How to Apply Machine Learning to Health Economics and Outcomes Research: Findings from the ISPOR Machine Learning Task Force

Author(s)

Discussion Leader: William Vincent Padula, PhD, MS, MSc, University of Southern California, Los Angeles, CA, USA
Discussants: William H. Crown, PhD, The Heller Graduate School of Social Policy and Management, Brandeis University, Waltham, MA, USA; Blythe Adamson, PhD, MPH, Flatiron Health, New York, NY, USA; David J. Vanness, PhD, Health Policy & Administration, Pennsylvania State University, University Park, PA, USA

Presentation Documents

PURPOSE: ISPOR convened a Task Force to establish best-practices for applications of machine learning in health economics and outcomes research (HEOR). First, we introduce machine learning that can support HEOR. Second, we review the PALISADE Checklist to guide machine learning applications in HEOR. Third, we provide case studies where machine learning could be useful over traditional HEOR approaches.

DESCRIPTION: Advances in machine learning offer tremendous potential benefits to patients. The Task Force identified five methodological areas where machine learning could enhance accuracy of HEOR findings that we will review: (1) improve cohort selection—identifying samples with greater specificity based on inclusion criteria; (2) identifying independent predictors and covariates of health outcomes that extend beyond covariates identified in the literature; (3) improve predictive analytics of health outcomes, including those that are high cost or life threatening; (4) improvements in causal inference using targeted maximum likelihood estimation or double-debiased estimation to produce reliable evidence more quickly and eliminating the need for costly, time-consuming randomized controlled trials; and (5) be applied to development of economic models to reduce structural, parameter and sampling uncertainty in cost-effectiveness analysis. Furthermore, we examine whether machine learning offers consistently interpretable and transparent solutions to healthcare analytics. Then we will review the PALISADE Checklist, which the Task Force developed as a guide for key considerations that machine learning can offer in balance with the need for transparency in findings that differentiate patient care in the future. Finally, we will provide handouts for a series of case studies with applications to the Checklist so that the audience can determine if machine learning provides enhancement over traditional HEOR methods. We will engage the audience directly in the last step using Poll Everywhere to solicit feedback from attendees regarding remaining gaps in the PALISADE Checklist and any barriers to adopting of machine learning in HEOR.

Conference/Value in Health Info

2022-05, ISPOR 2022, Washington, DC, USA

Code

103

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×