Development and External Validation of a Novel In-hospital Mortality Model after Pediatric Congenital Heart Surgery: A Multicenter Retrospective Cohort Study.

Summary

In 21,855 development and 5,221 external validation cases, legacy pediatric cardiac surgery mortality scores performed only moderately. A new six-variable model (age, height, oxygen support, prior cardiac operation, emergency surgery, STAT categories) achieved AUROCs of 0.864 (development) and 0.860 (external), with low Brier scores, indicating strong calibration and discrimination.

Key Findings

• Traditional scores (ABC, RACHS-1, STAT score/categories) showed only moderate AUROCs (0.685–0.808).
• Adding preoperative covariates improved AUROCs across models to 0.844–0.864.
• A new six-variable model achieved AUROC 0.864 (development) and 0.860 (external validation) with low Brier scores (0.00977 and 0.00654).

Clinical Implications

Integrating the six-variable model into preoperative assessment can improve resource allocation, ICU planning, and informed consent discussions in pediatric congenital heart surgery.

Limitations

• Retrospective design may introduce residual confounding and center-specific practice patterns.
• Model derived and validated within China; generalizability to other healthcare systems requires testing.

Future Directions

Prospective, multicountry validation and integration into clinical decision support; assessment of model impact on outcomes via implementation studies.