ORIGINAL ARTICLE: Associations of antidepressant medication with its various predictors including particulate matter: Machine learning analysis using national health insurance data

• Published in: Journal of psychiatric research Vol. 147; pp. 67 - 78
• Main Authors: Lee, Kwang-Sig, Kim, Geunyeong, Ham, Byung-Joo
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.03.2022
• Subjects: Adolescent; Adult; Aged; Air Pollutants - analysis; Air Pollution - analysis; Antidepressant; Antidepressive Agents - therapeutic use; Depressive Disorder, Major; Environmental Exposure - analysis; Humans; Machine Learning; Middle Aged; National Health Programs; Neighborhood condition; Particulate matter; Particulate Matter - analysis; Retrospective Studies; Socioeconomic satisfaction; Young Adult; Antidepressant; Socioeconomic satisfaction; Neighborhood condition; Particulate matter
• ISSN: 0022-3956; 1879-1379
• DOI: 10.1016/j.jpsychires.2022.01.011

This study uses machine learning and population-based data to analyze major determinants of antidepressant medication including the concentration of particulate matter under 2.5 μm (PM2.5). Retrospective cohort data came from Korea National Health Insurance Service claims data for 43,251 participants, who were aged 15–79 years, lived in the same districts of Seoul and had no history of antidepressant medication during 2002–2012. The dependent variable was antidepressant-free months during 2013–2015 and the 30 independent variables for 2012 were included (demographic/socioeconomic information, health information, district-level information including PM2.5). Random forest variable importance, the contribution of a variable for the performance of the model, was used for identifying major predictors of antidepressant-free months. Based on random forest variable importance, the top 15 determinants of antidepressant medication during 2013–2015 included cardiovascular disease (0.0054), age (0.0047), household income (0.0037), gender (0.0027), the district-level proportion of recipients of national basic living security program benefits (0.0019), district-level social satisfaction (0.0013), diabetes mellitus (0.0012), January 2012 PM2.5 (0.0011), district-level street ratio (0.0010), drinker (0.0009), chronic obstructive pulmonary disease (0.0008), district-level economic satisfaction (0.0006), exercise (0.0005), March 2012 PM2.5 (0.0005) and November 2012 PM2.5 (0.0004). Besides these predictors, smoker and district-level deprivation index are found to be influential most widely, given that they ranked within the top 10 most often in sub-group analysis. In conclusion, antidepressant medication has strong associations with neighborhood conditions including socioeconomic satisfaction and the seasonality of particulate matter. Strong interventions for these factors are really needed for the effective management of major depressive disorder. •The most comprehensive machine-learning analysis for antidepressant medication.•A population-based cohort of 43,251 participants, 30 predictors and 92 models.•The seasonality of particulate matter associates with antidepressant medication.•Neighborhood satisfaction associates with antidepressant medication.