Work

Longitudinal laboratory patterns and subsequent mortality in intensive care admissions with evidence of active abdominal aortic aneurysm: a MIMIC-IV landmark study

Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Intensive care unit (ICU) admissions involving abdominal aortic aneurysm (AAA) are heterogeneous, and conventional summary values may obscure temporal change. We evaluated whether ordered laboratory measurements identify reproducible patterns associated with subsequent mortality in admissions with active AAA evidence. Methods We conducted a retrospective MIMIC-IV v3.1 landmark study. Active AAA evidence required a principal diagnosis, rupture code, repair code, or text-supported rupture or repair. Admissions alive and hospitalized at ICU hour 72 were eligible. Six laboratory markers were summarized in three consecutive 24-hour windows. Twelve posterior-sampling iterative imputations were followed by Gaussian mixture models and consensus clustering. Subsequent mortality was analyzed using logistic regression with patient-clustered robust standard errors. Results Of 1,343 AAA-coded ICU admissions, 526 had active AAA evidence. After excluding 16 deaths and 59 live discharges before hour 72, 451 admissions from 433 patients remained. The selected solution comprised a lower-intensity pattern (n = 289) and a multisystem derangement pattern (n = 162), with a median adjusted Rand index of 0.926. Subsequent mortality was 3.5% and 12.3%, respectively. After adjustment for age, sex, and chronic kidney disease, the odds ratio was 3.79 (95% confidence interval, 1.40-10.25; P = .009). Sensitivity estimates were directionally consistent across alternative covariate, cohort, completeness, and marker specifications. In post hoc analyses, excluding creatinine disrupted class agreement (adjusted Rand index, -0.011) and attenuated the association (odds ratio, 1.65; 95% confidence interval, 0.67–4.06); additional adjustment for early creatinine yielded an odds ratio of 2.57 (95% confidence interval, 0.87–7.58). Conclusions Ordered 72-hour laboratory measurements identified two reproducible patterns associated with subsequent mortality. The classification was substantially anchored by renal function and should be viewed as exploratory rather than as an AAA-specific subtype or clinical prediction tool. Trial registration Clinical trial number: not applicable.
Language: English
Subfield: Surgery
Field: Medicine
Open Access status: green
OpenAlex ID: W7236447485
Cites: 20
Cited by:
Related to: 10