Modifying the PELOD-2 score to predict mortality in critically ill patients

Abstract

Background The PELOD-2 score, which has been widely used to predict multiple organ dysfunction, may be used to predict mortality. Nevertheless, blood gas analyses (BGA) and lactate measurements required for the PELOD-2 cannot be performed in most limited resource settings.
 Objective To evaluate the performance of modified PELOD-2, without BGA and lactate, to predict mortality in critically ill children. 
 Methods This retrospective cohort study in critically ill children admitted to the pediatric intensive care unit (PICU), Dr. Sardjito Hospital, Yogyakarta, was undertaken from January to December 2018. The modifications to the PELOD-2 score were PELOD-2A (without BGA), PELOD-2B (without lactate), and PELOD-2C (without BGA and lactate). The modified PELOD-2 scores were evaluated using receiver operating characteristic (ROC) curve for discrimination, and Hosmer-Lemeshow goodness-of-fit test for calibration.
 Results Of 130 subjects, 68 (52.3%) died. A PELOD-2 score cut-off of 6.5 and modified PELOD-2A, 2B, and 2C had sensitivities for predicting mortality of 73.5%, 67.7%, 70.6%, and 63.2%, respectively, and specificities of 75.8%, 77.4%, 77.4%, and 79%, respectively. The area under curve (AUC) of the PELOD-2 score was 78.3 (95%CI 70.5 to 86.2). The AUCs of the modified PELOD-2 scores ranged from 76.8 (95%CI 68.7 to 84.9) to 77.9 (95% CI 69.9 to 85.8). The positive predictive values of PELOD-2 and modified PELOD-2A, 2B, 2C were 76.9%, 76.7%, 77.4% and 76.8%, respectively. The Hosmer-Lemeshow goodness-of-fit test showed good calibration for PELOD-2 (x2=8.74; P=0.27) and modified PELOD-2A (x2=4.91; P=0.67).
 Conclusion The PELOD-2A, modified without BGA, can still predict mortality well in critically ill PICU patients when using a cut-off score ≥ 6.5.