Study group
The Fukushima Health Management Survey had been conducted on January 18, 2012 to evaluate the impact of radiation and determine the health status of Fukushima residents, considering the diffusion of radioactive substances and evacuation due to the FDNPS accident, subsequently helping with illnesses prevention, early illnesses detection, and early treatment provision aimed at maintaining and improving the future health of the residents.8 Individuals who completed the Fukushima Health Management Survey, including the Mental Health and Lifestyle Survey component, comprised those residing in any of the 13 municipalities (all areas within Hirono-machi, Naraha-machi, Tomioka-machi, Kawauchi-mura, Okuma-machi, Futaba-machi, Namie-machi, Katsurao-mura, and Iitate-mura, as well as parts of Tamura city, Minamisoma city, Kawamata town, and Date city) who had to be evacuated because of the Great East Japan Earthquake (registered residents).
A total of 180,604 individuals aged ≥15 years (individuals born before April 1, 1995) had been eligible for the 2011 edition of the registered questionnaire. Valid responses were obtained from 73,431 individuals (mean age, 56.4 years), with a response rate of 40.7%. After excluding 46,365 individuals under 65 years of age, 1,220 with an unknown fracture history, and 3,933 who already had a history of fractures in 2011, a total of 21,913 individuals aged ≥65 years (10,271 men; 11,642 women; mean age: 75.0 ± 6.9 years) comprised the sample for the present study.
Incidences of fractures were determined using the questionnaire on fractures from 2012 to 2016. Accordingly, 4,439 individuals were further excluded due to missing fracture data from 2012 to 2016 given that they had never responded to a questionnaire after 2011. Ultimately, 17,474 patients (8,336 men; 9,138 women; mean age, 74.3 ± 6.5 years; mean follow-up duration, 3.7 ± 1.5 person-years) were targeted (Figure 1).
Data regarding age, sex, physical factors (history of fractures, cancer, stroke, heart disease, diabetes, dyslipidemia, hepatic disorder, high blood pressure, and thyroid disease), social factors [experience of the earthquake, tsunami, and nuclear power plant accident (heard the explosion); need for assistance; change in employment status; and change in residence], psychological factors (history of mental illness and PCL), and lifestyle factors (history of smoking and drinking, sleep satisfaction levels, and exercise habits) obtained from the self-administered questionnaire items used in the 2011 Mental Health and Lifestyle Survey were herein analyzed.
Fracture determination
In the Mental Health and Lifestyle Survey, questions regarding the presence or absence of fractures differ depending on the year. Thus, incidences of fractures were determined by combining the questions.
The presence or absence of fractures in 2011 and 2012 was confirmed by responding to a question on “A history of fractures after age 50.” In 2013, apart from the above question, a combination of answers regarding whether “a fracture was diagnosed by a physician within the past year” had been used to determine the presence or absence of fractures. In 2014 and 2015, the presence or absence of fractures had been determined based on only the answer to “fractures within 1 year.” In 2016, the incidence of fractures had been determined by a question on “History of fractures after the age of 50.”
Definition of estimated fracture occurrence date and calculation of follow-up period
The questionnaire used herein could not determine the date on which the fracture occurred. As such, this study estimated fracture occurrence dates by identifying the midpoint between the date the questionnaire for the year no fracture occurred was filled out and the date the questionnaire for the year a fracture occurred was filled out or 6 months before the date the questionnaire for the year a fracture occurred was filled out.
A number of participants also had trouble completing the questionnaire, particularly with regard to information on the month and date of completion, making it impossible to calculate the follow-up period. The questionnaire was distributed by mail in February of the survey year (e.g., for 2011, the questionnaire would have been distributed by February 2012). A breakdown of the months in which the questionnaire had been filled out showed that approximately 77% to 87% were filled out in February for each year. Therefore, when information regarding the month in which they responded was missing, we assumed that they responded in February of that year. Moreover, when information regarding the date the questionnaire was answered was missing, we assumed that they responded to the questionnaire on the 15th of that month.
Evaluation of post-traumatic stress disorder symptoms
The presence or absence of PTSD symptoms was evaluated using PCL, 9 a self-administered questionnaire that obtained information regarding symptoms of PTSD according to the Diagnostic and Statistical Manual of Mental Disorders-IV diagnostic criteria. The reliability, validity, and diagnostic efficiency of the Japanese version of the PCL reported for the determining PTSD symptoms among residents who experienced the Fukushima nuclear accident had been previously estabsliehd.10 The respondents were asked to answer a total of 17 questions using a five-point Likert scale. Accordingly, individuals with higher total scores were strongly suspected to have PTSD. Moreover, a previous study had determined that a total PCL score ≥ 44 was the cut-off point for suspecting the presence of PTSD symptoms.11 In the present study, the total PCL score could be calculated only for those who answered 16 or more questions. When only 16 questions were answered, the average score for the 16 questions was assigned to the missing items to calculate the total score. Participants with total PCL scores ≥ 44 were defined as "with PTSD symptoms" and examined.
History of disease
Residents were asked whether they had a history of cancer, stroke, heart disease, hypertension, diabetes, dyslipidemia, hepatic disorder, thyroid disease, or mental illness.
Lifestyle
The questionnaire section on smoking habits comprised three choices: never smoked, former smoker, or current smoker. The section on drinking habits also comprised three choices: never drinks or rarely drinks (less than once a month), former drinker, or current drinker (more than once a month). The question regarding sleep satisfaction comprised four choices: satisfied with sleep, slightly unsatisfied with sleep, quite unsatisfied with sleep, and very dissatisfied with sleep or does not sleep at all. Furthermore, the question regarding exercise habits comprised four choices: almost daily, two to four times a week, approximately once a week, or almost never.
Experience of the Great East Japan Earthquake
The question regarding experience of the Great East Japan Earthquake involved individuals responding to whether they had experienced the earthquake, tsunami, and nuclear power plant accident (heard the explosion).
Need for assistance
The question regarding need for assistance involved individuals responding to whether they could eat, change clothes, use restrooms, and shop independently. Those who answered that assistance was required for any of the four items were defined as requiring assistance.
Changes in employment status
Regarding change in employment status (job change or unemployment) following the earthquake and accident, residents could respond with either “changed” or “unchanged.”
Changes in housing and evacuation
Residents could respond to the question regarding change in residence after the earthquake with one of the following answers: residing in a shelter, residing in temporary housing, renting a house or apartment, residing in a relative’s house, residing in their own house, or other (free to comment). Respondents who had lived in temporary or evacuation shelters immediately after the earthquake were defined as having changed their residence.
In addition, residents of Tamura city, Minamisoma city, Date city, and Kawamata town did not reside in a temporary or evacuation center in 2011 were defined as not having evacuated. Others were defined as having evacuated.
Statistical analysis methods
All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The Kaplan–Meier method and log-rank test were used to compare difference in the incidence of fractures based on questionnaire answers. Moreover, univariate and multivariate Cox proportional hazards models were used to obtain crude and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between each factor and fractures. Furthermore, multivariate Cox proportional hazards models for men and women were established to determine differences according to sex.
In this study, considering the possibility that differences in questioning for fractures in each year and incomplete tracking may introduce selection and information bias is necessary. Therefore, as a sensitivity analysis, we confirmed the robustness of the results by performing a Cox regression analysis limited to those without deficiencies in the fracture questionnaire for all years.
All data are presented as number of individuals (n), mean, standard deviation, median, 25th percentile, 75th percentile, or percentages. P < 0.05 indicated statistical significant.