What is the PRISM‑IV (Pediatric Risk of Mortality) score, how is it calculated, and how are its results interpreted?

PRISM IV Score Overview

The PRISM IV (Pediatric Risk of Mortality IV) score is a physiologically-based severity of illness scoring system that uses 17 physiologic variables collected during the first 2-4 hours of PICU admission to predict hospital mortality risk in critically ill children. 1

What PRISM IV Measures

PRISM IV quantifies physiologic derangement and, when combined with other independent variables, calculates expected mortality and morbidity risk for pediatric intensive care patients 1. The score serves as a critical tool for:

• Comparing quality of care between different PICUs
• Monitoring effects of practice changes within units over time
• Assessing relationships between illness severity, length-of-stay, and costs
• Evaluating the impact of resource allocation decisions 2

How PRISM IV is Calculated

Data Collection Period

• Physiologic data: Collected within 4 hours of PICU admission
• Laboratory data: Collected from 2 hours before to 4 hours after admission
• Key principle: Use the worst value for each variable during this window 1

Variables Included

The score incorporates 17 physiologic variables across multiple organ systems, with specific abnormal ranges for each. These include:

Neurologic/Mental Status:

• Pupillary reflexes
• Mental status assessment

Cardiovascular:

• Systolic blood pressure (age-adjusted ranges)
• Heart rate

Respiratory:

• PaO2/FiO2 ratio
• PaCO2
• Mechanical ventilation status

Metabolic/Laboratory:

• pH
• Total CO2
• Base excess
• Blood glucose
• Potassium
• Creatinine
• Blood urea nitrogen
• White blood cell count
• Platelet count
• Prothrombin time/INR

1

Additional Risk Factors

Beyond physiologic variables, PRISM IV incorporates:

• Age (neonates vs. older children)
• Admission source (emergency department, ward, operating room, etc.)
• Cardiopulmonary arrest within 24 hours before admission
• Cancer diagnosis
• Low-risk systems of primary dysfunction
• Neurologic vs. non-neurologic primary diagnosis 1

Calculation Method

The score uses logistic regression modeling with the collected variables to generate a probability of hospital mortality. Online calculators are available to compute both the raw PRISM IV score and the predicted mortality percentage 3.

Interpretation of Results

Performance Characteristics

In the 2015 validation study of 10,078 PICU admissions:

• Area under ROC curve: 0.88-0.90 (excellent discrimination)
• Overall mortality rate: 2.7% (range 1.3-5.0% across sites)
• Calibration: Good model fit demonstrated by Hosmer-Lemeshow testing (p=0.39-0.50) 1

Clinical Application

The PRISM IV score provides a probability of hospital death, NOT a definitive prediction for individual patients. Higher scores indicate greater physiologic derangement and higher mortality risk. The score is most useful for:

• Population-level comparisons: Comparing observed vs. expected mortality across units or time periods
• Quality improvement: Calculating standardized mortality ratios (SMR = observed deaths/expected deaths)
• Research stratification: Adjusting for illness severity in clinical studies

Important Limitations and Caveats

Age Restrictions:

• PRISM does not accurately predict mortality in infants less than 1 month of age 2
• For neonates, use CRIB II (Clinical Risk Index for Babies) instead 2

Special Populations:

• Cancer patients: Standard PRISM IV may underpredict mortality (SMR 1.8 in one study). A modified algorithm (PRISM 4-C) has been developed specifically for pediatric oncology patients with improved calibration 4
• The modified cancer-specific algorithm achieved an SMR of 1.02 vs. 1.8 for standard PRISM IV 4

Timing Considerations:

• PRISM IV is calculated from first 2-4 hours only of PICU admission 1
• Scores calculated at 24 or 72 hours may have different performance characteristics and showed poor predictive ability in at least one study 5
• For patients admitted for "optimizing" care before cardiac surgery or interventional catheterization, the data collection period has been specifically altered 1

Data Quality Requirements:

• Accurate data collection is critically important for valid predictions 2
• Resources must be allocated for trained personnel to collect and verify all variables 2
• Missing or incorrectly recorded values will compromise score accuracy

Common Pitfalls to Avoid

1. Do not use PRISM IV for individual prognostication - It predicts population-level risk, not individual patient outcomes
2. Do not apply to neonates under 1 month - Use CRIB II instead
3. Do not recalculate at 24 or 72 hours expecting the same predictive validity - The score is designed for admission data only
4. Do not ignore special populations - Consider modified algorithms for cancer patients
5. Do not use without proper training - Ensure staff collecting data understand variable definitions and timing requirements
6. Do not compare units without risk adjustment - Raw mortality rates are meaningless without accounting for case mix severity

Practical Use Algorithm

For standard PICU admissions (age >1 month, non-oncology):

• Collect worst physiologic values in first 4 hours
• Collect worst lab values from 2 hours pre- to 4 hours post-admission
• Calculate PRISM IV score using validated calculator
• Interpret as probability of hospital mortality
• Use for quality monitoring and benchmarking, not individual prognosis

For neonates (<1 month):

• Use CRIB II score instead 2

For pediatric oncology patients:

• Consider using PRISM 4-C (modified algorithm) for better calibration 4

The PRISM IV represents a significant improvement over earlier versions with enhanced data collection methods that minimize bias and errors, and the algorithm is now in the public domain for widespread use 1.