Establishment and Validation of a Machine-Learning Prediction Nomogram Based on Lymphocyte Subtyping for Intra-Abdominal Candidiasis in Septic Patients.

Summary

In 633 septic patients with intra-abdominal infection, the authors built and validated a prediction nomogram for intra-abdominal candidiasis using lymphocyte subtyping plus clinical variables. Machine-learning (random forest) guided variable selection; multivariable logistic regression underpinned the nomogram. High-dose corticosteroid exposure and CD4+ T-cell parameters emerged as important predictors; the model demonstrated good discrimination, calibration, and clinical utility.

Key Findings

• Prospective cohort of 633 septic patients with intra-abdominal infection enabled model development and validation.
• Random forest identified key immune and clinical predictors; multivariable logistic regression constructed the nomogram.
• High-dose corticosteroid exposure and CD4+ T-cell metrics were important predictors; the model showed good discrimination, calibration, and clinical usefulness.

Clinical Implications

Integration of lymphocyte subtyping into bedside risk models may refine antifungal stewardship, reduce delays in therapy, and prioritize diagnostics (cultures, imaging) for suspected IAC.

Limitations

• Abstract does not report external validation or specific performance metrics (e.g., AUC), limiting assessment of generalizability.
• Single-region cohort; performance in other settings and pathogen spectra remains to be tested.

Future Directions

External, multicenter validation; incorporation into EHR for real-time decision support; impact studies on antifungal timing, diagnostic yield, and outcomes.