To evaluate the performance of the Pediatric Index of Mortality 3 as mortality risk assessment model. DESIGN:This prospective study included all admissions 30 days to 18 years old for 12 months during 2016 and 2017. Data gathered included the following: age and gender, diagnosis and reason for PICU admission, data specific for the Pediatric Index of Mortality 3 calculation, PICU outcomes (death or survival), and length of PICU stay.SETTING: Nine units that care for children within tertiary or quaternary academic hospitals in South Africa. PATIENTS:All admissions 30 days to 18 years old, excluding premature infants, children who died within 2 hours of admission, or children transferred to other PICUs, and those older than 18 years old. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS:There were 3,681 admissions of which 2,253 (61.3%) were male. The median age was 18 months (interquartile range, 6-59.5 mo). There were 354 deaths (9.6%). The Pediatric Index of Mortality 3 predicted 277.47 deaths (7.5%). The overall standardized mortality ratio was 1.28. The area under the receiver operating characteristic curve was 0.81 (95% CI 0.79-0.83). The Hosmer-Lemeshow goodness-of-fit test statistic was 174.4 (p < 0.001). Standardized mortality ratio for all age groups was greater than 1. Standardized mortality ratio for diagnostic subgroups was mostly greater than 1 except for those whose reason for PICU admission was classified as accident, toxin and envenomation, and metabolic which had an standardized mortality ratio less than 1. There were similar proportions of respiratory patients, but significantly greater proportions of neurologic and cardiac (including postoperative) patients in the Pediatric Index of Mortality 3 derivation cohort than the South African cohort. In contrast, the South African cohort contained a significantly greater proportion of miscellaneous (including injury/accident victims) and postoperative noncardiac patients. CONCLUSIONS:The Pediatric Index of Mortality 3 discrimination between death and survival among South African units was good. Case-mix differences between these units and the Pediatric Index of Mortality 3 derivation cohort may partly explain the poor calibration. We need to recalibrate Pediatric Index of Mortality 3 to the local setting.
Background. Paediatric Index of Mortality (PIM) and PIM 2 scores have been shown to be valid predictors of outcome among paediatric intensive care unit populations in the UK, New Zealand, Australia and Europe, but have never been evaluated in the South African context. Objective. To evaluate the PIM and PIM 2 as mortality risk assessment models. Method. A retrospective audit of case records and prospectively collected patient data from all admissions to the Paediatric Intensive Care Unit (PICU) of Red Cross War Memorial Children's Hospital, Cape Town, during the years 2000 (PIM) and 2006 (PIM 2), excluding premature infants, children who died within 2 hours of admission, or children transferred to other PICUs. Results. For PIM and PIM 2 there were 128/962 (13.3%) and 123/1113 (11.05%) PICU deaths with expected mean mortality rates of 12.14% and 12.39%, yielding standardised mortality risk ratios (SMRs) of 1.1 (95% confidence interval (CI) 0.93-1.34) and 0.9 (95% CI 0.74-1.06), respectively. Receiver operating characteristic analysis revealed area under the curve of 0.849 (PIM) and 0.841 (PIM 2). Hosmer-Lemeshow goodness of fit revealed poor calibration for PIM (χ 2 =19.74; p=0.02) and acceptable calibration for PIM 2 (χ 2 =10.06; p=0.35). SMR for age and diagnostic subgroups for both scores fell within wide confidence intervals. Conclusion. Both scores showed good overall discrimination. PIM showed poor calibration. For PIM 2 both discrimination and calibration were comparable to the score derivation units, at the time of data collection for each. Calibration in terms of age and diagnostic categories was not validated by this study.
ObjectivesThe performance of mortality prediction models remain a challenge in lower- and middle-income countries. We developed an artificial neural network (ANN) model for the prediction of mortality in two tertiary pediatric intensive care units (PICUs) in South Africa using free to download and use software and commercially available computers. These models were compared to a logistic regression model and a recalibrated version of the Pediatric Index of Mortality 3.DesignThis study used data from a retrospective cohort study to develop an artificial neural model and logistic regression model for mortality prediction. The outcome evaluated was death in PICU.SettingTwo tertiary PICUs in South Africa.Patients2,089 patients up to the age of 13 completed years were included in the study.InterventionsNone.Measurements and Main ResultsThe AUROC was higher for the ANN (0.89) than for the logistic regression model (LR) (0.87) and the recalibrated PIM3 model (0.86). The precision recall curve however favors the ANN over logistic regression and recalibrated PIM3 (AUPRC = 0.6 vs. 0.53 and 0.58, respectively. The slope of the calibration curve was 1.12 for the ANN model (intercept 0.01), 1.09 for the logistic regression model (intercept 0.05) and 1.02 (intercept 0.01) for the recalibrated version of PIM3. The calibration curve was however closer to the diagonal for the ANN model.ConclusionsArtificial neural network models are a feasible method for mortality prediction in lower- and middle-income countries but significant challenges exist. There is a need to conduct research directed toward the acquisition of large, complex data sets, the integration of documented clinical care into clinical research and the promotion of the development of electronic health record systems in lower and middle income settings.
Mortality prediction models, which this author prefers to call outcome-risk assignment models, are ubiquitous in critical care practice. They are designed to assess the risk of dying for any given patient by assigning risk based on indices of physiological derangement, high-risk diagnoses or diagnostic categories, or the need for therapeutic intervention to support organ function. It is important to recognise that predictive models derived to date are not accurate enough to allow reliable prediction of individual patient outcome, and therefore cannot be used as criteria for admission to intensive care. Patients identified as high risk for mortality must be flagged for appropriate intensive intervention and close monitoring in order to reduce the risk of dying due to the current illness.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.