790 - Improving the Predictability of the Pediatric Resuscitation and Trauma Outcome (PRESTO) Model in Injury Patients in Tanzania

Publication Number: 790.5159

Joao Vitor Perez de Souza, Duke University School of Medicine, Durham, NC, United States; Elizabeth M. Keating, University of Utah, Salt Lake City, UT, United States; William Nkenguye, Kilimanjaro Christian Medical University College Kilimanjaro Clinical Research Institute, Moshi Municipal, Kilimanjaro, Tanzania; Happiness Duncan. Kajoka, Kilimanjaro Christian Medical University College, Moshi Urban, Kilimanjaro, Tanzania; Pollyana Coelho Pessoa Santos, University of North Carolina at Chapel Hill, Durham, NC, United States; Catherine Staton, Duke University School of Medicine, Durham, NC, United States; Blandina Theophil. Mmbaga, Kilimanjaro Clinical Research Institute, Moshi, Kilimanjaro, Tanzania; Francis M. Sakita, Kilimanjaro Christian Medical Centre, Moshi, Kilimanjaro, Tanzania; Joao Ricardo Nickenig Vissoci, Duke University School of Medicine, Durham, NC, United States

Background: Injuries are responsible for 950,000 pediatric deaths per year. The use of trauma prediction scores is helpful in determining the severity and prognosis of injury patients. The pediatric resuscitation and trauma outcome (PRESTO) score was developed as a simple score of mortality prediction in low and middle-income countries (LMICs). Using variables available at the bedside in resource-limited settings, PRESTO has been validated in three LMICs. In Tanzania, our team found that these variables were present for the majority of pediatric injury patients and the model performed well in predicting mortality, though limited by small sample size.

Objective: To improve the predictability of the PRESTO score for pediatric patients by increasing our sample size and including adult trauma patients at Kilimanjaro Christian Medical Centre (KCMC) in Tanzania.

Design/Methods: Data was collected November 2020-February 2024 from a pediatric injury registry, and April 2018-February 2024 from an adult injury registry. Missing data was addressed using multiple imputation with chained equations. The data was split into training and testing sets (75/25 ratio). Ten different machine learning algorithms were trained to predict in-hospital mortality. Clinical and demographic predictor variables collected in the Emergency Department were used as predictors, with adjustments for age-vital sign interactions. Model performance was evaluated through 10-fold cross-validation. Hyperparameters were optimized using grid search, and performance was assessed with the ROC-AUC, sensitivity and specificity.

Results: A total of 5,635 (911 pediatric and 4,724 adult) injury patients were included. Pediatric mortality was 6.8% while adult mortality was 4.4%. The three best-performing models were the Random Forest (RF), C5 Ruleset (C5), and Extreme Gradient Boosting (XGB) with ROC-AUCs of 0.92, 0.90, and 0.81. In the test set, a consistent pattern across the models was observed where the ROC-AUC for pediatric patients was worse than for adult patients. In detail, a difference of 0.08, 0.06, and 0.05 in favor of the adult patients was noted for the XGB, RF, and C5 models.

Conclusion(s): Adding adult data did not improve the predictability of the PRESTO score, supporting the idea that children need to be considered separately. These findings emphasize the importance of increasing the sample size of our pediatric injury registry to develop more accurate models for stratifying mortality risk. Such efforts will ensure that models are tailored to the unique needs of pediatric patients, particularly in LMICs, improving predictions and resource allocation.