Prism IV : verificação de índice de mortalidade pediátrico em uma unidade de terapia intensiva pediátrica do sul do Brasil

Abstract

Aims and Objectives: To evaluate the predictive capacity of Pediatric Risk of Mortality Score IV (PRISM IV) in a Pediatric Intensive Care Unit (PICU) in Southern Brazil. Secondarily compare this research to two other scores, Pediatric Index of Mortality (PIM) 2 and PIM 3 Methods: Longitudinal retrospective study, during one year, from January 1 to December 31, 2016. Children admitted to a PICU at a University Hospital in Southern Brazil. Patients older than 20 years and dying patients with vital signs incompatible with life after 2 hours of PICU admission were not included. Only the first entry into PICU during hospitalization was considered. Discrimination, calibration and comparison tests with other indexes and a PRISMIV / PIM2 concordance test were performed. The study was approved by the institution's ethics committee Results: There were 411 admissions in the year 2016, 378 patients were included in the study. Of these 378, 13 died, generating a mortality rate of 3.44%. PRISM IV estimated mortality of 3.18% with Standardised Mortality Ratio (SMR) of 1.08, zflora= -0.3. For the same sample, the PIM2 predicted a mortality of 2.78% and the PIM 3 of 2.51. In the same way, PIM 2 and PIM 3 presented SMR of 1.24 and 1.37, respectively, with zflora=-0.91 and -1.40. The Hosmer-Lemeshow (HL) adjustment test obtained a X² = 4,472 (p = 0.484) for PRISM IV. Similarly, the PIM 2 presented a good calibration with X² = 8,359 and p = 0.138. However, PIM 3 presented X²=16.013 and p = 0.007. The discrimination test with the area under the curve (AUC / ROC – Receiver Operating Charateristic) of PRISM IV was 0.811 (95% CI 0.695-0.928). Similarly, the area of the PIM 2 was 0.779 (95% CI 0.645-0.913) and the PIM 3 obtained 0.759 (IC 95% 0.621-0.898). Among the three prognostic scores there was no statistical difference. The disagreement between the results of PRISM IV and PIM 2 was small. Conlusion: PRISM IV presented good predictive capacity in the study population, demonstrating good calibration and discrimination. In the comparative analysis, a similar predictive capacity was observed between this PRISM IV and PIM 2, which was not confirmed with PIM 3. It is suggested that PRISM IV be a validated tool for this population.