The data can be used after approval of the Institutional Review Board and the Korea NHIS Big Data Operations Department (https://nhiss.nhis.or.kr/bd/ay/bdaya 001 iv.do). No separate ethics approval was required.

We used the national claims database of the National Health Insurance Service (NHIS) of Korea. The database contains the medical records of the entire Korean population covered by the obligatory NHIS and Medical Aid programs. Sociodemographic information and healthcare utilization information for inpatient and outpatient visits, including diagnoses, procedures, surgery, and prescription records, are available in the database. Diagnoses are coded according to International Classification of Disease-10th Revision-Clinical Modification (ICD-10-CM) nomenclature. The NHIS provides not only claims data but also health screening examination data. The NHIS provides a free biannual national health screening program (NHSP) to all beneficiaries aged ≥ 40 years, and the NHSP database includes patient history, physical examinations, anthropometric measurements, and questionnaire data.

In 2006, the Korean government implemented a registration program for rare intractable diseases, which provides disease-specific codes. The rare intractable disease code for PD is V124. In this study, we screened cases of newly-diagnosed PD patients (ICD-10 code G20; rare intractable disease registration code V124) between January 2008 and December 2017, excluding those diagnosed with secondary parkinsonism or atypical parkinsonism (ICD-10 codes G21-G23). We also excluded individuals diagnosed with PD before 2009, to include only new PD patients diagnosed after 2009 (washout period: 2008). Thereafter, we identified PD patients who had participated in the NHSP at least once within 2 years of their initial diagnosis. Individuals with missing data were excluded. Finally, we longitudinally followed these 32,419 PD patients to investigate the association between alcohol use and death. The patient enrollment flowchart is presented in Supplementary File 2.

As part of the NHSP, individuals who currently drink alcohol completed self-report questionnaires about their average alcohol consumption per day and week. Questions included: ‘How many days a week do you drink on average?’ and ‘How much do you usually drink a day (cups)?’ Although beverages may have different alcohol percentages, previous studies have shown that the alcohol content per cup is similar in Korea regardless of beverage type (Kim Y. G. et al. 2020b; Yoo et al. 2021). Therefore, each self-reported cup was converted to 8 g of alcohol. Based on the total alcohol intake per week from NHSP questionnaire within two years form PD diagnosis (baseline), participants were classified into four primary groups: heavy drinkers (more than 210 g of alcohol per week), moderate drinkers (105–209 g of alcohol per week), mild drinkers (less than 104 g of alcohol per week), and nondrinkers (individuals who were not currently drinking at baseline, including never- and former drinkers).

Changes in alcohol consumption behavior were assessed by identifying patterns of alcohol use within the two years before and after the PD diagnosis. Participants were categorized into three groups based on their drinking behavior over this period: constant drinkers (individuals who continued to drink alcohol both before and after their PD diagnosis), former drinkers (individuals who consumed alcohol before the PD diagnosis but stopped drinking afterward), and never-drinkers (individuals who have never consumed alcohol at any time).

The endpoint of this study was all-cause mortality in patients with PD until December 31, 2017. All-cause mortality was assessed using death records from the Korean NHIS database, which provides comprehensive data on patient outcomes, including mortality, for all registered individuals in Korea.

Demographic variables, such as age and sex, were based on data entered at the time of diagnosis. The financial income deciles of participants were categorized into two groups based on the NHI premium: low-income level (the lower 25% group) versus not low-income. The Charlson Comorbidity Index (CCI) and other comorbidities were identified based on the corresponding disease diagnostic codes. Comorbidities of participants in this study included hypertension (I10-I13, I15), dyslipidemia (E780-E785), pneumonia (J10-J18), osteoporosis (M80-M82), and depression (F23, F33). The levodopa equivalent daily dose (LEDD) was used to index disease severity. Anthropometric factors, including height, weight, and blood pressure, were noted. Body mass index (BMI) was calculated as weight divided by height (kg/m2). Plasma glucose, lipids, and hemoglobin levels were determined through venous sampling after fasting for > 8 h.

Information about lifestyle factors, including smoking and PA, were gleaned from self-report questionnaires. Smoking status was categorized based on the Centers for Disease Control and Prevention into the following three groups: current smoker (had smoked in their lifetime and currently smoked), ex-smoker (had smoked in their lifetime but did not currently smoke), and never-smoker. PA was classified into two groups by converting the intensity of PA and the number of weekly exercises per intensity level into a standard metabolic equivalent (MET): (1) light-intensity activities (< 600 MET × min/week) and (2) moderate/vigorous-intensity activities (≥ 600 MET × min/week) (Craig et al. 2003; Ainsworth et al. 2011; Lear et al. 2017).

The characteristics of participants according to alcohol consumption status were compared using analyses of variance (ANOVA) for continuous variables and Chi-square tests for categorical variables. Results are presented as means ± standard deviations for continuous variables and as frequencies and percentages for categorical variables. Survival analysis over time was performed through a log-rank test using Kaplan-Meier curves. Cox proportional hazards regression analyses were performed to estimate hazard ratios (HRs) and their 95% confidence intervals (CIs) for the association between alcohol consumption and all-cause mortality, with adjustment for covariates. We used three progressive models: Model 1 was unadjusted; Model 2 was adjusted for age, sex, income level, and residential area; and Model 3 was further adjusted for CCI, comorbidities, lifestyle factors (smoking status and PA), anthropometric data, and LEDD. To minimize abstainer bias in examining the association between alcohol consumption and mortality in PD, we performed a sensitivity analysis using never drinkers as the reference group. Statistical analyses were performed using SAS version 9.2 (SAS Inc., Cary, NC, USA). Statistical significance was set at P < 0.05.