Key result
The PRISM III and PIM III scores demonstrated adequate discrimination for predicting mortality (AUC-ROC 0.81 and 0.80, respectively) but poor calibration, significantly underestimating observed mortality.
Why the study?
The performance of PRISM III and PIM III indices and the factors affecting their performance had not been evaluated in a tertiary PICU in Saudi Arabia.
Do PRISM III and PIM III scores accurately predict mortality in pediatric intensive care unit patients in Saudi Arabia?
Observational (n=3,396)
No
Do PRISM III and PIM III scores accurately predict mortality in pediatric intensive care unit patients in Saudi Arabia?
PRISM III and PIM III scores demonstrated sufficient discrimination but poor calibration in a Saudi Arabian PICU, underestimating observed mortality and highlighting the need for local validation or development of new models.
Answers are grounded in this paper and cite their sources.
PRISM III shows better discrimination than PIM III in Saudi PICU; leaves open generalizability pending multicenter validation.
Objective: To assess the performance of the Pediatric Risk of Mortality III (PRISM III) and Pediatric Index of Mortality III (PIM III) indices in a tertiary pediatric intensive care unit (PICU) in Saudi Arabia and to identify the factors affecting the observed performance. Design: Retrospective, single-center study using data collected from the Virtual Pediatric Systems web-based database. Setting: King Fahad Medical City PICU, Saudi Arabia. Patients: All pediatric patients <14 years of age admitted between 1 January 2015, and 31 December 2019. Interventions: Comparison of PRISM III and PIM III performances in predicting mortality across different age groups, disease categories, and resuscitation decision statuses. Measurements: Normality of distribution was assessed using the Kolmogorov-Smirnov and Shapiro-Wilk tests. Patient characteristics were compared between survivors and non-survivors. The medians and ranges were calculated for continuous data, whereas frequencies and percentages were used for nominal data. The Mann-Whitney U test, Kruskal-Wallis test, and Chi-square test were used to compare the characteristics of survivors and non-survivors. Main Results: There was a significant difference between the predicted mortality and observed mortality in both the PRISM III and PIM III. Better discrimination was found after excluding do-not-resuscitate (DNR) patients. The worst calibration and discrimination were recorded for infants <12 months of age. The PRISM III performed significantly better in patients with metabolic/genetic and central nervous system illnesses. Non-DNR patients had a lower standardized mortality rate using the PRISM III and PIM III. The PRISM III and PIM III indices performed better in patients who died within the first week of admission. Conclusion: These models had sufficient discrimination ability and poor calibration. Since they were designed for particular patient characteristics and PICUs, further testing in different environments is necessary before utilization for planning and assessing performance. Alternatively, new models could be developed which are suitable for local PICUs.
No takes yet. Share an insight, caveat, or question.
Alkhalifah et al. (2022) conducted an observational in Critically ill pediatric patients (n=3,396). PRISM III and PIM III scores vs. Observed mortality was evaluated on Discrimination for mortality prediction (AUC-ROC) (AUC 0.81, 95% CI 0.79-0.84, p=<0.001). The PRISM III and PIM III scores demonstrated adequate discrimination for predicting mortality (AUC-ROC 0.81 and 0.80, respectively) but poor calibration, significantly underestimating observed mortality.
Synapse has enriched 10 related papers. Consider them for comparative context: