This study used nationwide cohort data obtained from the South Korean National Health Insurance Service (NHIS) (https://nhiss.nhis.or.kr/). The NHIS is a single universal insurer that covers nearly all South Koreans (approximately 50 million). The NHIS provides at least a biennial national health checkup for all South Koreans aged ≥ 40 years and all employees, regardless of age. Therefore, it includes comprehensive medical information about sociodemographic characteristics, health checkups (lifestyle and health examinations), and medical diagnosis and treatment based on the International Classification of Diseases, 10th revision (ICD-10) codes [15].

From this database, we initially identified 714,095 individuals who were newly diagnosed with dementia between January 1, 2008, and December 31, 2016. Dementia was defined based on ≥ 2 times/year of anti-dementia medication prescription (donepezil, rivastigmine, galantamine or memantine) under relevant ICD-10 codes (F00–F03, G30, or G31 for all-cause dementia). We further divided them into Alzheimer’s disease using F00 or G30 and vascular dementia using F01. In South Korea, anti-dementia medications can be prescribed when the test results of the Mini-Mental State Examination, Clinical Dementia Rating, or Global Deterioration Scale fulfill the dementia criteria [16, 17].

Among individuals newly diagnosed with dementia, we selected 165,430 individuals aged ≥ 40 years who underwent the NHIS health examination within 4 years after the diagnosis of dementia. We then excluded individuals with missing variables (n = 8,487) and those who died within 1 year of diagnosis (n = 10,111). Ultimately, 146,832 individuals with dementia (52,124 men and 94,708 women) were included in the analyses.

This study complied with the provisions of the Declaration of Helsinki and was approved by the Institutional Review Board of the Korea University Guro Hospital, Seoul, Korea (No. 2022GR0325). The requirement for written informed consent was waived since all data used in the analysis were anonymous and non-identifiable.

The participants were divided into normoglycemia and DM groups. Only type 2 DM was selected, which was defined as fasting plasma glucose (FPG) ≥ 126 mg/dL or ≥ 1 anti-diabetic medication prescription per year under ICD-10 codes E11–E14. Glycemic status was categorized into 4 groups as follows: (1) normoglycemia (FPG < 100 mg/dL without a history of claims for anti-diabetic medication and the ICD-10 codes E11–E14), (2) prediabetes (FPG 100–125 mg/dL without claims for anti-diabetic medication and the ICD-10 codes E11–E14), (3) new-onset DM (FPG ≥ 126 mg/dL without previous claims for anti-diabetic medication and the ICD-10 codes E11–E14), and (4) known DM (claims for anti-diabetic medication under the ICD-10 codes E11–E14).

In addition, the duration of diabetes was combined with the glycemic status. Based on this, individuals were divided into five groups as follows: (1) normoglycemia, (2) prediabetes, (3) new-onset DM, (4) DM < 5 years, and (5) DM ≥ 5 years.

The endpoint was all-cause mortality, which was assessed using nationwide death certificate data from the Korea National Statistical Office. The study participants were followed up from one year after the index date until the date of death or until December 31, 2019, whichever came first. The mean follow-up duration was 3.7 ± 2.1 (interquartile range: 2.0 ∼ 5.2) years.

Anthropometric measurements including height, weight, waist circumference, and blood pressure (systolic and diastolic) were measured by healthcare professionals. Body mass index (BMI) was calculated as weight divided by height in meters squared (kg/m2). Blood samples were collected after overnight fasting to measure the concentrations of FPG, total cholesterol, triglycerides, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, and creatinine.

Comorbidities were defined on the basis of a combination of health examination results and claims for medication prescriptions before the index date. Hypertension was defined as a systolic/diastolic blood pressure of ≥ 140/90 mmHg or ≥ 1 medication prescription per year under ICD-10 codes I10–I13 or I15. Dyslipidemia was defined as a total cholesterol concentration of ≥ 240 mg/dL or ≥ 1 medication prescription per year under ICD-10 code E78. Chronic kidney disease (CKD) was defined as estimated glomerular filtration rate < 60 mL/minute/1.73 m2 calculated using the modification of diet in renal disease equation.

Data on smoking status, alcohol consumption, and physical activity were collected using self-reported questionnaires. Smoking status was divided into two categories: ever smokers and never smokers. Alcohol drinkers were defined as individuals who consumed ≥ 1 g of average alcohol per day. Regular exercise was defined as vigorous exercise for ≥ 3 days per week or moderate exercise for ≥ 5 days per week.

Low income was defined as individuals at the lowest 25th percentile using the NHI premium as a proxy for income level, and eligible for medical aid. Place of residence was divided into two groups: urban (metropolitan and city) and rural areas.

Disability was defined based on national disability registration data, with a focus on disabilities resulting from brain impairment. This definition was used to assess the severity of dementia [18]. The number of anti-dementia medications was categorized into 1 to 4 according to the prescription within one year of dementia diagnosis.

Baseline characteristics according to glycemic status were presented as means ± standard deviation (SD) for continuous variables or numbers (percentages) for categorical variables. Continuous variables were compared using analysis of variance, and categorical variables were compared using the chi-squared test. The mortality rate was calculated by dividing the number of deaths by 1,000 person-years.

Survival probabilities according to glycemic status parameters were plotted using Kaplan-Meier curves and compared using log-rank tests. We performed multivariable Cox proportional hazards regression analyses to evaluate the associations between glycemic status parameters and the risk of all-cause mortality, and the results were reported as hazard ratios (HRs) and 95% confidence intervals (CIs). Three models were used: Model 1 was not adjusted for any variables; Model 2 was adjusted for age, sex, place of residence, income, smoking status, alcohol consumption, physical activity, BMI, hypertension, dyslipidemia, and CKD; and Model 3 was adjusted for disability and the number of anti-dementia medications in addition to the variables in Model 2. We performed a sensitivity analysis excluding participants who experienced all-cause mortality within 2 years of follow-up. Subgroup analyses were performed after stratified by all confounding variables.

All statistical analyses were performed using the SAS software (version 9.4; SAS Institute, Cary, NC, USA). Differences were considered statistically significant at P < 0.05.

Restrictions apply to the availability of all data analyzed in this study because they were used under license. Additional data are available through approval and oversight by the Korean NHIS.