Research Letter
April 9, 2025
Concordance of 30-Day Mortality and In-Hospital Mortality or Hospice Discharge After Sepsis
Hallie C. Prescott, Megan Heath, Namita Jayaprakash, et al
JAMA. Published online April 9, 2025. doi:10.1001/jama.2025.2526
Sepsis contributes to 1.7 million adult hospitalizations and 350 000 deaths annually in the United States.1To encourage improvements in sepsis management and outcomes, the Centers for Disease Control and Prevention (CDC) and the Centers for Medicare & Medicaid Services (CMS) are developing a risk-adjusted outcome measure for community-onset sepsis hospitalizations.2 Two candidate measures are (1) 30-day mortality after admission and (2) a composite outcome of in-hospital mortality or hospice discharge. While often preferred, 30-day mortality has drawbacks because postdischarge vital status data are not available in real time, causing a lag in outcomes assessment. Furthermore, 30-day mortality measures used in CMS programs rely on claims and social security data available only for Medicare and Medicaid beneficiaries, thus limiting the eligible population. To inform decision-making about which measure to implement for national benchmarking, we evaluated the 2 candidate measures in a multihospital cohort.
Methods
The Michigan Hospital Medicine Safety Consortium’s Sepsis Initiative (HMS-Sepsis) is a multihospital collaborative quality initiative funded by Blue Cross Blue Shield of Michigan. Professional abstractors enter data on a random sample of all-payer, community-onset adult sepsis hospitalizations into a central registry. Hospitalizations with diagnoses of sepsis or infection are manually reviewed for clinical evidence of infection and acute organ dysfunction to confirm eligibility.3 Outcomes through 90 days postdischarge are determined for all patients via chart review, public obituary data, and follow-up by telephone, email, and/or SMS messaging.
Using HMS-Sepsis registry data for patients discharged between November 2020 and July 2024, we assessed the concordance between 30-day and in-hospital mortality and between 30-day mortality and in-hospital mortality or hospice discharge, at both the encounter level (2 × 2 tables; Cohen κ) and the hospital level (Pearson correlation coefficient). We additionally evaluated whether concordance differed by hospitals’ use of (1) long-term acute care (LTAC) facilities or (2) post–acute care facilities (LTAC, subacute rehabilitation, or skilled nursing facilities), as measured by proportion of sepsis hospitalizations discharged to these facilities, since prior research has suggested that differential use of LTAC facilities may bias assessments of hospital mortality.4 Data were analyzed from October 21, 2024, to November 6, 2024. The study was deemed not regulated by the University of Michigan’s institutional review board.
Results
Among 37 200 patients hospitalized with community-onset sepsis at 67 hospitals, 4331 (11.6%) died in the hospital, 2805 (7.5%) were discharged to hospice, 8059 (21.7%) were discharged to a post–acute care facility (650 to an LTAC facility), and 7167 (19.3%) died within 30 days of admission. A total of 7136 (19.2%) died in the hospital or were discharged to hospice. Among the 2805 patients discharged to hospice, 1818 (64.8%) died within 30 days of admission, while 711 (25.3%) remained alive through day 90 after admission.
A total of 34 232 hospitalizations (92.0%) were classified concordantly by 30-day and in-hospital mortality (κ = 0.698), while 35 063 (94.3%) were classified concordantly by 30-day mortality and the composite outcome measure (κ = 0.815) (P < .001 for comparison) (Table). At the hospital level, there was numerically stronger correlation between 30-day mortality and the composite measure (r = 0.879) vs in-hospital mortality alone (r = 0.739), although the difference was not significant (P = .429) (Figure). Hospital-level use of LTAC facilities and post–acute care was weakly correlated with the difference between 30-day mortality and the composite measure (r = 0.357, P = .003; r = 0.239, P = .052) (Figure). The composite outcome was higher than 30-day mortality among hospitals with greater LTAC or post–acute care use.
Discussion
In this all-payer, multihospital cohort of patients with community-onset sepsis, the proportion of hospitalizations ending in death or hospice discharge was similar to 30-day mortality and substantially higher than in-hospital mortality alone. Concordance between this composite outcome and 30-day mortality was high and greater than in-hospital mortality alone at the encounter level.
These data suggest that the composite outcome of in-hospital mortality or hospice discharge may be a useful measure for national benchmarking of sepsis outcomes. While not identical to 30-day mortality, this composite outcome is available in real time for all patients and appears to have minimal systematic bias from variable hospital discharge practices.
