What are the current validated pediatric intensive care unit (PICU) severity‑of‑illness scores, such as PRISM‑IV, PIM‑3, and PELOD‑2?

Current Validated PICU Severity-of-Illness Scores

The currently validated pediatric intensive care unit severity-of-illness scores are PRISM-IV (Pediatric Risk of Mortality, version 4), PIM-3 (Pediatric Index of Mortality, version 3), and PELOD-2 (Pediatric Logistic Organ Dysfunction, version 2), with pSOFA (pediatric Sequential Organ Failure Assessment) emerging as an additional validated tool.

Overview of Validated Scoring Systems

PRISM-IV (Pediatric Risk of Mortality, Version 4)

• Calculates mortality risk using physiological variables collected over the first 24 hours of PICU admission 1
• Uses the worst values from 72 variables during the initial 24-hour period
• Recent validation studies demonstrate excellent discrimination with AUROC of 0.984 2 and good calibration in pediatric oncology populations 3
• Shows superior predictive accuracy compared to other scores in multiple recent studies 4, 2
• Requires accurate data collection by trained personnel 1

PIM-3 (Pediatric Index of Mortality, Version 3)

• Based on data collected at admission, requiring only 10 variables for risk calculation 1
• Data collected from 2 hours prior to 4 hours after admission 3
• Free accessibility and ease of use make it practical for resource-limited settings 5
• Recent studies show acceptable discrimination (AUROC 0.78-0.836) 3, 2 but variable calibration across different populations 5
• Does not correctly predict mortality in infants less than 1 month of age 1

PELOD-2 (Pediatric Logistic Organ Dysfunction, Version 2)

• Assesses multiorgan dysfunction rather than predicting mortality directly 4
• Evaluates six organ systems: respiratory, cardiovascular, neurologic, renal, hepatic, and hematologic 6
• Recent evidence demonstrates better discrimination power (AUROC 0.74-0.872) compared to earlier versions 4, 2
• Particularly useful for tracking disease progression during PICU stay 4
• Calculated at 24 hours after admission 2

pSOFA (Pediatric Sequential Organ Failure Assessment)

• Emerging as a reliable predictor with the best discrimination power in recent studies (AUROC 0.77) 4
• Allows continuous assessment throughout the PICU stay, tracking time course of patient condition 7
• Evaluates the same six organ systems as adult SOFA but adapted for pediatric populations
• Enables physicians to follow evolving disease processes in critically ill children 7

Key Clinical Considerations

Limitations Across All Scores

• None of these scores accurately predict mortality in neonates less than 1 month of age 1
• For newborns, use CRIB-II (Clinical Risk Index for Babies, updated 2003) instead 1
• Performance varies by geographic region and case mix—scores validated in developed countries may show different calibration in resource-limited settings 4, 5

Practical Application

• PRISM-IV requires 24-hour data collection, making it less useful for immediate triage decisions 1
• PIM-3 provides earlier risk stratification since it uses admission data only 1
• pSOFA and PELOD-2 are descriptive scores best used for tracking organ dysfunction rather than initial mortality prediction 4

Data Collection Requirements

• Accurate data collection is critically important for all scoring systems 1
• Sufficient resources must be available so that all information is collected and checked by a small number of trained, careful people 1
• Laboratory data should use the worst values within the specified time windows 3

Comparative Performance

Recent prospective studies comparing these scores demonstrate:

• PRISM-IV shows the best overall discrimination and calibration 2
• pSOFA and PELOD-2 are more reliable for predicting mortality than PIM-3 in some populations 4
• PIM-3 shows good discrimination but poor calibration in certain settings, particularly in India 5
• All three major scores (PRISM-IV, PIM-3, PELOD-2) show acceptable discriminative ability in pediatric oncology populations, but only PRISM-IV demonstrates good calibration 3

Clinical Utility Beyond Mortality Prediction

These scoring systems enable:

• Investigation of optimal ways to organize intensive care by comparing different units 1
• Monitoring effects of changes in practice by observing trends within units over time 1
• Assessment of relationships between severity-of-illness, length-of-stay, and cost 1
• Quality monitoring and comparative audit between institutions 1
• Stratification in clinical trials 7

Important Caveats

• The number of reported incidents does not correlate with true mortality—these scores are needed for quality evaluation 1
• Scores should not be used in isolation for individual patient management decisions 7
• Regional validation is essential before widespread implementation in new populations 4, 5
• Privacy concerns exist with third-party online calculators; open-source toolkits are now available for secure implementation 8