An exploratory routine-laboratory renal dysfunction-coagulopathy phenotype identifies high-risk ICU patients with pneumonia: MIMIC- IV derivation and fixed-centroid Chinese external validation
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Critically ill adults with pneumonia are clinically heterogeneous. Conventional severity scores summarize acute organ dysfunction but may not define clinically interpretable patient states. We investigated whether routine laboratory tests obtained early after ICU admission could identify reproducible phenotypes among ICU patients with pneumonia and whether these phenotypes could be externally validated in a Chinese ICU cohort. Methods We conducted a retrospective cohort study using MIMIC-IV as the derivation cohort and a single-center Chinese ICU cohort for fixed-centroid external assignment. Adult patients with pneumonia-related diagnoses, first ICU stay, and ICU length of stay of at least 24 hours were included in MIMIC-IV. The external cohort included adults with severe pneumonia admitted to the ICU of Longhua Hospital between January 1, 2024 and March 31, 2025. Eleven routine laboratory measures obtained during the first ICU day represented hematologic status, coagulation, renal function and electrolyte balance. K-means solutions from k = 2 to k = 6 were evaluated. The exploratory k = 3 solution used 20 random initializations, and repeated multistart analyses assessed sensitivity to initialization. Local patients were assigned without re-clustering by applying MIMIC-derived imputation, standardization and nearest-centroid rules. The primary outcome was in-hospital mortality. Secondary outcomes included CRRT/RRT and ICU length of stay. Results The MIMIC-IV derivation cohort included 16,933 ICU stays, with 3,687 in-hospital deaths. The external cohort included 412 patients, with 120 deaths. A three-phenotype exploratory solution was selected. Phenotype 1 showed a relatively preserved hematologic-renal profile and the lowest risk. Phenotype 2 showed cytopenic and electrolyte-shift features with intermediate risk. Phenotype 3 was characterized by renal dysfunction, coagulation abnormality and electrolyte derangement. In MIMIC-IV, mortality increased from 16.9% in phenotype 1 to 22.6% in phenotype 2 and 32.2% in phenotype 3. CRRT use increased from 2.2% to 5.2% and 22.5%, respectively. In the Chinese cohort, phenotype 3 also showed the highest mortality (50.0%) and CRRT use (63.9%). Compared with phenotype 1, phenotype 3 remained associated with mortality after adjustment for age, sex and SOFA score (OR 1.193, 95% CI 1.068–1.332). Bootstrap analysis identified phenotype 3 as the highest-mortality and highest-CRRT cluster in 99% of iterations. In 30 repeated analyses using 20 random initializations per analysis, all k = 3 solutions had an adjusted Rand index above 0.99 relative to the reference solution. Conclusions Early routine laboratory tests identified an exploratory renal dysfunction-coagulopathy high-risk phenotype among ICU patients with pneumonia. This phenotype was associated with higher mortality and markedly higher CRRT use in MIMIC-IV and showed directionally consistent fixed-centroid assignment in a Chinese ICU cohort. These results support further multicentre and prospective evaluation of routine-laboratory phenotyping; they do not establish phenotype-guided treatment.
Source: Research Square
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7204178164