Related Experiment Videos
Prospective evaluation of the Paediatric Risk of Mortality (PRISM) score
G Balakrishnan1, T Aitchison, D Hallworth
1Paediatric Intensive Care Unit, Royal Hospital for Sick Children, Glasgow.
Insights
The Paediatric Risk of Mortality (PRISM) score accurately predicts survival in pediatric intensive care but underpredicts deaths in post-cardiac surgery patients. A simplified 5-variable score shows similar performance.
Area of Science:
- Pediatric critical care medicine
- Clinical outcome prediction
- Healthcare quality and safety
Background:
- The Paediatric Risk of Mortality (PRISM) score is a widely used tool for predicting outcomes in pediatric intensive care units (PICUs).
- Prospective validation of the PRISM score is essential to ensure its continued accuracy and applicability across different patient populations and healthcare settings.
Purpose of the Study:
- To prospectively evaluate the performance of the admission day PRISM score in predicting mortality and survival in a pediatric intensive care unit.
- To assess the accuracy of the PRISM score across different patient subgroups, including operative versus non-operative cases and varying lengths of PICU stay.
- To explore the potential for a reduced variable set within the PRISM score for efficient and accurate outcome prediction.
Main Methods:
- Prospective study of 270 consecutive pediatric admissions (3 days to 18.6 years) to a PICU.
- Calculation of PRISM scores on admission and comparison with actual patient outcomes (survival or death).
- Analysis of score performance based on length of PICU stay, operative status, and a reduced set of five key variables.
Main Results:
- The PRISM score demonstrated high specificity (99%) for predicting survival but moderate sensitivity (48%) for predicting death.
- Prediction accuracy was highest for patients with shorter PICU stays (1-4 days).
- The score significantly underpredicted mortality in post-cardiac surgery patients (17% sensitivity) compared to non-operative patients (71% sensitivity).
- A simplified 5-variable PRISM score achieved similar sensitivity (41%) and specificity (99%) to the full 14-variable score.
- Six variable ranges showed different associations with non-survival compared to the original score, suggesting potential need for reweighting in the UK context.
Conclusions:
- The PRISM score's performance is largely institution-independent and effective for short-stay PICU patients.
- The PRISM score requires adjustment to improve its accuracy in predicting deaths among post-cardiac surgery patients.
- A reduced 5-variable set may be sufficient for reliable outcome prediction, potentially simplifying its application.
- Variable reweighting may be necessary for optimal PRISM score performance in UK PICUs.
Abstract:
The performance of the admission day Paediatric Risk of Mortality (PRISM) score for outcome prediction was assessed prospectively in 270 consecutive admissions, aged 3 days to 18.6 years, to a paediatric intensive care unit. Using a cut off of r = 0.00 (expected mortality = 50%), the overall sensitivity (correct prediction of death) was 48% while specificity (correct prediction of survival) was 99%, comparable with the original validation data of the score in the USA. Outcome prediction was most accurate when the stay in the paediatric intensive care unit was between one and four days. Sensitivity was appreciably lower for operative patients (17%) compared with non-operative patients (71%) because of a failure to predict deaths after cardiac surgery. The sensitivity (41%) and specificity (99%) using five variables (systolic blood pressure, Glasgow coma scale, carbon dioxide tension, and serum bicarbonate and serum calcium concentrations) was similar to that using all 14 variables. Six variable ranges related differently with non-survival compared with the score. It is concluded that the performance of the PRISM score is institution independent and good for short stay patients. It underpredicts deaths after cardiac surgery. Only five variables may be needed for satisfactory outcome prediction. Some of the variables need reweighting for paediatric intensive care units in the UK.