Health benefits of reduced patient cost sharing in Japan
Akihiro Nishi, J Michael McWilliams, Haruko Noguchi, Hideki Hashimoto, Nanako Tamiya & Ichiro Kawachi
Volume 90, Number 6, June 2012, 426-435A
Table 2. Effect of reduced cost sharing at age 70 yearsa on out-of-pocket medical spending and physical and mental health,b Japan, 2007
||Multiple imputation modelf
||Joint model with interaction
|Out-of-pocket medical spending||n||US$ per monthh||US$ per monthh||US$ per monthh||US$ per monthh||US$ per monthh||US$ per monthh||US$ per monthh||US$ per monthh|
|Eligible adultsa||8503||−24.93||3.61||< 0.001||−24.92||3.94||< 0.001||−25.25||3.59||< 0.001|
|Eligible adultsa with no reported health-care usei||2968||−3.40||4.19||0.417||−3.73||4.27||0.382||−4.19||4.58||0.361||−27.68||6.43||< 0.001|
|Eligible adultsa with reported health-care usei||5535||−32.72||5.29||< 0.001||−32.76||5.37||< 0.001||−32.44||4.61||< 0.001|
|Eligible adultsa with no reported health-care usei||2968||0.19||0.14||0.173||0.19||0.15||0.202||0.14||0.09||0.115||−0.11||0.13||0.419|
|Eligible adultsa with reported health-care usei||5535||−0.02||0.17||0.892||−0.04||0.18||0.799||0.02||0.09||0.857|
|Eligible adultsa||8503||0.58||0.20||0.004||0.56||0.20||0.007||0.66||0.18||< 0.001|
|Eligible adultsa with no reported health-care usei||2968||0.63||0.32||0.048||0.64||0.32||0.045||0.79||0.32||0.017||−0.29||0.42||0.502|
|Eligible adultsa with reported health-care usei||5535||0.43||0.27||0.111||0.39||0.28||0.156||0.50||0.23||0.032|
SE, standard error; US$, United States dollars.
a The cost-sharing rate decreased from 30% to 10% at the age of 70 years for individuals who had an annual taxable income under 12 000 United States dollars (US$).
b The effect of reduced cost sharing was evaluated using a regression discontinuity design and a generalized estimating equation method.
c The sample sizes include individuals with missing values, all of whom were included in the imputation model.
d We calculated β coefficients for the model: E(Y) = β0 + β1(age) + β2(age)2 + β3(age ≥ 70 dummy), where E(Y) is the expected value of the dependent variable, namely out-of-pocket medical spending, physical health or mental health, age is given in years and age ≥ 70dummy is a dichotomous indicator that equalled 1 if the participant was 70 years or older and 0 otherwise. No other control variables were used in the unadjusted model.
e The adjusted model included the control variables: gender, marital status, household size, occupational status, house ownership and the size of the settlement where the individual lived at the time of the Comprehensive Survey of People’s Living Conditions. Income was not included as it was used to determine whether an individual was eligible for reduced cost sharing. However, including income in the models did not substantially change the results.
f For multiple imputation, we used all the variables described in the methods section and an importance-resampling expectation-maximization algorithm with the assumption that values were missing at random. Five data sets generated from the multiple imputation were mobilized and pooled for point estimation. The same control variables included in the adjusted model were used in the multiple imputation and subsequent regression analyses. Since missing values were estimated and filled in, the standard errors in the imputation model should smaller than those in unadjusted or adjusted model.
g We calculated δ coefficients for the model: E(Y) = δ1(age) + δ2(age)2 + δ3(age ≥ 70dummy) + δ4(age) × (income eligibility or reported health-care use) + δ5(age)2 × (income eligibility or reported health-care use) + δ6(age ≥ 70dummy) × (income eligibility or reported health-care use) + Σδ(control variables), where E(Y) is the expected value of the dependent variable, namely out-of-pocket medical spending, physical health or mental health, age is given in years and age ≥ 70dummy is a dichotomous indicator that equalled 1 if the participant was 70 years or older and 0 otherwise. Coefficients were calculated between adults who were or were not eligible for reduced cost sharing and between eligible adults who did or did not report health-care use at the time of the survey. The main effect was measured by δ6.
h Spending in Japanese yen was converted into US$ using the rate that applied in July 2007 of approximately 120 Japanese yen per US$.
i Health-care use reported at the time of the Comprehensive Survey of People’s Living Conditions.
j Physical health status was evaluated on a scale of 0 to 16 points and was calculated as described in a footnote to Table 1
k Mental health status was evaluated on a scale of 0 to 24 points using a reversed version of Japanese Kessler-6 scale, such that 24 was best.