disenrollment, and enrollee characteristics between March and October 2020 compared to the same time in 2019 using full-state Medicaid populations from 6 states of a wide geographical region. The primary outcomes were Medicaid enrollment and disenrollment during the PHE. New enrollment included persons enrolled in Medicaid between March and October 2020 who were not enrolled in January or February, 2020. Disenrollment included persons who were enrolled in March of 2020 but not enrolled in October 2020. The study included 8.50 million Medicaid enrollees in 2020 and 8.46 million in 2019. Overall, enrollment increased by 13.0% (1.19 million) in the selected states during the PHE compared to 2019. New enrollment accounted for 24.9% of the relative increase, while the remaining 75.1% was due to disenrollment. A larger proportion of new enrollment in 2020 was among adults aged 27 to 44 (28.3% vs 23.6%), Hispanics (34.3% vs 32.5%) and in the financial needy (44.0% vs 39.0%) category compared to 2019. Disenrollment included a larger proportion of older adults (26.1% vs 8.1%) and non-Hispanics (70.3% vs 66.4%) than in 2019. Medicaid enrollment grew considerably during the PHE, and most enrollment growth was attributed to decreases in disenrollment rather than increases in new enrollment. Our results highlight the impact of coronavirus disease 2019 on state health programs and can guide federal and state budgetary planning once the PHE ends.

View details for DOI 10.1097/MD.0000000000032487

View details for PubMedID 36596028

Abstract

BACKGROUND: Early recognition of heart failure (HF) can reduce morbidity, yet HF is often diagnosed only after symptoms require urgent treatment.OBJECTIVES: The authors sought to describe predictors of HF diagnosis in the acute care vs outpatient setting within the Veterans Health Administration (VHA).METHODS: The authors estimated whether incident HF diagnoses occurred in acute care (inpatient hospital or emergency department) vs outpatient settings within the VHA between 2014 and 2019. After excluding new-onset HF potentially caused by acute concurrent conditions, they identified sociodemographic and clinical variables associated with diagnosis setting and assessed variation across 130 VHA facilities using multivariable regression analysis.RESULTS: The authors identified 303,632 patients with new HF, with 160,454 (52.8%) diagnosed in acute care settings. In the prior year, 44% had HF symptoms and 11% had a natriuretic peptide tested, 88% of which were elevated. Patients with housing insecurity and high neighborhood social vulnerability had higher odds of acute care diagnosis (adjusted odds ratio: 1.22 [95%CI: 1.17-1.27] and 1.17 [95%CI: 1.14-1.21], respectively) adjusting for medical comorbidities. Better outpatient quality of care (blood pressure control and cholesterol and diabetes monitoring within the prior 2 years) predicted a lower odds of acute care diagnosis. Likelihood of acute care HF diagnosis varied from 41% to 68% across facilities after adjusting for patient-level risk factors.CONCLUSIONS: Many first HF diagnoses occur in the acute care setting, especially among socioeconomically vulnerable populations. Better outpatient care was associated with lower rates of an acute care diagnosis. These findings highlight opportunities for timelier HF diagnosis that may improve patient outcomes.

View details for DOI 10.1016/j.jchf.2022.11.013

View details for PubMedID 36881392

View details for DOI 10.1038/s41591-022-01993-y

View details for PubMedID 36163298

Abstract

To develop deep learning models to recognize ophthalmic examination components from clinical notes in electronic health records (EHR) using a weak supervision approach.A corpus of 39,099 ophthalmology notes weakly labeled for 24 examination entities was assembled from the EHR of one academic center. Four pre-trained transformer-based language models (DistilBert, BioBert, BlueBert, and ClinicalBert) were fine-tuned to this named entity recognition task and compared to a baseline regular expression model. Models were evaluated on the weakly labeled test dataset, a human-labeled sample of that set, and a human-labeled independent dataset.On the weakly labeled test set, all transformer-based models had recall > 0.93, with precision varying from 0.815 to 0.843. The baseline model had lower recall (0.769) and precision (0.682). On the human-annotated sample, the baseline model had high recall (0.962, 95 % CI 0.955-0.067) with variable precision across entities (0.081-0.999). Bert models had recall ranging from 0.771 to 0.831, and precision >=0.973. On the independent dataset, precision was 0.926 and recall 0.458 for BlueBert. The baseline model had better recall (0.708, 95 % CI 0.674-0.738) but worse precision (0.399, 95 % CI -0.352-0.451).We developed the first deep learning system to recognize eye examination components from clinical notes, leveraging a novel opportunity for weak supervision. Transformer-based models had high precision on human-annotated labels, whereas the baseline model had poor precision but higher recall. This system may be used to improve cohort and feature identification using free-text notes.Our weakly supervised approach may help amass large datasets of domain-specific entities from EHRs in many fields.

View details for DOI 10.1016/j.ijmedinf.2022.104864

View details for PubMedID 36179600

Abstract

The lack of diversity, equity, and inclusion continues to hamper the artificial intelligence (AI) field and is especially problematic for healthcare applications. In this article, we expand on the need for diversity, equity, and inclusion, specifically focusing on the composition of AI teams. We call to action leaders at all levels to make team inclusivity and diversity the centerpieces of AI development, not the afterthought. These recommendations take into consideration mitigation at several levels, including outreach programs at the local level, diversity statements at the academic level, and regulatory steps at the federal level.

View details for DOI 10.1093/jamia/ocac156

View details for PubMedID 36048021

Abstract

Noncardia gastric cancer (NCGC) is a leading cause of global cancer mortality, and is often diagnosed at advanced stages. Development of NCGC risk models within electronic health records (EHR) may allow for improved cancer prevention. There has been much recent interest in use of machine learning (ML) for cancer prediction, but few studies comparing ML with classical statistical models for NCGC risk prediction.We trained models using logistic regression (LR) and four commonly used ML algorithms to predict NCGC from age-/sex-matched controls in two EHR systems: Stanford University and the University of Washington (UW). The LR model contained well-established NCGC risk factors (intestinal metaplasia histology, prior Helicobacter pylori infection, race, ethnicity, nativity status, smoking history, anemia), whereas ML models agnostically selected variables from the EHR. Models were developed and internally validated in the Stanford data, and externally validated in the UW data. Hyperparameter tuning of models was achieved using cross-validation. Model performance was compared by accuracy, sensitivity, and specificity.In internal validation, LR performed with comparable accuracy (0.732; 95% CI, 0.698 to 0.764), sensitivity (0.697; 95% CI, 0.647 to 0.744), and specificity (0.767; 95% CI, 0.720 to 0.809) to penalized lasso, support vector machine, K-nearest neighbor, and random forest models. In external validation, LR continued to demonstrate high accuracy, sensitivity, and specificity. Although K-nearest neighbor demonstrated higher accuracy and specificity, this was offset by significantly lower sensitivity. No ML model consistently outperformed LR across evaluation criteria.Drawing data from two independent EHRs, we find LR on the basis of established risk factors demonstrated comparable performance to optimized ML algorithms. This study demonstrates that classical models built on robust, hand-chosen predictor variables may not be inferior to data-driven models for NCGC risk prediction.

View details for DOI 10.1200/CCI.22.00039

View details for PubMedID 35763703

View details for Web of Science ID 000844401302060

Abstract

Purpose: Advances in artificial intelligence have produced a few predictive models in glaucoma, including a logistic regression model predicting glaucoma progression to surgery. However, uncertainty exists regarding how to integrate the wealth of information in free-text clinical notes. The purpose of this study was to predict glaucoma progression requiring surgery using deep learning (DL) approaches on data from electronic health records (EHRs), including features from structured clinical data and from natural language processing of clinical free-text notes.Design: Development of DL predictive model in an observational cohort.Participants: Adult patients with glaucoma at a single center treated from 2008 through2020.Methods: Ophthalmology clinical notes of patients with glaucoma were identified from EHRs. Available structured data included patient demographic information, diagnosis codes, prior surgeries, and clinical information including intraocular pressure, visual acuity, and central corneal thickness. In addition, words from patients' first 120 days of notes were mapped to ophthalmology domain-specific neural word embeddings trained on PubMed ophthalmology abstracts. Word embeddings and structured clinical data were used as inputs to DL models to predict subsequent glaucoma surgery.Main Outcome Measures: Evaluation metrics included area under the receiver operating characteristic curve (AUC) and F1 score, the harmonic mean of positive predictive value, and sensitivity on a held-out test set.Results: Seven hundred forty-eight of 4512 patients with glaucoma underwent surgery. The model that incorporated both structured clinical features as well as input features from clinical notes achieved an AUC of 73% and F1 of 40%, compared with only structured clinical features, (AUC, 66%; F1, 34%) and only clinical free-text features (AUC, 70%; F1, 42%). All models outperformed predictions from a glaucoma specialist's review of clinical notes (F1, 29.5%).Conclusions: We can successfully predict which patients with glaucoma will need surgery using DL models on EHRs unstructured text. Models incorporating free-text data outperformed those using only structured inputs. Future predictive models using EHRs should make use of information from within clinical free-text notes to improve predictive performance. Additional research is needed to investigate optimal methods of incorporating imaging data into future predictive models as well.

View details for DOI 10.1016/j.xops.2022.100127

View details for PubMedID 36249690

View details for Web of Science ID 000863680303793

Abstract

BACKGROUND: Blood transfusions are a vital component of modern healthcare, yet adverse reactions to blood product transfusions can cause morbidity, and rarely result in mortality. Therefore, accurate reporting of transfusion related adverse events (TRAEs) is paramount to improved transfusion practice. This study aims to investigate real-world data (RWD) on TRAEs by evaluating differences between ICD 9/10-based electronic health records (EHR) and blood bank-specific reporting.STUDY DESIGN AND METHODS: TRAE data were retrospectively collected from a blood bank-specific database between Jan 2015 and June 2019 as the reference data source and compared it to ICD 9/10 diagnostic codes corresponding to various TRAEs. Seven reactions that have corresponding ICD 9/10 diagnostic codes were evaluated: Transfusion related circulatory overload (TACO), transfusion related acute lung injury (TRALI), febrile non-hemolytic reaction (FNHTR), transfusion-related anaphylactic reaction (TRA), acute hemolytic transfusion reaction (AHTR), delayed hemolytic transfusion reaction (DHTR), and delayed serologic reaction (DSTR). These accounted for 33% of the TRAEs at an academic institution during the study period.RESULTS: Among 18637 adult blood transfusion recipients, there were 229 unique patients with 263 TRAE related ICD codes in the EHR, while there were 191 unique patients with 287 TRAEs identified in the blood bank database. None of the categories of reaction we investigated had perfect alignment between ICD 9/10 codes and blood bank specific diagnoses.DISCUSSION: Multiple systemic challenges were identified that hinder effective reporting of TRAEs. Identifying factors causing inconsistent reporting between blood banks and EHRs is paramount to developing effective workability between these electronic systems, as well as across clinical and laboratory teams.

View details for DOI 10.1111/trf.16880

View details for PubMedID 35437749

Abstract

PURPOSE: Cataract is a known effect of trabeculectomy (TE), but some surgeons are hesitant to perform combined phacoemulsification-TE (PTE) due to a risk of increased TE failure. Herein, we compare intraocular pressure (IOP) lowering between trabeculectomy (TE) and phacoemulsification-TE (PTE) and investigate factors that impact patient outcomes.METHODS: We performed a retrospective study of adults undergoing primary TE or PTE at our institution from 2010 to 2017. We used Kaplan-Meier survival analysis to investigate time to TE failure, and Cox proportional hazards modeling to investigate predictors of TE failure, defined as undergoing a second glaucoma surgery or using more IOP-lowering medications than pre-operatively.RESULTS: 318 surgeries (218 TE; 100 PTE) from 268 patients were included. Median follow-up time was 753days. Mean baseline IOP was 21.1mmHg. There were no significant differences in IOP between TE and PTE groups beyond postoperative year 1, with 28.9-46.5% of TE and 35.5-44.4% of PTE groups achieving IOP ≤10. Final IOP was similar in both groups (p=0.22): 12.41 (SD 4.18) mmHg in the TE group and 14.05 (SD 5.45) in the PTE group. 84 (26.4%) surgeries met failure criteria. After adjusting for surgery type, sex, age, race, surgeon, and glaucoma diagnosis there were no significant differences in TE failure.CONCLUSION: This study suggests there is no significant difference in the risk of TE failure in patients receiving TE versus those receiving PTE.

View details for DOI 10.1080/02713683.2022.2045611

View details for PubMedID 35317681

Abstract

The national increase in opioid use and misuse has become a public health crisis in the U.S. To tackle this crisis, the systematic evaluation and monitoring of opioid prescribing patterns is necessary. Thus, opioid prescriptions from electronic health records (EHRs) must be standardized to morphine milligram equivalent (MME) to facilitate monitoring and surveillance. While most studies report MMEs to describe opioid prescribing patterns, there is a lack of transparency regarding their data pre-processing and conversion processes for replication or comparison purposes.In this work, we developed Opioid2MME, a SQL-based open-source framework, to convert opioid prescriptions to MMEs using EHR prescription data. The MME conversions were validated internally using F-measures through manual chart review; were compared with two existing tools, as MedEx and MedXN; and the framework was tested in an external academic EHR system.We identified 232,913 prescriptions for 49,060 unique patients in the EHRs, 2008-2019. We manually annotated a sample of prescriptions to assess the performance of the framework. The internal evaluation for medication information extraction achieved F-measures from 0.98 to 1.00 for each piece of the extracted information, outperforming MedEx and MedXN (F-Scores 0.98 and 0.94, respectively). MME values in the internal EHR system obtained a F-measure of 0.97 and identified 3% of the data as outliers and 7% missing values. The MME conversion in the external EHR system obtained 78.3% agreement between the MME values obtained with the development site.The results demonstrated that the framework is replicable and capable of converting opioid prescriptions to MMEs across different medical institutions. In summary, this work sets the groundwork for the systematic evaluation and monitoring of opioid prescribing patterns across healthcare systems.

View details for DOI 10.1016/j.ijmedinf.2022.104739

View details for PubMedID 35325663

Abstract

Childhood obesity is a global public health issue, which has prompted governments to invest in prevention programmes. We aimed to investigate the effectiveness of parent-focused early childhood obesity prevention interventions globally.We did a systematic review and individual participant data meta-analysis. We searched databases and trial registries (MEDLINE, Embase, CENTRAL, CINAHL, PsycInfo, ClinicalTrials.gov, and WHO International Clinical Trials Registry Platform) from inception until Sept 30, 2024, for randomised controlled trials commencing before 12 months of age examining parent-focused behavioural interventions to prevent obesity in children, compared with usual care, no intervention, or attention control. Individual participant data were checked, harmonised, and assessed for integrity and risk of bias. We excluded trials that were quasi-randomised, investigated pregnancy-only interventions, or did not collect any child weight-related outcomes. The primary outcome was BMI Z score at age 24 months (±6 months). We did an intention-to-treat, two-stage, random effects meta-analysis to examine effects overall and for prespecified subgroups. We assessed certainty of evidence using Grading of Recommendations Assessment, Development, and Evaluation. This study is registered with PROSPERO, CRD42020177408.Of 19 990 identified records, 47 (0·24%) trials were completed and eligible. Of these, 18 (38%) assessed our primary outcome, BMI Z score. We obtained individual participant data for 17 (94%; n=9128) of these 18 trials (n=9383), representing 97% of eligible participants. Of these 9128 participants, 4549 (50%) were boys, 4415 (48%) were girls, and 164 (2%) had unknown sex. We found no evidence of an effect of interventions on BMI Z score at age 24 months (±6 months; mean difference -0·01 [95% CI -0·08 to 0·05]; high certainty evidence, τ2=0·01; n=6505; 2623 missing). Findings were robust to prespecified sensitivity analyses (eg, different analysis methods and missing data), and we found no evidence of differential intervention effects for prespecified subgroups including priority populations and trial-level factors.These findings indicate that examined parent-focused behavioural interventions are insufficient to prevent obesity at age 24 months (±6 months). This evidence highlights a need to re-think childhood obesity prevention approaches.Australian National Health and Medical Research Council.

View details for DOI 10.1016/S0140-6736(25)01144-4

View details for PubMedID 40945528

Abstract

Early childhood obesity prevention interventions that aim to change parent/caregiver practices related to infant (milk) feeding, food provision and parent feeding, movement (including activity, sedentary behaviour) and/or sleep health (i.e. target parental behaviour domains) are diverse and heterogeneously reported. We aimed to 1) systematically characterise the target behaviours, delivery features, and Behaviour Change Techniques (BCTs) used in interventions in the international Transforming Obesity Prevention for CHILDren (TOPCHILD) Collaboration, and 2) explore similarities and differences in BCTs used in interventions by target behaviour domains.Annual systematic searches were performed in MEDLINE, Embase, Cochrane (CENTRAL), CINAHL, PsycINFO, and two clinical trial registries, from inception to February 2023. Trialists from eligible randomised controlled trials of parent-focused, behavioural early obesity prevention interventions shared unpublished intervention materials. Standardised approaches were used to code target behaviours, delivery features and BCTs in both published and unpublished intervention materials. Validation meetings confirmed coding with trialists. Narrative syntheses were performed.Thirty-two trials reporting 37 active intervention arms were included. Interventions targeted a range of behaviours. The most frequent combination was targeting all parental behaviour domains (infant [milk] feeding, food provision and parent feeding, movement, sleep health; n[intervention arms] = 15/37). Delivery features varied considerably. Most interventions were delivered by a health professional (n = 26/36), included facilitator training (n = 31/36), and were interactive (n = 28/36). Overall, 49 of 93 unique BCTs were coded to at least one target behaviour domain. The most frequently coded BCTs were: Instruction on how to perform a behaviour (n[intervention arms, separated by domain] = 102), Behavioural practice and rehearsal (n = 85), Information about health consequences (n = 85), Social support (unspecified) (n = 84), and Credible source (n = 77). Similar BCTs were often used for each target behaviour domain.Our study provides the most comprehensive description of the behaviour change content of complex interventions targeting early childhood obesity prevention available to date. Our analysis revealed that interventions targeted multiple behaviour domains, with significant variation in delivery features. Despite the diverse range of BCTs coded, five BCTs were consistently identified across domains, though certain BCTs were more prevalent in specific domains. These findings can be used to examine effectiveness of components and inform intervention development and evaluation in future trials.PROSPERO registration no. CRD42020177408.

View details for DOI 10.1186/s12966-025-01708-9

View details for PubMedID 39910407

View details for PubMedCentralID 5773877

Abstract

Linking mothers to their newborns in health records is crucial for understanding the impact of policies, programs, and medical treatments on inter-generational health outcomes. While previous studies have used shared identifiers like names or addresses for linkage, such data are often unavailable in Medicaid records due to privacy concerns. We present a scalable framework and linking algorithm using Medicaid MAX and TAF claims data-lacking direct identifiers-that connects mothers and infants while ensuring privacy protection. Our method accommodates variations in Medicaid records over time and across states, supporting matches at different levels of stringency. Using data from all 50 states over 19 years, our algorithm linked 11.68 million mother-infant dyads, covering 68% of Medicaid-enrolled infants, over 30% of all U.S. infants. We provide our code to facilitate research on social determinants of health and the inter-generational effects of U.S. public policy.

View details for DOI 10.21203/rs.3.rs-5327524/v1

View details for PubMedID 39606483

View details for PubMedCentralID PMC11601815

Abstract

Misinformation about abortion is widespread and was exacerbated by the overturn of Roe v Wade. Young adults are among those facing the most direct impacts of new abortion restrictions and are more likely to access health information from online sources, where misinformation is prevalent. We explored how young adults perceive and evaluate abortion-related information in a time of heightened abortion restrictions.We conducted in-depth, semi-structured interviews with 25 young adults (aged 18-24 years, 56% assigned female at birth), recruited across 17 US states (44% living in states with restrictive abortion policies), between June and September 2022. We derived themes from the interviews using reflexive thematic analysis.While many participants were aware of and had personally encountered abortion misinformation, their susceptibility to false claims varied substantially based on their previous knowledge of abortion and exposure to anti-abortion rhetoric. Participants tended to reject some common myths regarding the medical risks of abortion (eg, association with breast cancer), while expressing a wider range of views regarding its impacts on fertility and mental health. When presented with contradictory sources of abortion information, most participants were unable to confidently reject the misleading source. Knowledge gaps left participants vulnerable to misinformation, while prior scepticism of anti-abortion rhetoric protected participants against misinformation.In this diverse national sample, young adults demonstrated a range of perceptions of abortion misinformation and approaches to identify it. These results lay the groundwork for future observational and experimental research in public health communication.

View details for DOI 10.1136/bmjsrh-2024-202498

View details for PubMedID 39500559

Abstract

To examine the association between social capital and household food insecurity among US families with newborns.This cross-sectional analysis used enrollment data from 881 newborn-caregiver dyads at six geographically-diverse US academic sites enrolled in the Greenlight Plus Trial, a comparative effectiveness trial to prevent childhood obesity. Ordinal proportional-odds models were used to characterize the associations of two self-reported measures of social capital: 1) caregiver social support and 2) neighborhood social cohesion, with household food insecurity after controlling for sociodemographic characteristics.Among 881 newborn-caregiver dyads (49% Hispanic, 23% non-Hispanic white, 17% non-Hispanic Black; 49% with annual household income <

Abstract

To examine whether a cultural adaptation of an early childhood obesity prevention program promotes healthy infant feeding practices.Prospective quasi-experimental study of a community-engaged multiphasic cultural adaptation of an obesity prevention program set at a federally qualified health center serving immigrant Chinese American parent-child dyads (N=298). In a group of historical controls, we assessed early infant feeding practices (breastfeeding, sugar-sweetened beverage intake) in 6-month-olds and then the same practices alongside early solid food feeding practices (bottle weaning, fruit, vegetable, sugary or salty snack consumption) in 12-month-olds. After implementation, we assessed these practices in an intervention cohort group at 6 and 12 months. We used cross-sectional groupwise comparisons and adjusted regression analyses to evaluate group differences.At 6 months, the intervention group had increased odds of no sugar-sweetened beverage intake (aOR: 5.69 [95% CI: 1.65, 19.63], p=0.006). At 12 months, the intervention group also had increased odds of no sugar-sweetened beverage intake (aOR: 15.22 [95% CI: 6.33, 36.62], p<0.001), increased odds of bottle weaning (aOR: 2.34 [95% CI: 1.05, 5.23], p=0.03), and decreased odds of sugary snack consumption (aOR: 0.36 [0.18, 0.70], p= 0.003). We did not detect improvements in breastfeeding, fruit, vegetable, or salty snack consumption.A cultural adaptation of a primary care-based educational obesity prevention program for immigrant Chinese American families with low-income is associated with certain healthy infant feeding practices. Future studies should evaluate cultural adaptations of more intensive interventions that better address complex feeding practices like breastfeeding and evaluate long-term weight outcomes.

View details for DOI 10.1016/j.acap.2024.06.005

View details for PubMedID 38880393

Abstract

The present study validated a newly developed easy-to-use observational instrument, the Health Environment Rating Scale-Early Childhood Consultation-Classroom version (HERS-ECC-C), to measure the quality of the classroom environment within early care and education centers participating in a mental health consultation program in a diverse area of the southeastern United States.Using a confirmatory factor analysis, three factors emerged capturing critical aspects of a high-quality classroom environment and demonstrated good reliability: (1) Supportive Practices, Positive Socioemotional Practices, and Classroom Management (alpha=.88), (2) Health and Family Communication (alpha=.79), and (3) Individualizing to Children's Needs (alpha=.80). Criterion-related validity was established through concurrent associations between the three HERS-ECC-C subscales and the domains of the Classroom Assessment Scoring System (CLASS) and predictive associations with the Childcare Worker Job Stress Inventory. The HERS-ECC-C Supportive Practices and Health and Family Communication subscales were associated with all three CLASS domains, and the Individualizing to Children's Needs subscale was associated with the CLASS Instructional support domain. Higher HERS-ECC-C subscale scores were associated with lower teacher-reported job stress. Findings provide initial evidence to support the use and continued development of the HERS-ECC-C as a tool to evaluate programs and classrooms engaged in mental health consultation professional development interventions.

View details for DOI 10.1002/imhj.22116

View details for PubMedID 38780350

Abstract

Background: Understanding how different populations respond to a childhood obesity intervention could help optimize personalized treatment strategies, especially with the goal to reduce disparities in obesity. Methods: We conducted a secondary analysis of the Greenlight Cluster Randomized Controlled Trial, a health communication focused pediatric obesity prevention trial, to evaluate for heterogeneity of treatment effect (HTE) by child biological sex, caregiver BMI, caregiver reported race and ethnicity, primary language, and health literacy. To examine HTE on BMI z-score from 2 to 24 months of age, we fit linear mixed effects models. Results: We analyzed 802 caregiver-child pairs, of which 52% of children were female, 58% of households reported annual family income of <

Abstract

To assess the impact of a school-based health intervention on adolescents' health knowledge, psychosocial assets and health behaviors, including comparisons of implementation mode: remote, hybrid or in-person. The Stanford Youth Diabetes Coaches Program, an 8-week, school-based health promotion and coaching skills program, was offered to adolescents (ages 14-18 years) from four low-income US communities. Mode of program implementation was remote, hybrid or in-person. Participants completed online pre- and postsurveys. Analysis included paired t-tests, linear regression and qualitative coding. From Fall 2020 to Fall 2021, 262 adolescents enrolled and 179 finished the program and completed pre- and postsurveys. Of the 179, 80% were female, with a mean age of 15.9 years; 22% were Asian; 8% were Black or African American; 25% were White; and 40% were Hispanic. About 115 participants were remote, 25 were hybrid and 39 were in-person. Across all participants, significant improvements (P < 0.01) were reported in health knowledge, psychosocial assets (self-esteem, self-efficacy and problem-solving) and health behaviors (physical activity, nutrition and stress reduction). After adjusting for sex and age, these improvements were roughly equivalent across the three modes of delivery. Participation was associated with significant improvements in adolescent health behaviors. Furthermore, remote mode of instruction was just as effective as in-person and hybrid modes.

View details for DOI 10.1093/her/cyae015

View details for PubMedID 38687641

Abstract

The American Academy of Pediatrics designed The Injury Prevention Program (TIPP) in 1983 to help pediatricians prevent unintentional injuries, but TIPP's effectiveness has never been formally evaluated. We sought to evaluate the impact of TIPP on reported injuries in the first 2 years of life.We conducted a stratified, cluster-randomized trial at 4 academic medical centers: 2 centers trained their pediatric residents and implemented TIPP screening and counseling materials at all well-child checks (WCCs) for ages 2 to 24 months, and 2 centers implemented obesity prevention. At each WCC, parents reported the number of child injuries since the previous WCC. Proportional odds logistic regression analyses with generalized estimating equation examined the extent to which the number of injuries reported were reduced at TIPP intervention sites compared with control sites, adjusting for baseline child, parent, and household factors.A total of 781 parent-infant dyads (349 TIPP; 432 control) were enrolled and had sufficient data to qualify for analyses: 51% Hispanic, 28% non-Hispanic Black, and 87% insured by Medicaid. Those at TIPP sites had significant reduction in the adjusted odds of reported injuries compared with non-TIPP sites throughout the follow-up (P = .005), with adjusted odds ratios (95% CI) of 0.77 (0.66-0.91), 0.60 (0.44-0.82), 0.32 (0.16-0.62), 0.26 (0.12-0.53), and 0.27 (0.14-0.52) at 4, 6, 12, 18, and 24 months, respectively.In this cluster-randomized trial with predominantly low-income, Hispanic, and non-Hispanic Black families, TIPP resulted in a significant reduction in parent-reported injuries. Our study provides evidence for implementing the American Academy of Pediatrics' TIPP in routine well-child care.

View details for DOI 10.1542/peds.2023-062966

View details for PubMedID 38557871

Abstract

To determine associations between sociodemographic and medical factors and odds of readmission after discharge from the neonatal intensive care unit (NICU) for infants with very low birth weight (VLBW, <1500g).Cohort study using linked data from the California Perinatal Quality Care Collaborative, California Vital Statistics, and the Child Opportunity Index 2.0 (COI). Infants with VLBW born from 2009 through 2018 in California were considered. Odds ratios of readmission within 30 days of discharge adjusting for infant medical factors, maternal sociodemographic factors, and birth hospital were calculated via multivariable logistic regression and fixed effect logistic regression models.42,411 infants met inclusion criteria. 8.5% of all infants were readmitted within 30 days of discharge. In addition to traditional medical risk factors, two sociodemographic factors were significantly associated with increased odds of readmission in adjusted models: payor other than private insurance for delivery [aOR =1.25 (95% CI 1.14-1.36)] and maternal education of less than high school degree [aOR = 1.19 (95% CI 1.06-1.33)]. Neighborhood COI was not associated with odds of readmission.Sociodemographic factors, including lack of private insurance and lower maternal educational attainment, are significantly and independently associated with increased odds of readmission after NICU discharge, in addition to traditional medical risk factors. Socioeconomic deprivation and health literacy may contribute to risk of readmission. Targeted discharge interventions focused on addressing social drivers of health warrant exploration.

View details for DOI 10.1016/j.jpeds.2024.114014

View details for PubMedID 38494087

Abstract

This study aimed to describe caregiver satisfaction with physician communication over the first two years of life and examine differences by preferred language and the relationship to physician continuity.Longitudinal data were collected at well visits (2 months to 2 years) from participants in a randomized controlled trial to prevent childhood obesity. Satisfaction with communication was assessed using the validated Communication Assessment Tool (CAT) questionnaire. Changes in the odds of optimal scores were estimated in mixed-effects logistic regression models to evaluate the associations between satisfaction over time and language, interpreter use, and physician continuity.Of 865 caregivers, 35% were Spanish-speaking. Spanish-speaking caregivers without interpreters had lower odds of an optimal satisfaction score compared with English speakers during the first 2 years, beginning at 2 months [OR 0.64 (95% CI: 0.43, 0.95)]. There was no significant difference in satisfaction between English-speaking caregivers and Spanish-speaking caregivers with an interpreter. The odds of optimal satisfaction scores increased over time for both language groups. For both language groups, odds of an optimal satisfaction score decreased each time a new physician was seen for a visit [OR 0.82 (95% CI: 0.69, 0.97)].Caregiver satisfaction with physician communication improves over the first two years of well-child visits for both English- and Spanish-speakers. A loss of physician continuity over time was also associated with lower satisfaction. Future interventions to ameliorate communication disparities should ensure adequate interpreter use for primarily Spanish-speaking patients and address continuity issues to improve communication satisfaction.Caregiver satisfaction with physician communication improves during the first two years of well-child visits and varies by language and with interpreter use and physician continuity.

View details for DOI 10.1016/j.acap.2024.03.004

View details for PubMedID 38458488

Abstract

BACKGROUND: The COVID-19 pandemic transformed the home lives of many families in the US, especially those with young children. Understanding the relationship between child and parent screen time and family stressors exacerbated by the pandemic may help inform interventions that aim to support early child development.OBJECTIVE: To assess the changing relationship between family screen time and factors related to pandemic-induced remote work and childcare/school closures.METHODS: Design, Setting, and Participants: In spring of 2021 we administered a survey, similar to one administered in spring of 2019, to a national sample of parents of young children (aged 6 to 60 months). Using iterative sampling with propensity scores, we recruited participants whose sociodemographic characteristics matched the 2019 survey. Participants were >18 years of age, proficient in English or Spanish, and residing in the US. Main Outcomes and Measures: The main outcomes were changes in child screen time (e.g., mobile phone, tablet, computer, television) and parenting technoference, defined as perceived screen-related interference with parent-child interactions. Additional survey items reported pandemic-related job loss, and changes to work hours, work location, caregiving responsibilities, daycare/school access, and family health and socioeconomic status.RESULTS: We enrolled 280 parents, from diverse backgrounds. Parents reported pandemic-related changes in child screen time (mean increase of 1.1 hour, SD 0.9), and greater parenting technoference (3.0 to 3.4 devices interfering per day; P=.01). Increased child screen time and parenting technoference were highest for parents experiencing job loss (mean change in child screen time 1.46 (SD 1.03); mean parenting technoference score 3.89 (SD 2.05)), second highest for working parents who did not lose their job (mean change in child screen time=1.02 (SD 0.83); mean parenting technoference score 3.37 (SD 1.94), and lowest for non-working parents (mean change in child screen time 0.68 (SD 0.66); mean parenting technoference score 2.66 (SD 1.70)), with differences significant at P<.01. School closure and job loss were most associated with increased child screen time during the pandemic after controlling for other stressors and sociodemographic characteristics (d=0.52, P<.001; d=0.31, P=.01). Increased child screen time and school closure were most associated with increased parenting technoference (d=0.78, P<.001; d=0.30, P=.01).CONCLUSIONS: Work and school changes due to the COVID-19 pandemic were associated with increased technology interference in the lives of young children. This study adds to our understanding of the interaction between technology use in the home and social factors that are necessary to support early child health and development. It also supports possible enhanced recommendations for primary-care providers and child-care educators to guide parents in establishing home-based "screen time rules" not only for their children but also for themselves.CLINICALTRIAL:

View details for DOI 10.2196/43315

View details for PubMedID 38446995

Abstract

Introduction: Despite increasing survival of children following hospitalization, hospitalization may increase iatrogenic risk for mental health (MH) disorders, including acute stress, post-traumatic stress, anxiety, or depression. Using a population-based retrospective cohort study, we assessed the rates of new MH diagnoses during the 12 months after hospitalization, including the moderating effects of ICU exposure.Study design/methods: This was a retrospective case control study using the Truven Health Analytics insurance database. Inclusion criteria included children aged 3-21 years, insurance enrollment for >12 months before and after hospital admission. We excluded children with hospitalization 2 years prior to index hospitalization and those with prior MH diagnoses. We extracted admission type, ICD-10 codes, demographic, clinical, and service coordination variables from the database. We established age- and sex-matched cohorts of non-hospitalized children. The primary outcome was a new MH diagnosis. Multivariable regression methods examined the risk of incident MH disorder(s) between hospitalized and non-hospitalized children. Among hospitalized children, we further assessed effect modification from ICU (vs. non-ICU) stay, admission year, length of stay, medical complexity, and geographic region.Results: New MH diagnoses occurred among 19,418 (7%) hospitalized children, 3,336 (8%) ICU-hospitalized children and 28,209 (5%) matched healthy controls. The most common MH diagnoses were anxiety (2.5%), depression (1.9%), and stress/trauma (2.2%) disorders. Hospitalization increased the odds of new MH diagnoses by 12.3% (OR: 1.123, 95% CI: 1.079-1.17) and ICU-hospitalization increased these odds by 63% (OR: 1.63, 95% CI: 1.483-1.79) as compared to matched, non-hospitalized children. Children with non-complex chronic diseases (OR: 2.91, 95% CI: 2.84-2.977) and complex chronic diseases (OR: 5.16, 95% CI: 5.032-5.289) had a substantially higher risk for new MH diagnoses after hospitalization compared to patients with acute illnesses.Conclusion: Pediatric hospitalization is associated with higher, long-term risk of new mental health diagnoses, and ICU hospitalization further increases that risk within 12 months of the acute episode. Acute care hospitalization confers iatrogenic risks that warrant long-term mental and behavioral health follow-up.

View details for DOI 10.3389/fped.2024.1344870

View details for PubMedID 38450296

View details for DOI 10.1016/j.contraception.2023.110204

View details for Web of Science ID 001114085100059

Abstract

BACKGROUND: While several studies examine the relationship between screen time and dietary practices in children and teenagers, there is limited research in toddlers. This study evaluates the association between television (TV) exposure and dietary practices in two-year-old children.METHODS: We conducted a cross-sectional, secondary data analysis from the Greenlight Intervention Study. Toddlers' daily TV watching time, mealtime TV, and dietary practices were assessed by caregiver report at the 24-month well child visit. Separate regression models were used and adjusted for sociodemographic/household characteristics and clinic site.RESULTS: 532 toddlers were included (51% Latino; 30% non-Latino Black; 59% ≤

Abstract

Objective: Child survival after intensive care unit (ICU) hospitalization has increased, yet many children experience acute stress that may precipitate mental/behavioral health comorbidities. Parents report stress after their child's hospitalization. Little is known about the individual and family characteristics that may moderate intergenerational relationships of acute stress. Design: Following ICU admission at a large academic medical center, a prospective cross-sectional cohort study assessed the associations between intergenerational characteristics and acute stress among children and families. Patients: Parent-child dyads (N = 88) were recruited from the pediatric ICU and pediatric cardiovascular ICU (CVICU) following ICU discharge. Eligible children were between 8 and 18 years old with ICU stays longer than 24 hours. Children with developmental delays were excluded. Caregivers were proficient in English or Spanish. Surveys were collected before hospital discharge. Measurements/Main Results: The primary outcome was "child stress" defined as a score≥17, measured by the Children's Revised Impact of Events Scale (CRIES-8). "Parent stress" was defined as an elevated composite score on the Stanford Acute Stress Reaction Questionnaire. We used validated scales to assess the child's clinical and family social characteristics. Acute stress was identified in 34 (39.8%) children and 50 (56.8%) parents. In multivariate linear regression analyses adjusting for social characteristics, parent stress was associated with increased risk of child stress (adjusted odds ratio 2.58, 95% confidence interval 0.69, 4.46, p < .01). In unadjusted analyses, Hispanic ethnicity was associated with greater child stress. In adjusted analyses, race, income, ICU length of stay, and language were not associated with child stress and did not moderate the parent-child stress relationship. Conclusions: Parent stress is closely correlated with child stress during ICU hospitalization. Hispanic ethnicity may be associated with increased risk for child stress, but further studies are required to define the roles of other social and clinical measures.

View details for DOI 10.1177/08850666231201836

View details for PubMedID 37743757

Abstract

Infants with high birthweight have increased risk for adverse outcomes at birth and across childhood. Prenatal risks to healthy food access may increase odds of high birthweight. We tested whether having a poor neighborhood food environment and/or food insecurity had associations with high birthweight.We analyzed cross-sectional baseline data in Greenlight Plus, an obesity prevention trial across 6 US cities (n=787), which included newborns with a gestational age greater than 34 weeks and a birthweight greater than 2500 grams. We assessed neighborhood food environment using the Place-Based Survey and food insecurity using the US Household Food Security Module. We performed logistic regression analyses to assess the individual and additive effects of risk factors on high birthweight. We adjusted for potential confounders: infant sex, race, ethnicity, gestational age, birthing parent age, education, income, and study site.Thirty-four percent of birthing parents reported poor neighborhood food environment and/or food insecurity. Compared to those without food insecurity, food insecure families had greater odds of delivering an infant with high birthweight (aOR 1.96, 95% CI: 1.01, 3.82) after adjusting for poor neighborhood food environment, which was not associated with high birthweight (aOR 1.35, 95% CI: 0.78, 2.34). Each additional risk to healthy food access was associated with a 56% (95% CI: 4%-132%) increase in high birthweight odds.Prenatal risks to healthy food access may increase high infant birthweight odds. Future studies designed to measure neighborhood factors should examine infant birthweight outcomes in the context of prenatal social determinants of health.

View details for DOI 10.1016/j.acap.2023.08.017

View details for PubMedID 37659601

Abstract

To assess if 100% fruit juice intake prior to 6 months is associated with juice and sugar-sweetened beverage (SSB) intake at 24 months and whether this differs by sociodemographic factors.We used longitudinal data from infants enrolled in the control (no obesity intervention) arm of Greenlight, a cluster randomized trial to prevent childhood obesity which included parent-reported child 100% fruit juice intake at all well child checks between two and 24 months. We studied the relationship between the age of juice introduction (before versus after six months) and juice and SSB intake at 24 months using negative binomial regression while controlling for baseline sociodemographic factors.We report results for 187 participants (43% Hispanic, 39% non-Hispanic Black), more than half (54%) of whom had reported 100% fruit juice intake before six months. Average 100% fruit juice intake at 24 months was greater than the recommended amount (of 4 oz) and was 8.2 oz and 5.3 oz for those who had and had not, respectively, been introduced to juice before six months. In adjusted models, early introduction of juice was associated with a 43% (95% CI: 5% to 96%) increase in juice intake at 24 months.100% fruit juice intake exceeding recommended levels at six and 24 months in this diverse cohort was prevalent. Introducing 100% fruit juice prior to six months may put children at greater risk for more juice intake as they age. Further research is necessary to determine if early guidance can reduce juice intake.

View details for DOI 10.1016/j.acap.2023.04.009

View details for PubMedID 37150479

Abstract

Background: Healthy lifestyle behaviors that can prevent adverse health outcomes, including obesity, are formed in early childhood. This study describes feeding, television, and sleep behaviors among one-year-old infants and examines differences by sociodemographic factors.Methods: Caregivers of one-year-olds presenting for well care at two clinics, control sites for the Greenlight Study, were queried about feeding, television time, and sleep. Adjusted associations between sociodemographic factors and behaviors were performed by modified Poisson (binary), multinomial logistic (multi-category), or linear (continuous) regression models.Results: Of 235 one-year-olds enrolled, 81% had Medicaid, and 45% were Hispanic, 36% non-Hispanic Black, 19% non-Hispanic White. Common behaviors included 20% exclusive bottle use, 32% put to bed with bottle, mean daily juice intake of 4.1±4.6 ounces, and active television time 45±73min. In adjusted analyses compared to Hispanic caregivers, non-Hispanic Black caregivers were less likely to report exclusive bottle use (odds ratio: 0.11, 95% confidence interval [CI] 0.03-0.39), reported 2.4 ounces more juice (95% CI 1.0-3.9), 124min more passive television time (95% CI 60-188), and 37min more active television time (95% CI 10-64). Increased caregiver education and higher income were associated with 0.4 (95% CI 0.13-0.66) and 0.3 (95% CI 0.06-0.55) more servings of fruits and vegetables per day, respectively.Conclusion: In a diverse sample of one-year-olds, caregivers reported few protective behaviors that reduce the risk for adverse health outcomes including obesity. Differences in behavior by race/ethnicity, income, and education can inform future interventions and policies. Future interventions should strive to create culturally effective messaging to address common adverse health behaviors.

View details for DOI 10.1016/j.obpill.2022.100051

View details for PubMedID 37990745

Abstract

OBJECTIVE: The purpose of this study is to understand how families from diverse sociodemographic backgrounds perceived the impact of the pandemic on the development of their children.METHODS: We used a multimethod approach guided by Bronfenbrenner's Ecological Systems Theory, which identifies 5 developmental systems (micro, meso, exo, macro, and chrono). Semistructured interviews were conducted in English or Spanish with parents living in 5 geographic regions of the United States between July and September 2021. Participants also completed the COVID-19 Exposure and Family Impact Survey.RESULTS: Forty-eight families participated, half of whose preferred language was Spanish, with a total of 99 children ages newborn to 19 years. Most qualitative themes pertained to developmental effects of the microsystem and macrosystem. Although many families described negative effects of the pandemic on development, others described positive or no perceived effects. Some families reported inadequate government support in response to the pandemic as causes of stress and potential negative influences on child development. As context for their infant's development, families reported a variety of economic hardships on the COVID-19 Exposure and Family Impact Survey, such as having to move out of their homes and experiencing decreased income.CONCLUSION: In addition to negative impacts, many parents perceived positive pandemic-attributed effects on their child's development, mainly from increased time for parent-child interaction. Families described economic hardships that were exacerbated by the pandemic and that potentially affect child development and insufficient government responses to these hardships. These findings hold important lessons for leaders who wish to design innovative solutions that address inequities in maternal, family, and child health.

View details for DOI 10.1097/DBP.0000000000001166

View details for PubMedID 36716765

View details for Web of Science ID 000928164500042

Abstract

INTRODUCTION: Social support is a modifiable social determinant of health that shapes breastfeeding outcomes and may contribute to racial and ethnic breastfeeding disparities. This study characterizes the relationship between social support and early breastfeeding.METHODS: This is a cross-sectional analysis of baseline data collected in 2019-2021 for an RCT. Social support was measured using the Enhancing Recovery in Coronary Heart Disease Social Support Instrument. Outcomes, collected by self-report, included (1) early breastfeeding within the first 21 days of life, (2) planned breastfeeding duration, and (3) confidence in meeting breastfeeding goals. Each outcome was modeled using proportional odds regression, adjusting for covariates. Analysis was conducted in 2021-2022.RESULTS: Self-reported race and ethnicity among 883 mothers were 50% Hispanic, 17% Black, 23% White, and 10% other. A large proportion (88%) of mothers were breastfeeding. Most breastfeeding mothers (82%) planned to breastfeed for at least 6 months, with more than half (58%) planning to continue for 12 months or more. Most women (65%) were confident or very confident in meeting their breastfeeding duration goal. In adjusted models, perceived social support was associated with planned breastfeeding duration (p=0.042) but not with early breastfeeding (p=0.873) or confidence in meeting breastfeeding goals (p=0.427). Among the covariates, maternal depressive symptoms were associated with lower breastfeeding confidence (p<0.001).CONCLUSIONS: The associations between perceived social support and breastfeeding outcomes are nuanced. In this sample of racially and ethnically diverse mothers, social support was associated with longer planned breastfeeding duration but not with early breastfeeding or breastfeeding confidence.

View details for DOI 10.1016/j.amepre.2022.10.002

View details for PubMedID 36460526

Abstract

Background Oral anticoagulation (OAC) reduces stroke and disability in atrial fibrillation (AF) but is underutilized. We evaluated the effects of a novel patient-clinician shared decision-making (SDM) tool in reducing OAC patient's decisional conflict as compared to usual care. Methods and Results We designed and evaluated a new digital decision aid in a multicenter, randomized, comparative effectiveness trial, ENHANCE-AF (Engaging Patients to Help Achieve Increased Patient Choice and Engagement for AF Stroke Prevention). The digital AF SDM Toolkit was developed using patient-centered design with clear health communication principles (e.g. meaningful images, limited text). Available in English and Spanish, the toolkit included the following: 1) a brief animated video; 2) interactive questions with answers; 3) a quiz to check on understanding; 4) a worksheet to be used by the patient during the encounter; and 5) an online guide for clinicians. The study population included English or Spanish speakers with non-valvular AF and a CHA2DS2-VASc stroke score ≥1 for men or ≥2 for women. Participants were randomized in a 1:1 ratio to either Usual Care (UC) or the SDM Toolkit. The primary endpoint was the validated 16-item Decisional Conflict Scale (DCS) at 1 month. Secondary outcomes included DCS at 6 months and the 10-item Decision Regret Scale (DRS) at 1 and 6 months as well as a weighted average of Mann-Whitney U-statistics for both DCS and DRS. A total of 1001 participants were enrolled and followed at 5 different sites in the United States between 12/18/19 and 8/17/22. The mean patient age was 69 ±10years (40% females, 16.9% Black, 4.5% Hispanic, 3.6% Asian), and 50% of participants had CHA2DS2-VASc scores ≥3 (M) or ≥4 (F). The primary endpoint at 1 month showed a clinically meaningful reduction in decisional conflict: a 7-point difference in median scores between the two arms (16.4 v 9.4; Mann-Whitney U-statistics=0.550; p-value=0.007). For the secondary endpoint of 1-month DRS, the difference in median scores between arms was 5 points in the direction of less decisional regret (p-value of 0.078). The treatment effects lessened over time: at 6 months the difference in medians was 4.7 points for DCS (p-value=0.060) and 0 points for DRS (p-value=0.35). Conclusions Implementation of a novel, Shared Decision-Making Toolkit (afibguide.com; afibguide.com/clinician) achieved significantly lower decisional conflict compared to usual care in patients with AF.

View details for DOI 10.1161/JAHA.122.028562

View details for PubMedID 36342828

Abstract

The first 1000 days of a child's life are increasingly recognized as a critical window for establishing a healthy growth trajectory to prevent childhood obesity and its associated long-term comorbidities. The purpose of this manuscript is to detail the methods for a multi-site, comparative effectiveness trial designed to prevent childhood overweight and obesity from birth to age 2 years.This study is a multi-site, individually randomized trial testing the comparative effectiveness of two active intervention arms: 1) the Greenlight intervention; and 2) the Greenlight Plus intervention. The Greenlight intervention is administered by trained pediatric healthcare providers at each well-child visit from 0 to 18 months and consists of a low health literacy toolkit used during clinic visits to promote shared goal setting. Families randomized to Greenlight Plus receive the Greenlight intervention plus a health information technology intervention, which includes: 1) personalized, automated text-messages that facilitate caregiver self-monitoring of tailored and age-appropriate child heath behavior goals; and 2) a web-based, personalized dashboard that tracks child weight status, progress on goals, and electronic Greenlight content access. We randomized 900 parent-infant dyads, recruited from primary care clinics across six academic medical centers. The study's primary outcome is weight for length trajectory from birth through 24 months.By delivering a personalized and tailored health information technology intervention that is asynchronous to pediatric primary care visits, we aim to achieve improvements in child growth trajectory through two years of age among a sample of geographically, socioeconomically, racially, and ethnically diverse parent-child dyads.

View details for DOI 10.1016/j.cct.2022.106987

View details for PubMedID 36323344

Abstract

Shared decision making (SDM) may result in treatment plans that best reflect the goals and wishes of patients, increasing patient satisfaction with the decision-making process.  There is a knowledge gap to support the use of decision aids in SDM for anticoagulation therapy in patients with atrial fibrillation (AF). We describe the development and testing of a new decision aid, including a multicenter, randomized, controlled, 2-arm, open-label ENHANCE-AF clinical trial (Engaging Patients to Help Achieve Increased Patient Choice and Engagement for AF Stroke Prevention) to evaluate its effectiveness in 1,200 participants.Participants will be randomized to either usual care or to a shared decision-making pathway incorporating a digital tool designed to simplify the complex concepts surrounding AF in conjunction with a clinician tool and a non-clinician navigator to guide the participants through each step of the tool. The participant-determined primary outcome for this study is the Decisional Conflict Scale, measured at 1 month after the index visit during which a decision was made regarding anticoagulation use. Secondary outcomes at both 1 and 6 months will include other decision making related scales as well as participant and clinician satisfaction, oral anticoagulation adherence, and a composite rate of major bleeding, death, stroke, or transient ischemic attack. The study will be conducted at four sites selected for their ability to enroll participants of varying racial and ethnic backgrounds, health literacy, and language skills. Participants will be followed in the study for 6 months.The results of the ENHANCE-AF trial will determine whether a decision aid facilitates high quality shared decision making in anticoagulation discussions for stroke reduction in AF. An improved shared decision-making experience may allow patients to make decisions better aligned with their personal values and preferences, while improving overall AF care.

View details for DOI 10.1016/j.ahj.2022.01.013

View details for PubMedID 35092723

Abstract

Undiagnosed concussions increase the risk of additional concussion and persistent symptoms from concussion. Because there are no reliable objective markers of concussion, self-reporting of subjective and non-visible symptoms are critical to ensuring proper concussion management. For this reason, educational interventions target concussion reporting, but the majority of studies have examined the efficacy of single educational interventions or compared interventions to one another. This randomized crossover study sought to identify whether there was benefit to administering multiple concussion education programs in tandem, back to back. The study randomized 313 male high school football players to first receive CrashCourse concussion education (CC) or Centers for Disease Control and Prevention video concussion education (CDC) followed by crossover with the other education. Athlete concussion-reporting intention, attitudes, subjective norms, perceived behavioral control, and enjoyment of education were assessed at baseline and after each intervention. There were statistically significant improvements across all measures, both after single intervention and crossover (all p < 0.001). Secondary analyses examining differences between education found that athletes reported higher enjoyment of concussion education immediately after participating in CC, as compared to CDC (p < 0.001). These findings demonstrate an additive benefit to implementing CC and CDC education in tandem, without decrement in enjoyment of concussion education after experiencing dual educations; in fact, enjoyment of concussion education improved after receiving education programs back to back. These educational programs appear to complement one another, and the results support the use of multi-modal concussion education to differentially target and maximize concussion reporting.

View details for DOI 10.1089/neur.2021.0033

View details for PubMedID 35018360

View details for PubMedCentralID PMC8742279

View details for DOI 10.1089/neur.2021.0033

View details for Web of Science ID 000729358800001

View details for DOI 10.1542/peds.2021-050448

View details for PubMedID 34433690

Abstract

BACKGROUND: There are few long-term studies of interventions to reduce in low socioeconomic status children with overweight or obesity. The Stanford GOALS trial evaluated a 3-year, community-based, multi-level, multi-setting, multi-component (MMM) systems intervention, to reduce weight gain among low socioeconomic status, Latinx children with overweight or obesity.METHODS: We did a two-arm, parallel group, randomised, open-label, active placebo-controlled trial with masked assessment over 3 years. Families from low-income, primarily Latinx communities in Northern California, CA, USA, with 7-11-year-old children with overweight or obesity were randomly assigned to a MMM intervention or a Health Education (HE) comparison intervention. The MMM intervention included home environment changes and behavioural counselling, community after school team sports, and reports to primary health-care providers. The primary outcome was child BMI trajectory over three years. Secondary outcomes included one- and two-year changes in BMI. This trial is registered with ClinicalTrials.govNCT01642836.FINDINGS: Between July 13, 2012, and Oct 3, 2013, 241 families were recruited and randomly assigned to MMM (n=120) or HE (n=121). Children's mean age was 9·5 (SD 1·4) years, 134 (56%) were female and 107 (44%) were male, and 236 (98%) were Latinx. 238 (99%) children participated in year 1, 233 (97%) in year 2, and 227 (94%) in year 3 of follow-up assessments. In intention-to-treat analysis, over 3 years, the difference between intervention groups in BMI trajectory was not significant (mean adjusted difference -0·25 [95% CI -0·90 to 0·40] kg/m2; Cohen's d=0.10; p=0·45). Children in the MMM intervention group gained less BMI over 1 year than did children in the HE intervention group (-0·73 [-1·07 to -0·39] kg/m2, d=0.55); the same was true over 2 years (-0·63 [-1·13 to -0·14] kg/m2; d =0.33). No differential adverse events were observed.INTERPRETATION: The MMM intervention did not reduce BMI gain versus HE over 3 years but the effects over 1 and 2 years in this rigorous trial show the promise of this systems intervention approach for reducing weight gain and cardiometabolic risk factors in low socioeconomic status communities.FUNDING: US National Institutes of Health.

View details for DOI 10.1016/S2213-8587(21)00084-X

View details for PubMedID 33933181

Abstract

CONTEXT: Concussion underreporting leads to delays in diagnosis and treatment, prolonging recovery time. Athletes' self-report of concussion symptoms therefore reduces risk.OBJECTIVE: Evaluate the efficacy of three concussion education programs in improving concussion-reporting intention.DESIGN: Randomized controlled clinical trial with assessment immediately and one-month after education.SETTING: Three high schools in California.PATIENTS OR OTHER PARTICIPANTS: 118 male football players were randomly assigned to receive concussion education via: CrashCourse (CC), Centers for Disease Control (CDC) video educational materials (Vi), or CDC written educational materials (Wr).MAIN OUTCOME MEASURES: Concussion-reporting intention was assessed at baseline, immediately after education, and at one-month follow-up. Secondary outcomes included concussion knowledge, attitudes, perceived reporting norms, and perceived behavioral control.RESULTS: Athletes across all educational formats had significant improvement in concussion-reporting intention immediately following education and at one-month follow-up (mean improvement 6.8% and 11.4%, respectively; p<0.001). Similar findings were observed across all education formats in secondary analyses examining knowledge, attitudes, and perceived behavioral control. However, there were significant differences by education and time (p=0.03). On post-hoc analysis, athletes who received CC had increased concussion-reporting intention immediately and at one-month (baseline=4.7, immediate=6.1, one-month=6.0; p=0.007 compared to significant increases only at one-month for CDC-Vi (baseline=4.3, immediate=5.2, one-month=5.8; p=0.001), and no significant improvement for CDC-Wr (p=0.10). Secondary analyses indicated significant differences between CC and both CDC interventions, in concussion knowledge and attitudes, immediately after education and at one-month. There were no significant differences in perceived behavioral control between-interventions or in perceived concussion-reporting norms across or between interventions.CONCLUSION: All athletes exhibited improved intent to report concussions, increased concussion knowledge, better concussion attitudes, and more perceived behavioral control, both immediately after education and at one-month follow-up. However, athletes randomized to CC reported greater intent to report concussion, more knowledge, and improved concussion-reporting attitudes, when compared to CDC-Vi and CDC-Wr.TRIAL REGISTRY: ClinicalTrials.gov trial ID number is XXX.

View details for DOI 10.4085/509-20

View details for PubMedID 33428746

Abstract

Undiagnosed concussions increase risk of additional injuries and can prolong recovery. Because of the difficulties recognizing concussive symptoms, concussion education must specifically target improving athlete concussion reporting. Many concussion education programs are designed without significant input from athletes, resulting in a less enjoyable athlete experience, with potential implications on program efficacy. Athlete enjoyment of previous concussion education programs moderates the improvement in concussion-reporting intention after experiencing the research version of CrashCourse (CC) concussion education. Prospective cohort study. Level of evidence: Level IV. Quantitative assessment utilizing ANOVA with moderation analysis of 173 male high school football players, aged 13 to 17, who completed baseline assessments of concussion knowledge, concussion reporting, and attitudes about prior educational interventions. Athletes were subsequently shown CC, before a follow-up assessment was administered assessing the same domains. At baseline, only 58.5% of athletes reported that they enjoyed their previous concussion education. After CC, athletes were significantly more likely to endorse that they would report a suspected concussion (from 69.3% of athletes to 85.6%; P<.01). Enjoyment of previous concussion education moderated concussion-reporting intention after CC (P=.02), with CC having a greater effect on concussion-reporting intention in athletes with low enjoyment of previous concussion education (b=0.21, P=.02), than on individuals with high enjoyment of previous concussion education (P=.99). Enjoyment of CC did not have a moderating effect on concussion-reporting intention. Athletes who previously did not enjoy concussion education exhibited greater gains in concussion-reporting intention than athletes who enjoyed previous education. Given the potential risks associated with undiagnosed concussions, concussion education has sought to improve concussion reporting. Because most athletes participate in concussion education programs due to league or state mandates, improving concussion-reporting intention in these low-enjoyment athletes is of particular relevance to improving concussion-reporting intention broadly.

View details for DOI 10.1177/00469580211022641

View details for PubMedID 34053328

View details for Web of Science ID 000672597101386

Abstract

Gastrostomy tube (G-tube) placement for children with neurologic impairment with dysphagia has been suggested for pneumonia prevention. However, prior studies demonstrated an association between G-tube placement and increased risk of pneumonia. We evaluate the association between timing of G-tube placement and death or severe pneumonia in children with neurologic impairment.We included all children enrolled in California Children's Services between July 1, 2009, and June 30, 2014, with neurologic impairment and 1 pneumonia hospitalization. Prior to analysis, children with new G-tubes and those without were 1:2 propensity score matched on sociodemographics, medical complexity, and severity of index hospitalization. We used a time-varying Cox proportional hazard model for subsequent death or composite outcome of death or severe pneumonia to compare those with new G-tubes vs those without, adjusting for covariates described above.A total of 2490 children met eligibility criteria, of whom 219 (9%) died and 789 (32%) had severe pneumonia. Compared to children without G-tubes, children with new G-tubes had decreased risk of death (hazard ratio [HR] 0.47, 95% confidence interval [CI] 0.39-0.55) but increased risk of the composite outcome (HR 1.21, CI 1.14-1.27). Sensitivity analyses using varied time criteria for definitions of G-tube and outcome found that more recent G-tube placement had greater associated risk reduction for death but increased risk of severe pneumonia.Recent G-tube placement is associated with reduced risk of death but increased risk of severe pneumonia. Decisions to place G-tubes for pulmonary indications in children with neurologic impairment should weigh the impact of severe pneumonia on quality of life.

View details for DOI 10.1177/08830738211000179

View details for PubMedID 33750232

Abstract

OBJECTIVES: This study examines the development of active television (TV) watching behaviors across the first 2 years of life in a racially and ethnically diverse, low-income cohort and identifies caregiver and child predictors of early TV watching.METHODS: We used longitudinal data from infants enrolled in the active control group (N = 235; 39% Latino; 29% black; 15% white) of Greenlight, a cluster randomized multi-site trial to prevent childhood obesity. At preventive health visits from 2 months to 2 years, caregivers were asked: "How much time does [child's first name] spend watching television each day?" Proportional odds models and linear regression analyses were used to assess associations among TV introduction age, active TV watching amount at 2 years, and sociodemographic factors.RESULTS: 68% of children watched TV by 6 months, and 88% by 2 years. Age of TV introduction predicted amount of daily active TV watching at 2 years, with a mean time of 93 minutes if starting at 2 months; 64 minutes if starting at 4 or 6 months; and 42 minutes if starting after 6 months. Factors predicting earlier introduction included lower income, fewer children in household, care away from home, male sex, and non-Latino ethnicity of child.CONCLUSIONS: Many caregivers report that their infants actively watch TV in the first 6 months of life. Earlier TV watching is related to sociodemographic factors yet predicts more daily TV watching at 2 years even controlling those factors. Interventions to limit early TV watching should be initiated in infancy.

View details for DOI 10.1016/j.acap.2020.11.002

View details for PubMedID 33161116

Abstract

Low-income racially and ethnically diverse children are at higher risk for obesity compared with their counterparts; yet, few studies have assessed their diet quality.The aim of the study was to evaluate the diet quality of a racially and ethnically diverse cohort of 2-year-olds using the Healthy Eating Index (HEI)-2010.We used 24-hour dietary recall data from caregivers of toddlers (24-34 months) at 4 pediatric resident clinics that participated in the Greenlight Study to calculate compliance with the Dietary Guidelines for Americans (DGA) using total HEI score (range 0-100) and 12 component scores.Participants (n = 231) were mostly Hispanic (57%) or non-Hispanic black (27%) and from low-income families. Mean HEI-2010 score was 62.8 (standard deviation [SD] 10.5). Though not significant, Hispanics had the highest HEI score. Toddlers of caregivers without obesity, older than 35 years and born outside the United States had higher HEI scores. Most had high HEI component scores for dairy, fruit, and protein foods, but few achieved maximum scores, particularly for whole grains (13%), vegetables (10%), and fatty acid ratio (7%).Despite scores reflective of DGA recommendations for fruit, dairy and protein foods, toddlers in this diverse sample had low quality diets as measured by the HEI, driven largely by low component scores for whole grains, vegetables, and ratio of unsaturated to saturated fatty acids.

View details for DOI 10.1097/MPG.0000000000002871

View details for PubMedID 33093378

View details for Web of Science ID 000587738600021

Abstract

Although pediatricians routinely counsel parents about preventing childhood injuries, we know little about parents' locus of control (LOC) in regards to preventing their children from being injured. We performed an observational analysis of sociodemographic differences in LOC for injury prevention, as measured by four items adapted from the Parental Health Beliefs Scales, in English- and Spanish-speaking parents of infants participating in the treatment arm of an obesity prevention study. First, we examined associations of parental LOC for injury prevention at the time their children were 2months old with parents' age, race/ethnicity, income, and education. Next, we analyzed time trends for repeated LOC measures when the children were 2, 6, 9, 12, and 24months old. Last, we examined the association between injury-related LOC items and children's injury (yes/no) at each time point. Of 452 parents, those with lower incomes had both lower internal and higher external LOC. Lower educational achievement was associated with higher external LOC. Both internal and external LOC scores decreased over time. Injuries were more common in children whose parents endorsed low internal and high external LOC. Future studies should examine whether primary care-based interventions can increase parents' sense of control over their children's safety and whether that, in turn, is associated with lower injury rates.Clinical Trial Registration: NCT01040897.

View details for DOI 10.1007/s10935-020-00615-y

View details for PubMedID 33104944

Abstract

OBJECTIVE: To determine whether proportion of breast versus formula feeding, and timing of complementary food introduction affect the odds of rapid gain in weight status in a diverse sample of infants.METHODS: Using data from Greenlight Intervention Study, we analyzed the effects of type of milk feeding (breastfeeding, formula, or mixed feeding) from the 2 to 6 month well visits, and the introduction of complementary foods before 4 months on rapid increase in weight-for-age z-score (WAZ) and weight-for-length z-score (WLZ) before 12 months using multivariable logistic regression models.RESULTS: Of the 865 infants enrolled, 469 had complete data on all variables of interest, and 41% and 33% of those infants had rapid increases in WAZ and WLZ, respectively. Odds of rapid increase in WAZ remained lowest for infants breastfeeding from 2 to 6 months (aOR 0.34; 95% CI: 0.17, 0.69) when compared to infants who were formula fed. Adjusted for feeding, introduction of complementary foods after 4 months was associated with decreased odds of rapid increase in WLZ (aOR 0.64; 95% CI: 0.42, 0.96).CONCLUSION: Feeding typified by predominant breastfeeding and delaying introduction of complementary foods after 4 months reduces the odds of rapid increases in WAZ and WLZ in the first year of life.

View details for DOI 10.1016/j.acap.2020.09.009

View details for PubMedID 32961335

Abstract

BACKGROUND: Low-income racially and ethnically diverse children are at higher risk for obesity compared with their counterparts; yet, few studies have assessed their diet quality.OBJECTIVE: To evaluate the diet quality of a racially and ethnically diverse cohort of 2-year-olds using the Healthy Eating Index (HEI)-2010.METHODS: We used 24-hour dietary recall data from caregivers of toddlers (24-34 months) at 4 pediatric resident clinics that participated in the Greenlight Study to calculate compliance with the Dietary Guidelines for Americans (DGA) using total HEI score (range 0-100) and 12 component scores.RESULTS: Participants (n = 231) were mostly Hispanic (57%) or non-Hispanic black (27%) and from low-income families. Mean HEI-2010 score was 62.8 (SD 10.5). Though not significant, Hispanics had the highest HEI score. Toddlers of caregivers without obesity, older than 35 years and born outside the U.S. had higher HEI scores. Most had high HEI component scores for dairy, fruit, and protein foods, but few achieved maximum scores, particularly for whole grains (13%), vegetables (10%), and fatty acid ratio (7%).CONCLUSION: Despite scores reflective of DGA recommendations for fruit, dairy and protein foods, toddlers in this diverse sample had low quality diets as measured by the HEI, driven largely by low component scores for whole grains, vegetables and ratio of unsaturated to saturated fatty acids.

View details for DOI 10.1097/MPG.0000000000002871

View details for PubMedID 32740532

Abstract

OBJECTIVE: Shared decision-making (SDM) may improve outcomes for children with medical complexity (CMC). CMC have lower rates of SDM than other children, but little is known about how to improve SDM for CMC. The objective of this study is to describe parent perspectives of SDM for CMC and identify opportunities to improve elements of SDM specific to this vulnerable population.METHODS: Interviews with parents of CMC explored SDM preferences and experiences. Eligible parents were ≥18 years old, English- or Spanish-speaking, with a CMC < 12 years old. Interviews were recorded, transcribed, and analyzed by independent coders for shared themes using modified grounded theory. Codes were developed using an iterative process, beginning with open-coding of a subset of transcripts followed by discussion with all team members, and distillation into preliminary codes. Subsequent coding reviews were conducted until no new themes emerged and existing themes were fully explored.RESULTS: We conducted interviews with 32 parents (27 in English, mean parent age 34 years, SD=7; mean child age 4 years, SD=4; 50% with household income <

Abstract

Background: Little is known about intended breastfeeding duration of women who initiate breastfeeding. We describe the association between intended and actual breastfeeding duration among low-income, diverse mothers who report maintaining breastfeeding for the first 2 months postpartum. Materials and Methods: We included mothers (64% Hispanic, 17% non-Hispanic black) participating in Greenlight, a cluster randomized childhood obesity prevention trial, who were providing breast milk at the 2-month preventive service visit and reported intended breastfeeding duration at this visit. Breastfeeding status was assessed at subsequent visits, up to 24 months. Poisson regression with a robust variance estimator was used to estimate risk ratios and 95% confidence intervals for meeting breastfeeding intentions. Covariates included race/ethnicity, income, receiving benefits from the Special Supplemental Nutrition Assistance Program for Women, Infants and Children (WIC), education, age, employment, depression, maternal obesity, U.S. born, whether infant was first born, and study site. Results: Median intended breastfeeding duration was 11.5 months (interquartile range [IQR]: 6-12) and median actual breastfeeding duration was 8.6 months (IQR: 4-14) (n=349). Approximately half (49%) met intended breastfeeding duration. Breastfeeding duration differed based on milk type provided at the 2-month visit in that mothers providing mostly or only breast milk had increased likelihood of meeting breastfeeding intentions. Regardless of milk type provided at 2 months, the longer a mother intended to breastfeed, the less likely she was to meet her breastfeeding intentions. Conclusions: In this diverse sample of women less than half met breastfeeding intentions despite maintaining breastfeeding for 2 months. Understanding factors that prevent mothers from attaining intended breastfeeding duration is critical to improving breastfeeding outcomes, especially in low income and ethnic minority populations.

View details for DOI 10.1089/bfm.2020.0025

View details for PubMedID 32357088

Abstract

We examined associations between household food insecurity status and parental feeding behavior, weight perception, and child weight status in a diverse sample of young children.Cross-sectional analysis of 2-year old children in Greenlight, a cluster randomized trial to prevent childhood obesity. The exposure was food insecurity, defined as a positive response to a validated screen. Outcomes were parent feeding behaviors/beliefs measured by the Child Feeding Questionnaire and child weight status. T-tests and linear regression were used to assess associations between food insecurity and each outcome. We adjusted for child sex, race/ethnicity, parent education, employment, site, number of children in the home, and WIC status.503 households (37%) were food insecure. After adjusting for covariates, parents from insecure households reported more pressuring feeding behaviors (mean factor score 3.2 compared to food secure parents mean factor score 2.9, p=0.01) and were more worried about their child becoming overweight (mean factor score 2.3 vs 2.0; p=0.02). No differences were observed in monitoring or restrictive feeding behaviors. After adjusting for covariates, there was no difference in weight status or prevalence of overweight/obesity of children or parents based on household food insecurity status.Parents from food insecure households reported more pressuring feeding behaviors. This finding underscores the need to address food insecurity and potentially prevent harmful effects on child feeding. Parents in food insecure households might benefit from linkage with resources and education to develop healthier feeding behaviors.

View details for DOI 10.1016/j.acap.2020.05.020

View details for PubMedID 32492577

View details for DOI 10.1542/peds.2018-2228

View details for Web of Science ID 000474923900009

Abstract

OBJECTIVE: Parent use of technology to manage child health issues has the potential to improve access and health outcomes. Few studies have examined how parent health literacy affects usage of Internet and cell phone technologies for health management.METHODS: Cross-sectional analysis of data collected as part of a randomized controlled experiment in 3 urban pediatric clinics. English- and Spanish-speaking parents (n=858) of children ≤8 years answered questions regarding use of and preferences related to Internet and cell phone technologies. Parent health literacy was measured using the Newest Vital Sign.RESULTS: The majority of parents were high Internet (70.2%) and cell phone (85.1%) utilizers (multiple times a day). 75.1% had limited health literacy (32.1% marginal, 43.0% low). Parents with higher health literacy had greater Internet and cell phone use (adequate vs. low: AOR=1.7[1.2-2.5]) and were more likely to use them for health management (AOR=1.5[1.2-1.8]); those with higher health literacy were more likely to use the Internet for provider communication (adequate vs. marginal vs. low: 25.0 vs. 18.0 vs. 12.0%, p=0.001) and health-related cell phone apps (40.6 vs. 29.7 vs. 16.4%, p<0.001). Overall preference for using technology for provider communication was high (70%) and did not differ by health literacy, although Internet and cell phone apps were preferred by higher literacy parents; no differences seen for texting.CONCLUSIONS: Health literacy-associated disparities in parent use of Internet and cell phone technologies exist, but parents' desire for use of these technologies for provider communication was overall high and did not differ by health literacy.

View details for PubMedID 30862511

View details for DOI 10.1213/ANE.0000000000004552

Abstract

Time spent commuting is associated with obesity. The objective of this study was to assess the relationship between neighborhood-level commute to work (CTW) times and self-reported health behaviors and food access.We conducted a cross-sectional analysis of caregivers with infants as part of the Greenlight Study, a multisite obesity trial in Chapel Hill, New York City, Nashville, and Miami. ZIP code-based commuting estimates were determined using the US Census American Community Survey. Self-reported health behaviors and food access data were collected by directed interview. Logistic and linear regression models were used to determine associations between neighborhood CTW times and health behaviors and food access.The average neighborhood CTW time for all ZIP codes was 29 minutes (n = 846). Caregivers in longer CTW time neighborhoods were more likely to endorse fewer food choices (adjusted odds ratio [AOR], 1.39; 95% confidence interval [CI], 1.15-1.69; P = .001) and difficulty accessing markets with fresh produce (AOR, 1.51; 95% CI, 1.02-2.25; P = .04). Neighborhood CTW time >30 minutes was associated with less caregiver physical activity (AOR, 0.58; 95% CI, 0.34-0.98; P = .044). Neighborhood CTW time was inversely related to infant television time (adjusted mean, 399 minutes/day for ≤30 minutes and 256 minutes/day for >30 minutes; P = .025). New York families in longer CTW neighborhoods were more likely to report difficulty accessing markets with fresh produce (AOR, 1.80; 95% CI, 1.03-3.14; P = .039).Neighborhood CTW time is associated with several self-reported health behaviors and perceived food access among caregivers with children. Neighborhood CTW times may represent city-specific features, including transportation infrastructure, which may impact the health of families.

View details for PubMedID 30041009

View details for PubMedID 30096446

Abstract

Food insecurity is associated with childhood obesity possibly mediated through caregiver feeding practices and beliefs. We examined if caregiver feeding practices differed by household food security status in a diverse sample of infants. We hypothesized that feeding practices differ based on food security status.Included in the baseline cross-sectional analysis of data from a randomized controlled trial to prevent obesity were 842 caregivers of 2-month-old infants presenting for well-child care at 4 academic institutions. Food insecurity exposure was based on an affirmative answer to 1 of 2 items in a 2-item validated questionnaire. Chi-square tests examined the association between parent feeding practices and food security status. Logistic regression adjusted for covariates. Differences in caregiver feeding practices by food security status and race/ethnicity were explored with an interaction term (food security status x race/ethnicity).Forty-three percent of families screened as food insecure. In adjusted logistic regression, parents from food-insecure households were more likely to endorse that "the best way to make an infant stop crying is to feed him or her" (adjusted odds ratio [aOR], 1.72; 95% confidence interval [CI], 1.28-2.29) and "when my baby cries, I immediately feed him or her" (aOR, 1.40; 95% CI, 1.06-1.83). Food-insecure caregivers less frequently endorsed paying attention to their baby when he or she is full or hungry (OR, 0.57; 95% CI, 0.34-0.96). Racial/ethnic differences in beliefs and behaviors were observed by food security status.During early infancy, feeding practices differed among caregivers by household food security status. Further research is needed to examine whether these practices are associated with increased risk of obesity and obesity-related morbidity.

View details for PubMedID 30248471

Abstract

OBJECTIVES: Black women are at increased risk of being called back for additional studies after a screening mammogram. With focus group input, we developed a brochure to improve awareness of the frequency of abnormal results. This study explored the brochure's acceptability and effect on understanding risk and breast cancer fears among black mammography patients at an urban safety-net breast imaging center in Miami, Florida.METHODS: A randomized controlled trial of the brochure (plus the standard result notification letter) versus usual care (standard notification letter alone). Black English-speaking women with an incomplete mammography result were randomized to the intervention or control group. Consenting participants completed a telephone questionnaire. Outcomes included awareness of result, anxiety level, and brochure acceptability. The chi2 or Fisher exact test was used and a univariate logistic regression was performed for intervention and control odds ratios.RESULTS: A total of 106 women were randomly selected to receive the brochure plus the letter or the letter alone. One chose to opt out; a minimum of three attempts were made to reach each of the remaining 105 women by telephone. Verbal communication was established with 59 of the randomized women, and 51 of those women agreed to participate in a survey to evaluate the brochure. There was no significant difference between the surveyed groups in knowledge of the result and follow-up plan. Surveyed intervention subjects were more likely to agree that "it is very common for women to have to follow up after a mammogram" (odds ratio [OR] 25.91, P = 0.029) and less likely to agree with the statement "getting a follow-up mammogram is scary" (OR 0.24, P = 0.021). Most intervention subjects said the pamphlet helped them understand their result "a lot" (79%, 19) and viewed it as "extremely" or "mostly" clear (96%, 23). Intervention subjects also voiced greater awareness of a telephone number they could call for more information about cancer (OR 11.38, P = 0.029).CONCLUSIONS: A culturally tailored brochure explaining the frequency of abnormal mammograms was well received by women at a large safety-net health system. Pilot testing suggests that it may improve patient perception of risk and awareness of informational resources. This strategy should be considered to enhance result communication.

View details for PubMedID 30608622

View details for DOI 10.14423/SMJ.0000000000000919

View details for Web of Science ID 000455050000001

Abstract

To assess parent decision-making regarding dosing tools, a known contributor to medication dosing errors, by evaluating parent dosing tool use, beliefs, and access, and the role of health literacy, with a focus on dosing cups, which are associated with an increased risk of multifold overdose.Cross-sectional analysis of data collected for randomized controlled study in 3 urban pediatric clinics. English/Spanish-speaking parents (n = 493) of children ≤8 years of age enrolled.reported tool use, beliefs, and access. Predictor variable: health literacy (Newest Vital Sign; limited [0-3], adequate [4-6]). Multiple logistic regression analyses conducted.Over two-thirds of parents had limited health literacy. Oral syringes (62%) and dosing cups (22%) were most commonly used. Overall, 24% believed dosing cups were the best tool type for dosing accuracy; 99% reported having access to ≥1 dosing tools with standard measurement markings. Parents with limited health literacy had greater odds of dosing cup use (limited vs adequate: aOR = 2.4 [1.2-4.6]). Parents who believed that dosing cups are best for accuracy had greater odds of dosing cup use (aOR = 16.3 [9.0-29.3]); this belief mediated health literacy-effects on dosing cup use.Factors associated with dosing tool choice, including parent health literacy and beliefs are important to consider in the design of interventions to reduce dosing errors; future larger-scale studies addressing this issue are needed.

View details for DOI 10.1016/j.jpeds.2019.08.017

View details for PubMedID 31604631

Abstract

To develop a nurse staffing prediction model and evaluate deviation from predicted nurse staffing as a contributor to patient outcomes.Secondary data collection conducted 2017-2018, using the California Office of Statewide Health Planning and Development and the California Perinatal Quality Care Collaborative databases. We included 276 054 infants born 2008-2016 and cared for in 99 California neonatal intensive care units (NICUs).Repeated-measures observational study. We developed a nurse staffing prediction model using machine learning and hierarchical linear regression and then quantified deviation from predicted nurse staffing in relation to health care-associated infections, length of stay, and mortality using hierarchical logistic and linear regression.We linked NICU-level nurse staffing and organizational data to patient-level risk factors and outcomes using unique identifiers for NICUs and patients.An 11-factor prediction model explained 35 percent of the nurse staffing variation among NICUs. Higher-than-predicted nurse staffing was associated with decreased risk-adjusted odds of health care-associated infection (OR: 0.79, 95% CI: 0.63-0.98), but not with length of stay or mortality.Organizational and patient factors explain much of the variation in nurse staffing. Higher-than-predicted nurse staffing was associated with fewer infections. Prospective studies are needed to determine causality and to quantify the impact of staffing reforms on health outcomes.

View details for DOI 10.1111/1475-6773.13249

View details for PubMedID 31869865

Abstract

Purpose Intensity of end-of-life care receives much attention in oncology because of concerns that high-intensity care is inconsistent with patient goals, leads to worse caregiver outcomes, and is expensive. Little is known about such care in those undergoing allogeneic hematopoietic cell transplantation (HCT), a population at high risk for morbidity and mortality. Patients and Methods We conducted a population-based analysis of patients who died between 2000 and 2013, within 1 year of undergoing an inpatient allogeneic HCT using California administrative data. Previously validated markers of intensity were examined and included: hospital death, intensive care unit (ICU) admission, and procedures such as intubation and cardiopulmonary resuscitation at end of life. Multivariable logistic regression models determined clinical and sociodemographic factors associated with: hospital death, a medically intense intervention (ICU admission, cardiopulmonary resuscitation, hemodialysis, intubation), and ≥ two intensity markers. Results Of the 2,135 patients in the study population, 377 were pediatric patients (age ≤ 21 years), 461 were young adults (age 22 to 39 years), and 1,297 were adults (age ≥ 40 years). The most common intensity markers were: hospital death (83%), ICU admission (49%), and intubation (45%). Medical intensity varied according to age, underlying diagnosis, and presence of comorbidities at time of HCT. Patients with higher-intensity end-of-life care included patients age 15 to 21 years and 30 to 59 years, patients with acute lymphoblastic leukemia, and those with comorbidities at time of HCT. Conclusion Patients dying within 1 year of inpatient allogeneic HCT are receiving medically intense end-of-life care with variations related to age, underlying diagnosis, and presence of comorbidities at time of HCT. Future studies need to determine if these patterns are consistent with patient and family goals.

View details for PubMedID 30183467

View details for Web of Science ID 000443532400001

Abstract

To ensure timely appropriate care for low-birth-weight (LBW) infants, healthcare providers must communicate effectively with parents, even when language barriers exist. We sought to evaluate whether non-English primary language (NEPL) and professional in-person interpreter use were associated with differential hospital length of stay for LBW infants, who may incur high healthcare costs. We analyzed data for 2047 infants born between 1 January 2008 and 30 April 2013 with weight <2500 g at one hospital with high NEPL prevalence. We evaluated relationships of NEPL and in-person interpreter use on length of stay, adjusting for medical severity. Overall, 396 (19%) had NEPL parents. Fifty-three percent of NEPL parents had documented interpreter use. Length of stay ranged from 1 to 195 days (median 11). Infants of NEPL parents with no interpreter use had a 49% shorter length of stay (adjusted incidence rate ratio (IRR) 0.51, 95% confidence interval (CI) 0.43⁻0.61) compared to English-speakers. Infants of parents with NEPL and low interpreter use (<25% of hospital days) had a 26% longer length of stay (adjusted IRR 1.26, 95% CI 1.06⁻1.51). NEPL and high interpreter use (>25% of hospital days) showed a trend for an even longer length of stay. Unmeasured clinical and social/cultural factors may contribute to differences in length of stay.

View details for DOI 10.3390/ijerph15081570

View details for PubMedID 30044374

View details for PubMedCentralID PMC6121500

View details for PubMedID 29966712

Abstract

The aim of this study was to assess: (1) the proportion of youth with special health care needs (YSHCN) with adequate transition preparation, (2) whether transition preparation differs by individual, condition-related and health care system-related factors, and (3) whether specific components of the medical home are associated with adequate transition preparation. The authors conducted a cross-sectional analysis of the 2009-2010 National Survey of Children with Special Health Care Needs, which surveyed a nationally representative sample of 17,114 parents of YSHCN ages 12 to 18 years. Adequate transition preparation was based on positive responses to questions about transition to an adult provider, changing health care needs, maintaining insurance coverage, and if providers encouraged YSHCN to take responsibility for health care needs. Weighted descriptive, bivariate and multivariate analyses were conducted to determine the association between patient and health care system factors and adequate transition preparation. Overall, 32.1% of YSHCN had adequate transition preparation. Older age, female sex, income ≤400% of the poverty level, lack of medical complexity, and having shared decision making, family-centered care, and effective care coordination were associated with increased odds of transition preparation. The majority of YSHCN do not receive adequate transition preparation and younger, male adolescents with medical complexity were less likely to receive transition preparation. Different patterns of disparities were identified for each subcomponent measure of transition preparation, which may help target at-risk populations for specific services. Efforts to improve transition preparation should leverage specific components of the medical home including care coordination, shared decision making, and family-centered care.

View details for PubMedID 29957127

Abstract

Studies of adolescent and young adult (AYA) oncology end-of-life care utilization are critical because cancer is the leading cause of nonaccidental AYA death and end-of-life care contributes significantly to health care expenditures. This study was designed to determine the quantity of and disparities in inpatient utilization in the last year of life of AYAs with cancer.The California Office of Statewide Health Planning and Development administrative discharge database, linked to death certificates, was used to perform a population-based analysis of cancer patients aged 15 to 39 years who died in 2000-2011. The number of hospital days and the inpatient costs were determined for each patient in the last year of his or her life, as were clinical and sociodemographic factors associated with high inpatient utilization. Admission patterns as death approached were also evaluated.The 12,883 patients were admitted for 40 days on average in the last year of life, and this cost

View details for PubMedID 28709388

Abstract

Children with medical complexity (CMC) are a subset of children and youth with special health care needs with high resource use and health care costs. Novel care delivery models in which care coordination and other services to CMC are provided are a focus of national and local health care and policy initiatives. Current models of care for CMC can be grouped into 3 main categories: (1) primary care-centered models, (2) consultative- or comanagement-centered models, and (3) episode-based models. Each model has unique advantages and disadvantages. Evaluations of these models have demonstrated positive outcomes, but most studies have limited generalizability for broader populations of CMC. A lack of standardized outcomes and population definitions for CMC hinders assessment of the comparative effectiveness of different models of care and identification of which components of the models lead to positive outcomes. Ongoing challenges include inadequate support for family caregivers and threats to the sustainability of models of care. Collaboration among key stakeholders (patients, families, providers, payers, and policy makers) is needed to address the gaps in care and create best practice guidelines to ensure the delivery of high-value care for CMC.

View details for PubMedID 29496972

Abstract

Private health insurance is associated with improved outcomes in patients with cancer. However, to the authors' knowledge, little is known regarding the impact of the Patient Protection and Affordable Care Act Dependent Coverage Expansion (ACA-DCE), which extended private insurance to young adults (to age 26 years) beginning in 2010, on the insurance status of young adults with cancer.The current study was a retrospective, population-based analysis of hospitalized young adult oncology patients (aged 22-30 years) in California during 2006 through 2014 (11,062 patients). Multivariable regression analyses examined factors associated with having private insurance. Results were presented as adjusted odds ratios and 95% confidence intervals. A difference-in-difference analysis examined the influence of the ACA-DCE on insurance coverage by race/ethnicity and federal poverty level.Multivariable regression demonstrated that patients of black and Hispanic race/ethnicity were less likely to have private insurance before and after the ACA-DCE, compared with white patients. Younger age (22-25 years) was associated with having private insurance after implementation of the ACA-DCE (odds ratio, 1.20; 95% confidence interval, 1.06-1.35). In the difference-in-difference analysis, private insurance increased among white patients aged 22 to 25 years who were living in medium-income (2006-2009: 64.6% vs 2011-2014: 69.1%; P = .003) and high-income (80.4% vs 82%; P = .043) zip codes and among Asians aged 22 to 25 years living in high-income zip codes (73.2 vs 85.7%; P = .022). Private insurance decreased for all Hispanic patients aged 22 to 25 years between the 2 time periods.The ACA-DCE provision increased insurance coverage, but not among all patients. Private insurance increased for white and Asian patients in higher income neighborhoods, potentially widening social disparities in private insurance coverage among young adults with cancer. Cancer 2018;124:110-7. © 2017 American Cancer Society.

View details for PubMedID 28940423

Abstract

To ensure timely appropriate care for low-birth-weight (LBW) infants, healthcare providers must communicate effectively with parents, even when language barriers exist. We sought to evaluate whether non-English primary language (NEPL) and professional in-person interpreter use were associated with differential hospital length of stay for LBW infants, who may incur high healthcare costs. We analyzed data for 2047 infants born between 1 January 2008 and 30 April 2013 with weight <2500 g at one hospital with high NEPL prevalence. We evaluated relationships of NEPL and in-person interpreter use on length of stay, adjusting for medical severity. Overall, 396 (19%) had NEPL parents. Fifty-three percent of NEPL parents had documented interpreter use. Length of stay ranged from 1 to 195 days (median 11). Infants of NEPL parents with no interpreter use had a 49% shorter length of stay (adjusted incidence rate ratio (IRR) 0.51, 95% confidence interval (CI) 0.43⁻0.61) compared to English-speakers. Infants of parents with NEPL and low interpreter use (<25% of hospital days) had a 26% longer length of stay (adjusted IRR 1.26, 95% CI 1.06⁻1.51). NEPL and high interpreter use (>25% of hospital days) showed a trend for an even longer length of stay. Unmeasured clinical and social/cultural factors may contribute to differences in length of stay.

View details for DOI 10.3390/ijerph15081570

View details for PubMedCentralID PMC6121500

Abstract

To compare the rates of shared decision making (SDM) reported by parents of children with medical complexity (CMC) with the rates of SDM reported by parents of noncomplex children with special health care needs (CSHCN).We examined the 2009-2010 National Survey of Children with Special Health Care Needs, a representative survey of 40 242 parents of CSHCN. CMC was defined as needing or using more medical care than usual, seeing 2 or more subspecialists, and positive response on at least 3 other items on the CSHCN screener. We identified 3 subgroups each of CMC and noncomplex CSHCN by sentinel diagnoses: asthma, seizures, and other diagnoses. SDM was defined as a binary composite variable, derived from 4 discrete items. We constructed 4 stepwise multivariable models to assess the relative odds of SDM, adjusted for sociodemographic characteristics (age, income, language, race, ethnicity, and marital status), behavioral comorbidity, family-centered care, and patient-centered medical home.The study population included 39 876 respondents. Compared with noncomplex CSHCN, CMC had a lower likelihood of SDM (aOR, 0.76; 95% CI, 0.64-0.91), which persisted in diagnostic subgroups: CMC with asthma (aOR, 0.67; 95% CI, 0.49-0.92) and CMC with other diagnoses (aOR, 0.74; 95% CI, 0.58-0.94), but not CMC with seizures (aOR, 0.95; 95% CI, 0.59-1.51).SDM is less common for CSHCN with complex needs than those without complex needs. Health system interventions targeting future-oriented care planning may improve SDM for CMC.

View details for DOI 10.1016/j.jpeds.2017.09.001

View details for PubMedCentralID PMC5732902

Abstract

Patients with a terminal illness should have access to their chosen location of death. Cancer is the leading cause of non-accidental death among adolescents and young adults (AYAs; those aged 15-39 years). Although surveys have suggested that a majority of these patients prefer a home death, to the authors' knowledge, little is known regarding their barriers to accessing their preferred location of death. As a first step, the authors sought to determine, across a large population, 20-year trends in the location of death among AYA patients with cancer.Using the Vital Statistics Death Certificate Database of the California Office of Statewide Health Planning and Development, the authors performed a retrospective, population-based analysis of California patients with cancer aged 15 to 39 years who died between 1989 and 2011. Sociodemographic and clinical factors associated with hospital death were examined using multivariable logistic regression.Of 30,573 AYA oncology decedents, 57% died in a hospital, 33% died at home, and 10% died in other locations (eg, hospice facility or nursing facility). Between 1989 and 1994, hospital death rates decreased from 68.3% to 53.6% and at-home death rates increased from 16.8% to 35.5%. Between 1995 and 2011, these rates were stable. Those individuals who were more likely to die in a hospital were those aged <30 years, of minority race, of Hispanic ethnicity, who lived ≤10 miles from a specialty center, and who had a diagnosis of leukemia or lymphoma.Overall, the majority of AYA cancer deaths occurred in a hospital, with a 5-year shift to more in-home deaths that abated after 1995. In-hospital deaths were more common among younger patients, patients of minority race/ethnicities, and those with a leukemia or lymphoma diagnosis. Further study is needed to determine whether these rates and disparities are consistent with patient preferences. Cancer 2017;123:4178-4184. © 2017 American Cancer Society.

View details for PubMedID 28700812

Abstract

Many adult patients with cancer who know they are dying choose less intense care; additionally, high-intensity care is associated with worse caregiver outcomes. Little is known about intensity of end-of-life care in children with cancer.By using the California Office of Statewide Health Planning and Development administrative database, we performed a population-based analysis of patients with cancer aged 0 to 21 who died between 2000 and 2011. Rates of and sociodemographic and clinical factors associated with previously-defined end-of-life intensity indicators were determined. The intensity indicators included an intense medical intervention (cardiopulmonary resuscitation, intubation, ICU admission, or hemodialysis) within 30 days of death, intravenous chemotherapy within 14 days of death, and hospital death.The 3732 patients were 34% non-Hispanic white, and 41% had hematologic malignancies. The most prevalent intensity indicators were hospital death (63%) and ICU admission (20%). Sixty-five percent had ≥1 intensity indicator, 23% ≥2, and 22% ≥1 intense medical intervention. There was a bimodal association between age and intensity: ages <5 years and 15 to 21 years was associated with intense care. Patients with hematologic malignancies were more likely to have high-intensity end-of-life care, as were patients from underrepresented minorities, those who lived closer to the hospital, those who received care at a nonspecialty center (neither Children's Oncology Group nor National Cancer Institute Designated Cancer Center), and those receiving care after 2008.Nearly two-thirds of children who died of cancer experienced intense end-of-life care. Further research needs to determine if these rates and disparities are consistent with patient and/or family goals.

View details for PubMedID 28963112

Abstract

Cancer is the leading cause of nonaccidental death among adolescents and young adults (AYAs). High-intensity end-of-life care is expensive and may not be consistent with patient goals. However, the intensity of end-of-life care for AYA decedents with cancer-especially the effect of care received at specialty versus nonspecialty centers-remains understudied.We conducted a retrospective, population-based analysis with the California administrative discharge database that is linked to death certificates. The cohort included Californians age 15 to 39 years who died between 2000 and 2011 with cancer. Intense end-of-life interventions included readmission, admission to an intensive care unit, intubation in the last month of life, and in-hospital death. Specialty centers were defined as Children's Oncology Group centers and National Cancer Institute-designated comprehensive cancer centers.Of the 12,938 AYA cancer decedents, 59% received at least one intense end-of-life care intervention, and 30% received two or more. Patients treated at nonspecialty centers were more likely than those at specialty-care centers to receive two or more intense interventions (odds ratio [OR], 1.46; 95% CI, 1.32 to 1.62). Sociodemographic and clinical factors associated with two or more intense interventions included minority race/ethnicity (Black [OR, 1.35, 95% CI, 1.17 to 1.56]; Hispanic [OR, 1.24; 95% CI, 1.12 to 1.36]; non-Hispanic white: reference), younger age (15 to 21 years [OR, 1.36; 95% CI, 1.19 to 1.56; 22 to 29 years [OR,1.26; 95% CI,1.14 to 1.39]; ≥ 30 years: reference), and hematologic malignancies (OR, 1.53; 95% CI, 1.41 to 1.66; solid tumors: reference).Thirty percent of AYA cancer decedents received two or more high-intensity end-of-life interventions. In addition to sociodemographic and clinical characteristics, hospitalization in a nonspecialty center was associated with high-intensity end-of-life care. Additional research is needed to determine if these disparities are consistent with patient preference.

View details for PubMedID 28829692

View details for PubMedCentralID PMC5598313

View details for DOI 10.1542/peds.2016-3237

View details for Web of Science ID 000404482500013

Abstract

Poorly designed labels and dosing tools contribute to dosing errors. We examined the degree to which errors could be reduced with pictographic diagrams, milliliter-only units, and provision of tools more closely matched to prescribed volumes.This study involved a randomized controlled experiment in 3 pediatric clinics. English- and Spanish-speaking parents (n = 491) of children ≤8 years old were randomly assigned to 1 of 4 groups and given labels and dosing tools that varied in label instruction format (text and pictogram, or text only) and units (milliliter-only ["mL"] or milliliter/teaspoon ["mL/tsp"]). Each parent measured 9 doses of liquid medication (3 amounts [2, 7.5, and 10 mL] and 3 tools [1 cup, 2 syringes (5- and 10-mL capacities)]) in random order. The primary outcome was dosing error (>20% deviation), and large error (>2× dose).We found that 83.5% of parents made ≥1 dosing error (overdosing was present in 12.1% of errors) and 29.3% of parents made ≥1 large error (>2× dose). The greatest impact on errors resulted from the provision of tools more closely matched to prescribed dose volumes. For the 2-mL dose, the fewest errors were seen with the 5-mL syringe (5- vs 10-mL syringe: adjusted odds ratio [aOR] = 0.3 [95% confidence interval: 0.2-0.4]; cup versus 10-mL syringe: aOR = 7.5 [5.7-10.0]). For the 7.5-mL dose, the fewest errors were with the 10-mL syringe, which did not necessitate measurement of multiple instrument-fulls (5- vs 10-mL syringe: aOR = 4.0 [3.0-5.4]; cup versus 10-mL syringe: aOR = 2.1 [1.5-2.9]). Milliliter/teaspoon was associated with more errors than milliliter-only (aOR = 1.3 [1.05-1.6]). Parents who received text only (versus text and pictogram) instructions or milliliter/teaspoon (versus milliliter-only) labels and tools made more large errors (aOR = 1.9 [1.1-3.3], aOR = 2.5 [1.4-4.6], respectively).Provision of dosing tools more closely matched to prescribed dose volumes is an especially promising strategy for reducing pediatric dosing errors.

View details for DOI 10.1542/peds.2016-3237

View details for PubMedID 28759396

View details for PubMedCentralID PMC5495522