Study design and setting
This study had a retrospective cohort design and utilized data from residents of a Japanese suburban city in the Tokyo metropolitan area. The population of this city was about 490,000 (data from 2018), and older adults aged ≥ 65 years accounted for 27.4%, which was comparable with the national average of 28.1% [22].
Data sources
We used three separate datasets with data from April 2007–March 2018, including 1) care needs certification for LTC insurance, 2) insurance premium levels, and 3) LTC insurance claims. These datasets are stored in each municipality across Japan to keep a record of the Japanese LTC insurance system. Each residents' data is stored in each dataset according to the type of information as follows.
The three datasets comprise:
-
1) Data on care needs certification for LTC insurance, including each individual’s certified levels and assessment results from 74 items on the severity of physical disabilities and cognitive impairments used to assess LTC needs [23, 24]. Each beneficiary’s certified level is determined by applying a computer-based algorithm to the test results of physical disabilities and cognitive impairments. Additionally, a review by a panel of specialists appointed by the local government is used as well [23]. We used these data to understand beneficiaries’ age, sex, certified levels, specific medical procedures, disability level, and different functional limitations.
-
2) Data on LTC insurance premium levels, which are classified from individual and household-level taxation. The LTC insurance premium levels are used to indicate household income, which we have used as a proxy for economic status.
-
3) Data on LTC reimbursement claims including information on the cost of housing adaptation grants and assistive devices. We utilized data on the implementation and costs of housing adaptation grants based on reimbursement claims.
We were granted access to these datasets in accordance with an agreement from a collaborative research project between this municipality and the Dia Foundation for Research on Ageing Societies, Tokyo. All data were linked using the pre-assigned anonymous identifying numbers.
Study sample
The study sample comprised residents aged ≥ 65 years old who had been certified for care support levels in LTC insurance for the first time by the local government between April 2010 and March 2018. In Japanese LTC insurance system, the eligibility criteria are for 65 years or older, or those who are 40–64 years of age with any of 16 specific diseases. These people can apply for the certification, and are classified into one of seven levels (care support level 1 or 2, care need levels 1–5); we focused only on older adults with care support level 1 or 2. We regarded certification as being the first for an individual if we found no existing certification data between April 2007 and March 2010, because care needs certification is generally reviewed every two years. From the initial sample (n = 11,229), we excluded those younger than 65 years old (n = 46), those lacking data on LTC insurance premium levels (n = 62), and those who were excluded from certification because they moved out of the municipality within 12 months of the certification or because of their death (n = 749). With these exclusions, 10,372 individuals were included in the final study sample.
Conceptual framework and variables
To select variables for the analysis of housing adaption grant use, we applied a conceptual framework based on Andersen’s behavioral model [25,26,27] and on previous studies [28,29,30,31,32,33,34], (see Fig. 1). Andersen’s frequently used behavioral model is useful to comprehensively elucidate the factors related to the process of health service utilization, including the utilization of LTC services [27]. It includes predisposing characteristics, enabling resources, and need factors as factors related to health service use [25]. Enabling resources are the conditions which permit an individual or family to act on a value to satisfy a need regarding health service utilization [25]. Need factors refer to the individual’s self-perceived illness or the severity of illness as clinically judged by professionals [25].
Predisposing characteristics
Age and sex were selected as predisposing factors based on the review of previous research [30,31,32]. Age was grouped into three categories: 65–74 years old, 75–84 years old, and ≥ 85 years old [35]. These variables were derived from care needs certification for LTC insurance data.
Enabling resources
Household income was selected as an enabling resource based on the review of previous research [28,29,30, 33, 34], because those with higher socioeconomic status could pay for services and access more information on them, and also socioeconomic status was related to greater use of health or long-term care services in previous studies [29, 30]. Household income levels were determined by each beneficiary’s LTC insurance premium level, which ranged from persons receiving public assistance (i.e., living below the poverty line as determined by the national government and receiving financial support for all medical and LTC service expenditures under public assistance programs) [36], level 1 (all members of the household have a total annual income ≤ 800,000 JPY [7,533 USD]) to level 18 (persons who have an individual total annual income ≥ 20,000,000 JPY [188,341 USD]). Level 5 (persons who are not taxed individually but have family members within the same household paying taxes) is the standard household income level where standard premium rates apply. Household income was categorized into four categories: receiving public assistance, low (levels 1–4), middle (level 5), and high household income (levels 6–18). This variable was taken from LTC insurance premium levels data.
Need factors
Based on the review of previous research, we selected specific medical procedures, disability levels, and functional limitations as need factors. Andersen [25] mentions two types of need factors: evaluated needs and perceived needs; we focused on evaluated needs in this study. Evaluated needs are objectively assessed, and are important because they may represent the most immediate cause of health service use [25]. All variables comprising need factors were taken from care needs certification for LTC insurance data.
Data regarding specific medical procedures during the last 14 days were collected in the assessment of care needs certification. The 12 types of medical procedures provided mainly by nurses included intravenous cannulation, total parenteral nutrition, dialysis, stoma, oxygen therapy, mechanical ventilation, treatment of bronchotomy, treatment for pain, tubal feeding, monitoring (e.g., blood pressure, heart rate, and oxygen saturation), treatment of pressure ulcers, and catheterization. Medical procedures were categorized as “received” if at least one of these types of medical procedures had been provided. Specific medical procedures were used because medical backgrounds might be important to decide the implementation of housing adaptation grants to recognize effects on their disabilities and prognosis, as with other LTC services [29, 31]. However, we did not have access to data on medical backgrounds, such as medical claims, and we only use medical procedures as a variable related to the medical background in the data on care needs certification.
Disability levels were also selected, referring to the previous research [29, 30, 32] and assessed using the “degree of independent daily living for older people with disabilities”; individuals were categorized as “independent” (independent or Level J1: some disabilities, but daily living is mostly independent, capable of going outdoors with means of transportation, etc.), “Level J2” (some disabilities, but daily living is mostly independent, capable of going out near home), “Level A1” (indoor living and predominantly independent, goes out with assistance, spends most time during the daytime out of bed), “Level A2” (Indoor living and predominantly independent, but unable to go out without assistance and does not go out frequently; repeating cycles of lying down and getting up from bed during the daytime), or “Level B1 or worse” (Level B: some assistance needed for indoor living, also lies in bed for much of the daytime, although sitting position is possible; Level C: bedridden all day, requires assistance with excretion/urination, meals, and dressing/undressing) [37].
Functional limitations related to housing accessibility are referred to in the previous reports [14, 38]. The detailed categories of functional limitations were cognitive function, visual impairment, hearing impairment, poor balance, upper extremity impairment, and lower extremity impairment. We used data on care needs certification for LTC insurance and applied these categories. The application process is shown in Supplementary Information 1. The application of these categories and confirmation of content validity was checked by a researcher with more than ten years of clinical experience as a physical therapist and a researcher who is the author of previous reports of functional limitations [14].
Cognitive function was assessed based on the nationally standardized methods designated by the Japanese Ministry of Health, Labour and Welfare, and was measured as the “degree of independent daily living for older adults with dementia” [37]. Cognitive function was categorized into “independent,” “Rank I” (the patient has some type of dementia, but is almost independent in terms of daily living at home and in society), or “Rank IIa or worse” (for example, Rank IIa: some daily life-disturbing symptoms, behaviors, and problems in communication are seen, but the individual can lead daily life independently if kept watch on by someone when outside the home; Rank M: marked psychiatric symptoms/related symptoms or serious physical disorders requiring expert management).
Visual impairment was ascertained from four ranked answers and categorized into “intact” or “visually impaired” (“can see figure to check visual acuity from 1 m distance” to “hardly see anything”). Hearing impairment was also ascertained from four ranked answers and categorized into “intact” or “hearing impaired” (“can hear as much as the regular volume of sound” to “hardly hear anything”). Poor balance was assessed as to whether the individuals can keep standing and divided into three categories: “can keep standing without some support,” “can keep standing with some support,” or “cannot keep standing.” Upper extremity impairment was assessed dichotomously based on whether the patient had paralysis in the upper extremities on the right or left side. Upper extremity impairment was categorized as “upper extremity impaired” if at least one item was applicable. Lower extremity impairment was assessed in the same way and was categorized as “lower extremity impaired” if at least one item was applicable.
Housing adaptation data
Implementation of housing adaptation was categorized into implemented or not implemented based on the existence of reimbursement claims data. We selected the study sample at the date of care needs certification, and we decided to implement or not implement housing adaptations during the 12-month follow-up. We defined the index year as the one containing the month with the date of care needs certification. The reason we focused on the 12 months after certification was that the maximum valid duration of the first certification was 12 months [39]. The costs were also used for the data on reimbursement claims data, and JPY was converted to USD using the exchange rate on March 31, 2018 (1 USD = 106.19 JPY) [40]. The cost included only the municipality’s payments, not individuals’ out-of-pocket payments.
Statistical analysis
To examine the associations between functional limitations related to housing accessibility with the implementation of the housing adaptation grants, we determined the number and percentage of individuals for each characteristic. These functional limitations were compared to implement or not implemented housing adaptations using the χ2 test.
Subsequently, we used a multivariable logistic regression analysis to estimate the likelihood of implementation of housing adaptations, illustrated by the odds ratio (OR) with a 95% confidence interval (95% CI). The dependent variables were the implementation of housing adaptation grants; independent variables were functional limitations related to housing accessibility. As covariates, we included predisposing characteristics (age and sex), enabling resources (household income level), and need factors (specific medical procedures and disability levels). In addition, there was concern that the results would change over time during these nine years, although the basic system of housing adaptation grants has not changed. We conducted a sensitivity analysis to check the difference in the first and last three years (2010–2012 and 2015–2018).
Later, we conducted a subgroup analysis and extracted only the sample who implemented housing adaptation grants to examine whether the costs were different by their functional limitations. The costs were compared using the Mann–Whitney U test and Kruskal–Wallis test. Multiple comparisons with the statistically significant independent variables were conducted using Dunn’s test with a Bonferroni adjustment [41]. SPSS version 25.0 (SPSS Inc., Chicago, IL) was used for all analyses, and the significance threshold was set at p = 0.05 (two-tailed).