Vitamin D and Healthcare Service Utilization in Children: Insights from a Machine Learning Approach

Abstract

Vitamin D deficiency and insufficiency is a global health issue: an association has been demonstrated between vitamin D deficiency and a myriad of acute and chronic illnesses. Data regarding vitamin D status among children hospitalized with non-critical illnesses are scanty. We aimed to: (1) identify profiles of children hospitalized due to non-critical illnesses, using vitamin D levels as the driving outcome; (2) assess the association between patient profiles and length of stay. The study included 854 patients (1–17 years old) who underwent blood tests for detecting vitamin D levels. A regression tree was used to stratify patients. The relationship between vitamin D levels and length of stay was explored using Poisson regression. The regression tree identified three subgroups. Group A (16%): African, North African, Hispanic, and Indian patients. Group B (62%): Caucasian and Asian patients hospitalized for respiratory, metabolic, ill-defined, infective, and genitourinary diseases. Group C (22%): Caucasian and Asian patients hospitalized for digestive, nervous, and musculoskeletal diseases, blood and skin diseases, and injuries. Mean serum vitamin D level (ng/mL) was 13.7 (SD = 9.4) in Group A, 20.5 (10.0) in Group B, and 26.2 (12.6) in Group C. Group B was associated with the highest BMI z-score (p p = 0.041). Mean length of stay was longer in Group A than in the other groups (p p = 0.024) in Group A and by 5% (p = 0.029) in Group B per 10 ng/mL increase in vitamin D level. We identified three subgroups of hospitalized children, defined according to ethnicity and discharge diagnosis, and characterized by increasing vitamin D levels. Vitamin D levels were associated with length of hospitalization.

Keywords

• 25(OH)D
• vitamin D deficiency
• paediatrics
• hospitalization
• regression tree
• international classification of diseases