Development and temporal-validation of prognostic models for 5-year risk of pneumonia, respiratory failure/collapse, and fracture among adults with cerebral palsy.

Abstract

Purpose: Pneumonia, respiratory failure, and fracture are common and highly burdensome for adults with cerebral palsy (CP). To date, there are no clinically friendly tools to assess individualized risk of these outcomes for adults with CP. The objective was to develop and validate prognostic models for 5-year risk of pneumonia, respiratory failure/collapse, and fracture for adults with CP.

Patients and methods: This single medical center retrospective cohort study used medical records from January 1, 2012 till June 1, 2024 from adults ≥18 years old with CP. The development cohort (n=1,520) included those with a start date of follow-up from January 1, 2015 till December 31, 2015 and evidence of being treated at the medical center for ≥3 years prior to day 0. The 5-year risk of outcomes was modelled using logistic regression and variations of the following predictors that were collected during the 3-year baseline: age, sex, Whitney Comorbidity Index, pneumonia, respiratory failure/collapse, fracture, and asthma/COPD. Discrimination (c-statistic) and calibration statistics assessed the model's performance. Decision curve analysis assessed the model's clinical utility. The models were validated in a temporal validation cohort, whose start date of follow-up was January 1, 2016 to May 31, 2019 (n=529).

Results: The prognostic models had good discrimination (c-statistic=0.76-0.78), good-to-excellent calibration, and demonstrated superior clinical utility in identifying true positives and true negatives. All models demonstrated temporal validation.

Conclusions: Prognostic models for 5-year risk of outcomes were developed and temporally validated for adults with CP using measures that can be easily collected from medical records.