Patrones de Niños con Condiciones Crónicas Complejas: Un Análisis de Clases Latentes

The objective of this study was to distinguish empirical classes among children with complex chronic conditions (CCCs) and to assess whether such classes can predict future health care use.

We analyzed claims data from children aged 1 to 18 years with a CCC who were continuously enrolled in a 10-state Medicaid database from 2017 to 2019. We performed a latent class analysis using demographic factors, clinical characteristics, and health care use patterns in 2017 and assessed the ability of the classes to differentiate health care spending and use in 2018 to 2019 using negative binomial and logistic regression.

We included 185 672 children with a CCC (52% male; median [IQR] age: 11 [5, 15] years). Eight indicator variables led to a 3-class solution (entropy = 0.83): Class 1 (9.1% of the cohort) was characterized by high neuro-disability, high technology dependence, and high multimorbidity; Class 2 (14.8%) had high neuro-disability and low technology dependence; and Class 3 (76.0%) had low neuro-disability and low technology dependence. Compared with children in Class 3, total spending in 2017 to 2018 was increased among both Class 1 and Class 2 (total spending rate ratio [RR] 6.9 [95% CI: 6.7-7.0] and RR 2.5 [95% CI: 2.5-2.6], respectively). The largest categories of subsequent spending were for inpatient care and outpatient specialist services among individuals in Class 1 and for outpatient drugs, outpatient specialists, and mental health for those in Class 2.

Children with CCCs can be categorized into meaningful classes based on readily available data with different patterns of future health care use and costs.