Informing Financial Investments in Health: Applications of Decision-Analytic Methods
Lofgren, Katherine T.
MetadataShow full item record
CitationLofgren, Katherine T. 2020. Informing Financial Investments in Health: Applications of Decision-Analytic Methods. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.
AbstractIn this dissertation, I use decision-analytic methods to investigate prospective financial allocation decisions in three health policy contexts: research, regulation, and insurance.
The first paper explores how clinicians and funders could plan randomized controlled trials (RCTs) using value of information (VOI) methods. We developed an accessible entry point for non-decision scientists to gain intuition on how a value-based study design may differ from a power-calculation (constrained on Type I/II error) required sample size. Using a case study of implantable cardiac devices, we demonstrate that VOI has the potential to prioritize the study of patient subpopulations that may otherwise be overlooked. There is a sustained interest in shifting away from statistical significance as a metric of meaning in research results and design. This work aims to make value-based study design an accessible alternative for those conducting and funding clinical research.
My second paper considers the potential pitfalls of regulatory decision making based on surrogate outcomes. The Food and Drug Administration’s (FDA) Accelerated Approval Pathway (AAP) was established in 1992 to get high-need patients new therapeutic options faster. To achieve this goal, the FDA approves AAP drug-indications on surrogate outcomes. However, unlike clinical outcomes (e.g. survival), surrogate outcomes are only suggestive of therapeutic benefit (e.g. cholesterol as a proxy for heart disease risk). Using a simulation study and a AAP case study, we examine how regulatory decision-making compares when surrogate vs. clinical outcomes are available. We find that regulators with a surrogate outcome that is a noisy proxy for clinical benefit, correctly estimate the likelihood the current regulatory decision is wrong, but overestimate the consequences of the possible decision error. When regulators have access to a surrogate outcome that is a biased (in favor of treatment effect) proxy of clinical benefit, the regulator both misjudges health consequences of decision errors and underestimates the likelihood the current decision is wrong. We show that confirmatory evidence must include the clinical outcome for the initial misjudgments in decision-making uncertainty and error consequences to be corrected.
Finally, the third paper examines how low- and middle-income countries can approach essential health benefits design using mathematical optimization to balance competing objectives. We estimate both expected health and financial risk protection benefits from 22 candidate interventions in the country-context of Ethiopia. Our hypothetical benefits package is assumed to guarantee universal population access and eliminate all household out-of-pocket health expenditures. We find that the ‘best buys’ often diverge depending on the objective. When both objectives are considered jointly using integer programming optimization, the optimal benefits package flexibly changes based on the relative importance of health and financial protection. Mathematical optimization provides a practical framework for policymakers to design universal health coverage polices that jointly prioritize multiple objectives within financial constraints.
Citable link to this pagehttps://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37365875
- FAS Theses and Dissertations