Background: Effective management of chronic conditions such as type 2 diabetes requires periodic clinical monitoring and frequent re-evaluation of treatment decisions over the course of the patient’s illness. To optimize and personalize care, the relevant treatment strategies to compare should not only adapt to the patient’s evolving clinical condition but should also integrate the levels of both adherence to treatment and the frequency of clinical monitoring. The comparison of such complex adaptive treatment strategies is particularly relevant to patients and stakeholders concerned with the burden imposed by both daily medication intake and frequent clinical monitoring. Because randomized trials are not well suited to produce such heterogeneous evidence, we need advanced analytic methods that can be applied to real-world observational data and emulate trial inferences. Existing causal inference methods are not well adapted to real-world data that are highly variable in timing and content such as electronic health record (EHR) data.
Objectives: To advance and adapt existing causal inference methods so they can not only be adequately applied to real-world observational data but also better inform the impact of both real-world treatment adherence and frequency of clinical monitoring in comparative effectiveness research (CER).
Methods: Using both simulated data and existing EHR data from a recent nonrandomized type 2 diabetes study, we will evaluate the applicability and practical performance of new causal inference tools in real-world CER. In particular, we will study their behaviors in studies with relatively small sample sizes and document the conditions under which these methods can provide valid CER evidence.
Patient and Stakeholder Engagement: Stakeholders will drive the development, illustration, and dissemination of the proposed methods by identifying areas within chronic disease management with high-priority CER questions. Insights from patients, practicing clinicians, clinical and operational health plan leaders, and CER researchers will help us understand the types of comparative effectiveness questions involving sequences of treatment and monitoring decisions that are of most value to healthcare decision makers, with a focus on the management of chronic diseases and type 2 diabetes.
Anticipated Impact: This project will improve the applicability of sophisticated causal methods with EHR data to compare complex treatment strategies that can better inform decision making by clinicians and patients for the management of chronic conditions. It will increase the transparency, relevance, validity, and precision of the evidence generated with these advanced causal inference methods.