A Bayesian Difference-in-Difference Framework for the Impact of Primary Care Redesign on Diabetes Outcomes. Issue 1 (1st January 2019)

• Record Type: Journal Article
• Authors: Normington, James; Lock, Eric; Carlin, Caroline; Peterson, Kevin; Carlin, Bradley
• Is Part Of: Statistics and public policy. Volume 6:Issue 1(2019)
• Page Start: 55
• Page End: 66
• Publication Date: 2019-01-01
• Subjects: Bayesian hierarchical modeling -- Diabetes -- Difference-in-differences -- Errors in covariates -- Patient-centered medical home -- Primary care redesign; Policy sciences -- Methodology -- Periodicals; Social sciences -- Statistical methods -- Periodicals; Medical statistics -- Methodology -- Periodicals; Statistics -- Periodicals; Medical statistics -- Methodology; Policy sciences -- Methodology; Social sciences -- Statistical methods; Statistics; Periodicals; 320.60727
• DOI: 10.1080/2330443X.2019.1626310
• Languages: English
• ISSNs: 2330-443X

Abstract

Abstract: Although national measures of the quality of diabetes care delivery demonstrate improvement, progress has been slow. In 2008, the Minnesota legislature endorsed the patient-centered medical home (PCMH) as the preferred model for primary care redesign. In this work, we investigate the effect of PCMH-related clinic redesign and resources on diabetes outcomes from 2008 to 2012 among Minnesota clinics certified as PCMHs by 2011 by using a Bayesian framework for a continuous difference-in-differences model. Data from the Physician Practice Connections-Research Survey were used to assess a clinic's maturity in primary care transformation, and diabetes outcomes were obtained from the MN Community Measurement program. These data have several characteristics that must be carefully considered from a modeling perspective, including the inability to match patients over time, the potential for dynamic confounding, and the hierarchical structure of clinics. An ad-hoc analysis suggests a significant correlation between PCMH-related clinic redesign and resources on diabetes outcomes; however, this effect is not detected after properly accounting for different sources of variability and confounding. Supplementary materials for this article are available online.