Factors influencing local residents’ attitudes toward mental health welfare centers Authors: Hyun Woo Jung, Jae Hyun Kim, Un Sun Chung, Won Jae Choi, Jung Jae Lee Published: 2026 ABSTRACT AbstractThis study examines the factors influencing the Not in My Backyard (NIMBY) attitude toward Mental Health Welfare Centers (MHWCs) in South Korea, integrating individual and regional determinants. Using data from the 2022 Mental Health Perception Survey and Korean Statistical Information Service, we apply a Markov Chain Monte Carlo generalized linear multilevel model to assess predictors of NIMBY attitudes. Regional analysis indicates that a higher proportion of old housing is associated with stronger NIMBY attitudes, whereas higher mortality rates are linked to lower opposition, suggesting that neighborhood conditions and community-level health vulnerability shape public acceptance of mental health services. At the individual level, older age, poor physical health, negative perceptions of mental illness, and exposure to MHWC advertisements were associated with higher NIMBY attitudes, while internet-based information and personal recommendations were linked to lower opposition. The study highlights the limitations of traditional public awareness campaigns and underscores the importance of tailored, trust-based communication strategies that consider local context. These findings provide policy-relevant insights for promoting community-based mental health care.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58253-z.Keywords: Not in My Backyard, Mental Health Welfare Centers, Multilevel Modeling, South KoreaSubject terms: Health care, Health humanities, Medical humanities, Psychology, Psychology Introduction South Korea has recorded one of the highest suicide rates among OECD countries over the past two decades1; however, the utilization rate of mental health services in Korea is only 12.1%, considerably lower than that of Canada (46.6%) and the United States (43.1%), and even lower than Japan (20%), which shares similar Northeast Asian cultural characteristics2. The most frequently reported reasons for not seeking mental health care despite experiencing mental illness include believing that the problem can be managed personally, fear of social stigma, and uncertainty about where to obtain help3. These responses may reflect the broader social and cultural context in which mental health services are perceived and accessed, potentially shaping public attitudes toward community-based mental health facilities4. Within this broader context, the acceptability of Mental Health Welfare Centers (MHWCs) among local residents plays a crucial role in the effective delivery of community-based mental health services. MHWCs serve not only as treatment facilities but also as community hubs that provide prevention programs, crisis intervention, and psychosocial support for residents5–7. However, negative perceptions of mental illness among community members may generate resistance to the establishment of such facilities, potentially hindering the expansion of community mental health services. Moreover, such resistance is not unique to mental health services and is widely recognized as part of a broader social phenomenon. This form of opposition has long been described as the NIMBY (Not in My Backyard) phenomenon. Previous research has identified various determinants of NIMBY attitudes toward community facilities, including perceived safety risks, stigma toward target populations, distrust in public institutions, perceived unfairness in facility placement, and potential impacts on property values4,8–11. These concerns often reflect broader social perceptions rather than direct experiences with the facilities themselves. Likewise, more recent studies suggest that public acceptance of socially sensitive infrastructure can change following major social or environmental events, indicating that contextual factors may shape residents’ attitudes beyond individual demographic characteristics12. From a theoretical perspective, these dynamics can be understood through an ecological framework, which conceptualizes individual attitudes as embedded within broader social and community environments13. Under this perspective, residents’ responses to community facilities are shaped not only by personal perceptions and information sources but also by structural characteristics of the surrounding community13. For example, neighborhood conditions such as a high proportion of old housing stock may exacerbate residents’ concerns about further deterioration in neighborhood image, perceived property values, and associated stigma. In addition, such conditions may reflect a higher concentration of long-term residents, whose stronger place attachment may lead to greater resistance to facilities perceived as degrading the character of their neighborhood14. Similarly, broader demographic and health indicators, such as regional mortality rates, may reflect community vulnerability and influence the perceived need for local health and social services15. Lastly, regional population size may capture broader urban–rural differences, including levels of social cohesion, perceived crowding, and competition for local resources, and may also be associated with residents’ NIMBY attitudes16. In Korea, the historical development of mental health care may further contribute to residents’ attitudes toward community-based mental health services. Unlike many Western countries that actively pursued deinstitutionalization beginning in the late twentieth century, Korea maintained an institutionalized psychiatric care system for a prolonged period, characterized by a relatively large number of psychiatric beds and long hospital stays17,18. Although the Mental Health and Welfare Act was revised in 2016 to strengthen patient rights and promote community-based care19,20, negative perceptions of mental illness among local residents remain prevalent21–23. These attitudes may contribute to resistance toward the establishment of mental health facilities within residential areas. Regional context may also play a decisive role in shaping attitudes toward mental health services. In Korea, substantial regional disparities exist in mental health outcomes; for instance, suicide rates in provincial areas often significantly exceed those in major metropolitan cities. Specifically, Chungcheongnam-do Province (CDP) has recorded the highest suicide rates among inland regions in recent years24, representing a critical area where the demand for mental health infrastructure is most urgent. Therefore, examining community attitudes within this specific region provides a vital case study for understanding how regional vulnerability and service needs intersect to influence public acceptance—a perspective that may be overlooked in aggregate national-level analyses. Therefore, this study aims to identify factors associated with NIMBY attitudes toward MHWCs by integrating both individual-level perceptions and regional-level contextual factors in CDP. Using data from the 2022 Mental Health Perception Survey and regional statistics from the Korean Statistical Information Service (KOSIS), we employ a Markov Chain Monte Carlo generalized linear multilevel model (MCMC GLMM) to examine determinants of residents’ attitudes toward the establishment of mental health centers. By simultaneously considering individual characteristics and regional conditions, this study seeks to provide a more comprehensive understanding of community acceptance of mental health services and to contribute evidence for policies aimed at promoting community-based mental health care. Study data and methods Data sources This study utilized individual-level data from the 2022 Mental Health Perception Survey (MHPS) and regional-level data from the Korean Statistical Information Service (KOSIS). The MHPS targeted residents aged 20 to 69 in CDP, South Korea, which consists of 15 regions, including 8 cities and 7 counties. MHPS achieved representativeness through proportional quota sampling based on the Ministry of the Interior and Safety’s Resident Registration Population Statistics, considering gender, age, and region. The MHPS sample size was 1,000, with 610 participants surveyed online and the remaining 390 through personal interviews. The MHPS was conducted by the Chungcheongnam-do Mental Health Welfare Center. The other dataset, KOSIS, managed by the Korean National Statistical Office, is the most comprehensive statistical database in Korea, providing 343 types of statistics produced by 87 organizations. It publishes 100 major indicators reflecting Korea’s economic, social, and health conditions, such as birth and death rates. Although KOSIS only provides macro-level summary statistics, it offered sufficient regional-level statistics for this study. Regional-level data extracted from KOSIS were linked to the regional variables in the MHPS dataset. The regional-level data consists of summary statistics for 15 administrative districts: Asan (AS) city, Boryeong (BR) city, Cheonan (CN) city, Gongju (GJ) city, Gyeoryong (GY) city, Nonsan (NS) city, Seosan (SS) city, Buyeo (BY) county, Cheongyang (CY) county, Dangjin (DJ) city, Geumsan (GS) county, Hongseong (HS) county, Seocheon (SC) county, Taean (TA) county, Yesan (YS) county. This study received approval from the Institutional Review Board of Dxxxxxx oooooooooo (approval number: DKUH 2025-02-014). Variables The dependent variable is the NIMBY attitude, assessed using a 5-point scale for the statement, “I am okay with a facility for people with mental illnesses being established in my neighborhood” (Table 1). Table 1Definitions of variables.VariableCategoryCodeDependent variableNot in my back yard attitude0: Strongly positive; 1: Positive; 2: Normal; 3: Negative; 4: Strongly negativeIndependent variableIndividual levelAge group0: 20–29; 1: 30–39; 2: 40–49; 3: 50–59; 4: 60–69Subjective mental health0: Good; 1: Normal; 2: BadSubjective physical health0: Good; 1: Normal; 2: BadInformation sourceInternet (0: No; 1: Yes)Advertisement (0: No; 1: Yes)Public release (0: No; 1: Yes)Promotional material (0: No; 1: Yes)Education or campaign (0: No; 1: Yes)Recommendation from an acquaintance (0: No; 1: Yes)Attitudes toward mental illnessCollaborationI can work with those who experienced mental health issues (0: Strongly positive; 1: Positive; 2: Normal; 3: Negative; 4: Strongly negative)PerceptionIndividuals with mental disorders are often perceived as more dangerous than those without (0: Strongly positive; 1: Positive; 2: Normal; 3: Negative; 4: Strongly negative)InclusivityAnyone can experience mental health issues (0: Strongly positive; 1: Positive; 2: Normal; 3: Negative; 4: Strongly negative)PreventionSuicide can be prevented through counseling and support (0: Strongly positive; 1: Positive; 2: Normal; 3: Negative; 4: Strongly negative)Region levelNumber of local residentsContinuous (log-transformed)Proportion of old housing in the local areaNumber of houses over 30 years old ÷ Total number of houses × 100; continuousMortality rateNumber of deaths ÷ Population × 100,000; continuous (log-transformed) Definitions of variables. The independent variables consist of two levels: individual level and regional level. At the individual level, variables include gender, age group, subjective mental health, subjective physical health, sources of information about the MHWCs, and attitudes toward mental health and psychiatric patients. Notably, the operationalization and categorization of these individual-level variables were consistent with the original survey instrument to preserve the integrity of the data as initially collected. Specifically, the category for MHWC-related information sources was assessed with the question, “Where have you obtained information about the Mental Health Welfare Center?” Respondents could choose from the following options: Internet (homepage, Social Networking Service (SNS) searching, Youtube, etc.), advertisements (newspaper, television, radio, public transit), public release by public institutions, promotional material (poster, leaflet, brochure), education or campaigns, and recommendations from acquaintances. Since multiple selections were allowed for this question, the data were transformed into dummy variables. Additionally, the survey included four statements to assess attitudes and perceptions regarding mental health and psychiatric patients: “Can work with people who have experienced mental health issues,” “Individuals with mental disorders are often perceived as more dangerous than those without,” “Anyone can experience mental health issues,” and “Suicide can be prevented through counseling and support.” Responses to these statements were measured on a 5-point scale. A preliminary reliability analysis for these items yielded a Cronbach’s alpha of 0.606, suggesting that they represent distinct conceptual dimensions of stigma rather than a single unidimensional scale. Therefore, instead of using a composite score, each item was treated as an independent predictor to capture the unique influence of specific perceptions on NIMBY attitudes. In fact, the study previously examined socioeconomic status factors, such as gender, educational level, job status, marital status, family type, religion, and household income. These variables were not retained in the final model for parsimony, although they were examined in preliminary analyses and did not show meaningful associations with NIMBY attitudes. This approach is consistent with prior studies suggesting that these factors may not have a substantial influence25. At the regional level, three variables were selected for multilevel modeling: the number of local residents, the proportion of old housing in the area, and the mortality rate. These variables were selected based on their theoretical relevance to regional context, reflecting demographic structure, neighborhood conditions, and population health status. Prior research has demonstrated that neighborhood environments influence individual attitudes and behaviors26, that physical and socioeconomic characteristics of residential environments are closely associated with social perceptions and community responses, particularly in relation to neighborhood disorder and stigma27–29, and that community-level health conditions, such as mortality, reflect broader contextual vulnerability shaped by regional environments17. A preliminary screening of available regional indicators was conducted to ensure empirical variability and to avoid multicollinearity; however, the final selection was not based solely on statistical significance. The regional-level data were integrated with individual-level data by matching administrative districts to individuals’ residential locations in the data integration process. Detailed calculations and coding for each variable are presented in Table 1. Statistical analyses This study applied multilevel linear regression analysis to data consisting of 1,000 individuals (level 1) nested within 15 regions (level 2) to hierarchically identify the determinant factors of NIMBY attitude. Traditional regression methods face challenges in distinguishing whether variations among individuals arise from individual-level differences or disparities between groups. This limitation violates the independence assumption required by standard regression approaches. Multilevel regression, however, addresses these issues by accounting for the hierarchical structure of the data. Four models were examined in this study. Model 1 was a null model that included only the intercept without any independent variables to establish a baseline for variation in NIMBY attitude across the two levels. Model 2 incorporated individual-level factors as predictors, building on the baseline of the null model. Model 3 introduced regional-level factors based on the null model. Lastly, all individual- and regional-level factors were included in Model 4. The sample size at the higher level of the hierarchy has been a persistent concern in multilevel modeling25. Maximum likelihood (ML) estimation methods tend to perform well with large sample sizes but exhibit suboptimal performance with smaller samples, particularly when the number of clusters is limited. A small sample size in multilevel models can lead to biased variance component estimates30. It is widely acknowledged that a minimum of 50 clusters is required to ensure reliable results in multilevel modeling25. To address the problem of small cluster sizes, MCMCglmm-based estimation methods have been proposed as a viable alternative to likelihood-based approaches, offering robust performance even in models with limited clusters30,31. Given the small number of regions (n = 15) in this study, Bayesian MCMCglmm estimation was chosen over maximum likelihood (ML) methods to reduce potential bias in estimating variance components associated with regional-level predictors32. Bayesian multilevel models were estimated using the MCMCglmm package in R. Weakly informative priors were specified for the variance components (R=list(V = 1, fix = 1), G=list(G1 = list(V = 1, nu = 0), G2 = list(V = 1, nu = 0))). Each model was estimated using a single Markov chain with 55,000 iterations, including a burn-in of 5,000 iterations and a thinning interval of 20, yielding an effective sample size of approximately 2,500 posterior samples. Because the models were estimated using a single Markov chain, Gelman–Rubin diagnostics were not applicable. Convergence was therefore evaluated based on effective sample size diagnostics and the acceptance ratios reported by the MCMCglmm output. Acceptance ratios for the liability set ranged between 0.2 and 0.5, indicating adequate mixing of the Markov chain. Missing values were handled using complete-case analysis within the MCMCglmm estimation procedure, whereby observations with missing values in the variables included in each model were excluded during model fitting. All data coding and descriptive analyses were performed using Stata/MP version 18.0 (StataCorp LP, College Station, TX, USA). Multilevel modeling with Bayesian MCMC glmm estimation was conducted in RStudio (RStudio, PBC, Boston, MA, USA) using the MCMCglmm package. The R code used for model estimation is provided in the Supplementary Material. Study results Descriptive statistics at the individual level The distribution of NIMBY attitudes among residents in CDP exhibited a normal distribution, with the majority of responses classified as ‘Normal’ (408, 40.8%) (Table 2). However, the distribution was slightly skewed toward the negative side, with ‘Negative’ responses accounting for 297 (29.7%) and ‘Positive’ responses for 184 (18.4%). Among the demographic characteristics, only age groups demonstrated significant differences. Regarding health status, both subjective mental health and physical health were found to be significantly associated with NIMBY attitudes. In the category of “sources of information on MHWCs,” individuals who received information through advertisements exhibited more negative attitudes (42.7%) compared to those who did not (26.9%). Lastly, within the category of opinions on mental health and psychiatric patients, the mean scores for “Collaboration,” “Inclusivity,” and “Prevention” increased significantly as NIMBY attitudes shifted from strongly positive to strongly negative, indicating that more negative views in these areas are associated with more negative NIMBY attitudes. Table 2Individual level of characteristics in relation to the NIMBY.Individual characteristicsNot in my back yard attitudeSPPositiveNormalNegativeSNChi2(AN)n (%)n (%)n (%)n (%)n (%)n (%)Age group (years)20–2913 (8.5)48 (31.4)53 (34.6)27 (17.7)12 (7.8)0.00030–3911 (6.3)46 (26.1)63 (35.8)35 (19.9)21 (11.9)40–496 (2.7)41 (18.5)106 (48.0)44 (19.9)24 (10.9)50–598 (3.4)28 (12.0)96 (41.0)95 (40.6)7 (3.0)60–695 (2.3)21 (9.7)90 (41.7)96 (44.4)4 (1.9)Subjective mental healthGood24 (4.1)102 (17.4)236 (40.3)197 (33.6)27 (4.6)0.000Normal11 (3.5)53 (16.8)140 (44.3)81 (25.6)31 (9.8)Bad8 (8.2)29 (29.6)32 (32.6)19 (19.4)10 (10.2)Subjective physical healthGood24 (4.8)91 (18.3)194 (39.0)164 (33.0)24 (4.8)0.001Normal9 (2.5)65 (18.0)160 (44.2)104 (28.7)24 (6.6)Bad10 (7.1)28 (19.8)54 (38.3)29 (20.6)20 (14.2)Information sourceInternetNo33 (4.1)137(16.9)335 (41.4)246 (30.5)57 (7.1)0.143Yes10 (5.2)47 (24.5)73 (38.0)51 (26.6)11 (5.7)AdvertisementNo41 (5.0)163 (19.8)337 (41.0)221 (26.9)60 (7.3)0.000Yes2 (1.1)21 (11.8)71 (39.9)76 (42.7)8 (4.5)Public releaseNo37 (4.7)147 (18.7)320 (40.6)228 (29.0)55 (7.0)0.66Yes6 (2.8)37 (17.4)88 (41.3)69 (32.4)13 (6.1)Promotional materialNo39 (4.4)165 (18.8)361 (41.1)249 (28.3)65 (7.4)0.052Yes4 (3.3)19 (15.7)47 (38.8)48 (39.7)3 (2.5)Education orcampaignNo38 (4.0)176 (18.6)386 (40.8)279 (29.5)67 (7.1)0.201Yes5 (9.3)8 (14.8)22 (40.7)18 (33.3)1 (1.9)Recommendation from an acquaintanceNo39 (4.2)171 (18.5)371 (40.2)277 (30.0)65 (7.1)0.599Yes4 (5.2)13 (16.9)37 (48.1)20 (25.9)3 (3.9)Attitudes toward mental illnessCollaborationMean2.4 (0.08)2.8 (0.04)3.4 (0.04)3.6 (0.07)4.4 (0.26)0.000PerceptionMean3.1 (0.08)3.2 (0.04)3.1 (0.05)3.1 (0.1)3.2 (0.29)0.423InclusivityMean2.9 (0.05)3.1 (0.04)3.4 (0.06)3.5 (0.11)4.2 (0.46)0.000PreventionMean2.8 (0.06)3.1 (0.04)3.4 (0.06)3.4 (0.12)4.0 (0.27)0.000Total43 (100.0)184(100.0)408(100.0)297(100.0)68(100.0) Individual level of characteristics in relation to the NIMBY. 137 (16.9) Education or campaign 184 (100.0) 408 (100.0) 297 (100.0) 68 (100.0) Descriptive statistics at the regional level Table 3 presents descriptive statistics for regional variables across 15 districts, consisting of 8 cities and 7 counties. The sample size data were extracted from the Mental Health and Psychological Survey (MHPS), while the remaining variables were sourced from KOSIS. All variables demonstrated variability across districts. The sample size corresponds to the number of local residents in each district. Among the districts, Cheonan (CN) city, the largest metropolitan area in CDP, had the largest population and sample size, while Cheongyang (CY) county had the smallest population and sample size. The proportion of old housing was generally higher in counties and lower in cities. Buyeo (BY) county exhibited the highest proportion of old housing at 49.2%, whereas Gyeoryong (GY) city had the lowest at 5.5%. Similarly, mortality rates were higher in counties and lower in cities. Seocheon (SC) county recorded the highest mortality rate at 1,877.6 per 100,000, while Gyeoryong (GY) city had the lowest at 507.1 per 100,000. Table 3Descriptive statistics of the regional variables.Sample sizeNumber of local residentsProportion of old housing in the local areaMortality rate(per 100,000)CityAS162363,19711.6%662.4BR4496,51132.6%1276.0CN330687,95113.7%560.8DJ75171,19218.0%897.4GJ45106,03835.1%1323.2GY2242,4165.5%507.1NS51116,85239.3%1421.3SS85177,94720.9%823.7CountyBY2661,48149.2%1849.8CY1329,76142.8%1697.7GS2252,39939.7%1527.8HS43100,33929.6%1193SC2049,50946.0%1877.6TA2760,65429.4%1227.3YS3576,96735.2%1531.9Note: AS: Asan City; BR: Boryeong City; CN: Cheonan City; GJ: Gongju City; GY: Gyeoryong City; NS: Nonsan City; SS: Seosan City; BY: Buyeo County; CY: Cheongyang County; DJ: Dangjin City; GS: Geumsan County; HS: Hongseong County; SC: Seocheon County; TA: Taean County; YS: Yesan County. Descriptive statistics of the regional variables. Note: AS: Asan City; BR: Boryeong City; CN: Cheonan City; GJ: Gongju City; GY: Gyeoryong City; NS: Nonsan City; SS: Seosan City; BY: Buyeo County; CY: Cheongyang County; DJ: Dangjin City; GS: Geumsan County; HS: Hongseong County; SC: Seocheon County; TA: Taean County; YS: Yesan County. Table 4 summarizes the distribution of “Not in My Back Yard” (NIMBY) attitudes across 15 districts, categorized into cities and counties. Neutral attitudes were the most frequently observed across most districts. In contrast, strongly positive (SP) and strongly negative (SN) attitudes showed relatively lower proportions overall. Cities generally displayed higher proportions of positive attitudes, whereas counties exhibited higher proportions of strongly negative and neutral attitudes. For instance, strongly positive attitudes were more prevalent in CN city (4.85%), AS city (6.17%), and DJ city (5.33%). Conversely, counties such as CY (0.0%) and SC (0.0%) had no respondents in the strongly positive category. Negative and strongly negative attitudes were more concentrated in certain districts. Notably, CN city recorded the highest proportion of strongly negative respondents (34.01%), followed by NS city (8.08%) and AS city (8.75%). Among counties, BY (5.72%) and HS (6.06%) demonstrated notable proportions of negative attitudes despite their smaller sample sizes. Overall, cities tended to exhibit more evenly distributed responses across categories, while counties displayed a stronger skew toward neutral and strongly negative responses. Table 4Regional attitude of NIMBY.Not in my back yard attitudeSPPositiveNormalNegativeSNTotaln (%)n (%)n (%)n (%)n (%)CityAS10 (23.26)34 (18.48)79 (19.36)26 (8.75)13 (19.12)162 (16.2)BR1 (2.33)5 (2.72)11 (2.7)21 (7.07)6 (8.82)44 (4.4)CN16 (37.21)55 (29.89)134 (32.84)101 (34.01)24 (35.29)330 (33.0)DJ4 (9.3)21 (11.41)34 (8.33)13 (4.38)3 (4.41)75 (7.5)GJ2 (4.65)10 (5.43)11 (2.7)17 (5.72)5 (7.35)45 (4.5)GY0 (0.0)8 (4.35)11 (2.7)3 (1.01)0 (0.0)22 (2.2)NS1 (2.33)9 (4.89)16 (3.92)24 (8.08)1 (1.47)51 (5.1)SS2 (4.65)18 (9.78)41 (10.05)21 (7.07)3 (4.41)85 (8.5)CountyBY1 (2.33)0 (0.0)7 (1.72)17 (5.72)1 (1.47)26 (2.6)CY0 (0.0)2 (1.09)4 (0.98)6 (2.02)1 (1.47)13 (1.3)GS1 (2.33)5 (2.72)9 (2.21)6 (2.02)1 (1.47)22 (2.2)HS1 (2.33)6 (3.26)13 (3.19)18 (6.06)5 (7.35)43 (4.3)SC0 (0.0)1 (0.54)7 (1.72)9 (3.03)3 (4.41)20 (2.0)TA2 (4.65)4 (2.17)12 (2.94)7 (2.36)2 (2.94)27 (2.7)YS2 (4.65)6 (3.26)19 (4.66)8 (2.69)0 (0.0)35 (3.5)Total43 (100.0)184 (100.0)408 (100.0)297 (100.0)68 (100.0)1000 (100.0)Note: AS: Asan City; BR: Boryeong City; CN: Cheonan City; GJ: Gongju City; GY: Gyeoryong City; NS: Nonsan City; SS: Seosan City; BY: Buyeo County; CY: Cheongyang County; DJ: Dangjin City; GS: Geumsan County; HS: Hongseong County; SC: Seocheon County; TA: Taean County; YS: Yesan County. Regional attitude of NIMBY. Note: AS: Asan City; BR: Boryeong City; CN: Cheonan City; GJ: Gongju City; GY: Gyeoryong City; NS: Nonsan City; SS: Seosan City; BY: Buyeo County; CY: Cheongyang County; DJ: Dangjin City; GS: Geumsan County; HS: Hongseong County; SC: Seocheon County; TA: Taean County; YS: Yesan County. Determinant factors of NIMBY attitude using multilevel linear regression Table 5 presents the factors associated with NIMBY attitudes across four successive multilevel models. Model 1 is a null model that excludes independent variables, serving as a baseline to evaluate the variance attributed to random effects, such as regional-level differences. This model provides a foundation for comparing subsequent models that include explanatory variables. Table 5Results of multilevel linear regression analysis for the NIMBY attitude.VariablesCategoriesNot in my back yard attitudeModel 1Model 2Model 3Model 4PM(95% CrI)PM(95% CrI)PM(95% CrI)PM(95% CrI)Age group(20–29)30–390.16 (-0.038, 0.389)0.167 (-0.04, 0.377)40–490.196 (-0.022, 0.396)0.204 (-0.001, 0.404)50–590.159 (-0.05, 0.396)0.159 (-0.061, 0.378)60–690.252 (0.019, 0.466)0.24 (0.009, 0.444)Subjective mental health (Good)Normal-0.056 (-0.21, 0.104)-0.041 (-0.227, 0.114)Bad-0.419 (-0.722, 0.111)-0.414 (-0.743, -0.144)Subjective physical health (Good)Normal0.042 (-0.123, 0.211)0.021 (-0.131, 0.205)Bad0.262 (-0.023, 0.506)0.246 (-0.022, 0.513)Information sourceInternet (No)Yes-0.151 (-0.321, 0.016)-0.15 (-0.317, 0.005)Advertisement (No)Yes0.207 (0.021, 0.38)0.207 (0.021, 0.387)Public release (No)Yes0.036 (-0.128, 0.194)0.02 (-0.153, 0.171)Promotional material (No)Yes-0.037 (-0.259, 0.158)-0.029 (-0.219, 0.192)Education or campaign (No)-0.15 (-0.451, 0.135)-0.149 (-0.448, 0.135)Recommendation from an acquaintance (No)Yes-0.222 (-0.449, 0.019)-0.221 (-0.455, 0.013)Attitudes toward mental illnessCollaborationContinuous0.37 (0.295, 0.448)0.37 (0.291, 0.452)PerceptionContinuous-0.129 (-0.209, -0.049)-0.124 (-0.204, -0.046)InclusivityContinuous0.022 (-0.065, 0.102)0.012 (-0.07, 0.099)PreventionContinuous0.177 (0.095, 0.269)0.18 (0.096, 0.262)Regional population (log)0.165(0.024, 0.318)0.004 (-0.138, 0.157)Old housing (%)0.05(0.028, 0.072)0.045 (0.022, 0.066)Mortality rate (log)-2.457(-3.83, -1.04)-2.465 (-3.888, -1.077)Constants3.20(3.07, 3.34)1.95 (1.62, 2.25)8.42(5.06, 11.9)8.136 (4.53, 11.60)ICCRegional level0.0030.002< 0.001< 0.001Region > Individual level< 0.001< 0.001< 0.001< 0.001Marginal R-square0.0010.0000.1080.108Conditional R-square0.2430.240.3160.316Note: Values are presented as posterior means (PM) with 95% credible intervals (CrI). pMCMC values are reported in Supplementary material. Models were estimated using Bayesian MCMC with 55,000 iterations, a burn-in of 5,000, and thinning of 20 (effective sample size ≈ 2,500 per chain). ICC = intraclass correlation coefficient. Acceptance ratios for the liability set ranged from 0.2 to 0.5. Regional population (log) denotes the log-transformed number of local residents. Old housing (%) denotes the proportion of houses older than 30 years in the local area. Mortality rate (log) denotes the log-transformed mortality rate per 100,000 population. Results of multilevel linear regression analysis for the NIMBY attitude. Age group (20–29) 0.165 (0.024, 0.318) 0.05 (0.028, 0.072) -2.457 (-3.83, -1.04) 3.20 (3.07, 3.34) 8.42 (5.06, 11.9) Note: Values are presented as posterior means (PM) with 95% credible intervals (CrI). pMCMC values are reported in Supplementary material. Models were estimated using Bayesian MCMC with 55,000 iterations, a burn-in of 5,000, and thinning of 20 (effective sample size ≈ 2,500 per chain). ICC = intraclass correlation coefficient. Acceptance ratios for the liability set ranged from 0.2 to 0.5. Regional population (log) denotes the log-transformed number of local residents. Old housing (%) denotes the proportion of houses older than 30 years in the local area. Mortality rate (log) denotes the log-transformed mortality rate per 100,000 population. However, the intra-class correlation (ICC) values across all four models were consistently very low, indicating that only a very small proportion of the variance in individual NIMBY attitudes was attributable to regional characteristics. Model 2 incorporates individual-level variables into the null model, while Model 3 includes only regional-level variables. A comparison of these models suggests that regional variables contribute to explaining variation in NIMBY attitudes. The results from Model 4, which includes both individual- and regional-level variables, reveal that at the individual level, age groups 40–49 (PM = 0.204; 95% CrI: -0.001, 0.404) and 60–69 (PM = 0.24; 95% CrI: 0.009, 0.444), poor physical health (PM = 0.246; 95% CrI: -0.022, 0.513), information from advertisements (PM = 0.207; 95% CrI: 0.021, 0.387), negative views about collaboration (PM = 0.37; 95% CrI: 0.291, 0.452), and negative perceptions of prevention (PM = 0.18; 95% CrI: 0.096, 0.262) were associated with a higher likelihood of NIMBY attitudes. Conversely, poor mental health (PM = -0.414; 95% CrI: -0.743, -0.144), information from the internet (PM = -0.15; 95% CrI: -0.317, 0.005), recommendations from acquaintances (PM = -0.221; 95% CrI: -0.455, 0.013), and positive perceptions (PM = -0.124; 95% CrI: -0.204, -0.046) were associated with a lower likelihood of NIMBY attitudes. At the regional level, the number of local residents was not statistically significant, unlike in Model 3. Additionally, an increase in the proportion of old housing in the area (PM = 0.045; 95% CrI: 0.022, 0.066) was associated with a rise in NIMBY attitudes, while an increase in the mortality rate (PM = -2.465; 95% CrI: -3.888, -1.077) was associated with a decrease in NIMBY attitudes. To ensure the robustness of our regional-level estimates given the 15 level-2 units, we performed sensitivity analysis (Table S5 in Supplementary Material). Discussion The purpose of this study was to identify the multi-level determinants influencing NIMBY attitudes toward MHWCs among local residents in CDP, South Korea. Using MCMC-based GLMM estimation, we examined how individual characteristics and regional contextual factors interact to shape these attitudes. A key finding of this study is that regional-level contextual factors, particularly neighborhood housing conditions and community health status, were associated with variations in NIMBY attitudes. In the Korean context, particularly in CDP, older housing is often associated with the potential for housing reconstruction projects. Areas with a higher concentration of housing over 30 years old are more likely to be considered for redevelopment or reconstruction, which may raise expectations of future property value increases33. In this context, the establishment of MHWCs may be perceived as undermining these expectations, thereby reinforcing resistance among residents27,33. In contrast, higher regional mortality rates were associated with lower NIMBY attitudes. This pattern may reflect the demographic reality of population aging and population decline in the region. As residents are more exposed to death within their communities, they may become more aware of the psychological and emotional challenges associated with these experiences, which may increase the perceived need for MHWCs28. Although the ICC values were very low, indicating that a small proportion of the variance in NIMBY attitudes was attributable to regional clustering, the conditional R-squared was substantially higher than the marginal R-squared. This suggests that the inclusion of random effects contributed meaningfully to the overall explanatory power of the model27,34. Taken together, these findings support the inclusion of both individual- and regional-level factors when examining NIMBY attitudes. At the individual level, the study identified factors including age, subjective mental health, subjective physical health, sources of information, and opinions on mental health and psychiatric patients. Firstly, older residents aged 60–69 were more likely to report higher NIMBY attitudes. This finding suggests that older residents may harbor more persistent stigma toward MHWCs, potentially shaped by earlier institutional models of mental health care. In South Korea, the deinstitutionalization of psychiatric hospitals was implemented relatively recently, only a decade ago20. Individuals who formed their perceptions during earlier periods may have developed more negative views toward community-based mental health facilities, and these perceptions may have persisted into later life. In addition, this study identifies a clear divergence in how subjective physical and mental health influence NIMBY attitudes. Poor physical health was associated with higher NIMBY attitudes, whereas poorer subjective mental health was linked to lower NIMBY attitudes. This divergence can be interpreted through land-use conflict theory, reflecting competing priorities for limited spatial resources27. Individuals with poor physical health may prioritize general medical infrastructure, such as hospitals or clinics, and thus perceive MHWCs as less aligned with their immediate needs. In contrast, those with poorer mental health are more likely to view MHWCs as accessible services for themselves. These findings suggest that NIMBY attitudes are shaped by the perceived alignment between individual health needs and the function of the facility. In this sense, perceived personal utility plays a central role in determining community acceptance35,36. The study also found that the source of information was significantly associated with NIMBY attitudes. Traditional strategies, including public releases, promotional materials, and mental health education, did not show significant effects, consistent with previous research indicating the limited impact of such conventional approaches37,38. In contrast, residents who obtained information through the internet or from acquaintances were more likely to exhibit lower NIMBY attitudes, suggesting the importance of accessible and trust-based communication channels39,40. However, these associations should be interpreted with caution due to potential endogeneity, as individuals with more favorable attitudes may be more likely to seek such information. Nevertheless, the findings highlight that communication credibility and interpersonal trust play a critical role in shaping public attitudes41,42. Existing opinions on mental health and psychiatric patients were also associated with NIMBY attitudes toward MHWCs. In particular, more negative attitudes toward interacting with individuals with mental illness and less favorable beliefs about mental health care were linked to stronger NIMBY attitudes, whereas more favorable perceptions were associated with lower NIMBY attitudes. These findings are consistent with prior research showing that public perceptions of psychiatric patients play a central role in shaping opposition to mental health facilities4,27,28,35. Together, these results suggest that social distance and misconceptions remain important factors influencing community acceptance. Overall, these results indicate that individual-level perceptions, alongside regional contextual factors, should be considered when examining NIMBY attitudes, particularly in the context of CDP, highlighting the need for policy approaches that address both social attitudes and local conditions. These findings indicate that policy interventions should be tailored to regional needs and local community characteristics. In areas with more deteriorated housing environments, where resistance may be stronger, emphasizing the broader community benefits of MHWCs may be important. In addition, in areas with higher mortality rates, the findings may reflect a greater need for mental health services, suggesting that these regions could be prioritized in the strategic placement of MHWCs. Moreover, traditional communication strategies—such as public releases, promotional materials, and mental health education—did not show significant effects, suggesting limited effectiveness relative to the resources invested. This highlights the need for more targeted and context-sensitive approaches. For older residents, who may hold more persistent stigma shaped by earlier institutional models of mental health care and are less familiar with internet use, interpersonal and trust-based channels may be effective. In contrast, for younger populations, who are more accustomed to internet use, social media and online campaigns may be more effective in improving acceptance. Limitations This study has several limitations. First, the findings may have limited generalizability, as the analysis was based on data from a single region. Although the NIMBY phenomenon is inherently context-specific, caution is required when applying these results to other regions or countries. Second, the relatively small number of regional units (n = 15) may constrain the robustness of multilevel modeling. Although Bayesian MCMC estimation was employed to mitigate this limitation, the possibility of biased variance estimates cannot be fully excluded. Third, the regional-level results were sensitive to variable selection. The inclusion or exclusion of certain variables influenced the magnitude and significance of others, reflecting the inherent instability of models with a limited number of higher-level units. Therefore, the interpretation of regional effects should be approached with caution. Conclusion This study identified key determinants of NIMBY attitudes toward MHWCs among local residents in CDP. Beyond individual characteristics, the findings highlight the importance of information sources, perceptions of mental illness, and regional contextual factors. In particular, NIMBY attitudes appear to be shaped by the alignment between individual needs and the perceived function of the facility, as well as by broader neighborhood conditions. These results underscore the importance of adopting communication strategies that emphasize credibility and lived experience, particularly through internet-based and peer-driven channels. Overall, this study provides empirical evidence to inform policies aimed at enhancing community acceptance of mental health services. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions H.W.J. designed the study, performed the formal analysis, and drafted the manuscript.J.-H.K., U.S.C., and W.J.C. reviewed and revised the manuscript.J.J.L. acquired funding, supervised the study, and reviewed and revised the manuscript.All authors read and approved the final manuscript. Funding This study was funded through a grant from the Korea Health Technology R&D Project, supported by the Korea Health Industry Development Institute (KHIDI) and the Ministry of Health & Welfare, Republic of Korea (Grant No. RS-2024-00438829). Data availability The authors will provide the raw data supporting the conclusions of this article upon request, with no unnecessary restrictions. Declarations Competing interests The authors declare no competing interests. Ethics approval and consent to participate This study involving human participants was reviewed and conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Dankook University (DKUH 2025-02-014). All participants provided written informed consent prior to their inclusion in the study. Role of the Funding Source The funding source had no role in the study design, data collection, analysis, interpretation of data, writing of the manuscript, or the decision to submit the manuscript for publication. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The authors will provide the raw data supporting the conclusions of this article upon request, with no unnecessary restrictions.