In-Hospital vs 30-Day Sepsis Mortality at US Safety-Net and Non-Safety-Net Hospitals.

Importance: In-hospital mortality of patients with sepsis is frequently measured for benchmarking, both by researchers and policymakers. Prior studies have reported higher in-hospital mortality among patients with sepsis at safety-net hospitals compared with non-safety-net hospitals; however, in critically ill patients, in-hospital mortality rates are known to be associated with hospital discharge practices, which may differ between safety-net hospitals and non-safety-net hospitals. Objective: To assess how admission to safety-net hospitals is associated with 2 metrics of short-term mortality (in-hospital mortality and 30-day mortality) and discharge practices among patients with sepsis. Design, Setting, and Participants: Retrospective, national cohort study of Medicare fee-for-service beneficiaries aged 66 years and older, admitted with sepsis to an intensive care unit from January 2011 to December 2019 based on information from the Medicare Provider Analysis and Review File. Data were analyzed from October 2022 to September 2023. Exposure: Admission to a safety-net hospital (hospitals with a Medicare disproportionate share index in the top quartile per US region). Main Outcomes and Measures: Coprimary outcomes: in-hospital mortality and 30-day mortality. Secondary outcomes: (1) in-hospital do-not-resuscitate orders, (2) in-hospital palliative care delivery, (3) discharge to a postacute facility (skilled nursing facility, inpatient rehabilitation facility, or long-term acute care hospital), and (4) discharge to hospice. Results: Between 2011 and 2019, 2 551 743 patients with sepsis (mean [SD] age, 78.8 [8.2] years; 1 324 109 [51.9%] female; 262 496 [10.3%] Black, 2 137 493 [83.8%] White, and 151 754 [5.9%] other) were admitted to 666 safety-net hospitals and 1924 non-safety-net hospitals. Admission to safety-net hospitals was associated with higher in-hospital mortality (odds ratio [OR], 1.09; 95% CI, 1.06-1.13) but not 30-day mortality (OR, 1.01; 95% CI, 0.99-1.04). Admission to safety-net hospitals was associated with lower do-not-resuscitate rates (OR, 0.86; 95% CI, 0.81-0.91), palliative care delivery rates (OR, 0.66; 95% CI, 0.60-0.73), and hospice discharge (OR, 0.82; 95% CI, 0.78-0.87) but not with discharge to postacute facilities (OR, 0.98; 95% CI, 0.95-1.01). Conclusions and Relevance: In this cohort study, among patients with sepsis, admission to safety-net hospitals was associated with higher in-hospital mortality but not with 30-day mortality. Differences in in-hospital mortality may partially be explained by greater use of hospice at non-safety-net hospitals, which shifts attribution of death from the index hospitalization to hospice. Future investigations and publicly reported quality measures should consider time-delimited rather than hospital-delimited measures of short-term mortality to avoid undue penalty to safety-net hospitals with similar short-term mortality.

A Novel Machine Learning Algorithm for Creating Risk-Adjusted Payment Formulas.

Importance: Models predicting health care spending and other outcomes from administrative records are widely used to manage and pay for health care, despite well-documented deficiencies. New methods are needed that can incorporate more than 70 000 diagnoses without creating undesirable coding incentives. Objective: To develop a machine learning (ML) algorithm, building on Diagnostic Item (DXI) categories and Diagnostic Cost Group (DCG) methods, that automates development of clinically credible and transparent predictive models for policymakers and clinicians. Design, Setting, and Participants: DXIs were organized into disease hierarchies and assigned an Appropriateness to Include (ATI) score to reflect vagueness and gameability concerns. A novel automated DCG algorithm iteratively assigned DXIs in 1 or more disease hierarchies to DCGs, identifying sets of DXIs with the largest regression coefficient as dominant; presence of a previously identified dominating DXI removed lower-ranked ones before the next iteration. The Merative MarketScan Commercial Claims and Encounters Database for commercial health insurance enrollees 64 years and younger was used. Data from January 2016 through December 2018 were randomly split 90% to 10% for model development and validation, respectively. Deidentified claims and enrollment data were delivered by Merative the following November in each calendar year and analyzed from November 2020 to January 2024. Main Outcome and Measures: Concurrent top-coded total health care cost. Model performance was assessed using validation sample weighted least-squares regression, mean absolute errors, and mean errors for rare and common diagnoses. Results: This study included 35 245 586 commercial health insurance enrollees 64 years and younger (65 901 460 person-years) and relied on 19 clinicians who provided reviews in the base model. The algorithm implemented 218 clinician-specified hierarchies compared with the US Department of Health and Human Services (HHS) hierarchical condition category (HCC) model's 64 hierarchies. The base model that dropped vague and gameable DXIs reduced the number of parameters by 80% (1624 of 3150), achieved an R2 of 0.535, and kept mean predicted spending within 12% ($3843 of $31 313) of actual spending for the 3% of people with rare diseases. In contrast, the HHS HCC model had an R2 of 0.428 and underpaid this group by 33% ($10 354 of $31 313). Conclusions and Relevance: In this study, by automating DXI clustering within clinically specified hierarchies, this algorithm built clinically interpretable risk models in large datasets while addressing diagnostic vagueness and gameability concerns.

Medicaid Expansion and Preventable Emergency Department Use by Race/Ethnicity.

INTRODUCTION: This study aimed to examine changes in emergency department (ED) visits for ambulatory care sensitive conditions (ACSCs) among uninsured or Medicaid-covered Black, Hispanic, and White adults aged 26-64 in the first 5 years of the Affordable Care Act Medicaid expansion. METHODS: Using 2010-2018 inpatient and ED discharge data from nine expansion and five nonexpansion states, an event study difference-in-differences regression model was used to estimate changes in number of annual ACSC ED visits per 100 adults ("ACSC ED rate") associated with the 2014 Medicaid expansion, overall and by race/ethnicity. A secondary outcome was the proportion of ACSC ED visits out of all ED visits ("ACSC ED share"). Analyses were conducted in 2022-2023. RESULTS: Medicaid expansion was associated with no change in ACSC ED rates among all, Black, Hispanic, or White adults. When excluding California, where most counties expanded Medicaid before 2014, expansion was associated with a decrease in ACSC ED rate among all, Black, Hispanic, and White adults. Expansion was also associated with a decrease in ACSC ED share among all, Black, and White adults. White adults experienced the largest reductions in ACSC ED rate and share. CONCLUSIONS: Medicaid expansion was associated with reductions in ACSC ED rates in some expansion states and reductions in ACSC ED share in all expansion states combined, with some heterogeneity by race/ethnicity. Expansion should be coupled with policy efforts to better link newly insured Black and Hispanic patients to non-ED outpatient care, alongside targeted outreach and expanded primary care capacity, which may reduce disparities in ACSC ED visits.