521 - Latent class analysis of medical needs and social drivers of health in bronchopulmonary dysplasia

Publication Number: 521.5659

Brandon M. Smith, Johns Hopkins University School of Medicine, Baltimore, MD, United States; Radhika Raghunathan, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, United States; Brianna Aoyama, Johns Hopkins University School of Medicine, Baltimore, MD, United States; Sara B. Johnson, Johns Hopkins University School of Medicine, Baltimore, MD, United States; Joseph Collaco, Johns Hopkins Children's Center, Baltimore, MD, United States

Background: Bronchopulmonary dysplasia (BPD) is a leading cause of morbidity for children born prematurely with known disparities by race and ethnicity. Beyond their initial hospital stay, children’s lung health is influenced by both complex medical needs and social drivers of health. Therefore, identification of risk factors in early childhood is key to optimizing lung function over the life course. Studies to date have identified medical and social factors that separately influence outcomes for children with BPD; however, few have assessed the interplay of these factors together using a person-centered approach.

Objective: Identify and characterize subgroups of children with BPD based on both medical needs and social drivers in early childhood.

Design/Methods: We utilized clinical registry data from our local BPD Patient Registry (2008-2024). Latent class analysis (LCA) was used to identify subgroups of children with BPD less than 3 years old, based on heterogeneity in medical needs (e.g., gestational age, birth weight, length of hospital stay, BPD severity, oxygen and medication use, medical technology, clinic visit number) and social drivers (e.g., household income, childhood opportunity index [COI]). One to six class latent models were fit, and fit statistics were examined to determine the optimal number of classes.

Results: Demographics are shown in Table 1. A four-class model was selected based on fit statistics (e.g., decreasing Bayesian Information Criterion, significant Lo-Mendell-Rubin test statistic) and class interpretability (e.g., substantive interpretation, class sizes). Indicator probabilities by class are shown in Table 2 and Figure 1. Classes were summarized as: 1) Severe BPD with high medical needs (e.g., severe BPD, extremely preterm, extremely low birth weight, medical technology and medication needs, range of income and COI); 2) Moderate BPD with low social risk (e.g., moderate BPD, high income and COI; 3) Moderate BPD with moderate social risk (e.g., moderate BPD, income and COI; 4) Moderate BPD with high medical needs and high social risk (e.g., moderate BPD, medical technology and medication needs, low income and COI).

Conclusion(s): Using a person-centered analytic approach, we identified four distinct classes of children with BPD, which may help to understand the trajectory of respiratory outcomes during childhood. Next steps include examining the association between latent classes and respiratory outcomes, with moderation by race and ethnicity, to inform tailored intervention strategies that can disrupt disparities in outcomes for children with BPD.

Table 1
Participant Demographics

Table 2
Probability of class selection for each medical need and social driver in four class model

Figure 1
Probabilities of class selection by medical need and social drivers in four class model. (BPD=bronchopulmonary dysplasia; COI=Childhood Opportunity Index)