BACKGROUND/RATIONALE:
Nearly all hospitals in the United States, including all VA Medical Centers (VAMCs), report mortality rates for hospitalized patients, and performance on these outcomes measures often carry high stakes. Hospital mortality constitutes two domains of the Strategic Analytic Information and Learning (SAIL) model that VA employs to evaluate the quality and efficiency of care provided across all VAMCs. Valid hospital outcome measures must adequately account for differences in clinical risk. Without adequate risk-adjustment, performance reports may erroneously penalize facilities that serve high-risk populations, or, even worse, incentivize facilities to admit low-risk patients. Much of the prior concern with risk-adjustment has involved the source of the data, the selection of appropriate covariates, or the optimal approach to statistical modeling. Substantially fewer studies have examined the role of socioeconomic status and other sociodemographic factors in risk-adjustment, though these factors predict worse post-discharge outcomes and vary markedly across facilities.

OBJECTIVE(S):
The overarching goal of this project is to develop and test novel risk-adjustment approaches that incorporate Veterans' sociodemographic characteristics into assessments of hospital mortality. Our aims are: 1) describe VAMC-level variations in the sociodemographic characteristics of Veterans hospitalized with heart failure and pneumonia; 2) assess the performance of risk-adjustment models that do and do not include sociodemographic characteristics; and 3) evaluate the impact of incorporating sociodemographic data on the relative performance of VA Medical Centers.

METHODS:
We propose a retrospective, observational study that will develop and compare alternative risk-adjustment models predicting mortality within thirty days of admission for heart failure and pneumonia. We will then test the performance of models that do and do not incorporate sociodemographic characteristics and assess the impact of including sociodemographic characteristics on profiling VAMC-level hospital mortality rates for heart failure and pneumonia. Aim 1 will assess how the sociodemographic characteristics of Veterans admitted with heart failure and pneumonia vary across VAMCs. Aim 2 will compare the existing claims-based VA/Centers for Medicare and Medicaid Services (CMS) risk-adjusted mortality models for heart failure and pneumonia with models that incorporate claims and novel sociodemographic data; and determine the contribution of sociodemographic characteristics to mortality models that include both claims-based diagnoses and clinical covariates derived from the VA's electronic health record. Aim 3 extends these analyses by determining whether relative quality rankings of VA medical centers change when sociodemographic factors are included in mortality risk-adjustment models.

FINDINGS/RESULTS:
We derived a risk-adjustment model of 30-day mortality that included clinical and sociodemographic covariates and compared the performance of these models to those that included claims-based covariates alone. The addition of novel clinical and sociodemographic covariates improved the performance of risk-adjustment models. In the pneumonia cohort, the c-statistics were 0.68 (claim-based model), 0.77 (claims-based plus clinical covariates), and 0.81 (claims-based, clinical and sociodemographic covariates). In the heart failure cohort, the c-statistics for risk-adjustment models were 0.68 (claim-based model), 0.78 (claims-based plus clinical covariates), and 0.78 (claims-based, clinical and sociodemographic covariates).

IMPACT:
The project will develop and test novel approaches to risk-adjustment that include Veterans' sociodemographic factors into assessments of VAMC-level hospital mortality. This contribution is significant because rigorous outcomes measurement is central to VA's strategy to improve care for Veterans, assess quality across sites, and benchmark performance to the private-sector. If risk-adjustment fails to account for sociodemographic determinants of mortality that are known to vary across VA providers, then excluding these factors may penalize VA sites that disproportionately serve vulnerable patients and generate incorrect inferences about the quality of VA care. Since VA uses performance results for accountability purposes and to make determinations about allocation of resources, it is essential that VA uses the most robust risk-adjustment methods available. Given emerging momentum to consider sociodemographic characteristics for risk-adjustment purposes, there is a pressing need to develop an empirical evidence base about the implications of such adjustments.

We will update this section when our initial set of analyses are completed and the results disseminated.

---

External Links for this Project

NIH Reporter

Grant Number: I01HX002104-01A1
Link: https://reporter.nih.gov/project-details/9188841

---

PUBLICATIONS:

Journal Articles

1. Trivedi AN, Jiang L, Johnson EE, Lima JC, Flores M, O'Toole TP. Dual Use and Hospital Admissions among Veterans Enrolled in the VA's Homeless Patient Aligned Care Team. Health services research. 2018 Dec 1; 53 Suppl 3:5219-5237. [view]
2. Trivedi AN, Jiang L, Silva G, Wu WC, Mor V, Fine MJ, Kressin NR, Gutman R. Evaluation of Changes in Veterans Affairs Medical Centers' Mortality Rates After Risk Adjustment for Socioeconomic Status. JAMA Network Open. 2020 Dec 1; 3(12):e2024345. [view]
3. Silva GC, Jiang L, Gutman R, Wu WC, Mor V, Fine MJ, Kressin NR, Trivedi AN. Mortality Trends for Veterans Hospitalized With Heart Failure and Pneumonia Using Claims-Based vs Clinical Risk-Adjustment Variables. JAMA internal medicine. 2020 Mar 1; 180(3):347-355. [view]
4. Silva GC, Jiang L, Gutman R, Wu WC, Mor V, Fine MJ, Kressin NR, Trivedi AN. Racial/Ethnic Differences in 30-Day Mortality for Heart Failure and Pneumonia in the Veterans Health Administration Using Claims-based, Clinical, and Social Risk-adjustment Variables. Medical care. 2021 Dec 1; 59(12):1082-1089. [view]

DRA: Health Systems Science
DRE: TRL - Applied/Translational
Keywords: Provider Performance Measures, Risk Adjustment
MeSH Terms: none