Health Equity and Hospital Readmissions: Does Inclusion of Patient Functional and Social Complexity Improve Predictiveness? (Journal of General Internal Medicine)

Medicare’s Hospital Readmissions Reduction Program (HRRP) was intended to encourage hospitals to improve care for older adults. However, the program has raised health equity concerns because its risk-adjustment model does not account for patient social complexity; the concern is that HRRP may aggravate health care disparities by penalizing financially challenged hospitals and reducing their resources to improve care. Authors examined 1,742 observations for fee-for-service Medicare beneficiaries ages ≥ 65 years from 2006 to 2012 Medicare hospital and linked Health and Retirement Study (HRS) data. Social support data were obtained from the HRS Psychosocial and Lifestyle Questionnaire (PLQ), given every other wave to a randomly chosen half of the full HRS sample. Outcomes were CMS unplanned hospital-wide 30-day readmissions.

Authors compared the discrimination of three models: (1) the CMS base model (adjusting for age, sex, and clinical risk factors), (2) the CMS base model additionally adjusting for patient sociodemographics, functional/health status, health behaviors, and social factors (“full model”), and (3) the CMS base model additionally adjusting for income and race (“proxy model”). Patient health and social factors and healthy behavior were modeled as four latent factors using factor analysis: (1) socioeconomic status was indicated by household income and wealth; (2) poor health and functioning by self-rated health, indices of chronic conditions and difficulties with activities of daily living (ADL) and instrumental ADLs, receipt of home care, cognitive impairment, and use of psychiatric medications; (3) negative social support was indicated by negative support from spouse, child, other family, and friends; and (4) healthy behaviors by regular vigorous, moderate, and light household physical activity. Predictiveness was measured using model concordance, or the c-statistic.

The sample had an average age of 74.8 years and was majority male. On average, respondents had two chronic conditions, with 35% reporting fair or poor health. Overall, 11.9% of respondents were readmitted. We observed limited differences in latent factors measuring health/functioning, social factors, and health behaviors by readmission status. In multivariable analyses, compared to the base CMS model, the full model had better predictiveness, but the proxy model did not.

• Geoffrey J. Hoffman
• Charleen Hsuan, JD, PhD
• Ninez A. Ponce, PhD, MPP
• et al