Construction of a Pediatrics Risk Score to Predict High Health Care Costs Among a Community Health Center Cohort

Area of Science:

• Pediatric Health Services Research
• Health Economics
• Clinical Informatics

Background:

• Effective risk-stratification is crucial for managing ambulatory pediatric populations.
• Identifying high-cost patients early can inform resource allocation and intervention strategies.
• Existing risk prediction models may not adequately capture the unique needs of diverse pediatric age groups.

Purpose of the Study:

• To develop and validate age-specific risk scores for predicting high healthcare costs in an urban pediatric population.
• To identify key predictors of high healthcare expenditure accessible within electronic health records.
• To create actionable tools for risk stratification in ambulatory pediatric care.

Main Methods:

• Retrospective cohort study of 8960 children (ages 1-18) at a community health center.
• Multivariable logistic regression and random forest modeling using pre-2017 electronic health record (EHR) data to predict 2017 costs.
• Development of three age-specific nomograms (1-5, 6-11, 12-18 years) to predict costs, validated with 2018 data.

Main Results:

• Prior healthcare utilization, specialty care in younger children, and behavioral health diagnoses in older children were significant predictors of high costs.
• Final age-specific models demonstrated good predictive performance (C-statistics ≥0.68) for both 2017 and 2018 cost data.
• The models successfully utilized readily available EHR data for risk prediction.

Conclusions:

• Age-specific prediction models using EHR data can effectively identify ambulatory pediatric patients at high risk for costly healthcare utilization.
• These models offer a feasible approach to risk stratification in pediatric primary care settings.
• Further strategies are needed to address healthcare utilization patterns among identified high-risk pediatric groups.