Study design
This study was a retrospective cohort study.
LTCI system
The use of LTCI services requires certification of care level. After an application is submitted by the elderly persons themselves or a family member, the municipality, as the insurer, orders two investigations for the applicant (Fig. 1). One of the investigations is a care need certification (nintei-tyousa), which is a visit by a care-related professional and an evaluation of their care needs by using a structured questionnaire. The other is a physician’s written opinion (shujii-ikensyo), which is a care evaluation by a primary care physician. Through the evaluation, a primary care physician scores activities of daily living and cognitive function by using a structured scale common to the whole country. Based on both investigations, a Care Need Certification Committee determines the care level of applicants. The care levels are divided into seven levels (support need levels 1–2 and care need levels 1–5). The higher the care level, the more services that are available to a certified person in a month. If the care needs vary due to a change of ADL or cognitive function, applicants can apply for a re-certification of care level. Out-of-pocket expenditures for LTCI users range from 10 to 30% according to income.
LTCI services are roughly divided four types: (1) home-visit services providing care by nurses, rehabilitators, or assistants in their private homes, (2) day services providing care in facilities during the day by rehabilitators or assistants, (3) short-stay services that consist of respite care for a short period, and (4) facility services that provide residence care to those who are unable to live at home. Re-certification is done when a disease progresses or there is a decline of ADL or cognitive function that requires increased services. Generally, users requiring a certain high care level are admitted to a facility.
Data on LTCI users
This study was conducted using a special sampling of certification data for long-term care and LTCI claim data (approval no. 0711-1). These data were stored in a LTCI comprehensive database which is administered by the Ministry of Health, Labour and Welfare (MHLW). This database maintains digitized claims of LTCI summarized monthly with details for all services used by each user. The MHLW has provided datasets extracted from this database for research institutes since 2018 based on expert council approval.
Data on municipalities
We compared the rurality of the residential area of LTCI users. Using the data on population and land area in each municipality published by the Statistics Bureau, Ministry of Internal Affairs and Communications, the population density of the municipality was collated with individual-level information [24].
Setting
The setting was Hiroshima, Okayama and Ehime prefectures. The worst damage from this disaster occurred in these prefectures: 212 out of 237 deaths, 8 out of 8 missing, 6603 out of 6767 houses completely destroyed, and 10,012 out of 11,243 houses were partially destroyed [5].
Participants and definition of disaster victims
Participants were certified users of the LTCI system in Hiroshima, Okayama and Ehime prefectures from May 2018 to June 2018. The observation period was from July 2018 to December 2018. Victims were defined as participants who changed to an exempt monthly fee for LTCI services after the disaster, reflecting the announcement by the MHLW that all victims of this disaster were exempt from LTCI service fees. This was applied even if they used LTCI services in another region from their registered home region. Local governments authorized a designation as a victim when a LTCI user’s house was completely or partially destroyed, burned down, or there was flooding of a floor, or similar damage, and/or when a main breadwinner died, became seriously injured, ill, or missing. Few people were exempted from the LTCI service fee prior to the disaster. It was reported that the number of deaths in three prefectures was 81 among people from the age of 20 to 69, who represent those most likely to be employed and the primary breadwinners. The number of missing was eight among all ages [25]. The number of deaths was 38% (81/212) for all deaths among all ages in the three prefectures [5]. In addition, the number of completely unemployed increased by three persons compared to before the disaster [26]. In contrast, there were 16,615 cases of destroyed homes in the three prefectures [5]. Therefore, we estimated the majority of the reasons for certification as a victim were due to home damage. Because this exemption excluded those who paid no out-of-pocket expenditures for the LTCI service fee, such as welfare recipients and A-bomb survivors, they were included in the non-victim group. Among LTCI users in the setting prefectures, there are 3475 people who changed status to become exempted from self-payment after the disaster [27]. Because we identified 3024 victims in this data set, the capture ratio for registration would be 87.0%.
We excluded certified users whose cognitive function were the worst on the rating scale from a physician’s written opinion, because a further cognitive decline could not be assessed.
Variables
The outcome variable was cognitive decline. A physician’s written opinion (shujii-ikensyo) include the dementia symptomatology assessment (DSA) to certificate the care level of LTCI insurance. This is a nationally standardized dementia scale to assess level of independence in cognitive functions (nintisyou-koureisya-no-seikatu-jiritudo) [12]. A care need certification examination also uses the same scale to assess dementia symptomatology. Although both investigations were conducted independently, the results showed a high correlation [12]. In addition to this, the DSA was proven to have high inter-rater reliability [28]. The level of dementia scale has high correlation with the Mini Mental State Examination and level I was equivalent to a 0.5 point on the Clinical Dementia Rating [29, 30]. We judged the decline of cognitive function when a primary care physician evaluated the DSA at the point of re-certification and there was a worsened level compared with the before result of DSA during the observation period. The DSA was evaluated with a re-certification of care level in the following cases.
-
1.
The valid certification period of care level ends. The period is generally one year.
-
2.
User applies for recertification of their care level due to worsening of their disease or increase in total amount of care needed.
We adopted the following variables as potential confounders: age classification, gender, the level of dementia scale before the disaster occurred, residential environment (home residents or facility residents), whether a participant used facilities that were shut down after the disaster, whether a participant was recertified and population density. We defined facilities that were shut down after the disaster as any care service that had users before the disaster and then changed status to having no users after the disaster during the observation period.
Ethical approval
Ethical approval was granted by the Ethics Committee for Epidemiological Research at Hiroshima University (Ref. no. E-1389).
Statistical analysis
We showed the baseline characteristics of victims and non-victims. We used a chi-square test for the discrete variables and Wilcoxon’s rank-sum test for ordinal variables and for continuous variables without a normal distribution.
Survival analysis was done using Kaplan-Meier analysis and a log-rank test to estimate the disaster risk. “Month = 0” was July 2018: which was the start of the assessment when the 2018 Japan Floods occurred. The Cox proportional hazards model was used to assess cognitive decline. To examine the effect of the disaster, multivariate analysis was conducted adjusting for age classification, gender, level of dementia scale before the disaster occurred, residential environment (home residents or facility residents), whether a participant used facilities that were shut down after the disaster and population density in addition to the crude model. After we confirmed that the interaction term between whether the users were victims and residential environment (home residents or facility residents) was statistically significant (p = 0.017), we stratified them for the analysis. Furthermore, we confirmed that there were no significant interaction terms between whether the users were victims and other covariates, including age classification.
In addition, we performed two sensitivity analyses. The first sensitivity analysis was the same Cox proportional hazards model restricted to only re-certified participants as in Sensitivity analysis 1. Sensitivity analysis 2 was the cox proportional hazards model restricted to only participants who were aged 85 years or older and we stratified the age code more precisely: 85–89, 90–94, and over 95. Cognitive function can rapidly decline in persons over 85 years old [31]. Because the proportion of people aged over 85 was higher in non-victims than victims, we conducted this analysis.
After a disaster occurs, simple estimation is important to approach a high-risk population. Therefore, we conducted a sub-group analysis. The residential environment (home or facility) and the level of DSA before the disaster were used for this grouping. Because a DSA level of 2b or lower allows for independent living with or without care support, the DSA level of 2b was adopted as the cutoff point. The four groups were as follows: (1) home residents who could live independently for the most activities of daily living with or without any care support (level of DSA ≤ IIb), (2) home residents who could not live independently for most activities of daily living without constant care support (level of DSA ≥ IIIa), (3) facility residents who could live independently for most activities of daily living with or without any care support (level of DSA ≤ IIb), and (4) facility residents who could not live independently for most activities of daily living without constant care support (level of DSA ≥ IIIa). We examined the hazard ratio of cognitive decline by the disaster in each group. The reference was non-victims in each group. After the Cox proportional hazard models, we confirmed the proportionality of the hazard assumption.
We performed all statistical analyses using STATA/MP version 16 (StataCorp, 2019).