Rosenberg Lab
OUR RESEARCH is focused on applied data science methodologies for clinical decision-making in cardiovascular disease and other applications. Specific topics include atrial fibrillation, drug-induced long-QT syndrome, preventive cardiology including lipid management, and programming of cardiac implantable electronic devices (e.g., pacemakers). We collaborate with several data science engineers, health services investigators, clinicians, and biostatisticians across multiple University of Colorado campuses, as well as outside institutions. Our team has experience in the analysis of data from wearable and implanted medical devices, genetic data, electronic health record data, and data from commercial wearable devices (Fitbit and AppleWatch). We have developed both web-based and smartphone prototype applications, with ongoing clinical testing focused on the N-of-1 trial approach. Additional information is available at analyzemydata.org or HTTPS://www.ncbi.nlm.nih.gov/myncbi/michael.rosenberg.2/bibliography/public/.
If you are interested in joining our team or collaborating, please contact: michael.a.rosenberg@cuanschutz.edu.
Michael Rosenberg, MD, FHRS, FACC, FAHA
Associate Professor
Medical Director, Clinical Sciences Program of CU Division of Cardiology
Medical Director, UCH ECG Laboratory
Cardiac Electrophysiology, Division of Cardiology
University of Colorado School of Medicine
Despite a number of innovations in the ablative treatment of atrial fibrillation, decisions around rhythm management remain a challenge due to a wide variation in the symptoms and presentation across patients. Such variability motivates approaches that can individualize treatment decisions, in a manner that also allows for the ever-expanding range of options that could be available as technology improves. The method we have applied to develop clinical decision support models to guide rhythm-management decisions is called reinforcement learning, which is a type of machine learning that allows the model to learn from experience to identify the optimal treatment approach for each individual patient.
Papers:
- Evaluation of Quantitative Decision‐Making for Rhythm Management of Atrial Fibrillation Using Tabular Q‐Learning
- Qualitative Evaluation of an Artificial Intelligence–Based Clinical Decision Support System to Guide Rhythm Management of Atrial Fibrillation: Survey Study
- Machine Learning Methodologies for Prediction of Rhythm-Control Strategy in Patients Diagnosed With Atrial Fibrillation: Observational, Retrospective, Case-Control Study
- Assessment of a Machine Learning Model Applied to Harmonized Electronic Health Record Data for the Prediction of Incident Atrial Fibrillation
- Applications of machine learning in decision analysis for dose management for dofetilide
Drug-induced prolongation of the QT interval (diLQTS) on an ECG is a common side-effect of many medications, and is the number one reason for FDA rejection of new compounds due to the risk of life-threatening arrhythmias. Methods to predict which individuals are at greatest risk of diLQTS, and guide decisions about how to manage risk, could have a significant impact on management of conditions across many clinical domains. Our team has studied risk of diLQTS using machine-learning methods, genetics, and through analysis of existing electronic health record tools created to prevent diLQTS. This work has led to greater understanding about both the risk of diLQTS, but also how human behavior factors into use of decision-support tools applied at scale across an entire healthcare system.
Papers:
- Interpretable Machine Learning Prediction of Drug-Induced QT Prolongation: Electronic Health Record Analysis
- A polygenic risk score for the QT interval is an independent predictor of drug-induced QT prolongation
- Prediction of Drug-Induced Long QT Syndrome Using Machine Learning Applied to Harmonized Electronic Health Record Data
- Validation of Polygenic Scores for QT Interval in Clinical Populations
- Assessing Prescriber Behavior with a Clinical Decision Support Tool to Prevent Drug-Induced Long QT Syndrome
The deluge of data that has emerged with advances in wearable and implanted devices has enormous potential to guide individualized decisions for patients. However, the process of extracting meaningful information from this data remains a major challenge for the field. Our team has examined data from both commercial and medical-grade devices to identify potential applications in both cardiology and health/wellness, which has included development of several applications for smartphone or web-based use. The insights from this work have led to projects centered on the N-of-1 approach to using each person’s data in a manner that can have the greatest impact on their health.
- Assessment of a Mobile Health iPhone App for Semiautomated Self-management of Chronic Recurrent Medical Conditions Using an N-of-1 Trial Framework: Feasibility Pilot Study
- Feasibility of Frailty Assessment Integrated with Cardiac Implantable Electronic Device Clinic Follow-up: A Pilot Investigation
- Device-measured physical activity data for classification of patients with ventricular arrhythmia events: A pilot investigation
- Burden of Premature Ventricular Complexes and Risk of Cardiomyopathy: A Cross-Sectional Study