Clinical Subphenotypes and Short-Term Outcomes in PICU Patients Receiving Continuous Blood Purification
Kaizong Huang, Jiahua Ke, Zhiyuan Wang, Xuhua Ge, Zhe Zhao, Feiyan Chen, Lei Zhang, Jingwen Ni, Jing Zhao, Senquan Liang, Shengxin Zhang, Qiang Qu, Hui Sun, Puyi Wang, Nan Gong, Hongxia Shen, Jiangwei Zhu, Hongwei Fan, Shujun LiOBJECTIVES:
To characterize clinical heterogeneity among PICU patients receiving continuous blood purification (CBP) and identify data-driven subphenotypes associated with distinct short-term outcomes.
DESIGN:
Multicenter, retrospective cohort study.
SETTING:
Fifteen PICUs (from January 2019 to August 2024).
PATIENTS:
Children (1 mo to 18 yr) receiving CBP during PICU hospitalization.
INTERVENTIONS:
None.
MEASUREMENTS AND MAIN RESULTS:
In this multicenter cohort of 598 CBP-treated children, unsupervised clustering at CBP initiation identified three reproducible subphenotypes using a seven-variable signature comprising aspartate aminotransferase, hemoglobin, oxygenation index, Pediatric Logistic Organ Dysfunction-2 score, procalcitonin, weight, and pH. These subphenotypes showed distinct multidomain physiologic profiles rather than a simple severity gradient. Subphenotype I had the most favorable short-term survival profile, subphenotype II had the poorest survival, and subphenotype III showed an intermediate pattern, with significant separation in 28-day survival across groups. Subphenotypes also differed in ICU length of stay, and exploratory analyses suggested subphenotype-dependent associations between high-volume hemofiltration exposure and ICU discharge. Cross-validated supervised models showed high discriminability of subphenotype assignment.
CONCLUSIONS:
CBP-treated children exhibited distinct subphenotypes at CBP initiation with divergent short-term prognoses, highlighting marked heterogeneity at this treatment stage. Subphenotype-based stratification may support risk characterization and trial enrichment and facilitate hypothesis generation regarding heterogeneity of treatment associations. Prospective validation is warranted.