Mechanistic Machine Learning
Advisor: Lena Mamykina, MS, PhD
The overarching goal of this research is to develop new approaches to facilitating clinical reasoning and decision-making through aligning clinicians’ cognitive models of patients’ pathophysiology, mechanistic models of human physiology, and empirical patient data. Reconciling differences between cognitive and mechanistic models requires deep understanding of medical cognition and clinical work, novel computational modeling, and innovative co-design approaches that will bring together cognitive scientists, data scientists, and experts in human-computer interaction and clinical decision support.
Differences in Presentation and Treatment of Acute Myocardial Infarction based on Gender and Race/Ethnicity
Advisor: Noémie Elhadad, PhD
We aim to measure and characterize differences in presentation and treatment of acute myocardial infarction (AMI) using structured electronic health record data and free-text from clinical notes. We are using empirical distributional semantics methods (e.g., word2vec and GloVe) to identify potential biases and empirical associations based on demographics. We are also analyzing clinical notes over time prior to and after initial diagnosis to understand differences in the trajectories of sign or symptom documentation for patients, lengths of time to treatment, and resulting impact on patient outcomes.
Recognizing and Normalizing Condition Concepts with Severity Modifiers
Advisor: Chunhua Weng, PhD
The presence of qualifiers such as severity or disease stage complicate the recognition and normalization of biomedical concepts from free text. A single concept can often be mapped to a terminology in multiple ways. Our problem is motivated by inconsistent recognition of disease conditions in Criteria2Query (C2Q), a tool for exploring clinical trial inclusion and exclusion criteria. We are developing a novel two-step, ontology-aware mapping scheme that is optimized for extracting disease concepts at a high level of specificity.
Projects Submitted for Funding
ARtificial Intelligence for the Study and Enhancement of Sexual and Gender Minority health
Co-investigator (PI: Zhou) – Direct Costs Requested – $275,000
The goals of the ARISE-SGM (ARtificial Intelligence for the Study and Enhancement of Sexual and Gender Minority health) research project are to leverage advances in artificial intelligence to address logistical challenges in obtaining representative samples of sexual and gender minority (SGM) populations and characterize comorbidity prevalence and potential health inequities associated with different segments of SGM populations. Original SRG Action: Impact Score:38 Percentile:24+
A Multicenter Randomized Controlled Trial of Pharmacist-Directed Transitional Care to Reduce Post-Hospitalization Utilization
National Institute on Aging (NIA) R01 AG 058911-01
Project Manager (Site PI: Schnipper)
Aim: To measure the impact of pharmacist-led discharge (PHARM-DC) interventions on post-discharge utilization within 30 days and identify patient subpopulations most likely to benefit from PHARM-DC.
Novel electronic health record phenotyping of LGBTQ intersectional identities and associated health disparities using natural language processing and machine learning
Co-Investigator (PI: Zhou)
Aim: To examine documentation of sexual orientation and gender identity (SOGI) in the electronic health record (EHR) and characterize LGBTQ health and potential health inequities.
Electronic Medication Adherence Reporting and Feedback During Care Transitions
Agency for Healthcare Research and Quality (AHRQ) R21 HS 024587-01
Project Manager/Statistical Programmer (PI: Schnipper)
Aim: To implement a smart pillbox intervention for patients discharged from the hospital to the community and to evaluate the effects of the intervention on post-discharge medication discrepancies, medication adherence, and chronic disease management.
Implementation of a Medication Reconciliation Protocol to Improve Patient Safety
Agency for Healthcare Research and Quality (AHRQ) R18 HS 023757-01A1
Project Manager/Statistical Programmer (PI: Schnipper)
Aim: The long-term objective of this research is to widely disseminate, implement, and evaluate sustainable medication reconciliation interventions that improve patient safety during care transitions.
Clinician-Patient Relationships: Boundaries, Barriers, Breakdowns
Gold Foundation/Lucian Leape Family Foundation
Project Director (PI: Schiff)
Aim: This study examines physician attitudes and practices related to caring acts that have been questioned as “inappropriate” or “unethical” crossing of professional-patient boundaries.
Understanding and Preventing Diagnostic Pitfalls
Risk Management Foundations of the Harvard Medical Institutions (CRICO)
Project Director (PI: Schiff)
Aim: The goal of this project is to identify leading diagnostic pitfalls that pose safety and malpractice risks, then design and test the accuracy of electronic screens for these pitfalls in the electronic health record.
Evaluating Variation in Overactive Bladder Medication Use and Other Treatment Options
Astellas Pharma US, Inc.
Statistical Programmer (PI: Loughlin)
Aim: To understand variation in overactive bladder treatment, medication adherence, and outcomes.
Are Physicians and Patients Adhering to the Published Guidelines for Fasting during Ramadan?
Martin P. Solomon Award for Primary Care Scholarship
Co-Investigator (PI: Ali)
Aim: To measure differences in rates of health care utilization for Muslim patients just prior to, during, and after Ramadan.
BWH DGIM Research Day
Brigham Research Institute/Office for Research Careers Microgrant Pilot Program
Principal Investigator ($1,000)
Aim: To develop the first annual Division-wide retreat for showcasing contributions and promoting career development of research assistants in the Division of General Internal Medicine and Primary Care at Brigham and Women’s Hospital.
Patient Outcomes Associated with Year-End Departure of Resident Primary Care Physicians
Support for Excellence in Educational Development (SEED) Grant, BWH Co-Investigator/Statistical Programmer (PI: Solomon)
Aim: To investigate whether acute care utilization is increased disproportionately among patients of graduating resident physicians compared to non-graduating residents.
Acute Respiratory Infection Telephone Management Service
Brigham and Women’s Physicians Organization Care Redesign Incubator and Startup Program
Data Manager (PI: Linder)
Aim: To develop, implement, and evaluate an acute respiratory infection telephone management service at one of the Brigham and Women’s Primary Care Practices.
Medical Malpractice and Patient Safety Protocol / Proactive Reducation of Outpatient Malpractice: Improving Safety, Efficiency, and Satisfaction (PROMISES)
Agency for Healthcare Research and Quality (AHRQ) 1R18HS019508-01
Project Manager (PI: Schiff)
Aim: To test the impact of quality improvement techniques to accomplish innovations and improvements in 3 high risk ambulatory malpractice areas: 1) medication management, 2) test ordering and results management 3) follow-up and referral management.
Use of Behavioral Economics to Improve Treatment of Acute Respiratory Infections
National Institutes of Health/University of Southern California 1RC4AG039115-01
Data Manager/Statistical Programmer (Site PI: Linder)
Aim: To evaluate the use of behavioral economic concepts such as enhanced defaults, alternative options, and social norms on inappropriate antibiotic prescribing for acute respiratory infections.
U.S. President’s Emergency Plan for AIDS Relief (PEPFAR)
Health Resources and Services Administration (HRSA) U51HA02522
Data Quality Assurance Manager / Data, Training, and Education Coordinator (PI: Kanki)
A $362 million grant for work in Nigeria, Botswana, and Tanzania to train health care workers, develop monitoring and evaluation systems, strengthen health care infrastructures, and collaborate with local hospitals and clinics that provide treatment for people living with HIV/AIDS.
Pharmacist Intervention to Decrease Medication Errors in Heart Disease Patients (PILL-CVD)
National Heart, Lung, and Blood Institute (NHLBI), R01HL089755
Research Assistant (Site PI: Schnipper)
Aim: To evaluate a literacy-focused program that provides educational assistance from pharmacists at the time of hospital discharge to people hospitalized with heart problems.
Influenza in the Partners Primary Care Practice-Based Research Network
Roche Laboratories, RLI 06001207
Research Assistant (PI: Linder)
Aim: To measure and evaluate antiviral and antibiotic prescribing for patients with influenza being treated in primary care clinics.
Pediatric quality improvement tool at 33 Harvard PEPFAR sites in Nigeria
Co-developed and evaluated a utility to extract information from electronic health record (EHR) data warehouse and generate measures based on 23 pediatric quality of care indicators. Programmed module to review continuity of care, drug therapy initiation, medication adherence, loss to follow-up, laboratory monitoring, disease screening based on clinical symptoms assessment or diagnostic evaluation, treatment failure, toxicity management, nutritional assessment, and treatment response.
Adult quality improvement tool at 33 Harvard PEPFAR sites in Nigeria
Co-developed and evaluated a utility to extract information from EHR data warehouse and generate measures based on 15 adult quality of care indicators. Module reviews continuity of care, drug therapy initiation, loss to follow-up, laboratory monitoring, disease screening based on clinical symptoms assessment, treatment failure, and treatment response.
Data quality assessment tool, version 2.0 at 33 Harvard PEPFAR sites in Nigeria
Instituted major modifications to and evaluated a pre-existing EHR utility to incorporate expanded data quality indicators. With this utility, central and site-level data managers are able to better monitor and improve data completeness, validity, accuracy, and currency.