Socioeconomic status and the likelihood of informal care provision in Japan: An analysis considering survival probability of care recipients.

Studies show that the burden of caregiving tends to fall on individuals of low socioeconomic status (SES); however, the association between SES and the likelihood of caregiving has not yet been established. We studied the relationship between SES and the likelihood of adults providing long-term care for their parents in Japan, where compulsory public long-term insurance has been implemented. We used the following six comprehensive measures of SES for the analysis: income, financial assets, expenditure, living conditions, housing conditions, and education. We found that for some SES measures the probability of care provision for parents was greater in higher SES categories than in the lowest category, although the results were not systematically related to the order of SES categories or consistent across SES measures. The results did not change even after the difference in the probability of parents' survival according to SES was considered. Overall, we did not find evidence that individuals with lower SES were more likely to provide care to parents than higher-SES individuals. Although a negative association between SES and care burden has been repeatedly reported in terms of care intensity, the caregiving decision could be different in relation to SES. Further research is necessary to generalize the results.

Introduction

Health policy researchers have taken considerable interest in the influence that socioeconomic status (SES) has on individual health. Studies have repeatedly indicated a positive association between SES and a broad spectrum of health outcomes. Recent studies have extended the analysis to long-term care, as the SES gradient in long-term care, including both care receipt and care provision, is an important consideration when designing a public long-term care system, in which equality is valued. Moreover, as in many European countries, several policy reforms have recently been implemented in Japan’s public long-term care system, mainly due to the state’s limited fiscal capacity. These policy reforms include adjustments in eligibility rules and an increase in the coinsurance rate for long-term care services. In addition, supply-side policies have been implemented, constraining the number of long-term care facilities covered by public insurance. This could lead to a rising demand for informal care in the future; therefore, it is necessary to understand the factors motivating the decision to provide informal care.

Studies on the relationship between SES and long-term care provision mainly focus on the association between SES and caregiving intensity, frequency, and/or duration [1-3]. These studies generally show a higher burden of care for caregivers with lower SES. However, the association between SES and an individual’s decision to provide informal care is not very clear, whereas the factors motivating people to provide informal care have attracted attention [4]. On this basis, this study aims to analyze the association between SES and the likelihood of individuals providing care for parents and the underlying mechanism, using the Japanese Study of Aging and Retirement (JSTAR) datasets.

When we study the relationship between SES and the decision to participate in informal care provision, one important aspect that needs to be considered is the relationship between SES and the survival probability of care recipients. A study showed that due to selective mortality health inequality at the population level converges with increasing age [5]. In addition, the SES gradient of life expectancy has been repeatedly confirmed in many countries [6-8], with one U.S. study as an exception [9]. The observed SES gradient of life expectancy is likely to affect the relationship between SES and the probability of informal care provision, as care is provided only for those who are alive. Hence, we considered the selection of survival in our analyses to explore the relationship between SES and the provision of long-term care. We did so by weighting the sample using the inverse probability of the predicted survival of the parents [10].

In this study, we focused on adult children who provided care for their parents. Generally, spouses are the most typical source of informal care provision [11]. At the same time, adult children frequently play an essential role in caregiving. As of 2016, either co-residing adult children or non-co-residing relatives, which presumably includes the children of care recipients, accounted for one-third of elderly care cases in Japan [12]. In addition, studies emphasize the importance of the role of adult children as caregivers for parents in the Japanese context with the cultural tradition of eldest sons and their spouses taking care of elderly parents [13].

As noted in the literature [14, 15], the decision to provide informal care is a complex process, and this study does not aim to describe the complete set of determinants or all dimensions of the decision-making process within families. Rather, this study aims to provide evidence of how the propensity for care provision differs according to socioeconomic status and attempts to understand the factors in this relationship. We limited our analysis to care at home for the study of the relationship between SES and informal care provision, as informal care provision in a residential setting presents more extensive options regarding the type of care provided.

Literature

Determinants of care provision

Several studies have investigated the determinants of the supply of long-term care within a family. Employment or previous employment is an important factor. A study found that employment reduces willingness to provide informal care [16]. Similarly, employment status has a significant deterrent effect on care provision, with different consequences according to the type of previous work, type of care, and gender [17]. Studies also report the opposite causal direction, whereby care provision affects employment [13, 18].

Another important aspect to consider regarding the determinants of care provision is access to formal care, in which the costs and quality of care are significant issues [13]. Relatedly, studies have attempted to estimate the monetary value of care [19, 20]. The difference in care provision by race or ethnicity is partly explained by the price differences experienced by each group [21]. A Japanese study found that the capacity of long-term care facilities has no effect on employment at the regional level, suggesting that access to facility care is less likely to be related to a balance between care provision and employment [22]. Access to formal care depends on the coverage benefits provided by public long-term care insurance (LTCI). The introduction of public LTCI has a positive impact on the employment of caregivers [23].

Some studies have assessed the heterogeneity in care provision by gender. The literature often emphasizes the role of women as caregivers. The difference in the gender proportion of those who report providing care depends on the age of the caregiver [24]. Men are more likely to share care responsibility with other informal caregivers in caring for spouses than are women [4].

SES and care burden

Some studies have focused on the association between SES and care burden among caregivers. These studies measure care burden by considering care time [1, 3], frequency [1, 15], duration, and the degree of dependence [1]. They generally find an SES gradient of care provision, indicating that individuals with lower SES engage more intensively in informal care than individuals with higher SES. Specifically, Saito et al. report, using Japanese data, that caregivers in the lowest 25% of the income quantile are 1.43 and 1.79 times more likely to provide care for 36 and 72 hours per week, respectively, compared to caregivers in the top 25% [3]. A clinical review showed that the risk factors associated with the burden of caregiving included low levels of educational attainment [2].

Two studies indicate that inequality exists in access to formal care based on individuals’ SES, suggesting a potential mechanism for explaining the relationship between SES and care burden [25, 26]. Another explanation is the difference in labor participation across SES groups. The introduction of public LTCI increased the labor participation rate of family caregivers among high-income households, possibly due to the greater opportunity costs of time [27]. Studies reveal that inequality in informal care provision varies across countries, with greater inequality in countries with lower public financing for long-term care [28].

Two studies were the closest to our study in terms of motivation. The first study used the two-part model to explore the extensive (likelihood) and intensive (magnitude) margins of informal care provision [29]. Another study reported that less-educated single women tended to be primary caregivers for parents with high disability levels [26]. We extend the analysis of the probability of care provision for both male and female caregivers using various SES measures.

Institutional background

Compulsory public LTCI was introduced in Japan in 2000 to support the lives of the elderly. The beneficiaries are individuals aged 40 or above, and the LTCI finances use both tax (50%) and insurance premiums from beneficiaries (50%). Services covered by the LTCI include care both at home and at a facility.

Benefits are provided only in kind. To receive a service, beneficiaries apply for an assessment to determine whether they need care and how much if they do. The assessment consisted of two steps: computer assessment and interviews by experts. Based on the assessment, individuals are categorized into eight groups from “independent” to “highest-level care requirement” according to their care requirements. At the time of our analysis, individuals pay 10% of coinsurance when they receive LTC services, and those who are on public assistance are free of charge. In addition, depending on the care requirement category, a different monthly cap is set for the total amount of home-care services, by which those with highest-level care requirements are allowed to use the most. Individuals could receive more care than the cap if they paid 100% of the charge. Services at home include helping in daily life, such as assistance in shopping, cleaning, bathing, and toileting. Services also cover nurse visits for health care and rehabilitation at home. Those who receive care at home could also visit a facility to receive daycare or short stay services.

Three major types of facilities are covered by the term “LTCI”: “LTC welfare facilities” for those who are going to stay at nursing homes for their lifetime, “LTC rehabilitation” for those who are supposed to be back home after a certain period, and “LTC medical facilities” for those who require medical care. In addition, there are private nursing homes available.

Data and methods

Data

In this study, we used data from the Japanese Study of Aging and Retirement (JSTAR), conducted by the Research Institute of Economy, Trade and Industry (RIETI), Hitotsubashi University, and the University of Tokyo [30, 31]. The JSTAR has a consistent set of validated questionnaires with the Health and Retirement Study in the United States; the Survey of Health, Ageing and Retirement in Europe; and the English Longitudinal Study of Aging in the UK so that we can pool the data or compare the results across surveys. The response rate was 60% [31]. We used data provided by RIETI for research purposes on an application basis. The survey included a comprehensive set of questions about an individual’s economic, social, and health conditions for those aged 50 to 75 years at baseline in Japan. In 2007, the survey covered five municipalities, and in 2009, two additional municipalities were included. In 2011, three cities were added to the survey’s coverage, resulting in a total of 10 municipalities. In each municipality, participants were randomly chosen based on household registration [31]. The JSTAR has a longitudinal structure in which individuals can be traced over a maximum of four time points. However, our analysis is based on the first wave for each municipality conducted in 2007, 2009, or 2011; thus, our dataset is cross-sectional. This is because the purpose of this study is to describe the association between SES and informal care provision, rather than to investigate the causal impact of one on the other. In addition, we observed few changes in SES and care provision over time, producing less meaningful results from a longitudinal analysis. Our dataset consists of 7,105 individuals who responded to the first wave of the JSTAR conducted in 2007, 2009, and 2011 where at least one SES measure, gender and age were observed [30, 31].

Methods

We conducted four analyses based on different samples. In each analysis, we used the maximum number of observations with no missing values for both the dependent and independent variables. First, to understand the association between SES and informal care provision for parents, we ran a logistic regression where the dependent variable is a binary variable that takes unity if a respondent cares for at least one parent and/or parent-in-law. The main independent variable is an SES measure, and we control for a set of respondent characteristics, as described in the Variables section. In this analysis, we restrict our sample to respondents of whom at least one parents is alive. This serves as the baseline analysis, in which we do not consider the survival probability of parents. Second, to test if there is an SES gradient in the survival of parents, we examined the relationship between SES and the selection of the target population. If we observed the SES gradient, we needed to take the gradient into consideration as it would affect the observed relationship between SES and the probability of care provision. Specifically, we tested the hypothesis that parents of those in lower SES groups are less likely to be alive at the time of the survey, after controlling for the age and other characteristics of respondents. We used all the respondents in this analysis. Third, to consider the difference in the probability of survival according to SES, we attempted to match SES distribution using the inverse probability weighting method between respondents whose parents were alive and those whose parents were deceased. The probability weighting method is often used to deal with biases shown by non-responses in a survey [10], and is applied to non-random attrition caused by health conditions in longitudinal studies [32]. We applied the method to handle non-responses regarding care provided to parents due to their demise. The process included two steps: in the first step, we obtained the probability of parents’ survival using a probit model, where the dependent variable is a binary that indicates whether the respondent’s parents are alive, and computed the predicted probability for each respondent. In the second step, we regressed the care provision binary on SES and a set of respondent characteristics using the inverse of the predicted probability obtained in the first step as a weight. By doing so, we provided a greater weight, in general, to low-SES individuals compared to the baseline regression conducted in the first analysis. We used all the individuals to calculate the weight in the first step, and the second step was based on respondents whose parents were alive. Fourth, we ran two additional logistic regressions to understand how SES is (un)related to informal care provision. We considered the following two potential mechanisms to explain the relationship between SES and provision of care to parents. The first mechanism is the care requirement of parents, and we thus examined whether SES is associated with the care requirement of parents. Therefore, the first logistic regression with a binary variable to indicate the care requirements of parents was performed to examine if a difference exists in their care needs according to SES. This analysis was based on respondents whose parents were alive. The second factor is a choice between home care and residential care. We ran the second logistic regression with a binary variable that takes unity if a parent requires care at home and zero if he/she receives care from a facility. This study examined how the choice between home care and institutional care differed across SES groups. This analysis was based on respondents whose parents required care.

We included one SES measure in one regression and ran a total of six regressions in each analysis. We used 5% as the threshold for significance. For statistical analysis, we used Stata version 16.1.

Variables

Dependent variables

In the main analysis, we evaluated the association between SES and informal care provision for parents; thus, our dependent variable is a binary variable that indicates whether a respondent cares for a parent. The variable takes unity if a respondent cares for at least one parent (either his or her own parents or a spouse’s parents). Unfortunately, the survey did not include questions about whether a respondent cares for parents. Thus, we produced a binary variable using the following two questions. The first question asked whether a respondent cares for family members or relatives. The second question queried the conditions of the respondent’s parents and parents-in-law (if a respondent was married) whether they required care. We then produced the care provision variable as follows. First, we produced a binary that takes unity if a respondent cares for family members or relatives, and his/her father receives home-based care, and zero otherwise. Second, we repeated the first step for the respondent’s mother, mother-in-law, and father-in-law, and produced at most four binary variables for each respondent. Third, we produced the final binary variable, which takes unity if at least one of the four variables takes unity, and zero if all four variables are zero. In the analysis, we used a narrow definition of care that is associated with physical care, although informal care may involve assistance with household chores and other daily activities [33].

In the analysis of survival, we used a binary to indicate unity if the parent was alive at the time of the survey, and zero otherwise. In the analysis of the mechanisms, we used the following two dependent variables. The first dependent variable indicated the need for care. The variable takes unity if the parent receives home-based or residential care, and zero if the parent does not require care. This variable was created based on the question about the status of the parent’s care requirements, with three possible answers: receiving care at home, receiving care at a facility, and no need for care. If the first two options are selected, then the care needs variable takes unity, and 0 otherwise. The second dependent variable takes unity if the parent receives home-based care (first option), and zero if the parent receives residential care (second option).

Independent variables

The main independent variable in our analysis was the respondents’ SES status. To date, many studies have used education and income as SES measures. We used the following comprehensive set of SES measures: household income, financial assets, monthly expenditure, economic conditions, housing conditions, and education. Household income and assets are calculated as the sum of the respondents’ and their spouses’ income and assets. Monthly expenditure refers to household expenditures. Household income, financial assets, and monthly expenditure are recorded as intervals in the dataset based on the respondents’ responses, and we replaced income by the median value of the interval. In the analysis, we used four categories defined by quantiles. We also used equivalent income and expenditure for a robustness check. The standard way of obtaining equivalent income or expenditure is to divide income or expenditure by household size. Unfortunately, in the JSTAR, there was no information on the number of household members in 2007. We thus conducted a robustness check with two alternative SES measures, using data from 2009 and 2011 (S1 Fig).

Economic conditions and housing conditions were recorded by the interviewer using a five-point scale: very poor, poor, fair, good, and very good. As the responses of “very poor” were limited, we combined “very poor” and “poor” and produced four categories, where 1 indicates “very poor” and “poor,” and 4 indicates “very good.” For education, we prepared the following categories: 1) less than high school, 2) high school degree, 3) some college, and 4) university degree or higher.

Six measures of SES differ in the number of observations due to the different degrees of non-response. The two variables, economic and housing conditions, are complete, and the other four measures are not. In particular, household income and financial assets are variables with a significant number of missing observations. We explored the characteristics of missing observations in two ways. First, we conducted Little’s test for income and financial assets, and the result shows that the missing values in the two variables are not missing completely at random (χ2 = 113.15; p<0.001). Second, given the result of the Little’s test, we test whether missing observations are due to respondents’ health status measured by self-rated health (S1 Table). We do not find a systematic relationship between self-rated health and non-response to household income nor financial assets.

We also added the following control variables: age, age squared, gender, marital status (married and not widowed or divorced = 1), working status (working = 1), and year dummies. Unfortunately, information on regions is not available for the security level of the dataset used for our analysis.

Results

SES and care provision

Those who provide care at home for any of their parents account for 7% of respondents, and the percentage is higher for own parents than for parents of spouses (Table 1). In the sample, the percentages of respondents whose father or father-in-law were alive were 12% and 11%, respectively. The corresponding figures for mothers and mothers-in-law are greater at 34% and 31%, respectively, likely reflecting longer female life expectancies on average. There are variations in the number of observations among SES measures. A significant proportion of information regarding household income and financial assets was not shared by the respondents, while information about living and housing conditions was comprehensive, since the interviewers completed this section of the questionnaire. Thus, these SES measures were used in a complementary fashion. Fifty-two percent of the respondents were female; the mean age was 63.2 years (SD 7.04), and 79% of respondents had spouses at the time of the survey. Fifty-five percent of the respondents were employed.

Table 1

Summary statistics.

		N	Mean	Std Dev	Min	Max
Outcome measures
	Providing informal care at home (among respondents whose parents are alive)
	To any of the parents	3,569	0.070	0.255	0	1
	To own parents	2,649	0.062	0.242	0	1
	To spousal parents	2,106	0.046	0.209	0	1
Parent’s status
	Father is alive	7,105	0.122	0.328	0	1
	Mother is alive	7,105	0.341	0.474	0	1
	Father of spouse is alive	6,293	0.111	0.314	0	1
	Mother of spouse is alive	6,293	0.307	0.461	0	1
SES measures
	HH income	3,419	2.413	1.137	1	4
	HH financial asset	4,169	2.345	1.193	1	4
	Monthly expenditure	6,002	2.422	1.071	1	4
	Living condition	7,105	2.315	0.723	1	4
	Housing condition	7,105	2.466	0.847	1	4
	Education	7,070	2.221	0.993	1	4
Control variables (Respondent’s characteristics)
	Female	7,105	0.519	0.500	0	1
	Age	7,105	63.200	7.037	50	77
	Married	7,100	0.792	0.406	0	1
	Working	7,067	0.551	0.497	0	1

Source. JSTAR datasets of 2007, 2009, 2011 surveys.

Note. HH = household. We used only the first wave. The number of observations for providing informal care at home is based on the number of individuals whose parents are alive, and the corresponding values for other variables are based on all the individuals. Variables related to spouses’ parents were defined only for married respondents. HH income, HH financial assets, and monthly expenditure are categorized based on quantiles, where 1 indicates the lowest 25%. For living and housing conditions, 1, 2, 3, and 4 indicate “very poor” or “poor,” “fair,” “good,” and “very good,” respectively, based on the assessment by interviewers. For educational attainment, 1, 2, 3, and 4 indicate “less than high school,” “high school graduate,” “some college,” and “university degree or higher,” respectively. “Married” indicates married and not divorced or widowed.

A few significant associations between SES and care provision for parents were observed, depending on SES measures (Fig 1). The most apparent tendency was found in living conditions, and individuals with higher SES were more likely to care for parents, driven by females rather than males. Specifically, females in very good condition were 3.2 times more likely to provide care for parents than those in poor condition. A qualitatively consistent association was observed for housing conditions, although the association was not statistically significant. It was also found that females with some college education were twice more likely to provide care than those with less than high school education. In the other three SES variables, we did not find a systematic relationship between SES level and care provision. We also show the breakdown of parents or spousal parents (S2 Fig) and the same analysis where self-rated health was additionally controlled for (S3 Fig). A few positive associations between SES and care provision were found in relation to own parents. Caring for parents-in-law was very limited among males, resulting in a large standard error and imprecise estimates.

Fig 1

Odds ratio of care provision for parents by SES.

The figure shows the odds ratio of care provision compared to the lowest SES category based on logistic regressions controlling for the age and age squared, gender (for all), marital status (married and not divorced or widowed = 1), and working status (working = 1) of the respondents. The bars represent the 95% confidence interval. The dependent variable is a binary variable that takes unity if a respondent provides care to any parent and zero if at least one parent is alive and the respondent does not provide care. Each panel shows a different SES measure. The sample of the analysis consisted of those whose parents were alive. The number of observations for all the respondents ranges from 1,708 (household income) to 3,536 (living and housing conditions).

SES and survival of parents

We found a weak positive association between SES and care provision for some SES measures in the previous analysis. One of the reasons for the positive association could be selection by attrition due to the SES gradient of survival. To investigate this possibility, we subsequently analyzed the association between the SES of respondents and the survival of parents.

Overall, we found a positive association between the survival of parents and SES of respondents (Fig 2). The association was more pronounced for mothers and mothers-in-law than for fathers and fathers-in-law. In the five measures of SES, mothers of respondents in the top-two categories were more likely to be alive than those in the lowest SES category, controlling for age of respondents and other characteristics. For mothers-in-law, a qualitatively consistent association was found. The association was less clear for fathers and fathers-in-law; however, a positive association was still found when monthly expenditure (for fathers) and living and housing conditions (for fathers-in-law) were used.

Fig 2

Odds ratio of survival of own parents and parents-in-law by SES.

The figure shows the odds ratio of survival based on logistic regressions controlling for age and age squared, gender, and marital status (for father and mother, married and not widowed or divorced = 1) of the respondents. The bars represent the 95% confidence interval. The dependent variable is a binary variable that takes unity if the father, mother, father-in-law, or mother-in-law is alive at the time of the survey, and zero otherwise. Each panel shows a different SES measure. The sample of the analysis consisted of all respondents in the dataset. The number of observations ranges from 3,030 (parents-in-law for household income) to 7,100 (parents for living and housing conditions).

In addition, a respondent in a higher SES category systematically showed a greater odds ratio of parents’ survival in many SES measures. For example, the fathers of respondents in the third and fourth quantiles of monthly expenditure were 1.4, and 1.6 times more likely to be alive, respectively, than the first quantile. Similarly, the mothers of respondents who attended college, and obtained a university or higher degree were 1.6, and 1.7 times more likely to be alive than the mothers of those with less than high school education. Given the positive association between the SES of respondents and the survival of parents, in the next subsection, we considered the selection of survival in the analysis of care provision.

SES and care provision considering selection

The analysis in the previous subsection suggests that the higher provision of care for parents among individuals with higher SES could be explained by a higher probability of parents surviving. Hence, we assessed the association by considering the selection by survival by weighting with the inverse probability weighting. After the inverse probability weighting was applied, the mean SES of respondents whose parents were alive became closer to that of those whose parents were deceased (see S2 Table).

The qualitative associations were not consistent across SES measures, and the difference from the lowest category in the predicted probability was mostly not significant (Fig 3). The only significant difference from the bottom category was in the 50–70% group among females for the analysis with household income. In addition, there was no clear relationship between SES level and the probability of providing care. Overall, we did not find a negative association between SES and caregiving, even after considering the survival probability of parents. In the analysis using equivalent household income, males in the top category had a higher probability of care provision and females in the top category had a lower probability of care provision than those in the bottom category; however, a systematic relationship was not observed (S1 Fig).

Fig 3

Predicted probability of care provision considering heterogeneous survival probability.

The figure shows the predicted probability of care provision to parents using the inverse probability weighting. The bars represent the 95% confidence interval. In the first stage, age, age squared, gender, marital status of respondents, and year effects are included as dependent variables. The first stage was based on all the respondents. In the second stage, working status was also included, and the sample of the analysis consisted of those whose parents were alive. Each panel shows a different SES measure. The number of observations ranges from 1,710 (household income) to 3,551 (living and housing conditions).

We could identify two reasons for a hypothetical higher care burden among lower-SES individuals. First, the parents of lower SES individuals were more likely to require care, and, second, they tended to receive care at home rather than at a facility. Our analysis did not find evidence of a systematic difference by SES categories in these two cases, and hence these results could explain why we did not find an SES gradient of caregiving in Japan at least partially (S3 Table, S4 Fig).

Discussion

We assessed the relationship between care provision for parents and SES using the JSTAR dataset. To our knowledge, this is the first study on the SES gradient of long-term care provision, considering the selection of parents. Unlike studies that examined the burden of caregiving among caregivers by SES, we did not find strong evidence of a higher burden of caregiving falling on individuals with lower SES in terms of a decision on whether one provides informal care. This result did not change after the survival probability of parents was considered.

We observed a few positive associations between SES and care provision in some SES measures in the baseline analysis, which is consistent with a previous study that found that wealthier British households were more likely to provide care [29]. The relationship almost disappeared when parents’ survival was considered, suggesting that the observed greater probability of caregiving among higher SES individuals was partly because their parents were more likely alive.

The finding that there was no clear association between SES and living conditions may come from the institutional background in the LTC system in Japan, where institutional long-term care services based on means testing are available for the financially disadvantaged. These facilities include moderately priced nursing for the elderly (“Keihi roujin home”) and care homes for the elderly (“Yougo roujin home”). The facilities are complementary to institutional care services covered by the public LTCI, where all the beneficiaries are eligible regardless of economics status or the availability of informal caregivers. We did not find any association between SES and living arrangements among the elderly in line with this explanation.

Studies found a greater burden of care among low-SES caregivers [1, 3] and use among low-income individuals [34]. This difference could be because our study focuses on the entry into informal care rather than care intensity. In Japan, the public LTCI was introduced in 2000, and individuals can avail themselves of the benefit of formal care services at home with 10% coinsurance. However, because of the out-of-pocket payment for receiving services, low-SES families may tend to avoid using formal care services intensively. This heterogeneity in the composition of formal and informal care by SES could be a reason why the study found that low-SES caregivers tend to spend more time in providing care than high-SES caregivers. However, the heterogeneity in the composition of two types of care is unlikely to affect the propensity of caregiving, which is the target of our analysis. This is because limited research supports the claim that formal care completely replaces informal care [35], and most studies show that the formal and informal care can be provided in a complementary manner at home [36] with differing degree of disability [37], type of care [38], and country [39].

To our knowledge, no study has analyzed the association between life expectancy of parents in the context of caregiving, although our result is consistent with the findings from the following two streams of literature. First, studies have repeatedly reported a positive relationship between SES and life expectancy [6-8]. Second, numerous studies have suggested an intergenerational transfer of socioeconomic status, such as wealth [40] and education [41]. Our findings demonstrate the relationship between the life expectancy of parents and SES of children. The observed relationship between parents’ life expectancy and SES of children adds a layer to the complexity of decision-making around informal care giving, in particular, with respect to the relationship with SES. Specifically, decisions on informal caregiving could differ according to caregivers’ SES, as the probability of parents’ being alive differs. This indicates the importance of considering the survival of parents in the analysis of SES and caregiving.

The idea of considering the selection by mortality in the analysis is similar to a study, in which non-responses due to death in a longitudinal study were adjusted using inverse probability weighted estimators when the relationship between SES and health was assessed [32]. In this study, non-response in health status by attrition in the later waves was handled using the inverse probability weighting, whereas non-response of care-giving status due to parent’s death was considered using the same method in our study.

We used six measures of SES to understand the overall view of the relationship between SES and care provision. Our results are not completely consistent across SES measures. We could offer two reasons for this inconsistency. First, for some SES measures, such as household income and financial assets, non-negligible missing observations in the dataset reduce the sample size in the analysis. If the reporting of income or financial assets by individuals is systematically related to SES status, it could affect the results and hence the difference across SES measures. Second, each SES measure can capture different aspects of informal care provision. For example, monthly expenditure could be a proxy measure of income based on the permanent income hypothesis, the economic theory that people consider not only their current income but also their lifetime income to determine their consumption, and based on the theory, monthly expenditure shows an individual’s SES. At the same time, monthly expenditure could be affected by long-term care expenses, while income would not, in which case a consistent result in the analysis using income and monthly expenditure is not expected.

One of the obvious limitations is that we are not able to identify other potential informal caregivers, particularly the siblings of respondents, as the JSTAR does not request the information. If other caregivers are (primary) caregivers, it would affect the probability of care for parents, but we were not able to control for the effect due to data limitations. In addition, information on parent’s attributes is unavailable in the dataset and an inclusion of variables regarding parents’ characteristics could further improve the calculation of the weight. Finally, the JSTAR is not nationally representative; therefore the results of the current study are not necessarily generalizable to the Japanese population. Further research is required to investigate whether the results can be generalized using a nationally representative dataset.

Conclusions

We assessed the relationship between care provision for parents and SES using the JSTAR dataset. Our analysis focused on the likelihood of provision of care, whereas previous studies mainly explored the differences in SES and the intensity of care provided by caregivers. Unlike previous studies that examined the burden of caregivers by SES, we did not find strong evidence that a higher burden of caregiving fell on individuals with lower SES. Although a negative association between SES and care burden has been repeatedly reported in terms of care intensity, it is important to note that the decision around caregiving could differ in relation to SES.

Additional analysis using equivalent income and expenditure.

(a) Odds ratio of providing care and survival of parents; (b) care provision considering selection by survival. In the analysis in Figs 1–3, we use household income and monthly expenditure as SES measures. Although the information on household size is available only for the 2009 and 2011 surveys, we conduct the same analysis using equivalent income and expenditure, whereby household income and monthly expenditure are divided by the square root of household size. Panel (a) shows the odds ratio of providing care (top) and survival (bottom) based on logistic regressions. Panel (b) shows the predicted probability of care provision to parents using inverse probability weighting. In the first stage, age, age squared, gender, marital status of respondents, and year effects were included as dependent variables. In the second stage, the working status was also included. In both panels, the bars represent the 95% confidence interval. Each panel shows a different SES measure. The sample used for each analysis was the same as in the main analysis. (a) The number of observations for all the respondents in the top is 900 and 1,459 for equivalent income and expenditure, and the corresponding number of observations in the bottom ranges from 1,521 (parents-in-law for equivalent income) to 2,835 (parents for equivalent expenditure). (b) The number of observations is 900 for equivalent income and 1,459 for equivalent expenditure.

(TIF)

Click here for additional data file.

Care provision for parents by SES, own parents, or parents-in-law (a) Own parents (b) Spouse’s parents. The figure shows the odds ratio of care provision compared to the lowest SES category based on logistic regressions controlling for the age and age squared, gender (for all), marital status (for own parents, married and not widowed or divorced = 1), working status (working = 1) of respondents. The bars represent the 95% confidence interval. The dependent variable is a binary variable that takes unity if a respondent provides care to any of the parents, and zero if at least one parent is alive and the respondent does not provide care. Estimates are missing if there are no providers in the SES category. Each panel shows a different SES measure. The sample of the analysis consisted of those whose parents were alive. (a) The number of observations for all the respondents ranges from 1,708 (household income) to 3,536 (living and housing conditions). (b) The number of observations for all the respondents ranges from 985 (household income) to 2,098 (living and housing conditions).

(TIF)

Click here for additional data file.

Care provision for parents by SES controlling self-rated health.

The figure shows the odds ratio of care provision compared to the lowest SES category based on logistic regressions controlling for the age and age squared, gender (for all), marital status (married and not divorced or widowed = 1), working status (working = 1), and four binary variables to show the self-rated health of the respondents. The bars represent the 95% confidence interval. The dependent variable is a binary variable that takes unity if a respondent provides care to any parent and zero if at least one parent is alive and the respondent does not provide care. Each panel shows a different SES measure. The sample of the analysis consisted of those whose parents were alive. The number of observations for all the respondents ranges from 1,708 (household income) to 3,536 (living and housing conditions).

(TIF)

Click here for additional data file.

Possible mechanisms: (a) odds ratio of parents’ care needs by SES; (b) odds ratio of parents receiving care at home compared to care at a facility by SES. Panel (a) shows the odds ratio of parents receiving care at home to receiving care at a facility, based on logistic regressions controlling for the age and age squared, gender, and marital status of the respondents. The coefficient shows the association between respondents’ SES and the care needs of their parents when compared to the lowest SES category. The analysis was based on respondents whose parents were alive, and the number of observations ranges from 358 (household income for father-in-law) to 2,394 (living and housing conditions for mother). The bar represents the 95% confidence interval. In the first stage, the age of respondents and marital status (for the analysis of parents), as well as year effects, were included as independent variables, and in the second stage, only the age of parents and year effects are included. Each panel shows a different SES measure. Panel (b) shows the odds ratio of parents receiving care at home to receiving care at a facility among those who require care, based on logistic regressions controlling for the age and age squared, gender, and marital status of the respondents. The analysis was based on respondents whose parents required care, and the number of observations ranges from 81 (household income for father-in-law) to 680 (living and housing conditions for mother). The bar represents the 95% confidence interval. The dependent variable is a binary variable that takes unity if the father, mother, father-in-law, or mother-in-law requires care at home at the time of the survey, and zero if he/she receives care at a facility. Each panel shows a different SES measure.

(TIF)

Click here for additional data file.

Relationship between missing observations in household income and financial assets and self-rated health.

(PDF)

Click here for additional data file.

Balance of SES measures with and without the inverse probability weighting.

(PDF)

Click here for additional data file.

Summary statistics for the outcome variables in the analyses for possible mechanisms.

(PDF)

Click here for additional data file.

22 Feb 2021

PONE-D-21-00550

Socioeconomic status and the likelihood of informal care provision in Japan: an analysis considering survival probability of care recipients

PLOS ONE

Dear Dr. Ibuka,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Apr 08 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Yu Mon Saw

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2.We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3.We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

4. We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere.

"An earlier version of the analysis is presented at the GSA conference in November 2020 and as a result the conference abstract appears in Innovating in Aging Vol.4 with a title of “Socioeconomic status and informal care provision in Japan”.  We conduct revisions and extension of the analyses and therefore we cannot identify analyses specifically. A copy of the conference abstract is submitted as a Related Manuscript file.  "

Please clarify whether this [conference proceeding or publication] was peer-reviewed and formally published. If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

5. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Additional Editor comments:

We have received comments from reviewers with contradicting opinions. One reviewer recommended 'rejection,' and one suggested major revisions. Please address each comment carefully. Please also take this opportunity to improve the paper as much as possible, including ethical issues and clearly stating methodology of the study. Please ensure that the paper is aligned with the journal's guidelines and free from grammatical errors and typos. We will decide on whether to consider the manuscript further upon receiving the revised manuscript.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors conducted a cross-sectional study to assess the relationship between the social economic status and informal care provision using the questionnaire survey in Japan. I agree with the importance of the topic, but this study has considerable problems mainly regarding the methodology and its consistency with the purpose of the study. Also, this manuscript partly failed to follow the PLOS ONE submission guidelines, such as the lack of ethics statement in the Methods section, which requires modification.

Major

1. Introduction, p.5, l.3-5. And Methods, p.8, l.16-p.10,l.6)

In the introduction section, the authors stated “this study aims to provide evidence of how the propensity for care provision differs by socioeconomic status and attempts to understand the factors in this relationship”, but it was not clear how this aim of this study led to the four analyses, mentioned in the Methods part (p.8).

Especially, the fourth analysis which “examined how the choice between home care and institutional care differed across SES.”(p.10, l.6), did not seem consistent with the purpose of the study, because the authors clearly mentioned that “we have limited our analysis to care at home” in the Introduction part(p.5, l.5). Please explain why all these four analyses were necessary with the view of the aim of the study.

2. Data, p.8, l.1

The authors reported 7,268 individuals as the study population, but lacked the information on i) how the authors recruited the participants (e.g., voluntarily joined, or randomly selected, etc), ii) the response rate, and iii) how the authors treated the participants with missing values. Please provide these information and discuss the generalizability of the result based on the information above.

3. Methods, p.8

Please provide the ethics statement in the Methods section. PLOS ONE submission guidelines require the authors to include ethics statements that specify the name of the approving institutional review board or equivalent committee(s) and whether informed consent was written or oral.

4. Discussion, p.20, l.1

A brief explanation of the “permanent income hypothesis” will be necessary for the readers to understand the authors’ interpretation of the result.

5. Methods, p.9, l.3. and Discussion, p.18. l.16-18.

The authors mentioned they “tested the hypothesis that parents of those in lower SES groups are less likely to be alive at the time of survey” (as their second analysis) in the Methods part, and in the Discussion part, they compared its results with other previous studies which evaluated the association between SES and life expectancy. However, what the authors evaluated was not the life expectancy since they did not seem to consider the age of parents. Please include the age of parents in the analysis, or reconsider the appropriateness of comparison of the results with other previous studies that measured life expectancy.

6. Discussion, p.19. l.3

The authors stated “Our finding adds a layer to the complexity of decision-making regarding informal care giving”, but this sentence is vague. Please explain more clearly what kind of layer this study adds, and how it should be interpreted or implemented.

Minor

1. Introduction, p.4, l.14

The authors stated “Generally, spouse tend to be the most typical source of informal care provision”, but it lacks the reference. Please add the reference.

2. Introduction, p.4, l.18

The authors argued that “studies emphasize the importance of adult children as caregivers for parents in the Japanese context”, but please explain more clearly what “the Japanese context” is.

3. Data, p.8, l.1-4.

There is no description about the basic information on JSTAR, such as where it was conducted. I suggest to add the term “in Japan”, at the end of the sentence in line 4, page 8 (…for those aged 50 years and 75 years at the baseline “in Japan”).

4. Data, p.8, l.1

Although the authors mentioned “JSTAR” stands for “Japanese Study of Aging and Retirement” in the introduction part (p.4, l.2), mentioning it again in the method part will help the readers. Please consider it.

5. Data, p.8, l.4. Table1.

The authors mentioned this dataset (JSTAR) included “those aged 50 years and 75 years at the baseline.”, but Table1 shows the youngest was 47. Please explain why it happened.

6. Data, p.8, l.2

The authors pointed “JSTAR is in harmony with the Health and Retirement Study in the US and the other surveys”, but the expressions i) “in harmony with” and ii) “other surveys” are vague. Please clarify the term i)“in harmony with”, for example, “they have the same validated questionnaire and we can compare the results”, if that is the case. Also, please specify ii)”other surveys”.

7. Methods, p.8

Please provide i)the threshold for significance, ii)the name and version of any software package used, because these information are required by the PLOS ONE Submission Guidelines.

8. Methods, p.9, l.2

The authors wrote “The main dependent variable is an SES measure”, but isn’t the SES independent variable, not the dependent variable? (as from the context of this manuscript, the dependent variable should be provision of care)

9. Independent variables, p.13, l.3-13.

This whole paragraph (The proportion of those who provide care in line 3 ~ the respondents were employed at the time of survey in line 13.) explains about the basic characteristics of the study population. I advise you to move this paragraph to the beginning of the results part.

10. Independent variables, p.13, l.3-6. Table1.

The authors described “The proportion of those who provide care at home for any of the parents is 0.07”, “the proportions of respondents whose father or father-in-law are alive are 0.12 and 0.11, respectively.”, and “The corresponding figures for mothers and mothers-in-law are by far higher at 0.33 and 0.31, respectively”. However, Table1 indicates these numbers represent mean. Is it mean or proportion? Please clarify it.

11. Independent variables, p.13, l.12-13.

The authors described “approximately half are female” and “More than half of the respondents were employed”, but please provide the number and the percentage for “approximately half” and “more than half”.

12. Independent variables, p.13, l.13.

The authors reported “the mean age is 63 years”, but please add the standard deviation when you report the mean

13. Discussion, p.19, l.14. (related to the major comment no.2)

The authors reported “there are non-negligible missing observations in the dataset that reduces the sample size in the analysis.”. If so, the author should explain how they treated missing values, and report how many study participants were excluded from the analysis.

Reviewer #2: Referee #?

PONE-D-21-00550

Comments on “Socioeconomic status and the likelihood of informal care provision in Japan: an

analysis considering survival probability of care recipients”

1. Overall evaluation

This study aims to investigate the association socioeconomic status (SES) and the likelihood of informal care provision in Japan, using the first wave of a unique longitudinal data, entitled “Japanese Study of Aging and Retirement (JSTAR)”. This study finds a new evidence that those categorized in lower SES are not necessarily more likely to provide informal long-term care (LTC) at home than those in better SES. Rather, regarding some SES measures such as self-rated living standard and female caregivers’ educational achievement, even positive correlation between SES and the probability of LTC provision were reported, after adjusting “selection bias”. The findings are not consistent to the results obtained by most previous studies.

The author(s) highlighted two principal contributions of this study; First one is focusing on the heterogeneity in the same choice set and finding that the heterogeneity would be unlikely to influence the probability of caregiving by informal/family caregivers; Second one is adjusting “selection bias” caused by parents’ death for evaluating the association of SES and LTC care provision.

I fully agree with the contributions and novelties highlighted by the author(s). Although the author(s) did not go that far in their interpretation, their findings would have a critical policy implication, such that the economic loss in the society may be more serious than we expected, because those with higher SES and higher opportunity costs are more likely to provide informal LTC.

However, I think there are several points that need to be improved in this study. I would like to provide both major and minor comments as follows. Hopefully, they are helpful for the author(s) to revise this paper.

2. Comments

2-1. Missing SES at random?

As the author(s) pointed out in Discussion (page 19), “If the reporting of income and financial assets by individuals is systematically related to SES status, it could affect the results and hence the difference across SES measures”, I doubt if income and financial assets (of which number of observations are extremely few, in particular) are missing at random. So, I suggest the author(s) to conduct a test Little’s test of missing totally at random.

Reference:

・Cheng Li. (2013). Little’s test of missing completely at random. The Stata Journal (2013)13, Number 4: 795–809.

・Little, R. J. A. (1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association 83: 1198–1202.

2-2. Health status of potential family caregivers

(1) This is also related to my first comment. I totally understand that this study focuses on the association of SES and the probability of providing informal care at home. Even so, since JSTAR surveys adults aged 50 and older, health status of potential family caregivers is a significant determinant of informal care provision, of which heterogeneity cannot be ignored. For example, health status of respondents could be mediator to associate the likelihood of answering questions with SES status. Similar to “selection” due to parents’ mortality adjusted by this study, potential family caregivers with lower health status are less likely to provide informal care to elderly persons and their lower health status could be associated with lower or missing SES.

Fortunately, JSTAR has various indicators of respondent’s health status, i.e. self-rated health status (SRH) and type of disease. Therefore, I would like to suggest the author(s) to test whether or not if a representative health status like SRH is statistically significantly correlated with the probability of missing SES.

(2) Further, regarding health status of potential family caregivers, I would like to suggest the author(s) to investigate the association of a representative indicator like SRH and the likelihood of LTC provision like SES measures. Or, the author(s) can include SRH as a control variable into regression analyses.

2-3. Heterogeneity in regions and timings of the survey

(1) This study utilizes the first wave of JSTAR datasets which surveyed 10 different municipalities and was conducted in various years, 2007, 2009, and 2011. First, the characteristics of these municipalities vary in terms of the size of population, the age structure, the fiscal scale, the geographic area, the supply of medical care and LTC, etc. I think the author(s) should include municipality dummies as covariates for adjusting these characteristics if the data are available, as these geographical characteristics might be correlated to the accessibility to formal LTC services, which would have a significant impact on the probability of demand for informal care by family members.

(2) Second, like geographical heterogeneity, the timings of the survey also vary. Although the author(s) confirmed in Data and Methods (page8), “we observed few changes in SES and care provision over time”, I still think that year dummies should include in regression analyses. For example, the global macro economy experienced Bankruptcy of Lehman Brothers in 2009 and we had the Great East Japan Earthquake in 2011, which would be influence people’s SES and thus various household-level decision makings. Further, I also suggest to include interactive terms of municipality and year dummies, because the impacts of these drastic events would vary across regions.

2-4. Japan’s LTC insurance

In order to discuss the findings, the author(s) explained the background of public LTC insurance briefly in Discussion section (p17-18). However, I think it would be helpful for international readers to understand the context of this study if the author(s) provide a brief summary for the background and system of Japan’s LTC insurance system in an additional section.

2-5. Policy implications

As I pointed out in the overall evaluation, I think this study would have a critical policy implication, such that the economic loss in the society may be more serious than we expected, because those with higher SES and higher opportunity costs are more likely to provide informal LTC. Since this study aims to describe the “correlation” between SES status and the likelihood of informal LTC provision, rather than identifying any causal relations, it is understandable why the author(s) did not go that far to interpret their findings. However, at least, the author(s) had better to mention what might happen to potential socio-economic surplus and/or social welfare if their findings are true.

2-6. Minor comments

(1) Dependent variables explained in page 11 (“the variable takes unity if the parent received home-based or residential care, and zero if the parent does not require care) are not shown in basic statistics (Table 1). The author(s) had better to show these, although they are not major dependent variables.

(2) Is the following sentence in page 12 necessarily described? “Income and expenditure could be measured at the individual level”. I think the author(s) have already explained this in the former section and this is a kind of redundant.

(3) In Table 1 (page 30), please clarify the numbers of observations (N) for outcome measures. They vary among “to any of the parents”, “to own parents”, and “to spousal parents”, because the author(s) extracted only those whose parents are alive? I might overlook the clarification in the manuscript. But, to me, it is unclear how to determine the number of samples to be regressed.

(4) Related to the above comment (3), please clarify the numbers of observations to be analyzed in all figures including the ones in Appendix. For example, are all the number of observations the same among different SES measures in Figure1 and 2? As long as I see Table 1, the number of available observations vary across these SES measures and so they must be different in Figure 1 & 2.

(5) Also, I am not clear if the author(s) run the regression including these SES variables simultaneously in the same regression estimation. Or, they put each of them in separate regression. Please specify in “Methods”.

(6) Regarding Table A1 in Appendix, please show test statistics how significantly statistically different the mean of each SES measure between two groups (parent are alive versus parents are not alive) is.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Comments for authors on PONE-D-21-00550.pdf

Click here for additional data file.

Submitted filename: Comments for Editors on PONE-D-21-00550.pdf

Click here for additional data file.

17 May 2021

Response to comments from the editor and reviewers

We are grateful for the valuable comments provided by the editor and the two reviewers. We believe that incorporating these comments has helped us to improve the quality of the paper. We hereby provide our detailed responses to each comment below.

Editor’s comments:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Thank you for the links. In the revised version, we have followed PLOS ONE’s style requirements.

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

We did not mean that the data was available upon request to us, and we meant that we are not able to share it with anyone. The JSTAR data we used were the secondary data, provided by the Research Institute of Economy, Trade and Industry (RIETI) of Japan, who conducted the survey in collaboration with Hitotsubashi University and the University of Tokyo. The data are owned by RIETI, and the use of the JSTAR dataset is approved by RIETI only to qualified researchers. We have signed the terms and conditions for data use, which includes the non-transfer of data to any other person/party. Therefore, under the contract with RIETI, we are unable to provide the dataset.

The detailed rules of JSTAR data use are found at: https://www.rieti.go.jp/en/projects/jstar/. The contact information of RIETI is found below.

Research Institute of Economy, Trade and Industry (Attn: JSTAR Team)

11th floor, Annex, Ministry of Economy, Trade and Industry (METI)

1-3-1, Kasumigaseki, Chiyoda-ku

Tokyo, Japan 100-8901

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. We will update your Data Availability statement on your behalf to reflect the information you provide.

Please find 2-a).

3. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match. When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

In the revised version, we have included the grant numbers for the awards for our study in the “Funding Information” section.

4. We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere. "An earlier version of the analysis is presented at the GSA conference in November 2020 and as a result the conference abstract appears in Innovating in Aging Vol.4 with a title of “Socioeconomic status and informal care provision in Japan”. We conduct revisions and extension of the analyses and therefore we cannot identify analyses specifically. A copy of the conference abstract is submitted as a Related Manuscript file. Please clarify whether this [conference proceeding or publication] was peer-reviewed and formally published. If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

To apply for a presentation at the GSA meeting, we submitted an abstract with less than 250 words. There was a peer-review process in the conference application to select presenters for the conference, but no additional review process was undertaken for a publication of the abstract. The Volume 4 Supplement 1 contains all the program abstracts from the GSA 2020 annual scientific meeting (https://academic.oup.com/innovateage/issue/4/Supplement_1). The publication is a conference abstract with less than 250 words. The current paper is the full paper with more than 8,000 words; therefore, it should be considered as a distinct work from the published conference abstract.

1. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information

We apologize for the missing information. We have included Supporting Information files at the end of the manuscript and have updated the in-text citations to correlate accordingly.

Additional Editor comments:

5. We have received comments from reviewers with contradicting opinions. One reviewer recommended 'rejection,' and one suggested major revisions. Please address each comment carefully. Please also take this opportunity to improve the paper as much as possible, including ethical issues and clearly stating methodology of the study. Please ensure that the paper is aligned with the journal's guidelines and free from grammatical errors and typos. We will decide on whether to consider the manuscript further upon receiving the revised manuscript.

We sincerely appreciate the opportunity to revise and resubmit the paper. We have carefully revised the paper according to comments and suggestions by the editor and reviewers. We have ensured that the paper conforms with the journal’s guidelines.

Reviewers' comments:

Reviewer: 1

The authors conducted a cross-sectional study to assess the relationship between the social economic status and informal care provision using the questionnaire survey in Japan. I agree with the importance of the topic, but this study has considerable problems mainly regarding the methodology and its consistency with the purpose of the study. Also, this manuscript partly failed to follow the PLOS ONE submission guidelines, such as the lack of ethics statement in the Methods section, which requires modification.

Major

1. Introduction, p.5, l.3-5. And Methods, p.8, l.16-p.10,l.6)

In the introduction section, the authors stated “this study aims to provide evidence of how the propensity for care provision differs by socioeconomic status and attempts to understand the factors in this relationship”, but it was not clear how this aim of this study led to the four analyses, mentioned in the Methods part (p.8).

Especially, the fourth analysis which “examined how the choice between home care and institutional care differed across SES.”(p.10, l.6), did not seem consistent with the purpose of the study, because the authors clearly mentioned that “we have limited our analysis to care at home” in the Introduction part(p.5, l.5). Please explain why all these four analyses were necessary with the view of the aim of the study.

Thank you for the comments. We apologize for the insufficient clarity regarding the relationship between the objective and four analyses. Although the four analyses have different roles, they are all important in our view. We start with the analysis based on the traditional approach. Then, to show why the traditional approach is not necessarily appropriate we conduct the second analysis. The third analysis is the central analysis where we propose the application of the inverse probability weighing method for the analysis of the relationship between SES and care provision. The fourth analysis is a supplementary analysis for the mechanism to understand the main result. We believe that this supplementary analysis is particularly important as the negative association between SES and care provision has been often reported in previous studies, and we need to explain why we do not find one.

To explain our analyses better, we have extended the purpose of the analyses as follows: “On this basis, this study aims to analyze the association between SES and the likelihood of individuals providing care for parents and the underlying mechanism, using the Japanese Study of Aging and Retirement (JSTAR) datasets.” (p3, third line from the bottom). In addition, we have added the purpose of each one of the four analyses in the Methods section on pages 10 and 11.

We agree with the reviewer that the fourth analysis uses those who receive care both at home and at a facility. We have revised the manuscript as follows: “We limited our analysis to care at home for the study of the relationship between SES and informal care provision.” (p5, L5)

2. Data, p.8, l.1

The authors reported 7,268 individuals as the study population, but lacked the information on i) how the authors recruited the participants (e.g., voluntarily joined, or randomly selected, etc), ii) the response rate, and iii) how the authors treated the participants with missing values. Please provide these information and discuss the generalizability of the result based on the information above.

We apologize for the lack of information regarding the data used. The survey is not originally ours, and we use the data of Japanese Study of Aging and Retirement (JSTAR), and the dataset provided by the Research Institute of Economy, Trade and Industry (RIETI), a government research institute of Japan. Furthermore, the participants are randomly selected in the municipality of the target, and the response rate is 60%. We excluded individuals with missing values for the variables used in each analysis. We have added these explanations in the Data section (p9 – p10, L5).

3. Methods, p.8

Please provide the ethics statement in the Methods section. PLOS ONE submission guidelines require the authors to include ethics statements that specify the name of the approving institutional review board or equivalent committee(s) and whether informed consent was written or oral.

As this analysis is based on secondary data, this study is exempted from the approval of intuitional review board.

4. Discussion, p.20, l.1

A brief explanation of the “permanent income hypothesis” will be necessary for the readers to understand the authors’ interpretation of the result.

Thank you for pointing this out. We have added an explanation of the permanent income hypothesis as follows: “the economic theory that people consider not only their current income but also lifetime income to determine their consumption” (p25, L9)

5. Methods, p.9, l.3. and Discussion, p.18. l.16-18.

The authors mentioned they “tested the hypothesis that parents of those in lower SES groups are less likely to be alive at the time of survey” (as their second analysis) in the Methods part, and in the Discussion part, they compared its results with other previous studies which evaluated the association between SES and life expectancy. However, what the authors evaluated was not the life expectancy since they did not seem to consider the age of parents. Please include the age of parents in the analysis, or reconsider the appropriateness of comparison of the results with other previous studies that measured life expectancy.

We agree that parents’ age is important for their life expectancy. In this study, the main target of the analysis is caregivers, and the purpose of the analysis regarding the survival of parents is to test if the probability of parents’ being alive would differ by SES for caregivers. To make a fair comparison across caregivers with different SES levels, we controlled for age of respondents. By doing so, we could interpret the result as the probability of parents being alive could differ by SES for a respondent at the same age.

Additionally, the dependent variable for this analysis is a binary variable with a value of one if a parent is alive, and zero otherwise. We can observe the age of parents so long as they are alive, but we are unable to do so if the parents are dead. If we include the age of parents in the regression, we will have to give up all the data where the dependent variable takes a zero value, which makes regression analysis impossible.

6. Discussion, p.19. l.3

The authors stated “Our finding adds a layer to the complexity of decision-making regarding informal care giving”, but this sentence is vague. Please explain more clearly what kind of layer this study adds, and how it should be interpreted or implemented.

We appreciate the comments. We have provided a comprehensive explanation as follows:

“Our findings demonstrate the relationship between the life expectancy of parents and SES of children. The observed relationship between parents’ life expectancy and SES of children adds a layer to the complexity of decision-making around informal care giving, in particular, with respect to the relationship with SES. Specifically, decisions on informal caregiving could differ according to caregivers’ SES, as the probability of parents’ being alive differs. This indicates the importance of considering the survival of parents in the analysis of SES and caregiving.” (P24, L6)

Minor

1. Introduction, p.4, l.14

The authors stated “Generally, spouse tend to be the most typical source of informal care provision”, but it lacks the reference. Please add the reference.

Thank you. We have added the reference.

2. Introduction, p.4, l.18

The authors argued that “studies emphasize the importance of adult children as caregivers for parents in the Japanese context”, but please explain more clearly what “the Japanese context” is.

Thank you for pointing this out. The Japanese context refers to the cultural tradition where eldest sons and their spouses tend to take care of old parents in the family. We have added the explanation. (p4, second line from the bottom)

3. Data, p.8, l.1-4.

There is no description about the basic information on JSTAR, such as where it was conducted. I suggest to add the term “in Japan”, at the end of the sentence in line 4, page 8 (…for those aged 50 years and 75 years at the baseline “in Japan”).

We have added “in Japan.” Thank you for the suggestion.

4. Data, p.8, l.1

Although the authors mentioned “JSTAR” stands for “Japanese Study of Aging and Retirement” in the introduction part (p.4, l.2), mentioning it again in the method part will help the readers. Please consider it.

We have written out the full name of the survey in the Methods section. Thank you for the suggestion.

5. Data, p.8, l.4. Table1.

The authors mentioned this dataset (JSTAR) included “those aged 50 years and 75 years at the baseline.”, but Table1 shows the youngest was 47. Please explain why it happened.

Indeed, the survey includes those who are younger than 50 years old. We had four of such people in the dataset. We suspect that it is due to response error by respondents, or different persons in the household responding to the survey. Nonetheless, these people were not the target subject of the survey, and therefore, we decided to exclude them from the revised version of the analysis. Thank you for pointing this out.

6. Data, p.8, l.2

The authors pointed “JSTAR is in harmony with the Health and Retirement Study in the US and the other surveys”, but the expressions i) “in harmony with” and ii) “other surveys” are vague. Please clarify the term i)“in harmony with”, for example, “they have the same validated questionnaire and we can compare the results”, if that is the case. Also, please specify ii)”other surveys”.

Thank you. Yes, the reviewer is right that the survey has the same validated questionnaire with RHS, SHARE, and ELSA, which makes the comparison of results possible across surveys. We have explained this in the text.

7. Methods, p.8

Please provide i)the threshold for significance, ii)the name and version of any software package used, because these information are required by the PLOS ONE Submission Guidelines.

Thank you. We have added the information in the Methods section.

8. Methods, p.9, l.2

The authors wrote “The main dependent variable is an SES measure”, but isn’t the SES independent variable, not the dependent variable? (as from the context of this manuscript, the dependent variable should be provision of care)

This was our mistake. Thank you for pointing this out. We have revised it to “independent variable.”

9. Independent variables, p.13, l.3-13.

This whole paragraph (The proportion of those who provide care in line 3 ~ the respondents were employed at the time of survey in line 13.) explains about the basic characteristics of the study population. I advise you to move this paragraph to the beginning of the results part.

We have moved the whole sentence to the beginning of Results section, following the reviewer’s comments.

10. Independent variables, p.13, l.3-6. Table1.

The authors described “The proportion of those who provide care at home for any of the parents is 0.07”, “the proportions of respondents whose father or father-in-law are alive are 0.12 and 0.11, respectively.”, and “The corresponding figures for mothers and mothers-in-law are by far higher at 0.33 and 0.31, respectively”. However, Table1 indicates these numbers represent mean. Is it mean or proportion? Please clarify it.

As the variables pointed out are all binary, the mean shows the proportion of those who have a unity variable. Therefore, we can interpret the mean as the proportion of the variables with a value of one. In the revised version, we have changed the numbers to those in percentage so that it would be consistent with the other descriptions.

11. Independent variables, p.13, l.12-13.

The authors described “approximately half are female” and “More than half of the respondents were employed”, but please provide the number and the percentage for “approximately half” and “more than half”.

We have provided the specific numbers for the description.

12. Independent variables, p.13, l.13.

The authors reported “the mean age is 63 years”, but please add the standard deviation when you report the mean.

We have added the standard deviation in the text.

13. Discussion, p.19, l.14. (related to the major comment no.2)

The authors reported “there are non-negligible missing observations in the dataset that reduces the sample size in the analysis.”. If so, the author should explain how they treated missing values, and report how many study participants were excluded from the analysis.

Thank you for providing this very helpful suggestion. The treatment of missing observations is definitely important for any analyses, and we have explained it. Our four analyses are based on different samples, and there are variations in the number of missing variables in SES. Therefore, we employ the following strategy to treat missing observations: first, we produce the largest dataset with 7,105 respondents where all the major variables are not missing (that is, at least one SES measure, gender, and age are observable). This is our original dataset. Second, in each analysis based on different subsamples from the original dataset, we use all the observations with no missing dependent and independent variable values. This way, we can take advantage of as much information as possible. We explained this in Methods section (P10, L3-9),

Reviewer’s comments:

Reviewer: 2

1. Overall evaluation

This study aims to investigate the association socioeconomic status (SES) and the likelihood of informal care provision in Japan, using the first wave of a unique longitudinal data, entitled “Japanese Study of Aging and Retirement (JSTAR)”. This study finds a new evidence that those categorized in lower SES are not necessarily more likely to provide informal long-term care (LTC) at home than those in better SES. Rather, regarding some SES measures such as self-rated living standard and female caregivers’ educational achievement, even positive correlation between SES and the probability of LTC provision were reported, after adjusting “selection bias”. The findings are not consistent to the results obtained by most previous studies.

The author(s) highlighted two principal contributions of this study; First one is focusing on the heterogeneity in the same choice set and finding that the heterogeneity would be unlikely to influence the probability of caregiving by informal/family caregivers; Second one is adjusting “selection bias” caused by parents’ death for evaluating the association of SES and LTC care provision.

I fully agree with the contributions and novelties highlighted by the author(s). Although the author(s) did not go that far in their interpretation, their findings would have a critical policy implication, such that the economic loss in the society may be more serious than we expected, because those with higher SES and higher opportunity costs are more likely to provide informal LTC.

However, I think there are several points that need to be improved in this study. I would like to provide both major and minor comments as follows. Hopefully, they are helpful for the author(s) to revise this paper.

We sincerely appreciate the useful comments. They are very helpful to improve our paper.

2. Comments

2-1. Missing SES at random?

As the author(s) pointed out in Discussion (page 19), “If the reporting of income and financial assets by individuals is systematically related to SES status, it could affect the results and hence the difference across SES measures”, I doubt if income and financial assets (of which number of observations are extremely few, in particular) are missing at random. So, I suggest the author(s) to conduct a test Little’s test of missing totally at random.

Thank you very much for the suggestion. We have conducted Little’s test for income and financial asset, and the results are reported (P15, L6) Based on Little’s test, household income and financial assets are not missing completely at random.

2-2. Health status of potential family caregivers

(1) This is also related to my first comment. I totally understand that this study focuses on the association of SES and the probability of providing informal care at home. Even so, since JSTAR surveys adults aged 50 and older, health status of potential family caregivers is a significant determinant of informal care provision, of which heterogeneity cannot be ignored. For example, health status of respondents could be mediator to associate the likelihood of answering questions with SES status. Similar to “selection” due to parents’ mortality adjusted by this study, potential family caregivers with lower health status are less likely to provide informal care to elderly persons and their lower health status could be associated with lower or missing SES.

Fortunately, JSTAR has various indicators of respondent’s health status, i.e. self-rated health status (SRH) and type of disease. Therefore, I would like to suggest the author(s) to test whether or not if a representative health status like SRH is statistically significantly correlated with the probability of missing SES.

Thank you very much for the suggestion. To determine whether missing observations in an SES measure are related to health status, we have employed a regression analysis where the dependent variable is a binary variable to indicate if an SES variable is missing and the dependent variable is self-rated health, and the result is shown in S1 Table in the Supporting Information. Our results do not show a systematic relationship between self-rated health and household income nor financial assets.

(2) Further, regarding health status of potential family caregivers, I would like to suggest the author(s) to investigate the association of a representative indicator like SRH and the likelihood of LTC provision like SES measures. Or, the author(s) can include SRH as a control variable into regression analyses.

We have also conducted the analysis by including series of dummy variables to show self-rated health (SRH) as control variables, and the result is shown in S3 Fig in the Supporting Information. The results did not change from those obtained from the analysis without SRH variables.

2-3. Heterogeneity in regions and timings of the survey

(1) This study utilizes the first wave of JSTAR datasets which surveyed 10 different municipalities and was conducted in various years, 2007, 2009, and 2011. First, the characteristics of these municipalities vary in terms of the size of population, the age structure, the fiscal scale, the geographic area, the supply of medical care and LTC, etc. I think the author(s) should include municipality dummies as covariates for adjusting these characteristics if the data are available, as these geographical characteristics might be correlated to the accessibility to formal LTC services, which would have a significant impact on the probability of demand for informal care by family members.

Thank you for the comments. We fully agree with the reviewer that it is better to control for municipality fixed effects. Unfortunately, the dataset we used for the analysis information on regions (municipalities) is masked. Moreover, the structure of our dataset allows us to include only either municipality fixed effects or year fixed effects, but not both. This is because we use a cross-sectional dataset which consists of the first waves of each municipality, and there is at a single year point data in a single municipality. Thus, in our analyses we add year dummies only, which is the best we could do given the constraint.

(2) Second, like geographical heterogeneity, the timings of the survey also vary. Although the author(s) confirmed in Data and Methods (page8), “we observed few changes in SES and care provision over time”, I still think that year dummies should include in regression analyses. For example, the global macro economy experienced Bankruptcy of Lehman Brothers in 2009 and we had the Great East Japan Earthquake in 2011, which would be influence people’s SES and thus various household-level decision makings. Further, I also suggest to include interactive terms of municipality and year dummies, because the impacts of these drastic events would vary across regions.

Thank you for the suggestion. In all the previous analyses, we included year dummies as control variables. For the reason stated in 2-3-(1), we are unable to include municipality dummies or interaction terms between municipality and year dummies.

2-4. Japan’s LTC insurance

In order to discuss the findings, the author(s) explained the background of public LTC insurance briefly in Discussion section (p17-18). However, I think it would be helpful for international readers to understand the context of this study if the author(s) provide a brief summary for the background and system of Japan’s LTC insurance system in an additional section.

Thank you very much for the suggestion. I have now included a new section of the institutional background starting on page 7. I hope this is helpful for readers.

2-5. Policy implications

As I pointed out in the overall evaluation, I think this study would have a critical policy implication, such that the economic loss in the society may be more serious than we expected, because those with higher SES and higher opportunity costs are more likely to provide informal LTC. Since this study aims to describe the “correlation” between SES status and the likelihood of informal LTC provision, rather than identifying any causal relations, it is understandable why the author(s) did not go that far to interpret their findings. However, at least, the author(s) had better to mention what might happen to potential socio-economic surplus and/or social welfare if their findings are true.

We truly appreciate the comments. We have seriously considered the comment by reviewers, and decided not to include the interpretation in Discussion section. This is because the results are mainly found among females, and SES of females and contribution to the economy are not directly related to each other in Japan. In our data, except for education, all SES variables are measured at the household level. This means, for example, a wife in higher economic conditions could not necessarily contribute to the economy more than a corresponding person in low economic conditions as she may be in a better position because of other family member who earns a lot. In addition, household wives may anyway not work regardless of caregiving status. Perhaps we could say more about males as the majority of them work, however, we did not find a positive association for males.

2-6. Minor comments

(1) Dependent variables explained in page 11 (“the variable takes unity if the parent received home-based or residential care, and zero if the parent does not require care) are not shown in basic statistics (Table 1). The author(s) had better to show these, although they are not major dependent variables.

Since the analysis is independent from the main analysis, we have prepared a new table to show summary statistics in S3 Table. Notice that all the variables therein are binaries, and we show the proportion of the respondents only.

(2) Is the following sentence in page 12 necessarily described? “Income and expenditure could be measured at the individual level”. I think the author(s) have already explained this in the former section and this is a kind of redundant.

We have removed the sentence following the reviewer’s suggestion.

(3) In Table 1 (page 30), please clarify the numbers of observations (N) for outcome measures. They vary among “to any of the parents”, “to own parents”, and “to spousal parents”, because the author(s) extracted only those whose parents are alive? I might overlook the clarification in the manuscript. But, to me, it is unclear how to determine the number of samples to be regressed.

Yes, the reviewer is right. The population for the statistics is those people of whom at least one of the parents is alive. We have clarified this in the footnote of Table 1.

(4) Related to the above comment (3), please clarify the numbers of observations to be analyzed in all figures including the ones in Appendix. For example, are all the number of observations the same among different SES measures in Figure1 and 2? As long as I see Table 1, the number of available observations vary across these SES measures and so they must be different in Figure 1 & 2.

Thank you for pointing this out. We have now added the number of observations in each analysis in the appendices of the figures. As the number of observations depend on the measure of SES and the target of analysis, we have shown them as a range.

(5) Also, I am not clear if the author(s) run the regression including these SES variables simultaneously in the same regression estimation. Or, they put each of them in separate regression. Please specify in “Methods”.

We include one SES measure in one regression and run six regressions for each analysis. We have clarified this in the Methods section (P12, L10).

(6) Regarding Table A1 in Appendix, please show test statistics how significantly statistically different the mean of each SES measure between two groups (parent are alive versus parents are not alive) is.

Thank you for the suggestion. We have now included test statistics in S2 Table (former Table A1).

Submitted filename: 210422ResponseLetter.docx

Click here for additional data file.

8 Jun 2021

PONE-D-21-00550R1

Socioeconomic status and the likelihood of informal care provision in Japan: An analysis considering survival probability of care recipients

PLOS ONE

Dear Dr. Ibuka,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jul 23 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Yu Mon Saw

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: (No Response)

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Comments to the authors

The authors revised the manuscript to meet my previous comments. However, the following concern remain, which should be addressed.

1. Discussion part, p.22, l.11-12, Conclusion p.26, l.14-15

The authors mentioned “we did not find strong evidence of a higher burden of caregiving falling on individuals with lower SES.” in the discussion part.

However, the authors reached the conclusion that says “Although a negative association between SES and care burden has been repeatedly reported in terms of care intensity, it is important to note that the decision around caregiving could differ in relation to SES”.

Reviewer #2: The author(s) clarified all my questions and revised following my comments as possible as they can. However, the limitation of the data seems to be quite serious that the author(s) should clarify the further analyses to be necessary.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

27 Jun 2021

Response to comments from the editor and reviewers

We are grateful for the valuable comments provided by the editor and the two reviewers. We believe that incorporating these comments has helped us to improve the quality of the paper. We hereby provide our detailed responses to each comment below. We have also made minor revisions for better readability and these changes are tracked.

Reviewers' comments:

Reviewer 1:

The authors revised the manuscript to meet my previous comments. However, the following concern remain, which should be addressed.

1. Discussion part, p.22, l.11-12, Conclusion p.26, l.14-15

The authors mentioned “we did not find strong evidence of a higher burden of caregiving falling on individuals with lower SES.” in the discussion part.

However, the authors reached the conclusion that says “Although a negative association between SES and care burden has been repeatedly reported in terms of care intensity, it is important to note that the decision around caregiving could differ in relation to SES”.

We apologize for the unclear explanation. We revised the first sentence as follows:

“Unlike studies that examined the burden of caregiving among caregivers by SES, we did not find strong evidence of a higher burden of caregiving falling on individuals with lower SES in terms of a decision on whether one provides informal care.”

Reviewer 2:

The author(s) clarified all my questions and revised following my comments as possible as they can. However, the limitation of the data seems to be quite serious that the author(s) should clarify the further analyses to be necessary.

Thank you for pointing this out. We expanded the limitation according to reviewer’s comments. Also, we stated that further research is necessary in the abstract.

Submitted filename: 210623ResponseLetter.docx

Click here for additional data file.

2 Aug 2021

Socioeconomic status and the likelihood of informal care provision in Japan: An analysis considering survival probability of care recipients

PONE-D-21-00550R2

Dear Dr. Ibuka,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Yu Mon Saw

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: They follow thatall comment

It can be published

They followed our comment properly

This is a valuable your paper

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

5 Aug 2021

PONE-D-21-00550R2

Socioeconomic status and the likelihood of informal care provision in Japan: An analysis considering survival probability of care recipients

Dear Dr. Ibuka:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Yu Mon Saw

Academic Editor

PLOS ONE