Validation Of The Pediatric Resuscitation And Trauma (presto) Outcome Model In A Middle-income Country
Michael D Traynor, Jr.1, Etienne St. Louis2, Ahmed S Alsayed1, Matthew C Hernandez1, *Denise B Klinkner1, Robert Baird2, Dan Poenaru2, Victor Y Kong3, *Chirstopher R Moir1, Martin D Zielinski1, Damian L Clarke3
1Mayo Clinic, Rochester, MN;2McGill University Health Centre, Montreal, QC, Canada3University of KwaZulu Natal, Durban, South Africa
Background: The pediatric resuscitation and trauma (PRESTO) outcome model was developed for risk-adjusted comparisons of post-injury mortality in low- and middle-income countries (LMICs). While PRESTO is tailored to LMICs, it has not been validated outside of high-income countries (HICs). We sought to validate PRESTO using data from an MIC trauma registry.
Methods: We included children (age<15) admitted to a trauma center in South Africa from December 2012-January 2019. We excluded patients missing variables necessary for PRESTO calculation. Primary outcome was in-hospital death. ROC analyses determined the optimal decision threshold in the MIC setting. The previously HIC validated optimal threshold was compared to the MIC validated threshold using area under the ROC curves (AUROC).
Findings: Of 1,160 injured children, 987 (85%) had complete data for calculation of PRESTO. Median age was 7 (IQR: 4, 11) and 67% were male. Mortality was 2% (n=23). Mean predicted mortality was 0.5% (range: 0-25.7%, AUROC 0.93). The HIC validated threshold had a sensitivity of 26.1% and specificity of 99.7% while the MIC threshold showed a sensitivity of 82.6% and specificity of 89.4%. The MIC threshold yielded superior discrimination (AUROC 0.86 [CI: 0.78,0.94]) compared to the previously established HIC threshold (0.63 [CI: 0.54, 0.72], p<0.0001), Figure 1.
Conclusions: PRESTO can be applied in MIC settings and discriminates between children at risk for in-hospital death following trauma. Further research should help clarify optimal decision thresholds for quality improvement and benchmarking in LMIC settings.
Back to 2019 Abstracts