COVID-19-related role disruption and alcohol use disorder risk among adults in South Korea Authors: Hyun Jun Lee, Gyeong-Min Lee, Ji-Sung Hong, Jae-Hyun Kim, Jung Jae Lee Published: 2026 ABSTRACT This cross-sectional study examined whether COVID-19–related disruption of occupational, social/leisure, and family roles was associated with alcohol use disorder (AUD) risk among adults in South Korea. We analyzed data from 936 adults aged 20–69 years who participated in the 2022 Chungnam Provincial Mental Health Survey. COVID-19–related disruption was assessed across three role domains and categorized as low, moderate, or high. AUD risk was defined using the Korean version of the Alcohol Use Disorders Identification Test. Multivariable logistic regression was used to estimate adjusted odds ratios (aORs), controlling for sociodemographic characteristics and depressive symptoms. Overall, 129 participants (13.8%) screened positive for AUD risk. The crude prevalence of AUD risk increased across disruption categories: 11.0% in the low-disruption group, 12.3% in the moderate-disruption group, and 20.8% in the high-disruption group. In the fully adjusted model, high disruption was associated with higher odds of AUD risk compared with low disruption (aOR = 1.93, 95% CI: 1.10–3.39; p = 0.021), whereas moderate disruption was not statistically significant. These findings suggest that substantial pandemic-related role disruption may be associated with increased AUD risk, although causal inference is limited by the cross-sectional design.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-55907-w. Introduction The COVID-19 pandemic has led to substantial changes in healthcare systems and public health policies worldwide. In South Korea, the government implemented a phased social distancing policy that included restrictions on business hours and limitations on private gatherings1. These measures resulted in significant changes in daily life, including the transition from in-person to remote learning and a widespread decrease in social outings. These shifts disrupted daily routines and led to reduced social connectedness, heightened emotional stress, financial difficulties, and increased levels of depression2. Work environments also shifted to remote formats; however, the lack of organizational support and intensified surveillance under performance-based corporate cultures often led to elevated job-related fatigue and stress3. Restrictions on social activity, combined with increased stress and uncertainty, may have influenced patterns of alcohol consumption. The uncertainty and anxiety caused by these societal changes may have increased alcohol use, and in such environments, drinking could have more harmful effects on individuals4. According to the 2021 National Alcohol Consumption Survey by the Korea Health Promotion Institute, prolonged periods of the pandemic were associated with increases in both the frequency and quantity of alcohol consumption. The proportion of individuals who reported a decrease in alcohol intake dropped, while those reporting more frequent drinking increased5. Significant changes also occurred in drinking settings and behaviors. The percentage of people drinking alone more than doubled—from 12.6% before the pandemic to 29.2% afterward—and home drinking surged from 23.3% to 70.7%5. Solitary drinking is known to contribute to binge drinking and hinder moderation, especially among adolescents and young adults, where it may lead to risk of alcohol use disorder (risk of AUD)6. Data from the National Center for Mental Health in Korea show that both current drinking rates and high-risk drinking rates continued to rise from 2019 to 2022, with a particularly noticeable increase among women7. Additionally, the proportion of problematic drinkers—individuals involved in offenses such as drunk driving, harassment, or violence—remained above 18% throughout the pandemic and peaked at 22% in June 20227. In contemporary society, alcohol is often used both as a social facilitator—in workplace gatherings and among acquaintances—and as a means of coping with stress or alleviating distress8. During the COVID-19 era, strict infection control policies such as social distancing restricted interpersonal contact and made face-to-face interaction difficult. As a result, workplaces and universities shifted to remote operations, drastically reducing opportunities for social connection. These changes increased feelings of loneliness and isolation9. Although the government’s distancing policies aimed to suppress viral transmission, they unintentionally imposed psychological burdens on individuals confined at home2. Concerns over post-pandemic life and financial instability, particularly job loss, have led to rising levels of stress, depressive symptoms, and increased alcohol consumption4. Even in high-income countries like those in North America and Europe, alcohol consumption rose significantly during the pandemic, despite restrictions on bar and restaurant access. This suggests that increases in alcohol use may occur regardless of environmental limitations on drinking venues10. Previous studies indicate that the absence of vaccines during the early pandemic necessitated strict distancing measures, which led to economic insecurity. This in turn was associated with psychological issues such as depression and stress, and higher frequency of alcohol consumption11. Beyond economic activity, reductions in religious and leisure activities also negatively impacted subjective well-being12. Most existing research on risk of AUD during the COVID-19 period has focused on the effects of social isolation and increased stress levels, often analyzing changes in alcohol consumption due to lockdown policies13,14. However, limited research has examined how restricted social activity during the pandemic may be directly associated with risk of alcohol use disorder. Therefore, the present study aimed to analyze how disruptions in work roles, leisure and hobby activities, and family responsibilities during the COVID-19 pandemic were associated with the risk of AUD. The findings are intended to provide insights for policy decisions during public health crises and suggest strategies to improve individual health and well-being during future pandemics. Methods Data source and study participants This study utilized data from the 2022 Mental Health Awareness Survey conducted by the Chungcheongnam-do Metropolitan Mental Health Welfare Center. The purpose of the survey was to assess the mental health status and awareness of mental health issues among residents of Chungcheongnam-do, South Korea, with the ultimate goal of providing foundational data for improving public mental health and awareness within the local community. The survey targeted adults aged 20 to 69 years residing in all 16 cities and counties of Chungcheongnam-do. A proportional stratified sampling method was employed based on the 2022 October-end resident registration population statistics published by the Ministry of the Interior and Safety, considering gender, age, and regional population distribution. A total of 1,000 valid responses were collected through a structured questionnaire using both web-based surveys and one-on-one interviews conducted between November 5 and November 20, 2022. Participants who did not respond to key variables regarding disruption of social activities due to COVID-19 or alcohol use were excluded from the analysis. Additionally, individuals with missing data on general characteristics, including age, gender, residential area, education level, employment status, monthly income, and depression scale scores, were also excluded. After applying these criteria, a total of 936 participants were included in the final analysis. A complete-case analytic approach was used. Participants with missing information on COVID-19-related disruption, alcohol use, sociodemographic covariates, or depressive symptoms were excluded from the analysis. Although the original survey used proportional stratified sampling by sex, age, and region, individual-level sampling weights were not available in the analytic dataset; therefore, all analyses were conducted without survey weights. Independent variables The independent variable in this study was the degree of disruption in social activities due to the COVID-19 pandemic. This was measured using three items that asked participants to rate how much the pandemic interfered with their (1) occupational roles, (2) social and leisure activities, and (3) family life and responsibilities. The question posed was: “Since the outbreak of COVID-19, to what extent has your current situation interfered with your work, social/leisure activities, and family or household roles?” Responses were rated on a three-point scale, where “Not at all” and “Slightly” were assigned 0 points, “Moderately” was assigned 1 point, and “Severely” or “Very severely” were assigned 2 points. This categorization was determined a priori to represent meaningful levels of cumulative role disruption. A score of 0 represented no meaningful disruption across all assessed domains, scores of 1–3 represented partial or moderate disruption, and scores of 4–6 represented substantial disruption affecting multiple domains or severe disruption in at least two domains. Dependent variables The dependent variable in this study was risk of AUD, which was assessed using the Korean version of the Alcohol Use Disorders Identification Test (AUDIT-K). The AUDIT-K consists of 10 items covering alcohol consumption, alcohol dependence symptoms, and alcohol-related problems. Each item is scored from 0 to 4, resulting in a total score ranging from 0 to 40. In accordance with Korean standard criteria, participants were classified as being at risk of AUD if they scored ≥ 20 for men or ≥ 10 for women, reflecting sex-specific differences in alcohol-related harm. Because the AUDIT-K is a screening instrument rather than a diagnostic interview, the outcome is interpreted as risk of AUD rather than clinically diagnosed AUD15,16. Control variables Covariates included sociodemographic and health-related characteristics: age, sex, residential area, marital status, educational attainment, employment status, household income, and depressive symptoms. Age was modeled in five categories (20–29, 30–39, 40–49, 50–59, and 60–69 years). Sex was male or female. Residential area was classified as urban (city/district) or rural (county). Marital status was married or unmarried (including divorced, separated, or widowed). Educational attainment was ≤high school or ≥college. Employment status was employed or unemployed. Household income was grouped into tertiles (low, middle, high). Depressive symptoms were assessed using the Korean version of the Patient Health Questionnaire-9 (PHQ-9), a 9-item self-report instrument assessing depressive symptoms over the previous two weeks. Each item is scored from 0 to 3, yielding a total score ranging from 0 to 27, with higher scores indicating more severe depressive symptoms. Consistent with established screening criteria, scores of 10 or higher were classified as elevated depressive symptoms17, and a binary indicator was included in the regression models. Analytical approach and statistics Descriptive statistics were used to summarize participant characteristics, and chi-square tests were used to compare the distribution of risk of AUD across participant characteristics and COVID-19-related disruption categories. Multivariable logistic regression was used to estimate adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for the association between COVID-19-related disruption and risk of AUD. The fully adjusted model included age, sex, residential area, marital status, educational attainment, employment status, household income, and depressive symptoms. To improve transparency, we additionally examined the distribution of covariates across disruption categories and conducted stepwise logistic regression models. Stratified analyses were performed by residential area and age group (< 50 vs. ≥50 years), and formal interaction terms were tested to assess whether the association differed by residential area or age group. Sensitivity analyses were conducted by modeling the disruption score as a continuous variable, using a binary high-disruption indicator, and excluding depressive symptoms from the adjustment set. All analyses were conducted as unweighted complete-case analyses because individual-level sampling weights were not available. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). A two-tailed p-value of less than 0.05 was considered statistically significant. Results Table 1 summarizes participant characteristics and the unadjusted distribution of risk of AUD. Among 936 participants, 129 (13.8%) screened positive for risk of AUD. In the crude distribution, the prevalence of risk of AUD increased across COVID-19-related disruption categories: 11.0% in the low-disruption group, 12.3% in the moderate-disruption group, and 20.8% in the high-disruption group (p = 0.003). Risk of AUD also differed significantly by age, sex, marital status, household income, and depressive symptoms. Table 1General characteristics of the study population by risk of AUD.VariablesTotalRisk of AUDp-valueNoYes N % N % N %Total93610080786.212913.8Disruption of social activity due to COVID-19 High21222.616879.24420.80.003 Moderate42445.337287.75212.3 Low30032.126789.03311.0Age 20–2913214.19168.94131.1< 0.0001 30–3916017.112980.63119.4 40–4920722.117082.13717.9 50–5923425.022596.293.8 60–6920321.719294.6115.4Sex Male49052.444290.2489.8< 0.0001 Female44647.636581.88118.2Residency Region Urban75780.964985.710814.30.376 Rural17919.115888.32111.7Marital status Single (including separated, divorced, bereavement)23825.417774.46125.6< 0.0001 Married69874.663090.3689.7Education ≤ High school47450.641687.85812.20.165 ≥ College46249.439184.67115.4Employment status Unemployed18319.616087.42312.60.595 Employed75380.464785.910614.1Family income Low849.06375.02125.00.002 Moderate48051.341185.66914.4 High37239.733389.53910.5Depression status No81687.274391.1738.9< 0.0001 Yes12012.86453.35646.7Note. AUD, alcohol use disorder. Risk of AUD was defined using sex-specific AUDIT-K cutoff scores. Table 1General characteristics of the study population by risk of AUD.VariablesTotalRisk of AUDp-valueNoYes N % N % N %Total93610080786.212913.8Disruption of social activity due to COVID-19 High21222.616879.24420.80.003 Moderate42445.337287.75212.3 Low30032.126789.03311.0Age 20–2913214.19168.94131.1< 0.0001 30–3916017.112980.63119.4 40–4920722.117082.13717.9 50–5923425.022596.293.8 60–6920321.719294.6115.4Sex Male49052.444290.2489.8< 0.0001 Female44647.636581.88118.2Residency Region Urban75780.964985.710814.30.376 Rural17919.115888.32111.7Marital status Single (including separated, divorced, bereavement)23825.417774.46125.6< 0.0001 Married69874.663090.3689.7Education ≤ High school47450.641687.85812.20.165 ≥ College46249.439184.67115.4Employment status Unemployed18319.616087.42312.60.595 Employed75380.464785.910614.1Family income Low849.06375.02125.00.002 Moderate48051.341185.66914.4 High37239.733389.53910.5Depression status No81687.274391.1738.9< 0.0001 Yes12012.86453.35646.7Note. AUD, alcohol use disorder. Risk of AUD was defined using sex-specific AUDIT-K cutoff scores. General characteristics of the study population by risk of AUD. Note. AUD, alcohol use disorder. Risk of AUD was defined using sex-specific AUDIT-K cutoff scores. Table 2 presents the results of the fully adjusted multivariable logistic regression model. COVID-19-related disruption was significantly associated with risk of AUD overall (p = 0.049). Compared with the low-disruption group, participants in the high-disruption group had higher adjusted odds of risk of AUD (aOR = 1.93, 95% CI: 1.10–3.39; p = 0.021). The moderate-disruption group did not differ significantly from the low-disruption group (aOR = 1.17, 95% CI: 0.70–1.97; p = 0.545). Other factors associated with risk of AUD included age, sex, employment status, and depressive symptoms. Table 2Factors associated with risk of AUD.VariableaOR95% CIp-valueCOVID-19-related disruption0.049 LowReference Moderate1.170.70–1.970.545 High1.931.10–3.390.021Age, years0.0003 20–295.722.26–14.45< 0.001 30–393.391.43–8.050.006 40–493.101.37–7.010.007 50–590.790.31–2.030.628 60–69ReferenceSex< 0.001 MaleReference Female2.901.85–4.54< 0.001Residential area0.993 UrbanReference Rural1.000.57–1.750.993Marital status0.236 MarriedReference Single/separated/divorced/widowed1.390.81–2.400.236Education0.137 ≥ CollegeReference ≤ High school1.430.89–2.300.137Employment status0.012 UnemployedReference Employed2.081.18–3.690.012Household income0.821 HighReference Low1.220.59–2.550.589 Moderate1.140.70–1.860.590Depressive symptoms< 0.001 NoReference Yes4.692.88–7.65< 0.001Note. High COVID-19-related disruption was associated with higher adjusted odds of risk of AUD compared with low disruption (aOR = 1.93, 95% CI: 1.10–3.39; p = 0.021), whereas moderate disruption was not statistically significant (aOR = 1.17, 95% CI: 0.70–1.97; p = 0.545). Table 2Factors associated with risk of AUD.VariableaOR95% CIp-valueCOVID-19-related disruption0.049 LowReference Moderate1.170.70–1.970.545 High1.931.10–3.390.021Age, years0.0003 20–295.722.26–14.45< 0.001 30–393.391.43–8.050.006 40–493.101.37–7.010.007 50–590.790.31–2.030.628 60–69ReferenceSex< 0.001 MaleReference Female2.901.85–4.54< 0.001Residential area0.993 UrbanReference Rural1.000.57–1.750.993Marital status0.236 MarriedReference Single/separated/divorced/widowed1.390.81–2.400.236Education0.137 ≥ CollegeReference ≤ High school1.430.89–2.300.137Employment status0.012 UnemployedReference Employed2.081.18–3.690.012Household income0.821 HighReference Low1.220.59–2.550.589 Moderate1.140.70–1.860.590Depressive symptoms< 0.001 NoReference Yes4.692.88–7.65< 0.001Note. High COVID-19-related disruption was associated with higher adjusted odds of risk of AUD compared with low disruption (aOR = 1.93, 95% CI: 1.10–3.39; p = 0.021), whereas moderate disruption was not statistically significant (aOR = 1.17, 95% CI: 0.70–1.97; p = 0.545). Factors associated with risk of AUD. Note. High COVID-19-related disruption was associated with higher adjusted odds of risk of AUD compared with low disruption (aOR = 1.93, 95% CI: 1.10–3.39; p = 0.021), whereas moderate disruption was not statistically significant (aOR = 1.17, 95% CI: 0.70–1.97; p = 0.545). Table 3 presents stratified analyses by residential area and age group. In region-stratified analyses, high disruption was not statistically significant among urban residents (aOR = 1.66, 95% CI: 0.89–3.07; p = 0.110), whereas the estimate among rural residents was large but imprecise (aOR = 7.13, 95% CI: 1.46–34.68; p = 0.015). In age-stratified analyses, high disruption was associated with higher odds of risk of AUD among adults younger than 50 years (aOR = 2.04, 95% CI: 1.10–3.79; p = 0.024), but not among adults aged 50 years or older (aOR = 1.43, 95% CI: 0.35–5.79; p = 0.618). Formal interaction tests did not show statistically significant effect modification by residential area (p = 0.284) or age group (p = 0.811). Table 3Subgroup analyses of COVID-19-related disruption and risk of AUD.SubgroupDisruption categoryaOR95% CIp-valueResidential area UrbanModerate vs. low1.140.65–2.010.645 UrbanHigh vs. low1.660.89–3.070.110 RuralModerate vs. low1.860.43–8.010.406 RuralHigh vs. low7.131.46–34.680.015Age group Age < 50 yearsModerate vs. low1.110.62–1.980.734 Age < 50 yearsHigh vs. low2.041.10–3.790.024 Age ≥ 50 yearsModerate vs. low1.210.36–3.990.759 Age ≥ 50 yearsHigh vs. low1.430.35–5.790.618Models were adjusted for sex, age, marital status, education, employment status, household income, and depressive symptoms, except for the stratifying variable. Age-stratified models were adjusted for sex, residential area, marital status, education, employment status, household income, and depressive symptoms. Formal interaction tests did not show statistically significant interaction by region (p = 0.284) or age group (p = 0.811). Table 3Subgroup analyses of COVID-19-related disruption and risk of AUD.SubgroupDisruption categoryaOR95% CIp-valueResidential area UrbanModerate vs. low1.140.65–2.010.645 UrbanHigh vs. low1.660.89–3.070.110 RuralModerate vs. low1.860.43–8.010.406 RuralHigh vs. low7.131.46–34.680.015Age group Age < 50 yearsModerate vs. low1.110.62–1.980.734 Age < 50 yearsHigh vs. low2.041.10–3.790.024 Age ≥ 50 yearsModerate vs. low1.210.36–3.990.759 Age ≥ 50 yearsHigh vs. low1.430.35–5.790.618Models were adjusted for sex, age, marital status, education, employment status, household income, and depressive symptoms, except for the stratifying variable. Age-stratified models were adjusted for sex, residential area, marital status, education, employment status, household income, and depressive symptoms. Formal interaction tests did not show statistically significant interaction by region (p = 0.284) or age group (p = 0.811). Subgroup analyses of COVID-19-related disruption and risk of AUD. Models were adjusted for sex, age, marital status, education, employment status, household income, and depressive symptoms, except for the stratifying variable. Age-stratified models were adjusted for sex, residential area, marital status, education, employment status, household income, and depressive symptoms. Formal interaction tests did not show statistically significant interaction by region (p = 0.284) or age group (p = 0.811). Supplementary Table S1 shows the distribution of covariates across COVID-19-related disruption categories. Age, employment status, and depressive symptoms differed significantly across disruption levels. Supplementary Table S2 presents stepwise logistic regression models, which showed that the association between high disruption and risk of AUD remained in the same direction across crude, partially adjusted, and fully adjusted models. Supplementary Table S3 presents sensitivity analyses. The results were generally consistent with the main findings when the disruption score was modeled continuously, when high disruption was compared with low or moderate disruption combined, and when depressive symptoms were excluded from the adjustment set. Discussion In response to the rapid global spread of COVID-19, governments worldwide implemented strict lockdown measures. These interventions were accompanied by abrupt changes in individuals’ work environments and leisure activities, as well as changes in family dynamics and daily routines. The potential health consequences of such lockdown measures have become a prominent topic in recent public health research18,19. In particular, studies have explored increases in alcohol and substance misuse during the pandemic period20,21. While there has been considerable interest in understanding the social and economic disruptions caused by COVID-19 and changes in alcohol consumption during the lockdown period, few studies have directly examined the link between these factors. This study aimed to assess how pandemic-related disruptions to occupational roles, family responsibilities, and leisure activities were associated with the risk of alcohol use disorder (AUD). The results of the present study indicate a significant association between COVID-19–related social activity disruption and the prevalence of AUD risk. Individuals who reported a high level of disruption showed higher adjusted odds of screening positive for risk of AUD compared with those reporting low disruption. Sudden changes in daily routines—such as economic hardship and the resulting decline in social status within work and family contexts—may elevate stress, contribute to depressive symptoms, and ultimately increase alcohol consumption, as supported by previous studies22. These findings can be interpreted through the lens of stress-coping theory, which posits that individuals experiencing role disruption and loss of social structure may turn to alcohol as a maladaptive coping mechanism to manage psychological distress23,24. Furthermore, the disruption of social capital—defined as the networks of relationships and norms that facilitate collective action—may have reduced individuals’ access to social support during the pandemic, thereby increasing vulnerability to harmful drinking behaviors25. Sociodemographic factors were also associated with risk of AUD. In the region-stratified analysis, the estimate among rural residents experiencing high disruption was large, but the confidence interval was wide, indicating substantial statistical uncertainty due to the smaller rural subgroup and limited number of AUD cases26. In contrast, the association among urban residents was not statistically significant. Moreover, the formal interaction test by residential area was not statistically significant27. Therefore, the region-stratified findings should be interpreted cautiously and should not be considered definitive evidence that the association differs between urban and rural residents. Future studies with larger and more geographically diverse samples are needed to clarify whether residential context modifies the relationship between pandemic-related role disruption and risk of AUD28,29. Younger age groups showed higher adjusted odds of risk of AUD compared with adults aged 60–69 years. Moreover, the association between social activity disruption and AUD risk was particularly pronounced among participants under 50, suggesting that younger adults may be more susceptible to the psychological and social impacts of the pandemic. Employment status was another significant factor. The odds of AUD risk were significantly higher among employed individuals than the unemployed, potentially reflecting increased work-related stress during the pandemic. Prior research has also reported that employed individuals consumed more alcohol than the unemployed during this period, underscoring the role of workplace stress as a key risk factor for problematic alcohol use30. Age-stratified analyses suggested that the association between high COVID-19-related disruption and risk of AUD was more evident among adults younger than 50 years. This may reflect greater disruption to work, social, and leisure roles among younger and middle-aged adults during the pandemic. However, the formal interaction test by age group was not statistically significant, indicating that subgroup differences should be interpreted cautiously rather than as definitive evidence of age-related effect modification31. These findings offer important guidance for policy interventions aimed at mitigating alcohol-related harms, particularly through the implementation of targeted strategies that consider the specific vulnerabilities of different age groups and employment contexts32. In addition to preventing and treating AUD risk, the expansion of telehealth-based substance use disorder services during public health crises may help maintain access to care for individuals at risk33. Given the cross-sectional nature of this study, causality cannot be inferred. It is possible that individuals with higher alcohol use or at risk for AUD experienced greater disruption in their social activities, suggesting potential bidirectionality between alcohol use and social disruption. Limitations Several limitations should be noted. First, because this study used a cross-sectional design, temporal ordering and causality cannot be established. Second, both COVID-19-related disruption and alcohol use were assessed using self-reported survey responses, which may be subject to recall bias and social desirability bias. Third, the AUDIT-K is a screening instrument and does not provide a clinical diagnosis of AUD; therefore, the outcome should be interpreted as risk of AUD rather than confirmed AUD. Fourth, although we adjusted for major sociodemographic characteristics and depressive symptoms, residual confounding by unmeasured factors such as prior drinking patterns, social support, occupation type, household composition, and other substance use may remain. Fifth, the analysis was conducted without survey weights because individual-level sampling weights were not available, and the sample was drawn exclusively from Chungcheongnam-do, which may limit generalizability to other regions of Korea with different cultural and demographic characteristics. Future studies should expand the scope of analysis to include diverse regions across the country. Conclusion This study found that high COVID-19-related disruption of occupational, social/leisure, and family roles was associated with increased odds of risk of AUD among adults in Chungcheongnam-do, South Korea. Stratified analyses suggested that this association may be more evident among adults younger than 50 years, although formal interaction testing did not confirm significant age-related effect modification. Region-stratified findings, particularly among rural residents, should be interpreted cautiously because of imprecise estimates and the absence of statistically significant interaction by residential area. These findings highlight the need for public health strategies that maintain social connectedness, support individuals experiencing substantial role disruption, and identify those vulnerable to high-risk drinking during periods of widespread social disruption. Further longitudinal studies are needed to confirm these findings across diverse regions and populations and to clarify the temporal relationship between social disruption and risk of AUD. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1