Prevention Services via Public Long-term Care Insurance can be Effective Among a Specic Group of Older Adults: A City Claim Data Analysis

Background (LTCI) Methods We conducted an analysis using the following population-based longitudinal data in Kashiwa City between April 2012 and March 2015: Data of National Health Insurance and LTCI claims, the survey for certication of LTCI, the register, and premium tier classication. All data was linked using the pre-assigned anonymous identifying numbers. We analyzed the Cox regression model using the time for the deteriorations of long-term care levels as an outcome and the use of preventive care services as the primary exposure among participants aged 75 years or older, who had either support levels 1 or 2 at the beginning of this analysis. The study was further stratied by both age and initial support level.


Abstract Background
To evaluate the effects of prevention services provided by long-term care insurance (LTCI) for older adults who require support from LTCI in Kashiwa City, Japan

Methods
We conducted an analysis using the following population-based longitudinal data in Kashiwa City between April 2012 and March 2015: Data of National Health Insurance and LTCI claims, the survey for certi cation of LTCI, the register, and premium tier classi cation. All data was linked using the preassigned anonymous identifying numbers. We analyzed the Cox regression model using the time for the deteriorations of long-term care levels as an outcome and the use of preventive care services as the primary exposure among participants aged 75 years or older, who had either support levels 1 or 2 at the beginning of this analysis. The study was further strati ed by both age and initial support level.

Results
The nal analysis included 1,289 participants. The primary result showed, among all participants, that preventive service was not effective (hazard ratio 0.96, 95% con dence interval 0.78-1. 19). In our subanalysis, the preventive service was effective in avoiding deteriorations only among those aged 85 and older with support level 1 (HR 0.65, 95% CI 0.43-0.97) out of four groups.

Conclusions
The preventive services of LTCI in Kashiwa City showed a signi cant effect on the deterioration among subjects aged 85 or older, whose disabled level were low (support level 1). Our results suggest that the prevention services provided by LTCI may not be effective for all older individuals; to provide these services e ciently, local governments, as insurers of LTCI, will need to identify the speci ed groups that may bene t from the preventive services.

Background
Many societies are rapidly aging. Japan began to see a super-aging population earlier than other countries and is expected to see a peak in their aged population in 2025, 1 when one third of individuals in Japan will be aged over 65. In 2000, long-term care insurance (LTCI) was introduced in Japan to be used as a public insurance system to provide care for disabled older adults, both at home and in facilities. 2 3 As Japanese society has had an increasing number of aging adults, the number of disabled people in LTCI has also increased greatly. In 2006, the number of individuals on LTCI had increased to 2.17 million, which is approximately 200% more from when the initiative was announced. 4 In particular, increases in the numbers of people with mild disabilities was noticeable. In 2006, the government established prevention services designed to decrease the number of people becoming disabled. 5 Prevention services include day care (rehabilitation), day services (support for daily living), and also visiting services. These services are arranged for each person based on an assessment of their disabilities. In the regular reforms, provision of prevention services was gradually left in the hands of local municipalities 6 and offered prevention services may be changed or limited. Therefore, discussion is necessary to plan and implement future prevention services and determine how resources will be concentrated.
The effects of prevention services have been evaluated in several studies. Many showed the effects of speci c programs such as muscle strength rehabilitation and locomotive function improvement programs. [7][8][9][10][11] To evaluate the outcomes, these studies used speci c physical functions such as the Timed Up Go test 7-9 11 12 or time standing on one leg. 7-10 For discussing public health policies, indicators are expected to address items such as the level of certi ed care need in LTCI to be supported.
However, few studies used that level as an outcome [12][13][14][15][16] and their results were controversial. [13][14][15][16] Therefore, we aimed to evaluate the effect of prevention services on disabled adults using the level of certi ed care need in LTCI. To offer further insight into resource investment, we showed the difference in the preventive effects on age and the initial certi ed level.

Study design and data collection
We conducted a population-based longitudinal analysis in Kashiwa City, Prefecture of Chiba, Japan. In this study, we used data of the National Health Insurance (NHI), LTCI claims, survey for certi cation of LTCI, registry, and insurance fee level for Kashiwa City A between April 2012 and March 2015. Kashiwa City is located in the Kanto region and has a population of 400,000 and an aging rate of 24.4%, which is lower than the national average of 26.6% in 2015. The survey for certi cation of LTCI collects items to determine the certi cation of LTCI, and is conducted by a surveyor who is commissioned by the local government to visit the subject and observe them directly. This survey was conducted almost nationwide with some differences, and has been statistically processed. Each data was linked using the ID number assigned for the study. This individual linked dataset was used in some previous studies. [17][18][19][20] The data was provided by Kashiwa City to the University of Tsukuba for research purposes, and the data provided and used for research purposes was approved by the Medical Ethical Review Board of University of Tsukuba (approved number 1448).

Participants
Participants were individuals who were 75 years old or older living in Kashiwa City and who rst received the LTCI certi cation from July 2012 to March 2014. They were certi ed as support level 1 or 2 included in the study. Their status as a resident of Kashiwa City was con rmed from the Basic Resident Register data before the rst certi cation that they required long-term care. In addition, in order to demonstrate that the use of preventive care services for six months after the initial certi cation was necessary, it was also con rmed that participants were both alive and had remained at support level 1 or 2 during the six months after their certi cation. Those who had been hospitalized during the 6 months were excluded.
The degree to which individuals need LTCI services indicates their need for care in daily life; this relates to the amount of LTCI services and available services they will use. 21 In 2006, the certi ed levels for LTCI were revised and divided into seven categories: support level 1-2 and care level 1-5, with care level 5 requiring the most intense amount of care. 5 The clinical picture of an individual requiring level 5 nursing care is almost bedridden. On the other hand, individuals with level 1 support can perform most of the basic activities of daily living on their own, but need to remain active to prevent the deterioration of their current state, increasing their need for nursing care. This category does require some support, though. Those who were certi ed as support level 2 require more assistance to complete activities of daily living, and some nursing care is also required. It is expected in the objective of the policy that their condition will remain the same or improve using LTCI services. 6 Outcome The outcome variable was the time (unit: days) from the date on which the participant rst certi ed for LTCI use, for either support level 1 or 2, until their state deteriorated to care level 1 or greater. Care level 1 indicates the elderly who need more services than watching-over care, especially those with limited mobility. For those who did not show any deterioration, the end date of the validity period in their observed nal certi cation of support level 1 or 2 was considered as the observation end date and censored. The Basic Resident Register data was also used to con rm that no participants had died or moved before the end of the observation period.

Exposure
Exposure was based on the use of preventive services for six months, including the rst month when they were certi ed as the eligible LTCI user, which was designated as support level 1 or 2. Those who used prevention services at least once were included in the exposure group. Long-term care prevention service is an LTCI service that can be used by those who have been identi ed as support level 1 or 2 and aims to ensure that the participant's independence in daily life does not deteriorate. The content of visiting care which include bathing, nursing, or outpatient care and rehabilitation. Coordination of preventive care services is performed by specialized staff commissioned by the local government. 5 However, the use of the service is based on its need by an individual or their family, and the service will not be provided unless applied for. For this reason, it is predicted that the use of the service is in uenced by individual characteristics including economic status, and it is necessary to adjust with covariates.

Covariates
Covariates were used to indicate subject severity. Covariates were set based on the use of medical services in the three months before the rst certi cation month. The Charlson Comorbidity Index (CCI) scores calculated from the illness information included in the medical claim data was treated as a covariate. 22 We used the revised CCI scores as validated for Japanese administrative data. 23 In addition, the degree of independence in daily activities and in cognitive function shown in the survey data for LTCI certi cation were treated as covariates, and were found to be predictors of deterioration in the previous analysis. 24 The degree of independence in daily activities or in cognitive function is widely used in Japan to evaluate the elderly activity (Appendix 1). These two indexes are created while the investigator objectively determines the degree of independence regarding the state of the elderly with disabilities at the care site. In the index for elderly daily activities evaluation, the judgment criteria are the following four levels: Rank J (living independently), Rank A (semi bedridden), Rank B (bedridden), and Rank C (bedridden). For our model, we de ned the disabled status was dependent when the degree of independence in daily activities was Rank A or severer. On the other hand, the degree of independence in cognitive function includes nine stages of daily life independence of elderly people with dementia. The dependence for demented status in our model was de ned dependent when that cognitive index was Rank or more dependent. For covariates in the socio-economic situation, we used the insurance fee of LTCI. The fee level is calculated based on the taxation of the insured households, with tax-exempt households designated as low-income households. The previous study showed that income level in uences service use. 19

Statistical analysis
We showed the characteristics of participants among preventive services users and controls with statistical tests. Chi-square tests were performed on the binary variables and Wilcoxon's sum rank test were applied to the ordered variable (the CCI). We drew the survival curves showing the deterioration from support level 1 or 2 to care level 1 or worse with Kaplan-Meier estimates. The differences of the curves were tested with log-rank tests. We analyzed the Cox regression model using the time from the rst day of certi cation for LTCI to when the participant's condition had deteriorated to certi ed care level 1 or worse as an outcome variable. The exposure was the use of preventive care services, adjusted for covariates.
Primary analysis was conducted in all subjects. These models were also tested in four groups categorized according to age (< 85, >=85) and the rst support level (support level 1, 2) as sub-analysis. STATA Version 14.2 (Stata-Corp LLC, College Station, Texas, United States of America) was used for the analysis. The statistical signi cance level was a two-sided p-value of less than 5%.

Results
The study subjects included 1,289 participants (Fig. 1), of which 578 participants (44.8%) were preventive service users ( Table 1). The number of users tends to be higher among women and participants with low household incomes. During the observation period, 348 (27.0%) of all participants showed a deterioration in their certi cated level to care level 1 or worse. The median time to deterioration was 212 days, with a minimum of 8 days and a maximum of 692 days. In the control group, the mean observation period was 522 days (standard deviation 266).  Table 2). Older age (age > = 85), more severely disabled (support level 2), and dependency for demented status were included as covariates in the Cox regression model and were signi cant factors of deterioration in the primary analysis. However, in the sub-analyses in which we performed Cox regression strati ed by age and initial care need status, the preventive service showed effectiveness (hazard ratio 0.65, 95% con dence interval 0.43-0.97) in only one group (i.e., age > = 85 & support level 1) of the four groups (Fig. 2).

Discussion
In the present study, we analyzed the effects of prevention services to prevent or impede the deterioration in the level of certi ed care need of LTCI in Japan. The prevention services were effective only in those aged over 85 and needing mild care. In the other groups, use of the prevention service did not maintain the level of certi ed care needed. Previous studies examined the effects on only speci c aged groups, 15 and did not conduct analysis with varying aged groups. Our results showed the effects of preventive services in comparison with varied age and status of disability.
Our ndings indicate that prevention services may not be effective for everyone. Prevention service of LTCI in Japan was initially started as a "preventive care bene t" based on uniform national standards and rewards. From April 2015 onward, the two preventive services (visiting and outpatient care) were transferred to the "Community support program" operated by local governments. For these two services, the nancial structure is the same as that for preventive bene t. However, each local government can decide its own standards and rewards. Services can be operated in various forms such as services with more exible standards and services by residents, including volunteers. For local governments, it is necessary to consider more effective resource allocation in the future when available resources become limited. Our results can allow governments to consider which groups or characteristic segments should be targeted for effective intervention. We demonstrated the potential effects of preventive services for some groups of people. The local government as an insurer of LTCI must identify the groups that may bene t from preventive services in their areas.
Older age was a factor of deterioration in previous studies. 15 25 Especially in some studies with multivariable analysis, older age was a signi cant factor of deterioration in the adjusted model, 15 which is consistent with our results in the primary analysis (Table 2). No previous study has analyzed the effects to avoid their deterioration in the older-aged group. In our results, preventive services were effective only in older age and for mildly disabled individuals. This signi cant effect of prevention services seemed to depend on the potential characteristics among those who needed care services for the rst time. The subjects aged 85 and over without certi cation were expected to spend their lives with small disabilities or problems in health. In that group (age > = 85 and support level 1), there are many persons who are potentially better than those in other groups 26 Therefore, it seems that this group was likely to have an effect of preventive care.
The strengths of our study are based on using public claim data. Using public administrative data allowed us to follow the entire population of one city, unlike some previous studies that adopted a prepost intervention of the speci c groups as their study design. 8 9 13 In this study, we collected information on the use of medical services and co-morbidities for adjustment using NHI claim data. Further, we extracted the status of cognitive impairment from the certi cation survey data, which included the factors of deterioration in the previous analysis. 24 This variable collection across data became possible only by linking several administrative data between individuals. Our study was the rst one to use linked administrative data for evaluating preventive services.

Limitation
There were four limitations to be considered in the present study. Firstly, we de ned the preventive services provided only during the six months after participants' initial certi cation of LTCI. The effect of preventive service provided at any time was not examined correctly. Therefore, there was some limitation in the time of service provision. We focused on the rst duration after rst certi cation because we expected that during this period, their needs would be new, thus motivating them for service use. Even for local government experts, such a duration of service use enables accurately considering subjects' needs.
Second, this study focuses on only one location. Therefore, the effects of the characteristics and resource allocation of the participants based on the location must be considered. However, this study is the rst to evaluate preventive services based on the certi ed level of LTCI using insurance claim data. Our study reveals useful methods and ndings for future local government-based interventions.
Third, the rst certi cation of LTCI was de ned based on the local government records during the observed period. The individual identi cation depended on the insured person number that was applied by the local government. The insured person number was changed when the person moved from one location to another. Therefore, it was impossible to identify whether the person had LTCI certi cation before moving. This limitation also affects the use of the claim data, even if it was nationwide. Such restrictions on individual tracking are di cult unless individual numbers are assigned to each citizen, and the countermeasures will be an issue for future studies.
Fourth, those who used prevention services at least once were included in the exposure group regardless of the frequency, types, or combinations of services, making it di cult to determine whether and how these components contributed to the observed effects. Future research should be conducted to examine the effects of such service usage in details.
Finally, the effects of unknown confounding cannot be denied. We tried to gather covariates using o cial data from the local government; nonetheless, there remains the possibility of handling other covariates by linking more existing data. This necessitates future research involving analysis of more covariates for precise results.

Conclusion
In conclusion, the preventive services of LTCI in Japan showed a signi cant effect on the deterioration only among subjects aged 85 or older and whose overall support needs were low. To achieve an e cient allocation of limited LTCI resources, it is important for policy makers to identify the population to be targeted for prevention services.

Availability of data and materials
No additional data are available due to data protection requirements.

Funding
The study has been supported nanced by a grant-in-aid from the Ministry of Health, Labour and Welfare; Health and Labour Sciences Research Grant, Japan; Comprehensive Research on Aging and Health (H30-choju-ippan-007) and Statistics and Information (H30-toukei-ippan-007). The funder had no role in the design of the study, collection, analysis, or interpretation of data, or in writing the manuscript.
Authors' contributions TI, TM, NS, and NT contributed in the study design, analysis, and writing of the article. KI and SY facilitated the collection of the municipality data. All authors critically revised the manuscript and approved the nal manuscript.

Competing interests
Takahiro Mori: The joint appointment as an associate professor at the University of Tsukuba was sponsored by FAST DOCTOR CO., LTD in the 2020 nancial year (i.e., April 2020 to the present). FAST DOCTOR CO., LTD has not played any role in the conduct of this study.
The other authors declare no con ict of interest.