Latent Classes in Long-Term Functional Status in Children With Recurrent Critical Illness

To identify and describe trajectories of functional status in children with repeated PICU admissions and to describe clinical and demographic characteristics of each trajectory.

Retrospective, cross-sectional study.

Twenty-four U.S. PICUs participating in the Virtual Pediatric Systems, LLC database.

Patients younger than 21 years old discharged from an index admission between April 2017 and December 2020, followed through December 2022.

None.

Functional status, a multidimensional marker of a child's physiologic and cognitive status, was measured using the Functional Status Scale (FSS). FSS scores were collected by trained abstractors to reflect pre-illness baseline and status at PICU discharge for each encounter. We performed latent trajectory analysis to examine longitudinal patterns in patient FSS categories and identify trajectory classes, with classes selected by the minimum Bayesian Information Criterion. We performed Kruskal-Wallis rank-sum test and chi-square testing to assess differences between groups. The analytic cohort included 6038 children with greater than 1 PICU encounter. Using a linear trajectory model, we identified six latent classes: three stable over time (good [46.6%], moderate [18.7%], and severe dysfunction [15.5%]), two with worsening over time (good/mild dysfunction with significant decline [7.9%], good/mild dysfunction with moderate decline [5.7%]), and one with improvement over time (mild dysfunction with improvement [5.6%]). In univariable analysis, trajectory classes differed based on age, race and ethnicity, diagnosis category, procedural exposure, severity of illness, and length of stay for the index admission. These associations were also present when assessed using multinomial regression and random forest classification.

In this multi-institutional database study using FSS scores to describe a contemporary, heterogeneous patient population with recurrent PICU-based observations of functional status, we identified six classes of outcome trajectories with different clinical and demographic characteristics.