Funding: Agency for Health Care Research and Quality, HHS, 2010-2016 (2 R01 HS 018334 04)
?Value-based purchasing? is a quality improvement strategy that links payment with healthcare outcomes, by paying less or not at all for poor outcomes. The Centers for Medicare and Medicaid Services (CMS) seeks to decrease the rate of hospital-acquired complications (HAC) and readmissions by holding hospitals financially accountable using risk-adjusted rates. CMS risk-adjustment models for outcomes of mortality and readmission include patient characteristics from routine administrative discharge data (e.g., diagnosis codes) with age and gender as the only socio-demographic variables. Research suggests other important patient characteristics such as functional status, mobility and level of social support also impact patients? risk for readmission and certain complications (e.g., pressure ulcers). To date, variables such as functional status, mobility and social support have not been included in risk-adjustment models because they are not available in routine discharge data; also, socio-demographic variables (e.g., income or education, which may relate to a patient?s ability to maintain functional status and secure social support) have not been included in risk-adjustment for outcomes due to concerns that adjusting for such factors would be akin to condoning poor care delivered to vulnerable patients. In order to determine how much socio-demographic factors relate as risks for poor hospital outcomes and readmissions (as intrinsic patient factors compared to factors extrinsic to patient and a function of the hospital), a more robust patient-specific data source is required than routine discharge data. To address this question, we will utilize a unique data source to extend our prior work examining the impact of value-based purchasing programs (including non-payment of HACs) on vulnerable patients and hospitals; we will use the nationally representative Health and Retirement Study (HRS) (with detailed data such as a patient?s functional status, mobility, social support, income and educational level) linked to patient-specific Medicare claims data. Our specific aims are:
1. To assess change in performance of our recently constructed risk-adjusted model for complications of pressure ulcers and urinary track infections as HACs after enhancement with HRS patient-specific measures (e.g., functional status, mobility, social support).
2. To assess change in performance of CMS?s risk-adjustment models for readmission (for pneumonia, heart failure, myocardial infarction) after enhancement with HRS patient-specific measures.
3. To evaluate the performance of the HRS-variable enhanced risk-adjustment models for HACs and readmission after replacing some HRS variables with census derived, zip-code level variables (such as median level of education, and income).
4. Using statewide Medicare claims data; to evaluate the performance of risk-adjustment models for HACs and readmission enhanced by census-data derived zip-code level socio-demographic variables.