Use of Latent Class Analysis to Predict Intensive Care Unit Admission and Mortality in Children with a Major Congenital Anomaly

Abstract

To define major congenital anomaly subgroups and assess outcome variability based on defined subgroups. This population-based cohort study used registries in Denmark for children born with a major congenital anomaly between January 1997 and December 2016, with follow-up until December 2018. We performed a latent class analysis (LCA) using child and family clinical and sociodemographic characteristics present at birth, incorporating additional variables occurring until age of 24 months. Cox proportional hazards regression models estimated hazard ratios (HRs) of pediatric mortality and intensive care unit (ICU) admissions for identified LCA classes. The study included 27,192 children born with a major congenital anomaly. Twelve variables led to a 4-class solution (entropy=0.74): (1) children born with higher income and fewer comorbidities (55.4%), (2) children born to young mothers with lower income (24.8%), (3) children born prematurely (10.0%), and (4) children with multiorgan involvement and developmental disability (9.8%). Compared with those in Class 1, mortality and ICU admissions were highest in Class 4 (HR=8.9, 95% confidence interval [CI]=6.4-12.6 and HR=4.1, 95% CI=3.6-4.7, respectively). More modest increases were observed among the other classes for mortality and ICU admissions (Class 2: HR=1.7, 95% CI=1.1-2.5 and HR=1.3, 95% CI=1.1-1.4, respectively; Class 3: HR=2.5, 95% CI=1.5-4.2 and HR=1.5, 95% CI=1.3-1.9, respectively). Children with a major congenital anomaly can be categorized into meaningful subgroups with good discriminative ability. These groupings may be useful for risk-stratification in outcome studies.