Social determinants of health and risk-adjusted sepsis mortality in the nationwide Veterans Affairs Healthcare System

Abstract: Importance: Traditional risk prediction and risk adjustment models have focused on clinical characteristics, but accounting for social determinants of health (SDOH) and complex health conditions could improve understanding of sepsis outcomes and our ability to predict outcomes, treat patients, and assess quality of care. Objective: To evaluate the impact of SDOH and health scales in sepsis mortality risk prediction and hospital performance assessment. Design: Observational cohort study. Setting: One hundred twenty-nine hospitals in the nationwide Veterans Affairs (VA) Healthcare System between 2017 and 2021. Participants: Veterans admitted through emergency departments with community-acquired sepsis. Exposures: Individual- and community-level SDOH (race, housing instability, marital status, Area Deprivation Index [ADI], and rural residence) and two health scales (the Care Assessment Need [CAN] score and Claims-Based Frailty Index [CFI]). Main outcomes and measures: The primary outcome was 90-day mortality from emergency department arrival; secondary outcomes included 30-day mortality and in-hospital mortality. Results: Among 144,889 patients admitted to the hospital with community-acquired sepsis, 139,080 were men (96.0%), median (IQR) age was 71 (64-77) years, and median (IQR) ADI was 60 (38-81). Multivariable regression models had good calibration and discrimination across models that adjusted for different sets of variables (e.g., AUROC, 0.782; Brier score, 1.33; and standardized mortality rate, 1.00). Risk-adjusted hospital performance was similar across all models. Among 129 VA hospitals, three hospitals shifted from the lowest or highest quintile of performance when comparing models that excluded SDOH to models that adjusted for all variables. Models that adjusted for ADI reported odds ratios (CI) of 1.00 (1.00-1.00), indicating that ADI does not significantly predict sepsis mortality in this cohort of patients. Conclusion and relevance: In patients with community-acquired sepsis, adjusting for community SDOH variables such as ADI did not improve 90-day sepsis mortality predictions in mortality models and did not substantively alter hospital performance within the VA Healthcare System. Understanding the role of SDOH in risk prediction and risk adjustment models is vital because it could prevent hospitals from being negatively evaluated for treating less advantaged patients. However, we found that in VA hospitals, the potential impact of SDOH on 90-day sepsis mortality was minimal.

Read the full article
Report a problem with this article

Related articles

  • More for Researchers

    Identifying opioid relapse during COVID-19 using natural language processing of nationwide Veterans Health Administration electronic medical record data

    Abstract: Novel and automated means of opioid use and relapse risk detection are needed. Unstructured electronic medical record data, including written progress notes, can be mined for clinically relevant information, including the presence of substance use and relapse-critical markers of risk and recovery from opioid use disorder (OUD). In this study, we used natural language processing (NLP) to automate the extraction of opioid relapses, and the timing of these occurrences, from veteran patients' electronic medical record. We then demonstrated the utility of our NLP tool via analysis of pre-/post-COVID-19 opioid relapse trends among veterans with OUD. For this demonstration, we analyzed data from 107,606 veterans OUD enrolled in Veterans Health Administration, comparing a pandemic-exposed cohort (n = 53,803; January 2019-March 2021) to a matched prepandemic cohort (n = 53,803; October 2017-December 2019). The recall of our NLP tool was 75% and our precision was 94%, demonstrating moderate sensitivity and excellent specificity. Using the NLP tool, we found that the odds of opioid relapse postpandemic onset were proportionally higher compared to prepandemic trends, despite patients having fewer mental health encounters from which to derive instances of relapse postpandemic onset. In this research application of the tool, and as hypothesized, we found that opioid relapse risk was elevated postpandemic. The application of NLP Methods: to identify and monitor relapse risk holds promise for future surveillance, risk prevention, and clinical outcome research.

    Mental Health Peer reviewed