Boussard Lab Research

Active Research Areas

In the Boussard Lab we focus on generating and evaluating artificial intelligence (AI) tools to aid in clinical decision support, to develop evidence for clinical guidelines and to inform policy. We develop novel machine learning/deep learning methods and natural language processing (NLP) techniques to derive insights from electronic health records (EHRs), administrative claims, disease registries, and patient-reported health outcomes. The AI tools are utilized for a variety of healthcare applications, including medical diagnostics, treatment protocol development, risk stratification, and clinical trial recruitment. Our research advances the state of the art in biomedical informatics to benefit clinical and translational research as well as patient care. We work in a highly collaborative environment with lab members from a broad range of background and expertise such as computer scientists, statisticians, and medical professionals.

Bias and Fairness

The development of fair artificial intelligence (AI), particularly machine learning (ML) algorithms, to inform decision making and improve health outcomes is a topic of great interest. Too often we hear of AI algorithms being used at point of care that are trained on small, unrepresentative datasets which may perpetuate biases and cause harm to certain populations. A comprehensive framework to assess and mitigate these issues is needed. We aim to assess and develop the equality and equity of data-driven clinical decision support tools and provide evidence that can guide standards related to reporting and technology deployment.

Guiding Principles to Address the Impact of Algorithm Bias on Racial and Ethnic Disparities in Health and Health Care

Promoting Equity In Clinical Decision Making: Dismantling Race-Based Medicine

Natural language processing to identify reasons for sex disparity in statin prescriptions

Picture a data scientist: a call to action for increasing diversity, equity, and inclusion in the age of AI

Health Policy Informatics & Population Health

Healthcare reform has led to US initiatives designed to improve the quality, efficiency, safety and effectiveness of healthcare care delivery. A major focus of our group is to generate Insight into the quality of healthcare delivery, using a broad range of clinical data, including electronic medical records (EMR), administrative data and healthcare registries.

Fraud, Waste, and Abuse

Genetic Testing: Fraud, Waste, and Abuse

COVID white paper: Fraud, Waste and Abuse

Payment Reform

Effect of Medicare's Nonpayment Policy on Surgical Site Infections Following Orthopedic Procedures.

Performance Measures in Neurosurgical Patient Care: Differing Applications of Patient Safety Indicators.

The Affordable Care Act reduces emergency department use by young adults: evidence from three States.

Quality Measurement

Association between patient‐initiated emails and overall 2‐year survival in cancer patients undergoing chemotherapy: Evidence from the real‐world setting

Utilization of Prostate Cancer Quality Metrics for Research and Quality Improvement: A Structured Review

A natural language processing algorithm to measure quality prostate cancer care.

EHRs and Quality of Care: An Observational Study Modeling Impact on Mortality, Readmissions, and Complications.

Patient Safety

Evaluating patient safety indicators in orthopedic surgery between Italy and the USA.

Patient safety in plastic surgery: identifying areas for quality improvement efforts.

The TRANSFORM Patient Safety Project: a microsystem approach to improving outcomes on inpatient units.

Comparative Effectiveness

Utilization and effectiveness of multimodal discharge analgesia for postoperative pain management

Interfacility transfer and mortality for patients with ruptured abdominal aortic aneurysm.

The Effect of Moving Carpal Tunnel Releases Out of Hospitals on Reducing United States Health Care Charges.

Methodologies

We pursue research in the application of data science to healthcare across the from individual patient trajectories, to entire populations, to health systems. We work on developing cutting-edge methodologies to derive insights from multi-modal digital data sources. Our team includes expertise in data mining, machine learning, deep learning, biostatistics, economics, and medicine. Some examples of our work are listed below.

NLP Knowledge Discovery
Measuring quality-of-care in treatment of young children with attention-deficit/hyperactivity disorder using pre-trained language models

Predictive Analytics
Predict, diagnose, and treat chronic kidney disease with machine learning: a systematic literature review.

Postoperative Pain Management

In the context of generating real-world data to curb the opioid epidemic, we aim to provide risk-stratification tools to identify patients at high risk for adverse pain outcomes after surgery, including prolonged opioid use and transition to chronic pain. Vulnerable populations, such as depressed, obese, and diabetics, suffer from significant pain management disparities, which can affect recovery and lead to serious adverse events. We develop machine learning methods based on clinical phenotypes extracted from real-world EHR data. We also focus on validating these models with external data using Veterans Health Administration and sharing them with the OMOP community.

Predicting premature discontinuation of medication for opioid use disorder from electronic medical records

Preoperative Versus Perioperative Risk Factors for Delayed Pain and Opioid Cessation After Total Joint Arthroplasty: A Prospective Cohort Study

Prediction of opioid-related outcomes in a medicaid surgical population: Evidence to guide postoperative opiate therapy and monitoring

Changes in postoperative opioid prescribing across three diverse healthcare systems, 2010-2020

Prior Research

Archive

Bias & Fairness

We aim to assess and develop the equality and equity of data-driven clinical decision support tools and provide evidence that can guide standards related to reporting and technology deployment.

Peeking into the black box, the fairness and generalizablity of a MIMIC-III benchmarking tool.

MINIMAR (MINimum Information for Medical AI Reporting)

Bias at warp speed: how AI may contribute to the disparities gap in the time of COVID-19

Reporting of demographic data and representativeness in machine learning models using electronic health records

Assessment of a Clinical Trial–Derived Survival Model in Patients With Metastatic Castration-Resistant Prostate Cancer

Putting the data before the algorithm in big data addressing personalized healthcare

COVID 19

Our contribution to the COVID-19 crisis, we provide systematic information about quality of care, risk scores of bad outcomes and inequities among communities. Our first focus was to establish a machine learning model to predict the need of invasive mechanical ventilation (IMV) among hospitalized patients. Barriers to accurate, unbiased prediction of COVID-19 disease trajectory include lack of readily available patient data to train and test models in the early outbreak of the disease and no public datasets available. We also evaluate the racial/ethnic inequities for COVID-19 testing patterns and positivity rates at Stanford Healthcare Alliance system.

Health management via telemedicine: Learning from the COVID-19 experience.

Learning From Past Respiratory Infections to Predict COVID-19 Outcomes: Retrospective Study

Learning from Past Respiratory Failure Patients to Triage COVID-19 Patient Ventilator Needs: A Multi-Institutional Study

Methodologies

We pursue research in the application of data science to healthcare across the from individual patient trajectories, to entire populations, to health systems. Past examples are below.

AI in Medicine

Digital twins for predictive oncology will be a paradigm shift for precision cancer care.

Real world evidence in cardiovascular medicine: assuring data validity in electronic health record-based studies

Predicting inadequate pain management in depressed patients: a machine learning approach

NLP Knowledge Discovery

An Automated Feature Engineering for Digital Rectal Examination Documentation using Natural Language Processing

Weakly supervised natural language processing for assessing patient-centered outcomes following prostate cancer treatment

Mining Electronic Health Records to Extract Patient-Centered Outcomes Following Prostate Cancer Treatment

Comparison of orthogonal NLP methods for clinical phenotyping and assessment of bone scan utilization among prostate cancer patients

Dr. Hernandez-Boussard discusses the application of machine learning and natural language processing (NLP) to Electronic Health Record (EHR) data at an NCI webinar. Check out the YouTube video here.