Research Mission
Our mission is to apply data-driven methods and leverage cutting-edge artificial intelligence (AI) tools to accurately measure and improve the quality of care provided by pediatricians to children with developmental, behavioral, and mental health conditions.
Background Information
Developmental, behavioral, and mental health conditions affect approximately one in five children. Given the high prevalence of these conditions, primary care pediatricians play a central role in their assessment and treatment. However, significant quality gaps and disparities in care remain.
The growing availability of large-scale electronic health record (EHR) databases offers an opportunity to utilize existing patient data to assess quality of care and gain deeper insights into the factors that influence outcomes for conditions such as attention deficit hyperactivity disorder (ADHD), autism spectrum disorder, anxiety, and depression.
Additionally, the rapid advancement of artificial intelligence (AI) and natural language processing (NLP) technologies such as large language models now enables the efficient extraction, analysis, and interpretation of vast amounts of clinical data. These tools have the potential to automate manual chart reviews, a typically time-consuming process where physicians review diagnoses, clinical notes, treatment plans, and test results. Additionally, since most clinical data is stored in unstructured formats like free-text notes, large language models can efficiently analyze this information at scale in particular. By quickly processing large-scale clinical data, AI models can help pediatricians efficiently assess care quality and deliver evidence-based care for children with developmental, behavioral, and mental health conditions.
At the Advanced Informatics for Mental Health Lab, we:
1. Investigate the feasibility and potential of using AI tools to accurately assess and enhance the quality of care for children receiving treatment for developmental, behavioral, and mental health conditions in primary care and specialist care settings
2. Conduct population health and health services research utilizing electronic health records and qualitative methods to identify trends and explore the impact of demographic, family, and other factors on the management and delivery of care for pediatric developmental, behavioral, and mental health conditions
3. Aim to optimize operational aspects and work closely with clinicians to refine clinical workflows, improve care processes, and enhance overall service delivery through quality improvement and implementation science
Recent Publications
Understanding ADHD Diagnosis and Medication Use Patterns Among Preschoolers
In our multi-institutional study of over 700,000 children titled "ADHD Diagnosis and Timing of Medication Initiation Among Children Aged 3 to 5 Years" , we investigated how often preschool-age children are diagnosed with ADHD in primary care and how quickly they begin medication. We found that many young children start medication soon after their diagnosis, with differences based on factors like race, ethnicity, and insurance. These findings highlight the need to better understand early prescribing patterns for ADHD and potential barriers to non-pharmacological treatments such as behavioral interventions.
Read the full article in JAMA Network Open.
Using LLMs to Monitor ADHD Medication Side Effects
In our publication titled “Applying Large Language Models to Assess Quality of Care: Monitoring ADHD Medication Side Effects”, we explored how a large language model (LLM) could help measure how well pediatricians follow guidelines for checking side effects in children taking ADHD medications. By analyzing over 15,000 clinical notes from a community-based primary care network, we found that the LLM accurately identified when side effects were mentioned in medical records. This research highlights how LLMs can support pediatricians in improving care for children with ADHD by ensuring medication side effects are monitored more consistently.
Read the full article in Pediatrics.
Leveraging NLP models to Measure Parent Training in Behavioral Management for ADHD
In our publication titled “Measuring quality-of-care in treatment of young children with attention-deficit/hyperactivity disorder using pre-trained language models”, we explored how natural language processing (NLP) models can help measure how well pediatricians follow evidence-based guidelines when treating young children with ADHD. We used data from electronic health records (EHRs) of children aged 4-6 years who were diagnosed with ADHD in a primary healthcare network. By analyzing over 1,000 clinical notes, we found that a pre-trained NLP model (BioClinicalBERT) could accurately identify whether pediatricians recommended parent training in behavioral management as an evidence-based first-line treatment for ADHD.
Read the full article in Journal of the American Medical Informatics Association (JAMIA).