1049 — Impact of Risk-adjustment for Socioeconomic Status on Veterans Affairs Medical Center Mortality Rates

Lead/Presenter: Amal Trivedi,  COIN - Providence
All Authors: Trivedi AN (Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI), Jiang L (Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI), Silva G (Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI) Wu W-C (Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI) Mor V (Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI) Fine MJ (Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, Pittsburgh, PA) Kressin NR (Center for Healthcare Organization and Implementation Research, VA Boston Healthcare System, Boston, MA) Gutman R ((Center of Innovation in Long-Term Services and Supports for Vulnerable Veterans, Providence, RI)

Objectives:
All Veterans Affairs Medical Centers (VAMCs) report risk-adjusted mortality rates for hospitalized patients using a CMS-developed claims-based model that does not include demographic factors other than age and sex. We evaluated the impact of adjusting for socioeconomic factors on VAMCs' 30-day mortality rates in heart failure and pneumonia.

Methods:
We assessed the performance of models that included CMS' claims-based covariates (28 comorbid conditions), clinical covariates (vital signs, laboratory values, and ejection fraction), and socioeconomic factors (census block-level neighborhood disadvantage, race/ethnicity, enrollment priority, Medicaid enrollment, homelessness, nursing home use, and rural residence). We fitted hierarchical random effects logistic regression models to estimate all 132 VAMCs hospital's risk-standardized 30-day mortality rates in heart failure and pneumonia after inclusion of claims-based, clinical, and socioeconomic covariates.

Results:
The addition of socioeconomic factors to CMS' model only modestly increased the c-statistic from 0.68 to 0.69 for 30-day mortality in heart failure and from 0.71 to 0.72 for 30-day mortality in pneumonia. VAMCs' mortality rates were highly correlated in models that included and did not include socioeconomic factors, with Spearman correlations of 0.98 for both heart failure and pneumonia. Using CMS' model for heart failure, VAMCs in the lowest, middle three, and highest quintiles had mean mortality rates of 6.0%, 7.2% and 8.8%, respectively. After adding socioeconomic covariates, the mortality rates for these hospitals were 6.1%, 7.2% and 8.6%. The mean change in rank after socioeconomic adjustment was 0.4 ranking positions (range -8 to +10) among hospitals in the highest quintile of mortality in heart failure and 0.9 ranking positions (range -11 to +17) among VAMCs in the lowest quintile. Similar findings were observed for mortality rankings in pneumonia, and for analyses that included clinical covariates or were limited to Veterans age 66+ enrolled in Medicare.

Implications:
The inclusion of socioeconomic factors did not meaningfully change VAMCs' risk-adjusted 30-day mortality rates following admission for heart failure and pneumonia.

Impacts:
Although there is vigorous policy debate about whether and how to adjust outcome measures for patients' socioeconomic characteristics, such adjustments are unlikely to have substantial impact on assessments of hospital mortality in VA.