Terminal admission phenotypes and inpatient mortality in gastrointestinal malignancies: A national analysis of U.S. hospitalizations, 2018–2023.

Abstract

e23231 Background: Inpatient mortality in gastrointestinal (GI) malignancies is often reported as a binary outcome, limiting insight into acute pathways preceding death. We characterized terminal admission phenotypes among hospitalized adults with GI cancers and quantified their national prevalence, resource utilization, and variation by cancer site and hospital characteristics. Methods: We conducted a survey-weighted cross-sectional analysis of the National Inpatient Sample (NIS), 2018–2023. Adult hospitalizations (≥18 years) with any-diagnosis GI malignancy (ICD-10-CM C15–C26) were included; terminal admissions were defined by in-hospital death. Terminal hospitalizations were hierarchically classified into mutually exclusive phenotypes using diagnosis-based groupings and utilization proxies: sepsis/infection, respiratory failure/ARDS, bleeding/anemia, cachexia/malnutrition/dehydration, hepatic failure, bowel obstruction/perforation, procedure-associated, ICU escalation without procedure, and other/unclassified. Cancer site was assigned using the principal diagnosis when available, with any-diagnosis fallback. Outcomes included phenotype prevalence, length of stay, hospitalization cost, ICU escalation, and procedure utilization. All estimates incorporated NIS stratification, clustering, and discharge weights, with 95% confidence intervals (CI). Results: GI malignancies accounted for 1.98% of U.S. hospitalizations (weighted mean 0.0198, 95% CI 0.0196–0.0200), representing 221,090 weighted inpatient deaths from 44,218 unweighted terminal admissions. Terminal phenotypes were dominated by sepsis/infection (51.9%, 95% CI 51.3–52.5), followed by respiratory failure/ARDS (12.8%, 95% CI 12.5–13.2), bleeding/anemia (10.1%, 95% CI 9.83–10.41), cachexia/malnutrition/dehydration (7.91%, 95% CI 7.62–8.22), and hepatic failure (5.94%, 95% CI 5.72–6.17). Distributions varied by cancer site, with hepatic failure prominent in hepatocellular/liver cancer (14.9%, 95% CI 14.1–16.0) and respiratory failure/ARDS in esophageal cancer (18.0%, 95% CI 16.9–19.2). Mean LOS and costs were highest for sepsis/infection (9.28 days, 95% CI 9.10–9.46; $43,357, 95% CI $42,041–$44,672). ICU escalation was common in sepsis/infection (42.9%, 95% CI 42.3–43.6) and bleeding/anemia (45.5%, 95% CI 44.0–47.0), but infrequent in cachexia/malnutrition/dehydration (6.86%, 95% CI 6.01–7.72). Conclusions: Inpatient mortality among hospitalizations involving GI malignancies reflects a limited set of terminal admission phenotypes, most often acute infectious, respiratory, bleeding, and nutritional complications. These phenotypes show distinct utilization patterns and vary by cancer site and hospital type, supporting a phenotype-based framework for identifying potentially modifiable acute-care pathways.