Bulletin of the World Health Organization

Health benefits of reduced patient cost sharing in Japan

Akihiro Nishi a, J Michael McWilliams b, Haruko Noguchi c, Hideki Hashimoto d, Nanako Tamiya e & Ichiro Kawachi a

a. Department of Society, Human Development and Health, Harvard School of Public Health, 677 Huntington Avenue, Boston, Massachusetts, MA 02115, United States of America (USA).
b. Department of Health Care Policy, Harvard Medical School, Boston, USA.
c. National Institute of Population and Social Security Research, Tokyo, Japan.
d. Department of Health Economics and Epidemiology Research, University of Tokyo, Tokyo, Japan.
e. Department of Health Services Research, University of Tsukuba, Ibaraki, Japan.

Correspondence to Akihiro Nishi (e-mail: anishi@hsph.harvard.edu).

(Submitted: 02 October 2011 – Revised version received: 16 December 2011 – Accepted: 09 January 2012 – Published online: 12 February 2012.)

Bulletin of the World Health Organization 2012;90:426-435A. doi: 10.2471/BLT.11.095380

Introduction

Asking patients to share the cost of health care is a measure frequently used to reduce demand and control the health-care budget.15 Moreover, patient cost sharing can also help finance universal health-care systems by raising additional revenue.6 However, there is still some debate about the level of cost sharing that is best able to control costs while ensuring access to health care, preventing impoverishment from out-of-pocket spending and avoiding adverse health effects.7,8 Several studies have quantified the effects of health insurance coverage and cost sharing on the use of important health services, but few have investigated their influence on health outcomes.914 Most studies were conducted in the United States of America, where universal insurance coverage has not been achieved.9,11,12,1520 The extent to which health outcomes are influenced by cost sharing under national health insurance programmes in other countries is unclear. In this study, we assessed how health outcomes in elderly adults were affected by a change in cost sharing under the universal national health insurance programme in Japan.

Japanese national health insurance provides universal coverage for inpatient, outpatient, dental and pharmaceutical services. There is a fixed cost-sharing rate that varies with age and a general upper limit on monthly out-of-pocket spending for all age groups.21 In 2007, the proportion of Japan’s total national health expenditure that was publicly financed through a tax transfer and social health insurance programme was 81.9%, which ranks among the highest rates for countries in the Organisation for Economic Co-operation and Development’s Asia–Pacific region.22,23 The remainder is private spending by households, which comprises out-of-pocket medical spending for services that may or may not be covered (e.g. non-prescribed over-the-counter medications) by the universal programme.22

Although the exact proportion of patient cost sharing under the public insurance scheme is fixed across services, the level varies by age and income. In particular, in 2001 the Japanese government introduced a system of two different cost-sharing rates: 30% for adults aged less than 70 years and 10% for those aged 70 years or older who had a relatively low income. The rate remained 30% for older individuals with a sufficiently high income.24 We took advantage of this natural experiment to investigate the effect of the change in patient cost sharing at the age of 70 years on physical and mental health.

Methods

Study population

We obtained data from the 2007 Comprehensive Survey of People's Living Conditions – a nationally-representative, cross-sectional survey administered by the Japanese Ministry of Health, Labour and Welfare.25 In total, 2000 sampling units were randomly selected from all 47 prefectures in Japan. All individuals and households in each unit were asked to complete a self-administered questionnaire in July 2007. We restricted our study population to 10 906 individuals aged between 64 and 75 years (i.e. 70 ± 6 years). We excluded individuals who were hospitalized or institutionalized (n = 408) and those who were eligible for free health care (n = 205). For the remaining 10 293, we performed multiple imputation to address potential bias due to missing data (e.g. data on general health status were missing for 13.3%).

Since 2001, the cost-sharing rate is normally reduced from 30% to 10% in the first month after people turn 70 years of age, provided their annual taxable income is below 12 000 United States dollars (US$), which was equivalent to 1.45 million Japanese yen in July 2007.24 Overall, 83.3% of the study population was eligible for the reduction in 2007.

Study variables

We analysed three dependent variables: out-of-pocket medical spending, physical health and mental health. Out-of-pocket spending included all expenditure on medical services in the month preceding the survey. We used a 16-point summary physical health measure developed for the Comprehensive Survey of People’s Living Conditions.26,27 A summary physical health score was derived for each individual from answers to questions on general health status, mobility, self-care and activities of daily living, and pain, as described in a footnote to Table 1. Mental health status was assessed using a Japanese version of the Kessler-6 24-point scale, which has been validated as a strong predictor of mental disorders, excluding substance use disorder, that can be identified using the Structured Clinical Interview of the Diagnostic and statistical manual of mental disorders, fourth edition and that have a Global Assessment of Functioning score less than 60.2831 We reversed the original Kessler-6 scale such that 24 was the best score and 0, the worst.

For modelling purposes, we created a dichotomous indicator, designated age ≥ 70dummy: the indicator equalled 1 if the participant was 70 years of age or older and 0 otherwise. A second dichotomous indicator, designated income eligibility, equalled 1 if the participant was eligible for the lower cost-sharing rate because of his or her taxable income, and 0 otherwise.

To adjust for potential confounders, we considered the following covariates in the regression models: gender, marital status, household size, occupational status, house ownership and the size of the settlement where the individual lived. In addition, we carried out separate analyses for participants who did and those who did not report health-care use under the public insurance programme at the time of the survey for at least one chronic illness that had been diagnosed at a clinic or hospital.

Statistical analysis

The effect of the reduction in the cost-sharing rate at the age of 70 years on out-of-pocket medical spending, physical health and mental health was quantified by estimating the discontinuity in the age trend of these outcomes using a regression discontinuity design.17,32 We assumed that out-of-pocket spending and physical and mental health scores would vary smoothly with age in the absence of the rate reduction and attributed any abrupt deviation in the age trend at the age of 70 years to that reduction. Deviations were quantified, first, by estimating the difference between the value of a variable for individuals aged 69 years and 11 months and the value for those in the first month of their 70th year and, second, by determining how much of that difference could not be explained by the trend observed from the age range of 64 to 69 years to the age range of 70 to 74 years.

We fitted the following unadjusted linear model for the three study outcomes:

(1)


where E(Y) is the expected value of the dependent variable, age is given in years and age ≥ 70dummy is as described above. Thus, β3 provides a measure of the discontinuity in E(Y) around the age of 70 years and can be interpreted as the effect of the cost-sharing rate reduction. To determine whether the effect observed varied by eligibility for the rate reduction, we fitted separate models for non-eligible and eligible adults.

In secondary analyses, we fitted joint models with interaction terms that compared discontinuities between non-eligible and eligible adults and derived differential discontinuities for eligible adults. These differential discontinuities provided more robust estimates of the effects of cost sharing by controlling for changes in the health of non-eligible adults between the ages of 69 and 70 years. However, our study sample had limited statistical power for estimating these differential effects. In addition, to determine how health-care needs affected outcomes in eligible adults, we fitted separate and joint models for eligible adults who were or were not receiving health care at the time of the survey.

To check a key assumption of the regression discontinuity design, we confirmed that important characteristics of the study participants did not change around the age of 70 years.17,32 Specifically, we used the models described above to look for age discontinuities in sociodemographic factors among eligible adults, such as gender, marital status, household size, occupational status, house ownership, taxable income and the size of the settlement where the individual lived. All P-values for these factors were found to be greater than 0.25, which confirms that the characteristics did not abruptly change at the age of 70 years.

Furthermore, we adjusted the above models for these sociodemographic characteristics and used multiple imputation to address bias that could have been introduced by missing data on outcome variables and covariates, as explained in a footnote to Table 2.33 Therefore, our main results were adjusted for sociodemographic characteristics and multiple imputation.

To understand the clinical significance of the average change in mental health status associated with reduced cost sharing, we used logistic regression to estimate the associated change in the prevalence of mild mental illness. We used a validated threshold for the diagnosis of mental illness (i.e. any mood or anxiety disorder in the Diagnostic and statistical manual of mental disorders, fourth edition) in Japan: a score of 19 or less on our reversed Kessler-6 scale.30

We carried out several sensitivity analyses. First, we progressively narrowed the age cohort to individuals aged between 67 and 72 years and then to those between 69 and 70 years. Second, we estimated discontinuities across 11 different dichotomous indicators for age thresholds ranging from 65 to 75 years. We used generalized estimating equations to account for correlated observations within households when estimating standard errors and conducting two-sided tests.34,35 Analyses were performed using Stata version 12.0 (StataCorp. LP, College Station, USA) and R version 2.13.0 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Significant differences in sociodemographic characteristics and health-care needs were found between participants aged from 64 to 69 years and those aged from 70 to 75 years (Table 1). Table 2 shows that adjustment for sociodemographic covariates and multiple imputation to address bias due to missing data did not substantially alter the results of the analysis. Although out-of-pocket medical spending by adults who were not eligible for lower cost sharing did not change significantly at the age of 70 years (β3 = −3.16 US$/month; P = 0.696; Table 2, multiple imputation model), spending by eligible adults decreased significantly and abruptly (β3 = −25.25 US$/month; P < 0.001), as shown in Fig. 1. There was no significant discontinuity in physical health score at the age of 70 years in either group (Fig. 2, and Table 2). However, the differential discontinuity in physical health score for eligible adults (δ6 = +0.27; P = 0.084, joint model with interaction after imputation) showed a relative improvement that approached statistical significance (Table 2). On the other hand, there was a significant improvement in mental health status in the eligible group at the age of 70 years (β3 = 0.66; P < 0.001; Table 2, multiple imputation model) but not in the non-eligible group (β3 = 0.03; P = 0.932), as shown in Fig. 3. This improvement was significant even among participants who reported no health-care use at the time of the survey (β3 = 0.79; P = 0.017), as shown in Fig. 4.

Fig. 1. Effect of reduced cost sharing at age 70 yearsa on out-of-pocket medical spending,b Japan, 2007

US$, United States dollars.
The vertical bars represent standard errors.
a The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
b Out-of-pocket medical spending includes expenditure both on items covered by public health-care insurance and items that are not covered, such as over-the-counter medicines.
c Spending in Japanese yen was converted into US$ using the rate that applied in July 2007 of approximately 120 Japanese yen per US$.

Fig. 2. Effect of reduced cost sharing at age 70 yearsa on physical health,b Japan, 2007

The vertical bars represent standard errors.
a The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
b Physical health status was evaluated on a scale of 0 to 16 points and was calculated as described in a footnote to Table 1.

Fig. 3. Effect of reduced cost sharing at age 70 yearsa on mental health,b Japan, 2007

The vertical bars represent standard errors.
a The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
b Mental health was evaluated on a scale of 0 to 24 points using a reversed version of Japanese Kessler-6 scale, such that 24 was best.

Fig. 4. Effect of reduced cost sharing at age 70 years on the mental healtha of eligible individualsb using and not using health-care,c Japan, 2007

The vertical bars represent standard errors.
a Mental health was evaluated on a scale of 0 to 24 points using a reversed version of Japanese Kessler-6 scale, such that 24 was best.
b The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
c Health-care use reported at the time of the 2007 Comprehensive Survey of People’s Living Conditions.

When eligible participants were divided into two subgroups with different health-care needs, the decrease in out-of-pocket spending at the age of 70 years was found to be greater among adults who reported health-care use at the time of the survey than in those who did not (P < 0.001 for interaction, Table 2), as shown in Fig. 5. In contrast, the effect of reduced cost sharing on mental health status was similar in the two subgroups (P = 0.502 for interaction), as was the effect on physical health status (P = 0.419 for interaction), as shown in Fig. 6.

Fig. 5. Effect of reduced cost sharing at age 70 years on out-of-pocket medical spending of eligible individualsa using and not using health-care,b Japan, 2007

US$, United States dollars.
The vertical bars represent standard errors.
a The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
b Health-care use reported at the time of the 2007 Comprehensive Survey of People’s Living Conditions.
c Spending in Japanese yen was converted into US$ using the rate that applied in July 2007 of approximately 120 Japanese yen per US$.

Fig. 6. Effect of reduced cost sharing at age 70 years on the physical healtha of eligible individualsb using and not using health-care,c Japan, 2007

The vertical bars represent standard errors.
a Physical health status was evaluated on a scale of 0 to 16 points and was calculated as described in a footnote to Table 1.
b The cost-sharing rate decreased from 30% to 10% at the age of 70 years for eligible individuals, who had an annual taxable income under 12 000 United States dollars (US$).
c Health-care use reported at the time of the 2007 Comprehensive Survey of People’s Living Conditions.

Logistic regression analysis showed that the prevalence of mild mental illness decreased significantly at the age of 70 years among eligible participants (odds ratio, OR: 0.718; P = 0.009), but not among non-eligible participants (OR: 0.984; P = 0.960). When we incrementally narrowed the age range of participants included in the analysis from the initial 64 to 75 years, the estimated effect of reduced cost sharing on mental health became progressively larger: β3 was 0.66 (P < 0.001) for the age range 64 to 75 years, 0.69 (P = 0.022) for the range from 67 to 72 years and 1.19 (P = 0.047) for the range from 69 to 70 years. Finally, of the 11 dichotomous indicators for the different age thresholds investigated, only that for a threshold of 70 years was associated with a significant discontinuity in mental health scores.

Discussion

In this nationally representative study, we found that a reduction in health-care cost sharing from 30% to 10% at the age of 70 years was associated with a significant decrease in out-of-pocket medical spending and a significant improvement in mental health in Japanese adults. There was also a relative improvement in physical health in those eligible for the reduction by comparison with those who were not eligible, but it was not significant. Nevertheless, the relative improvement in physical health was similar to that observed in a larger longitudinal study,18 which suggests that our study was not adequately powered to detect this effect. Furthermore, the sharp regression discontinuity design captured only the immediate effect of the cost sharing reduction and may have missed improvements that become manifest over time.

Our findings are largely consistent with previous studies of the effect that providing Medicare coverage in the United States after the age of 65 years has on health-care utilization and health outcomes: coverage was associated with an abrupt increase in health-care utilization and improvements in physical and mental health, mortality and disparities in disease control.9,1419 In addition, higher cost sharing among insured adults has been linked to less use of clinical services and adverse effects on health outcomes.8,9,1420,36

Our findings on out-of-pocket medical spending are consistent with the results of the RAND Health Insurance Experiment,12 which reported that reduced cost sharing substantially increased both total health-care utilization and total health-care spending. Although we could not measure total health-care spending directly, we were able to estimate it in the following way. Data from a separate nationally representative survey indicate that out-of-pocket spending on services included in Japan’s universal insurance coverage programme constitutes 70% of all out-of-pocket medical spending by elderly Japanese adults whose health care is covered by the programme, while spending on services that are not covered (e.g. over-the-counter medications) accounts for the remaining 30%.37 If total health-care spending is unchanged by reduced cost sharing, out-of-pocket spending by eligible adults should decrease to one third (i.e. from 30% to 10%). However, the observed value of out-of-pocket medical spending by eligible adults aged between 70 and 71 years was 29% higher than expected, which suggests that both total health-care spending and utilization had increased substantially and that this may have influenced mental and physical health.

The mental health gains we observed suggest that reduced cost sharing helped alleviate depression and anxiety, which are known to be the largest contributors to disability globally.38 Surprisingly, the improvement in mental health was even observed in individuals who were not using health-care services, which suggests that reduced cost sharing may have affected mental health in several ways. Individuals may have benefited from increased use of health-care services or from the greater sense of security provided by more affordable insurance coverage.

Policy implications

One topic for debate is whether, from a societal or governmental perspective, the gain in mental health linked to reduced cost sharing is worth the associated increase in health-care spending.3941 Fostering a sense of security is one of the core functions of universal health insurance coverage and our findings suggest that the Japanese government has ameliorated the burden of mental illness among the elderly by reducing their cost-sharing rate.2,42,43 However, because of fiscal constraints, in April 2008 the Japanese government decided to raise the cost-sharing rate from 10% to 20% after the age of 70 years, perhaps in 2012 or 2013.24,44 This decision is likely to become a major issue in the next national election. Closer follow-up of the effect of this policy change on population health and medical expenditure would provide the information policy-makers in medium- to high-income countries need to design a system of universal health insurance coverage that incorporates the optimum cost-sharing rate for elderly populations.

Limitations

A lack of survey data on chronic conditions prevented us from being able to identify individuals with treatable chronic conditions, such as cardiovascular disease and diabetes, whose health status may have been particularly affected by reduced cost sharing.14,15,18

In addition, because our study used a cross-sectional rather than a longitudinal design, the results may have been affected by selection bias. For example, in general, individuals who were not using health care between the ages of 64 and 69 years were more likely to use health care after the age of 70 years following the cost sharing reduction.12 However, those who did not start using health care after the age of 70 years were likely to be healthier than those who did. The effect would be an abrupt change in the composition of cross-sectional comparison group samples, which would probably lead to an underestimate of the mental and physical health benefits for the group using health care and an overestimate of the benefits for those not using health care. Thus, although strong conclusions can be drawn from our analysis about changes in mental health for the entire study sample, the results obtained by stratifying participants by use of health care should be interpreted with caution.

Finally, the fact that our study did not have a randomized experimental design makes it impossible to infer a cause and effect relationship when interpreting the findings. In particular, because eligibility for the cost-sharing reduction was determined by income, it was not possible to match the income distribution in eligible and non-eligible groups. Nevertheless, we used a strong regression discontinuity design to control for all observed and unobserved characteristics that trended smoothly with age in our study sample.32,45,46 Furthermore, our analysis included difference-in-difference comparisons that adjusted the effect of cost sharing on eligible adults for changes in physical and mental health observed in the non-eligible, or control, group.

In summary, a reduction in health-care cost sharing at the age of 70 years was associated with lower out-of-pocket medical spending and improved mental health in Japanese adults. This finding suggests that governments in medium- and high-income countries with universal health insurance coverage can reduce the burden of depression and anxiety by modifying cost sharing. The proposed increase in patient cost sharing currently under consideration by the Japanese government deserves close scrutiny because it may have an adverse effect on the mental, and possibly physical, health of older adults.


Acknowledegments

Akihiro Nishi is supported by the Nakajima Foundation. We thank Gary King at Harvard University, John Z Ayanian and Sae Takada at Harvard Medical School and Keiko Sakurai at the University of Tokyo.

Funding:

We were funded by the Japanese Ministry of Health, Labour and Welfare (H22-Policy-033), the Bill & Melinda Gates Foundation, the China Medical Board, the Beeson Career Development Award Program (NIA K08 AG038354) and the American Federation for Aging Research.

Competing interests:

None declared.

References

Share