Predicting the Risk of Mortality in Children using a Fuzzy-Probabilistic Hybrid Model

Area of Science:

• Pediatric Intensive Care
• Medical Informatics
• Fuzzy Logic Systems

Background:

• Pediatric Intensive Care Unit (PICU) mortality risk is typically assessed using objective data scales.
• Subjective human perception may offer valuable prognostic information for mortality risk.
• Fuzzy logic provides a framework for incorporating subjective data into predictive models.

Purpose of the Study:

• To develop a mathematical model predicting PICU mortality risk based on staff's subjective perceptions.
• To evaluate the accuracy of this subjective model against established validated scales.
• To explore the potential of combining subjective and objective data for improved prognostication.

Main Methods:

• A prospective observational study was conducted in two PICUs over two years.
• A fuzzy set program was used for staff to record subjective mortality risk assessments (short and long range).
• Pediatric Index of Mortality 2 (PIM2) and TISS28 were calculated; logistic regression and cross-validation were used for analysis.

Main Results:

• 599 children were included; 24 non-survivors in total.
• The best subjective model (MID - midpoint of short range) achieved 92.6% accuracy.
• The PIM2 model achieved 86.3% accuracy; a hybrid model combining MID and PIM2 reached 96.4% accuracy.

Conclusions:

• Subjective staff assessment is a valuable tool for estimating mortality risk in PICU, comparable to validated scales.
• A hybrid model integrating fuzzy logic-based subjective data with objective scales like PIM2 significantly improves mortality prediction accuracy.
• This approach offers a promising avenue for enhancing prognostic capabilities in pediatric intensive care.