Project Summary
Healthcare delivery systems are facing enormous organizational and economic challenges in the era of accountable care organizations and learning healthcare systems. Incentivizing the quality of care and the efficiency of preventative and therapeutic interventions means that new medical or management strategies are continuously being considered for their ability to improve patient outcomes. Therefore, healthcare delivery systems are frequently considering well-motivated changes, and need to develop evaluation strategies that can formally assess their impact. Development, dissemination, and adoption of contemporary research methods that can rigorously evaluate candidate treatment strategies delivered at the level of the patient, provider, or clinic are needed to ensure that valid and reproducible learning routinely occurs within the healthcare system, leading to the creation of best practices and improved public health.
Our proposed research will provide new design and analysis tools for group-randomized studies. In particular, we will guide the choice of design and choice of parameter of interest through the development of new causal inference ideas that explicitly recognize patient–provider dyads, which reflect the impact of both patient-specific and context-specific factors on ultimate health outcomes. We will also carefully evaluate the assumptions that are necessary for the use of standard mixed model analysis methods for inference about the causal parameters specified by explicit averages of counterfactual dyads. Given development of a conceptual framework we will then focus on developing software tools to plan a cluster-randomized trial with emphasis on longitudinal designs such as the stepped wedge design. Finally, we will develop study-monitoring strategies for stepped wedge designs that allow formal evaluation of interim signals for efficacy or safety.
The direct goals of the proposed research are to provide the methods necessary for improved clinical research designs with robust inference methods. We aim to enable the creation of reproducible and generalizable knowledge from within learning healthcare systems, and to improve the process of healthcare delivery and the associated patient outcomes.
Juul SE, Mayock DE, Comstock BA, Heagerty PJ. Neuroprotective potential of erythropoietin in neonates; design of a randomized trial. Matern Health Neonatol Perinatol. 2015 Dec 2;1:27. doi: 10.1186/s40748-015-0028-z. eCollection 2015. Review. PubMed PMID: 27057344; PubMed Central PMCID: PMC4823689. (Abstract only available)
Project Details
Information