Innovation and Methods for Learning Health System (IM4LHS)

The IM4LHS (Innovation and Methods for Learning Health System) lab is dedicated to advancing study designs, and statistical and machine learning methods in the context of learning health system (LHS) by promoting continuous improvement in healthcare delivery and patient outcomes. By collaborating with clinicians and administrators at the Center for Patient Safety Science (Center for Patient Safety Science | Vagelos College of Physicians and Surgeons) at Columbia University Irving Medical Center and New York Presbyterian Hospital, we aim to leverage the vast data from electronic health records to generate actionable insights to inform healthcare and operational decisions and improve patient outcomes.  Our two key focus areas are: integration of randomized clinical trials and risk prediction models within LHS. Embedding advanced methodologies in these areas will allow the LHS to become more adaptive, evidence-based, and patient-centered, resulting in higher-quality care and more efficient healthcare delivery.

Randomized Clinical Trials:

Randomized clinical trials can be used to rigorously assess the impact of interventions such as clinical decision support and quality improvement strategies on patient outcomes, clinical workflows, and healthcare efficiency.  By minimizing bias, these trials provide robust evaluations of interventions, allowing health systems to make data-driven decisions. The IM4LHS lab is particularly interested in the development and implementation of novel clinical trial methodologies such as pragmatic trials, adaptive designs, and sequential randomization within healthcare settings. These innovative and efficient strategies support rapid learning and adaptation, ensuring that patients benefit from the most effective and up-to-date care. Embedding randomized trials into healthcare systems establishes a dynamic cycle of continuous improvement and innovation to support a Learning Health System.

Risk Prediction Modeling:

In Learning Health Systems (LHS), risk prediction models enable personalized care by identifying individuals at high risk for adverse events, allowing for early decisions and tailored interventions. Our lab is particularly focused on developing risk prediction models that support clinical decision-making tools, enhancing the system’s ability to predict and respond to individual patient needs. Through continuous data feedback from clinical outcomes, these models are regularly refined to improve accuracy and responsiveness to evolving healthcare trends and patient demographics. The integration of risk prediction models within a learning health system can foster better care coordination, improved patient outcomes, and increased healthcare efficiency

Grant Support

Columbia-Cornell-Einstein EQUIP+ Center for Learning Health System Science (P30HS029763)
Columbia Clinical and Translational Science Award (UL1TR001873; TL1TR001875)

Lab Members

Team Members

  • Shing Lee, PhD

    • Co-director
    • Professor, Department of Biostatistics

  • Min Qian, PhD

    • Co-Director
    • Associate Professor, Department of Biostatistics

  • Jason Adelman, MD, MS

    • Associate Professor of Medicine (in Biomedical Informatics)
    • Associate Dean for Quality and Patient Safety
    • Director, Center for Patient Safety Science

  • Bin Cheng, PHD MS

    • Professor of Biostatistics at the Columbia University Medical Center

  • Gregory W. Hruby, PhD

    • Director of Data Science, Center for Patient Safety Science
    • Lead Data Scientist, New York-Presbyterian Hospital

  • Elly Kipkogei

    • PhD student, Department of Biostatistics

  • Benjamin L. Ranard, MD

    • Assistant Professor, Department of Medicine
    • Deputy Director, Center for Patient Safety Science