Impact of type of reconstructed residence on social participation and mental health of population displaced by disasters

After disasters, people are often forced to reconstruct or move to new residences. This study aimed to reveal the association between the types of reconstructed residences and psychosocial or psychiatric conditions among the population. A total of 1071 adult residents in a coastal town, whose houses were destroyed by the tsunami caused by the Great East Japan Earthquake, enrolled in the study five years after the disaster. The type of reconstructed post-disaster residences (reconstructed on the same site/disaster-recovery public condominium/mass-translocation to higher ground/privately moving to remote areas) and the current psychosocial indicators were investigated. The results revealed that individuals living in public condominiums showed significantly worse scores on the Lubben Social Network Scale-6 (p < 0.0001) and the Center for Epidemiologic Studies Depression Scale (p < 0.0001), and slightly worse scores on the Kessler Psychological Distress Scale (p = 0.035) and the Impact of Event Scale-Revised (p = 0.028). Lower psychosocial indicator scores in the public condominium group were more remarkable in younger adults aged < 65 years. Insomnia evaluated using the Athens Insomnia Scale was not different among the four residential types. In summary, residents moving into disaster-recovery public condominiums are likely to have less social interaction, be more depressed, and may need additional interventions.


Self-reported psychosocial indicators.
To evaluate the current psychosocial and psychiatric conditions of the displaced population, valid scores without missing data for each sub-item of the evaluated psychosocial indicators were obtained from 1048 (97.9%) of the eligible participants for the LSNS-6, 1032 (96.4%) for the K6, 799 (74.6%) for the CES-D, 971 (90.7%) for the IES-R, and 1041 (97.2%) for the AIS. General social interaction activities, estimated from the scores of the LSNS-6, were significantly lower (i.e., worse) among those who moved into disaster-recovery public condominiums (p < 0.0001, analysis of variance [ANOVA]). Of the evaluated psychiatric indicators, the scores of the CES-D (p < 0.0001, ANOVA) were significantly higher (i.e., worse) among those who moved into disaster-recovery public condominiums. The scores of K6 (p = 0.0345, Kruskal-Wallis test) and IES-R (p = 0.0281, Kruskal-Wallis test) were slightly higher (i.e., worse) in the public condominium group; however, the difference was not statistically significant (p < 0.01) in this study. Meanwhile, the AIS scores (p = 0.7572, ANOVA) did not differ according to the type of reconstructed residence. Box and whisker plots for these measures by the types of residence are depicted in Fig. 2 www.nature.com/scientificreports/ plot derived from discriminant analysis that employed scores of the five self-reported psychosocial indicators as covariates and current type of reconstructed residences as the categorical variable is depicted in Fig. 3. The obtained canonical plot showed a separation of the 95% confidence ellipse for public condominium group from ellipses for each of the other three types of current residences. Next, to adjust for the possible confounding effects of the current socio-economic factors on the self-reported psychosocial indicators, an analysis of covariance (ANCOVA), by employing the previous history of living in post-disaster prefabricated temporary housings, the current employment status, and the number of family members per household as covariates, was performed for each psychosocial indicator. The resulting scores of the LSNS-6 (p < 0.0001) and CES-D (p < 0.0001) remained significantly worse in the public condominium group. The scores of K6 (p = 0.0941) and IES-R (p = 0.0244) were slightly higher (i.e., worse) in the public condominium group; however, the difference was not statistically significant. The AIS scores (p = 0.5677) did not differ between the four residential types.
Correlation matrix within the five psychosocial indicators. To investigate the backgrounds behind the absence of difference in the AIS scores between the four types of reconstructed residence, correlation matrices within the five self-reported psychosocial indicators were built for the whole study participants and for those living in disaster-recovery public condominiums. The matrices are shown in Table 2. For both populations, the scores of the LSNS-6 showed no or only weak correlations with the other self-reported indicators. The overall patterns of the correlations among the five indicators were largely the same between the whole study participants and those living in public condominiums. Flow diagram of the study design. Shichigahama Town was one of the closest coastal cities or towns to the epicenter, where more than 10% of the residents lost their homes because of the tsunami. All 2343 residents whose homes were destroyed were initially recruited, of whom 58.2% responded and participated in the present study. Illustrations of the four different types of reconstructed residences five years after the earthquake and tsunami are shown, which were classified into: (1) those which did not move (reconstructed on the same site), (2) disaster-recovery public condominium, (3) mass-translocation to higher ground, and (4) private movement to remote areas. GEJE 2011 Great East Japan Earthquake. www.nature.com/scientificreports/

Demographics and scores of psychosocial indicators.
To evaluate the correlation between each of the evaluated background variables (i.e., sex, age, number of households, employment status, change in income) and the obtained scores of the self-reported psychosocial indicators, Spearman's correlation coefficients (ρ) between the background data and the scores of the five self-reported psychosocial indicators were evaluated. The correlation matrix is shown in Table 3. None of the evaluated background factors were significantly correlated with the evaluated psychosocial or psychiatric measures.

Socio-economic factors and scores of psychosocial indicators.
To evaluate the impact of employment status on the outcomes of this study, the current psychosocial and psychiatric conditions were further evaluated between the four types of residences and the current employment status. The results showed that the associations between psychosocial disturbances and life in disaster-recovery public condominiums were remarkable among the participants who were currently employed, as shown in Table 4. Based on this finding, Spearman's correlation coefficients between employment status (currentt employment, income change) and other background demographics (age, sex, and number of people in the household) were evaluated. The current employment status was found to be significantly correlated with males (ρ = + 0.24, p < 0.0001) and age (ρ = − 0.55, p < 0.0001). The current employment rate in the total enrolled younger adults aged < 65 years was 80.2% (n = 477/595), whereas in older adults aged ≥ 65 years, it was 26.5% (n = 120/453). The aforementioned results imply that the negative psychosocial impact of living in disaster-recovery public condominiums may be more remarkable in the younger age group with a higher employment rate. Based on these findings, comparisons of the scores for the evaluated psychosocial indicators between the four types of residences were performed after stratifying the cohort into younger adults (aged < 65 years at the time of the survey) and older adults (aged ≥ 65 years at the time of the survey). The results showed worse scores in participants living in the disaster-recovery public condominiums for LSNS-6, K6, CES-D, and IES-R, which were all more remarkable in younger adults aged < 65 years than in older adults aged ≥ 65 years, as shown in Table 5.

Discussion
In this cross-sectional study with self-reported questionnaires, including indicators for social interaction activities and mental health conditions, the association between the current type of reconstructed residence and concurrent scores of the indicators was evaluated. The results indicated that moving into disaster-recovery public condominiums was associated with lower levels of social interaction and elevated rate of depressive state. This tendency was observed regardless of the current employment status or the age group, with a cutoff age of 65 years; Table 1. Demographic data, social interaction, and mental health conditions by the current residential types 5 years after the GEJE. The p-values are the results of the analysis of variance, Kruskal-Wallis test, or chi-square test. AIS Athens Insomnia Scale, CES-D Center for Epidemiologic Studies Depression Scale, GEJE 2011 Great East Japan Earthquake, IES-R Impact of Event Scale-Revised, K6 Kessler Psychological Distress Scale, LSNS-6 Lubben Social Network Scale. *Mean ± standard deviation. † Median and interquartile range (25-75 percentile). ‡ Significantly higher or lower in the post-hoc test.  www.nature.com/scientificreports/ however, it was especially remarkable among the younger adult disaster-affected population aged < 65 years. As the social isolation is an established risk of several mental health problems such as depression, the achieved results were reasonable findings 19,20 .
The results regarding moving into disaster-recovery public condominiums being associated with lower communication levels and worse mental health conditions may be explained by several hypotheses. A conceivable theory is that the majority of the residents in the sea-side area affected by the tsunami used to live in the old community and were accustomed to having close life-long relationships with neighbors 21 , which were suddenly destroyed by the disaster. They were unexpectedly forced to separate and relocate several times after the GEJE 22 . These social separation and isolation may have been accelerated by the environment of the public condominium, with its higher levels of security and privacy. Generally, residents living in a house are easily identifiable as being at home or not, and neighbors are able to communicate more easily with the residents 22 . In contrast, people who move to public condominiums find it difficult to be identifiable as being at home, and neighbors may refrain from casually visiting the residents. While the structure of public condominiums may be beneficial for maintaining privacy and security, it may deprive occasions of the social interactions.
Another finding to be discussed was the absence of difference in the AIS scores between the four types of reconstructed residences. To investigate the backgrounds, the correlation matrices within the five self-reported psychosocial indicators were evaluated in the whole participants and those living in public condominiums; however, the obtained patterns of the correlations were largely the same. This finding suggests that the association between sleep disturbance and other mental health disturbances may not be changed by a life in a public condominium. A possible hypothesis to explain the aforementioned finding may be that the discriminatory value of the AIS scores could be weak to show the small difference between the current types of residences, if any, as the range of the scores of AIS is relatively narrow with a positively-skewed distribution. However, the predictive and discriminatory values of AIS with a cutoff score of 6 for the patients with insomnia has been well established 23 , and such systematic problems based on the diagnostic characteristics of the AIS seems to be less likely. Another possible theory may be that the people with worse psychosocial or psychiatric conditions might have failed to answer to the questionnaires of AIS with missing or incomplete sub-item scores. Another possibility would be that there was virtually no difference in the prevalence of insomnia between the four residential types. This possibility seems to be reasonable, as the causal relationship or directionality between social isolation and sleep disturbance has not yet been established 24 . www.nature.com/scientificreports/ A limitation of this study was that the exact amount of income of the disaster-affected population was not analyzed. It is likely that the amount of income may have influenced the decision to either reconstruct houses or move into more affordable public condominiums. Certainly, the relationship between lower socio-economic status and poor mental health conditions is known 25 . In this study, the current employment status and changes in the amount of income did not correlate with the current scores of the mental health-related questionnaire; however, this factor warrants further investigation. Another limitation was that this study did not investigate the  www.nature.com/scientificreports/ details of the stressful experience or the caused economic and human damages by the disaster in each participant, which could have also influenced the current psychosocial and psychiatric conditions among the population.
In conclusion, disaster-affected residents who moved into disaster-recovery public condominiums, especially younger adults aged < 65 years with a high employment rate, are more likely to be affected by the potential negative psychosocial impact of living in disaster-recovery public condominiums. Effective intervention by national and local governments may be helpful to facilitate social interactions and better physical and mental health among the disaster-affected population who have moved into disaster-recovery public condominiums.

Methods
Study design. This study was a questionnaire-based cross-sectional observational survey performed five years after the occurrence of the GEJE on March 11, 2011. Adult residents of Shichigahama Town aged ≥ 18 years at the time of the study, whose houses were destroyed by the tsunami, were initially recruited in November 2016. The level of damage to the residences was uniformly and objectively judged by local government officials based on the standard building damage evaluation criteria defined by the Japanese government. Paper-based questionnaires to investigate the current type of reconstructed residence, current social interaction activities, and current www.nature.com/scientificreports/ psychosocial and mental health conditions were mailed to the candidates. From the respondents, individuals with missing data regarding the type of reconstructed residence were excluded from the analysis by applying the listwise deletion method. Missing data from other respondents with valid data regarding their current type of residence were handled by applying the pairwise deletion method in each statistical analysis.
Indicator for the present social interaction activities. The levels of social networking and social interaction activities were assessed using the LSNS-6 to evaluate general social networking, along with seven self-reported questionnaires that were originally designed to evaluate specific social interaction activities. The LSNS-6 was originally developed by Lubben et al. in 2003 to evaluate social network level 14,26 . The scale comprises six self-reported questions, three of which are about relationships with family or relatives, while the other three relate to relationships with unrelated friends or acquaintances in the local community. It is scored from 0 to 30; a higher score indicating better social relationships. Scores < 12 suggest social isolation 27,28 .
Indicators for the current mental health conditions. The current mental health conditions of the participants five years after the disaster were evaluated using the following questionnaires.
K6. The K6 assesses the psychological stress response of participants in the last 30 days. It comprises six categorical self-reported questions 15,29 . The higher the score, the more likely the subjects are to suffer from mood and anxiety disorders, including depressed states. In the present study, the standard K6 cutoff score of 13 or higher was applied to estimate the presence of psychological distress 30,31 .

CES-D.
The CES-D was developed by the National Institute of Mental Health as a simple screening tool for patients with depression in epidemiological surveys 16,32 . This scale comprises 20 self-reported questions about depression-related physical and mental conditions in the last week of the participant. It is scored from 0 to 60; a higher score suggesting a more severe depressive state, with a cutoff score of ≥ 16 33,34 .

IES-R.
The original version of the IES was developed by Horowitz et al. in 1979 35 , and later revised (IES-R) 17 .
The scale evaluates the severity of a post-traumatic stress reaction in the participant's last week. It comprises 22 self-reported questions-eight questions for evaluating intrusion-based symptoms, eight questions about avoidance-based symptoms, and six questions about hyperarousal-based symptoms. It is scored from 0 to 88, a higher score suggesting a higher likelihood of experiencing a post-traumatic stress reaction and a cutoff score being ≥ 25 17,36 . www.nature.com/scientificreports/ AIS. The AIS was developed by the World Health Organization as a simple and reliable assessment tool for insomnia 18,37 . The scale comprises eight self-reported questions. It is scored from 0 to 24; a higher score indicating more severe insomnia. Usually, subjects with scores ≥ 6 are suspected to be suffering from insomnia 23,38 . Background demographic data. In addition to the above self-reported psychosocial indicators, we collected the following demographic data regarding the social and familial backgrounds of the participants: age, sex, number of family members per household, and current employment status at the time of the investigation. Furthermore, information about whether the amount of income decreased after the disaster was also investigated.
Statistical analysis. Comparisons of quantitative data with normal distributions between the four groups by the type of the current reconstructed residence were performed using the ANOVA, followed by the Tukey-Kramer post-hoc test. Comparisons of the self-reported psychosocial indicators between the different types of current residences were further performed using an ANCOVA by employing the potential covariates with p < 0.10 in the univariate analyses (i.e., the current employment status, number of family members per household) to adjust for the current socio-economic factors. Comparisons of quantitative variables with non-normal distributions were performed using the Kruskal-Wallis test, followed by the Scheffe post-hoc test. Qualitative data were compared using the chi-square test. Because multiple comparisons were simultaneously performed in this study, statistical significance was set at p < 0.01. Discriminant analysis was performed using the scores of the five self-reported psychosocial indicators as covariates and current type of reconstructed residences as the categorical variable, and a canonical plot of the psychosocial indicators by the current type of residences was depicted. Statistical analyses were conducted using SPSS Statistics Base 22 software (IBM, Armonk, NY, USA) and R Statistical Software (version 4.0.5; R Foundation, Vienna, Austria). Discriminant analysis was performed using JMP Pro 16.0 (SAS Institute Inc., NC, USA).
Ethics approval, consent to participate. This study was conducted in accordance with the current version of the Declaration of Helsinki, as revised in 2013. All participants gave written informed consent. The study was approved by the ethics committee of Tohoku University Graduate School of Medicine (approval number: THK-2011482).