ICU-onset atrial fibrillation surrogate, thrombocytopenia, and short-term outcomes in patients with cancer requiring intensive care: a retrospective ICU cohort study.

backgroundCancer and atrial fibrillation frequently coexist. Intensive care adds hemodynamic stress, organ failure, and competing bleeding risk. Whether platelet-defined thrombocytopenia modifies the association between an ICU-onset atrial fibrillation surrogate and short-term outcomes in patients with cancer remains uncertain.

methodsWe performed a secondary analysis of a retrospective ICU cohort derived from MIMIC-IV version 3.1, a de-identified critical care database with date-shifted patient timelines. Adult ICU stays with solid or hematologic malignancy were included. The exposure was an ICU-onset AF surrogate rather than adjudicated incident AF. Early thrombocytopenia was defined from the minimum platelet count within 48 h and categorized as > = 150, 100-149, 50-99, and < 50 × 10^9/L. The primary outcome was in-hospital death. Secondary outcomes were 28-day death, major bleeding identified by ICD coding, and ischemic stroke identified by ICD coding. Multivariable logistic regression tested AF-by-platelet interaction. Sensitivity models added available cancer-site, cardiovascular comorbidity, major surgery, structured anticancer medication indicators, and anticoagulation dose-class variables.

resultsAmong 9,640 cancer ICU stays, 1,788 (18.5%) met criteria for the ICU-onset AF surrogate. In-hospital death occurred in 1,724 stays (17.9%), 28-day death in 2,394 (24.8%), ICD-coded major bleeding in 1,001 (10.4%), and ICD-coded ischemic stroke in 410 (4.3%). In primary complete-case interaction models (n = 9,505), the AF-by-platelet interaction for in-hospital death was borderline (P = 0.069), with the largest mortality signal in the platelet < 50 × 10^9/L stratum (adjusted OR 1.61, 95% CI 1.09-2.37). This signal persisted after expanded adjustment (adjusted OR 1.54, 95% CI 1.04-2.27) and anticoagulation dose-class adjustment (adjusted OR 1.50, 95% CI 1.01-2.22). For ICD-coded major bleeding, the interaction was significant in the primary model (P = 0.040) and remained similar after expanded adjustment (P = 0.050); the platelet 50-99 × 10^9/L stratum showed lower adjusted odds, a finding interpreted as exploratory.

conclusionsIn this cancer ICU cohort, an ICU-onset AF surrogate was associated with short-term outcome patterns that differed across platelet strata. Severe thrombocytopenia marked the subgroup with the clearest mortality signal. Because AF, bleeding, and stroke were defined from structured data rather than adjudicated events, the findings should be interpreted as hypothesis-generating.