Acute kidney injury and in-hospital mortality in critically ill patients with gynecologic malignancies: a retrospective cohort study across three intensive care databases

Abstract Background Acute kidney injury (AKI) affects more than half of critically ill patients and is independently associated with mortality, yet evidence in women with gynecologic malignancies admitted to the intensive care unit (ICU) remains scarce. The only prior multicenter study in this population relied on database-recorded (ICD-coded) AKI, an exposure definition prone to misclassification. We evaluated the association between standardized, KDIGO-based AKI staging and mortality in a large three-database cohort. Methods This retrospective cohort study used MIMIC-IV (v3.0), MIMIC-III (v1.4), and eICU-CRD (v2.0; https://physionet.org/content/eicu-crd/2.0/). We included adult women with gynecologic malignancies during their first ICU admission (n = 203, 151, and 463, respectively; total 817). Exposure was the maximum KDIGO creatinine-based AKI stage within 7 days of ICU admission (severe AKI = stage ≥ 2). The primary outcome was in-hospital mortality. Database-specific multivariable logistic regressions adjusted for age, cancer type, non-renal illness severity, diabetes, chronic kidney disease, and sepsis were combined by DerSimonian-Laird random-effects pooling. Sensitivity analyses addressed AKI definition (urine-output-inclusive), urine output (UO)–creatinine concordance, and missing data via multiple imputation by chained equations (MICE, m = 20). Results Any AKI occurred in 209/796 patients (26.3%) and severe AKI in 55 (6.9%). In-hospital mortality increased monotonically across KDIGO stages 0–3 (10.2%, 21.4%, 31.6%, 41.2%; adjusted p-trend = 2.7×10⁻⁶). Severe AKI was associated with in-hospital mortality (pooled adjusted OR 3.78, 95% CI 1.62–8.80; p = 0.002; I²=24%), as was any AKI (OR 2.70, 1.29–5.67; p = 0.009). Severe AKI was also associated with ICU mortality (OR 3.22, 0.96–10.76) and 28-day mortality (OR 2.78, 1.28–6.04). Estimates were robust to MICE (severe AKI OR 3.48, 1.55–7.82) and persisted under the full KDIGO definition in MIMIC (OR 2.19, p = 0.017), whereas urine-output-inclusive definitions in eICU attenuated the associations, likely reflecting liberal UO thresholds and sparse, informatively missing UO documentation. Conclusions In this three-database cohort of critically ill patients with gynecologic malignancies, severe AKI defined by standardized creatinine-based KDIGO staging showed a strong, dose-dependent association with short-term mortality. Standardized AKI staging identifies a high-risk population that may benefit from intensified monitoring and nephrotoxin stewardship.