Participants
Shelter group
Surveys were conducted with residents living in temporary housing shelters in four municipalities in three coastal areas of the Tohoku region of Japan that were severely damaged by the March 2011 earthquake, tsunami, and nuclear power plant meltdowns. Shelters were selected based on a communication with municipalities in each area. Two shelters were selected in one city of Iwate prefecture (pre-disaster population ~23,000) which lost 7.8% of its population in the tsunami (1,601 died and 207 were lost); interviews were conducted in June-August 2014. In Miyagi prefecture, all six temporary housing residences of one town were selected. The town had a population of 20,419 in 2010; 79 died and 2 were lost due to the tsunami (0.4% in all). The survey was also conducted from June to August 2014. In Fukushima prefecture, two municipalities in the coastal area were selected. One had a population of about 71,000, with 1,105 causalities (1.6%) due to the tsunami; three temporary housing residences were selected as study sites. The other had a population of about 21,000, with 541 causalities (2.6%); all residents were evacuated because of the Fukushima Dai’ichi Nuclear Plant accident. Two temporary residences were selected as study sites. The Fukushima surveys were conducted from October 2013 to February 2014. All residents aged 20 or over living in each temporary housing were approached first by a flier, followed by a visit of survey staff. We limited our sample to adults aged 20 or over, because the instrument used in the study was designed to apply to adults. The Committees of Ethics in Research of Human Subjects also allowed us to contact to obtain direct informed consent only for adults aged 20 or over.
General population
A total of 37 area units were randomly selected from East Japan (excluding the Kanto area), and 1,850 residents aged 20–74 years (50 residents per area unit) were randomly selected based on the population registry. These areas were at least 20 kilometres away from the disaster area and not directly damaged by the earthquake or tsunami, or forced to evacuate because of the meltdowns. These residents were contacted first by a mailed invitation latter, followed by a visit of survey staff. The survey was conducted between August and October, 2014 as part of the World Mental Health Japan Second Survey (WMHJ2) [31].
Data Collection and Ethical Consideration
Participants were the subjects of face-to-face, computer-assisted interviews. The instrument used in this study was the Japanese translation of the World Health Organization Composite International Diagnostic Interview (CIDI), version 3.0 [32, 33] developed for trained lay interviewers. The principal investigator (NK) obtained the permission to use the CIDI 3.0 in the surveys from the WHO World Mental Health Survey Executive Committee (see the supplementary file). Participation was voluntary, and participants were assured of anonymity and confidentiality. Written consent was obtained from each respondent. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The Committees of Ethics in Research of Human Subjects of the Graduate School of Medicine of The University of Tokyo approved the study protocol and informed consent procedure (No. 10131-(7)).
Assessment of mental disorders
The CIDI assessment included six DSM-IV [34] diagnoses analysed in this report: major depressive episode (MDE), manic or hypomania episode (MAN), generalized anxiety disorder (GAD), panic disorder (PD), post-traumatic stress disorder (PTSD), and alcohol use disorder (AUD; including both alcohol abuse and dependence). We selected these six disorders from 12 mental disorders assessed in the WMHJ2 survey [31], while we excluded phobic disorders because these disorders have an onset in an early life [35], dysthymia because the diagnosis requires longer duration criteria (i.e., two or more years) [34] and may not be relevant for this 3-year retrospective recall survey, and drug use disorder because of its very low prevalence in Japan [31]. The hierarchy rule was not applied: for instance, GAD was diagnosed even if MDE was present. For MDE, the exclusion of post-bereavement depression (the criteria E) was not used. This was because of lack of precise information to apply the hierarchy rule, as well as for the sake of simplicity. Prevalence of any of the six mental disorders was labelled as ANY.
All diagnoses were assessed for the respondent’s lifetime, and then, the age of first onset was asked. For mood (MDE and MAN) and anxiety disorders (GAD, PD, and PTSD), respondents from the disaster area were asked additional questions about the cause of the episode and whether the onset was before or after the disaster. Time of onset of each disorder was then classified into before the Great East Japan Earthquake (hereafter the disaster), the same year of the disaster (2011) (1st year), the 2nd year (2012), and the 3rd year (2013) after the disaster.
For disorders that first occurred after the disaster, duration (years) was calculated based on questions on the maximum length of duration for the new onset disorder. If the disorder was active during the 30 days before the survey, the disorder was classified as on-going; otherwise, the respondent was judged as having remitted.
Disaster experiences
Respondents were asked about three types of disaster experiences: personal injury (any vs. none), bereavement of family/relative(s) or friend(s)/acquaintance(s), and house damage (any vs. none). Most previous studies of post-disaster mental health used one bereavement variable combining bereavement of family and friends [36-39]. In this study, in order to investigate individual and additive effects of bereavement of family and friends, the bereavement was classified into four groups: no bereavement, bereavement of family/relative, bereavement of friends/acquaintances, and both. Perceived radiation risk was determined from responses to a single item: “For several months after the accident of the Fukushima Dai’ichi Nuclear Power Plant, how much were you worried about the possibility that you and your family members were exposed to radioactivity? (Not at all, a little, to some extent, much, or very much)” Those who responded “much” or “very much” were classified as high perceived radiation risk.
Seeking treatment
The CIDI included a standard set of questions on seeking treatment for problems with mental health and substance [40]. Medical treatment was defined as a visit to psychiatrists or other medical doctors. The proportion of these who sought medical treatment (either psychiatrists or other medical doctors) was calculated among respondents who had a post-disaster mental disorder.
Demographic variables
Demographic variables assessed during the interview included sex (male vs female), age (20–39, 40–64, and 65+ years old), marital status (married, divorced/separated, widowed, and never married), education (less than high school and high school or higher), and activities of daily living (ADLs) (limited and not limited).
Statistical analyses
Respondents who had experience of any of the six mental disorders in their lifetime prior to 2011 were excluded from the analyses to investigate the new onset of any of the mental disorders. First, cumulative incidences (i.e., proportions of respondents who had a disorder before a certain point including those who remitted after the onset) of ANY and specific mental disorders among were compared between the shelter sample and the general population sample. The discrete-time proportional hazard model applying the logistic regression [41] was used to estimate odds ratios (ORs) of the incidence in the shelter group compared to the general population, adjusting for sex, age, and education, because the time intervals were discrete. Speed of remission from ANY post-disaster mental disorder was compared between the two samples by using the Kaplan-Meier survival curve (log-rank test for a statistical significance). The median duration (in years) and 95% CI of the individual mental disorders were also calculated.
In the shelter group, sociodemographic and disaster-related factors associated with ANY new onset post-disaster mental disorder were investigated with using the discrete-time proportional hazard model. Disorder-specific analyses were conducted for individual mental disorders with 10 or more cases. Among respondents who developed post-disaster mood or anxiety disorders in the shelter group, the association between each of socio-demographic and disaster-related factors and remission from the mental disorder was investigated by using the Kaplan-Meier method with log-rank test.
All variables used in these analyses had no missing response. All statistical analyses were conducted with IBM SPSS Statistics version 22.