John Giardina, doctoral student in Health Policy concentrating in Decision Science, successfully defended his thesis, “Accounting for Heterogeneity in Health Decision Analysis.” Overseen by Ankur Pandya (chair), Karen Sepucha, Michael Chernew, and Nicolas Menzies, his dissertation examined ways to incorporate differences across disease risk, treatment effects, and preferences into decision models and analyses.
In his first chapter, Giardina introduced a method to improve the parameterization of chronic disease microsimulation models to account for individuals’ changes in risk factors over time, which could lead to improved personalized screening and prevention strategies. These strategies are often based on risk factors like blood pressure and cholesterol that change as people age, so the relationship between these changes and the probability of disease events like a heart attack need to be incorporated into each cycle of the model. Giardina used a joint longitudinal and time-to-event model to capture this dynamic relationship and showed that this approach improved the validity of an ischemic stroke microsimulation model compared to existing modeling methods.
In his second chapter, Giardina investigated whether heterogeneous treatment effects could be used to successfully personalize blood pressure treatment. Previous research has estimated how the impact of intensive blood pressure treatment differs across individuals, often using machine learning methods to generate specific predictions of how treatment will affect cardiovascular disease risk for each individual in a population. Assessing the accuracy of these predictions is difficult, however, since researchers cannot simultaneously observe what will happens if a particular individual receives both intensive and standard treatment. To avoid this issue, Giardina assessed whether outcomes would be improved on average if these predictions of treatment effect were used to make treatment decisions. He found that the decision objective affects whether the predictions of individual-level treatment effects would be helpful in decision-making – except for particular decision contexts, the estimates were not accurate enough to improve decisions.
Finally, in this third chapter, Giardina discussed modeling work he did in 2022 to estimate off-ramps and on-ramps for masking in elementary schools in response to SARS-CoV-2. He emphasized that a one-size-fits-all approach to masking guidelines is not feasible, since each community faces different risks, needs, and goals. Giardina discussed how the modeling approach he implemented with his co-authors generated recommendations that could be adapted to the needs and preferences of different school districts and communities.
Following his graduation, Giardina will be joining the faculty at the Medical Practice Evaluation Center at Massachusetts General Hospital and Harvard Medical School.