Measuring Emergency Care Survival: The Implications of Risk Adjusting for Race and Poverty

Kimon L H Ioannides; Avi Baehr; David N Karp; Douglas J Wiebe; Brendan G Carr; Daniel N Holena; M Kit Delgado

doi:10.1111/acem.13485

Measuring Emergency Care Survival: The Implications of Risk Adjusting for Race and Poverty

Acad Emerg Med. 2018 Aug;25(8):856-869. doi: 10.1111/acem.13485.

Authors

Kimon L H Ioannides^{1

2}, Avi Baehr^{2

3}, David N Karp⁴, Douglas J Wiebe^{4

5}, Brendan G Carr⁶, Daniel N Holena^{4

7

5}, M Kit Delgado^{4

8

5}

Affiliations

¹ Department of Emergency Medicine, Temple University Hospital, Philadelphia, PA.
² Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA.
³ Department of Emergency Medicine, Denver Health and Hospital Authority, Denver, CO.
⁴ Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania, Philadelphia, PA.
⁵ Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia, PA.
⁶ Department of Emergency Medicine, Sidney Kimmel Medical College, Thomas Jefferson University, Philadelphia, PA.
⁷ Department of Surgery, University of Pennsylvania, Philadelphia, PA.
⁸ Center for Emergency Care Policy and Research, Department of Emergency Medicine, University of Pennsylvania, Philadelphia, PA.

Abstract

Objectives: We determined the impact of including race, ethnicity, and poverty in risk adjustment models for emergency care-sensitive conditions mortality that could be used for hospital pay-for-performance initiatives. We hypothesized that adjusting for race, ethnicity, and poverty would bolster rankings for hospitals that cared for a disproportionate share of nonwhite, Hispanic, or poor patients.

Methods: We performed a cross-sectional analysis of patients admitted from the emergency department to 157 hospitals in Pennsylvania with trauma, sepsis, stroke, cardiac arrest, and ST-elevation myocardial infarction. We used multivariable logistic regression models to predict in-hospital mortality. We determined the predictive accuracy of adding patient race and ethnicity (dichotomized as non-Hispanic white vs. all other Hispanic or nonwhite patients) and poverty (uninsured, on Medicaid, or lowest income quartile zip code vs. all others) to other patient-level covariates. We then ranked each hospital on observed-to-expected mortality, with and without race, ethnicity, and poverty in the model, and examined characteristics of hospitals with large changes between models.

Results: The overall mortality rate among 170,750 inpatients was 6.9%. Mortality was significantly higher for nonwhite and Hispanic patients (adjusted odds ratio [aOR] = 1.27, 95% confidence interval [CI] = 1.19-1.36) and poor patients (aOR = 1.21, 95% CI = 1.12-1.31). Adding race, ethnicity, and poverty to the risk adjustment model resulted in a small increase in C-statistic (0.8260 to 0.8265, p = 0.002). No hospitals moved into or out of the highest-performing decile when adjustment for race, ethnicity, and poverty was added, but the three hospitals that moved out of the lowest-performing decile, relative to other hospitals, had significantly more nonwhite and Hispanic patients (68% vs. 11%, p < 0.001) and poor patients (56% vs. 10%, p < 0.001).

Conclusions: Sociodemographic risk adjustment of emergency care-sensitive mortality improves apparent performance of some hospitals treating a large number of nonwhite, Hispanic, or poor patients. This may help these hospitals avoid financial penalties in pay-for-performance programs.

Abstract

Grants and funding