Estimating Clustering in Survey Responses Among Primary Care Providers and Practice Staff: Initial Evidence from the Comprehensive Primary Care (CPC) Initiative

• Activity Number: 188
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Derekh Cornwell*+ and Jared Coopersmith
• Companies: Mathematica Policy Research and Mathematica Policy Research
• Keywords: Sample Design ; Clustering ; Healthcare evaluations

Abstract

Reliable survey data on health care providers and staff underpins evaluations of health care interventions. A central component in the design of these surveys is the extent to which providers' and staff survey responses are correlated within medical practices. Failure to account for this clustering can lead to a sample design that is under-powered for the intended analysis. While estimates of clustering for patient outcomes at the practice level are available, analogous estimates for survey-based outcome measures for providers and staff are unavailable.

We address this gap by providing estimates of clustering effects within practices and regions, among 1,300 primary care providers and 1,700 practice staff who completed a survey for the Comprehensive Primary Care initiative, a large-scale health care evaluation sponsored by the Centers for Medicare & Medicaid Services. The estimated intra class correlation coefficients (ICCs) are calculated for selected measures of interest in health care provider surveys, such as job satisfaction and burnout. Incorporating more accurate ICC estimates into future practice-based survey designs will improve precision when determining sample sizes.