RISK OF DYSGLYCEMIA FOLLOWING MENTAL DISORDERS AND ASSESSMENT OF RISK HETEROGENEITY: A MILITARY CONSCRIPTION COHORT STUDY OF YOUNG KOREAN MALES
Author(s)
Jimin Do, MS1, Siin Kim, MS, PharmD, PhD2, Miryoung Kim, PhD3, Hae Sun Suh, MA, MS, PhD4.
1Department of Regulatory Science, Graduate School, Kyung Hee University, Seoul, Korea, Republic of, 2College of Pharmacy, Kyungsung University, Busan, Korea, Republic of, 3College of Pharmacy, Sunchon National University, Suncheon-si, Korea, Republic of, 4College of Pharmacy, Kyung Hee University, Seoul, Korea, Republic of.
1Department of Regulatory Science, Graduate School, Kyung Hee University, Seoul, Korea, Republic of, 2College of Pharmacy, Kyungsung University, Busan, Korea, Republic of, 3College of Pharmacy, Sunchon National University, Suncheon-si, Korea, Republic of, 4College of Pharmacy, Kyung Hee University, Seoul, Korea, Republic of.
OBJECTIVES: To evaluate the association between mental disorder and dysglycemia in young Korean males and identify high-risk subgroups.
METHODS: A nationwide retrospective cohort study linked Military Manpower Administration(MMA) conscription examination records—covering a near-complete census of Korean males—with nationwide claims data(index period 2019-2021; follow-up until 2024). The exposed group comprised individuals with a first mental disorder diagnosis(ICD-10:F-codes) following a two-year diagnosis-free period; the first diagnosis was the index date. Controls had no mental disorder diagnosis throughout the observation period and were assigned corresponding index dates within the same age and MMA examination-year strata. Individuals with prior diabetes or abnormal glucose(ICD-10:E10-E14,R73) or fasting glucose >125mg/dL within two years before the index were excluded. Dysglycemia was defined as the first diagnosis of T2DM(ICD-10:E11) or abnormal glucose(ICD-10:R73) after the index date. Index year-stratified Cox models estimated adjusted hazard ratio(HR); Kaplan-Meier restricted mean survival time(RMST) analysis was performed. A Causal Survival Forest(CSF) incorporating demographic, healthcare utilization, and baseline clinical/laboratory variables(e.g.,BMI, glucose, lipid, liver enzymes) estimated average and individual-level RMST differences and characterized risk heterogeneity; individuals were stratified into conditional average treatment effect(CATE) quintiles to identify high-risk subgroups.
RESULTS: A total of 29,291 exposed and 698,074 controls were included(mean follow-up:1,649days). Cumulative incidence was 3.56% in the exposed and 1.87% in the control group. Mental disorder was significantly associated with increased dysglycemia risk(adjusted HR:1.70, 95% CI:1.59 to 1.81). The CSF estimated an average RMST difference of −7.77days(95% CI −9.26 to −6.28), indicating earlier dysglycemia onset in the exposed group, consistent with the Kaplan-Meier RMST estimate(mean difference:−9.3days, p<0.001). The highest CATE quintile had higher baseline BMI, blood pressure, GGT, and ALT than the lowest quintile.
CONCLUSIONS: Mental disorder was associated with increased dysglycemia risk in young Korean males. Causal machine learning identified clinically meaningful individual-level risk heterogeneity, supporting risk-stratified metabolic monitoring after mental disorder diagnosis.
METHODS: A nationwide retrospective cohort study linked Military Manpower Administration(MMA) conscription examination records—covering a near-complete census of Korean males—with nationwide claims data(index period 2019-2021; follow-up until 2024). The exposed group comprised individuals with a first mental disorder diagnosis(ICD-10:F-codes) following a two-year diagnosis-free period; the first diagnosis was the index date. Controls had no mental disorder diagnosis throughout the observation period and were assigned corresponding index dates within the same age and MMA examination-year strata. Individuals with prior diabetes or abnormal glucose(ICD-10:E10-E14,R73) or fasting glucose >125mg/dL within two years before the index were excluded. Dysglycemia was defined as the first diagnosis of T2DM(ICD-10:E11) or abnormal glucose(ICD-10:R73) after the index date. Index year-stratified Cox models estimated adjusted hazard ratio(HR); Kaplan-Meier restricted mean survival time(RMST) analysis was performed. A Causal Survival Forest(CSF) incorporating demographic, healthcare utilization, and baseline clinical/laboratory variables(e.g.,BMI, glucose, lipid, liver enzymes) estimated average and individual-level RMST differences and characterized risk heterogeneity; individuals were stratified into conditional average treatment effect(CATE) quintiles to identify high-risk subgroups.
RESULTS: A total of 29,291 exposed and 698,074 controls were included(mean follow-up:1,649days). Cumulative incidence was 3.56% in the exposed and 1.87% in the control group. Mental disorder was significantly associated with increased dysglycemia risk(adjusted HR:1.70, 95% CI:1.59 to 1.81). The CSF estimated an average RMST difference of −7.77days(95% CI −9.26 to −6.28), indicating earlier dysglycemia onset in the exposed group, consistent with the Kaplan-Meier RMST estimate(mean difference:−9.3days, p<0.001). The highest CATE quintile had higher baseline BMI, blood pressure, GGT, and ALT than the lowest quintile.
CONCLUSIONS: Mental disorder was associated with increased dysglycemia risk in young Korean males. Causal machine learning identified clinically meaningful individual-level risk heterogeneity, supporting risk-stratified metabolic monitoring after mental disorder diagnosis.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EPH165
Topic
Epidemiology & Public Health, Methodological & Statistical Research
Disease
Diabetes/Endocrine/Metabolic Disorders (including obesity), Mental Health (including addiction)