Current trend analysis of healthcare utilization in Congenital Heart Disease (CHD) is flawed: what 50 years of longitudinal pediatric CHD data and segmented regression modeling reveal

Abstract Introduction Congenital heart disease (CHD) is the most common congenital anomaly with a rising prevalence. Trend analyses are often employed to evaluate the impact of this growing population on healthcare use and to guide service planning. Although nonlinearity is common in such trends, current methods rely on linear regression analysis. However, applying a linear model to a nonlinear trend can bias results, obscure trend changes, and lead to inaccurate estimates. Purpose This study aims to determine whether significant nonlinearity exists in hospitalization trends among the pediatric Swedish CHD population, focusing on both initial repair and follow-up care. Methods This retrospective longitudinal study uses Swedish national registry data to analyze live births with a registered CHD diagnosis between 1970 and 2017. Cases (n=67,814) were stratified into the following pediatric age groups: infancy (1 year), children (1–9 years), and adolescence (10–17 years), where initial repair for most lesions predominantly occurred in infancy. Trends for average annual number of hospitalizations were analyzed using the segmented regression modeling technique which models the data into piecewise linear relationships. The method used in this paper focused on computer identified breakpoints within the trend data. Results A simple linear regression line was not appropriate for any of the age cohort groups, regardless of CHD complexity. One significant breakpoint was detected for all age groups across the general CHD population, where the infant category had two breakpoints (1992 and 2003). Hospitalization increased until 1979 among adolescents, followed by children in 1985 and infants in 1992. The most drastic reduction occurred in infant hospitalization from over 2.5 per year in 1992 to 1 in 2003, and has been steady since. Conversely, children faced on average 2 hospitalizations a year until 1985, and have thereafter steadily decreased towards zero. Similarly, increasing hospitalizations were seen in adolescents till 1979, where it is now trending towards zero per year as well. Conclusion(s) Non-linearity was found among this population asserting that using linear regression will produce inaccurate or misleading results. Hospitalization peaked around the 1980s and has since been steadily decreasing between 50-100% across all pediatric age categories. This study is the first to identify specific years of comparability issues in utilization trends and how these trends have evolved over time, supporting long-held clinical insights and wisdom.