Descriptive Statistics
Table 1 presents the descriptive statistics for each of the three waves of the CHARLS. The percentage of household members above 75 years of age was 16% in 2011, 23% in 2013 and 30% in 2015, and nearly 65% to 70% of the household members were living with a spouse. By 2015, 1.15% of respondents had no insurance, with UEBMI and UBMI accounting for 13.69% and 84.91% of the respondents, respectively. Over half (67%) of the households were in rural areas. Of the households, 25.59%, 34.87% and 39.53% lived in the eastern, middle and western zones, respectively. For the needs-based factors, the inpatient rate of the sample was 15%, 22% and 26% in 2011, 2013 and 2015, respectively. Over the study period, the overwhelming majority of households (83%) reported having a chronic condition and at least 30% of households had a disabled member. These summary data are consistent with the results of the fifth NHSS in 2013 [4].
Table 1. Household descriptive statisticsa (N=2790)
Variables
|
Definition
|
2011
|
2013
|
2015
|
Mean(SDb)
|
Mean(SD)
|
Mean(SD)
|
Health expenditure per month
|
Continuous variable
|
34 (109)
|
62(273)
|
77(301)
|
Total expenditure
|
Continuous variable
|
314(357)
|
404(693)
|
448(709)
|
Non-food expenditure
|
Continuous variable
|
144(219)
|
213(518)
|
250(593)
|
Predisposing factors
|
Household member age above 75
|
|
0.16(0.36)
|
0.23(0.42)
|
0.30(0.46)
|
Marriage
|
0=No spouse 1=With spouse
|
0.70(0.46)
|
0.65(0.47)
|
0.65(0.48)
|
Enabling factors
|
Household size
|
Continuous variable
|
2.54(1.83)
|
4.28(1.58)
|
2.38(1.20)
|
Economic status
|
1=Poorest 2=Poor 3=Average 4=Richer 5=Richest
|
2.57(1.39)
|
2.50(1.36)
|
2.37(1.35)
|
Insurance typec
|
0=NI 1=UEBMI 2=UBMI 3=OI
|
1.81(0.51)
|
1.80(0.49)
|
1.84(0.40)
|
Area
|
0=Urban 1=Rural
|
0.67(0.49)
|
0.67(0.47)
|
0.67(0.47)
|
Zone
|
0=Eastern 1=Middle 2=Western
|
1.14(0.80)
|
1.14(0.80)
|
1.14(0.80)
|
Need factors
|
Inpatient or not
|
0=No 1=Yes
|
0.15(0.36)
|
0.22(0.36)
|
0.26(0.44)
|
Chronic disease or not
|
0=No 1=Yes
|
0.83(0.37)
|
0.83(0.38)
|
0.87(0.34)
|
Disability or not
|
0=No 1=Yes
|
0.30(0.46)
|
0.45(0.50)
|
0.46(0.50)
|
Outpatient or not
|
0=No 1=Yes
|
0.30(0.46)
|
0.32(0.47)
|
0.30(0.46)
|
Note: (a) Sorting was performed according to the CHARLS survey data; (b) The standard deviation is shown in parentheses; (c) Abbreviations: NI, no insurance; UEBMI, Urban Employee Basic Medical Insurance; UBMI, Unified Basic Medical Insurance for urban residents without formal employment and rural residents; OI, other insurance.
Incidence and Intensity of Catastrophic Health Expenditure (CHE)
Table 2 summarizes the incidence (Hc) and intensity (O and MPO) of CHE for the 3 survey years. The results were calculated using the commonly recommended cut-off points of 10% and 40% associated with total and non-food expenditure, respectively [29, 31].
Table 2. Incidence and intensity of CHE among elderly households in China, from 2011 to 2015a
Year
|
2011
|
2013
|
2015
|
Out-of-pocket health care spending as a share of total expenditure (cut-off point =10%)
|
Head count(SE)
|
29.92(0.008)
|
36.52(0.007)
|
39.42(0.009)
|
p-valueb
|
|
<0.001
|
<0.001
|
Overshoot(SE)
|
6.85(0.003)
|
9.87(0.004)
|
11.45(0.003)
|
p-valueb
|
|
<0.001
|
<0.001
|
Mean positive overshoot
|
22.89
|
27.03
|
31.25
|
Out-of-pocket health care spending as a share of non-food expenditure (cut-off point =40%)
|
Head count(SE)
|
20.86(0.007)
|
27.63(0.008)
|
31.00(0.009)
|
p-valueb
|
|
<0.001
|
<0.001
|
Overshoot(SE)
|
3.12(0.002)
|
5.06(0.002)
|
8.75(0.003)
|
p-valueb
|
|
<0.001
|
<0.001
|
Mean positive overshoot
|
14.96
|
18.31
|
28.23
|
Note: (a) Presented as % unless otherwise indicated; (b) Statistical testing was conducted by comparing the year-specific "head count" with the equivalent value in 2011.
The incidence of CHE (accounting for >10% of total expenditure) continued to rise during the study period, from 29.92% (95% CI: 28.23% to 31.63%) in 2011 to 39.43% (95% CI: 37.61% to 41.24%) in 2015. In contrast, when the incidence of CHE was defined as health expenditure exceeding 40% of non-food expenditure, the incidence of CHE also continued to rise during the whole study period, from 20.86% (95% CI: 19.35% to 22.37%) in 2011 to 31.00% (95% CI: 29.28% to 32.72%) in 2015.
When intensity of CHE was assessed as the value of the overshoot, regardless of which measurement method was adopted, we found that the intensity of CHE also increased. When based on total household expenditure for the whole study period, the overshoot (O) increased from 6.85% (95% CI: 6.27% to 7.45%) in 2011 to 11.45% (95% CI: 10.67% to 12.23%) in 2015. When based on household non-food expenditure, the overshoot (O) also increased, from 3.12% (95% CI: 2.71% to 3.53%) in 2011 to 8.75% (95% CI: 8.14% to 9.36%) in 2015.
In the case of the mean positive overshoot (MPO), it was interesting to note that in 2011, those spending more than 10% of their total expenditure on healthcare spent on average 32.89% (10%+22.89%) of their total expenditure on healthcare. This proportion grew over the study period and reached 41.25% (10%+31.25%) by 2015. When healthcare expenditure was considered as a share of non-food expenditure, the level of this mean positive overshoot was very similar. In 2011, those spending more than 40% of their non-food expenditure on healthcare spent on average 54.96% (40%+14.96%) of their non-food expenditure on healthcare, and this grew to 68.23% (40%+28.23%) by 2015. Consequently, the intensity of CHE grew over the study period.
Determinants of CHE
Table 3 presents the results of the logistic regression analysis of the longitudinal data and the cross-sectional data from the 2015 CHARLS wave based on two different denominators: the determinants of CHE at 10% of total expenditure and the determinants of CHE at 40% of non-food expenditure. For the fixed effects model, variables such as gender and region were automatically deleted because they did not change over the survey period. We use the cross-sectional data from the 2015 CHARLS wave to analyse the impact of zone and area separately.
Table 3. Determinants of the prevalence of catastrophic health expenditure using a panel logistic regression model
|
(1)c
|
(2)d
|
(3)e
|
Variables
|
Odds Ratio
|
Odds Ratio
|
Odds Ratio
|
Predisposing factors
|
|
|
|
Household has a member older than 75 years
(compared with younger than 75)
|
|
|
|
Household has a member older than 75 years
|
1.370a
|
1.162
|
1.061
|
|
(0.107)e
|
(0.438)
|
(0.552)
|
Married (compare with no spouse)
|
|
|
|
Living with spouse
|
1.731***
|
1.776***
|
1.767***
|
|
(0.004)
|
(0.003)
|
(<0.001)
|
Enabling factors
|
|
|
|
Zone (compared with eastern area)
|
|
|
|
Middle area
|
|
|
1.363***
|
|
|
|
(0.008)
|
Western area
|
|
|
1.317**
|
|
|
|
(0.017)
|
Area (compared with urban)
|
|
|
|
Rural
|
|
|
0.825*
|
|
|
|
(0.078)
|
Insurance status (compared with no insuranceb)
|
|
|
|
UEBMI
|
1.010
|
1.955
|
0.791
|
|
(0.982)
|
(0.106)
|
(0.592)
|
UBMI
|
1.305
|
1.503
|
0.664
|
|
(0.353)
|
(0.166)
|
(0.323)
|
OI
|
2.498
|
2.199
|
0.249
|
|
(0.214)
|
(0.270)
|
(0.216)
|
Economic status (compared with poorest)
|
|
|
|
Poor
|
0.502***
|
0.583***
|
0.527***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Average
|
0.345***
|
0.449***
|
0.351***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Richer
|
0.198**
|
0.327***
|
0.310***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Richest
|
0.079***
|
0.188***
|
0.116***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Household size (compared with fewer than 4 members)
|
|
|
|
Household size more than 4
|
0.970
|
0.871
|
0.899
|
Need factors
|
(0.763)
|
(0.179)
|
(0.596)
|
Impatient compared with not impatient
|
|
|
|
Inpatient
|
1.793***
|
1.763***
|
2.641***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Chronic disease compared with no chronic disease
|
|
|
|
Chronic disease
|
1.043
|
1.191
|
2.056***
|
|
(0.910)
|
(0.626)
|
(<0.001)
|
Disability compared with no disability
|
|
|
|
Disability
|
1.926***
|
2.424***
|
1.127
|
|
(0.005)
|
(<0.001)
|
(0.189)
|
Outpatient compared with not outpatient
|
|
|
|
Outpatient
|
2.706***
|
2.424***
|
3.294***
|
|
(<0.001)
|
(<0.001)
|
(<0.001)
|
Note: (a) * p<0.1, ** p<0.05, *** p<0.01; (b) Abbreviations: UEBMI, Urban Employee Basic Medical Insurance; NI, no insurance; UBMI, Unified Basic Medical Insurance for urban residents without formal employment and rural residents; OI, other insurance; (c) CHE was defined as 10% of total expenditure; (d) CHE was defined based on 40% of non-food expenditure; (e) The cross-sectional 2015 CHARLS data for logistic regression, and CHE was defined as 40% of total non-food expenditure; (f) P-value in parentheses.
In column (1), where CHE was defined as 10% of total expenditure, age and household size were not significant determinants of CHE. Compared with not living with a spouse, living with a spouse increased the prevalence of CHE by approximately 1.73 times (p=0.004). Compared with those with NI, who had to pay the total cost of health care services out-of-pocket, those with UBMI, UBMI and OI were estimated to be 1.01 (95% CI: 0.42 to 2.41; p=0.982), 1.31 (95% CI: 0.74 to 2.29; p=0.353) and 2.49 (95% CI: 0.59 to 10.57; p=0.214) times more likely to experience CHE, respectively, though none of the differences were statistically significant. This means that the various types of health insurance did not significantly reduce CHE. The socioeconomic status of households was another key driver of CHE. Compared with the poorest group, the richest group was 0.08 (95% CI: 0.04 to 0.13; p<0.001) times and the medium and the richer groups were 0.35 (95% CI: 0.25 to 0.47; p<0.001) and 0.20 times (95% CI: 0.14 to 0.28; p<0.001), respectively, more likely to experience CHE. The poor group was 0.50 times (95% CI: 0.39 to 0.64; p<0.001) more likely than the poorest group to experience CHE. Regarding the need factors, those who used inpatient services in the last review year and those who used outpatient services in the last review month were 1.79 (95% CI: 1.42 to 2.27; p<0.001) and 2.25 (95% CI: 2.19 to 3.34; p<0.001) times more likely than those who did not use such services to experience CHE. Those whose households had member(s) with a disability were 1.93 (95% CI: 1.22 to 3.04; p=0.005) times more likely to experience CHE than households with no disabled persons. Last, households with a member with a chronic disease were only 1.04 (95% CI: 0.51 to 2.15; p=0.91) times more likely than households without a member with chronic disease to experience CHE; this difference was not significant.
In column (2), where CHE was defined based on 40% of non-food expenditure, most of the results were similar, though the effect of socioeconomic status was slightly larger than that in column (1). For example, the richest group was 0.19 times (95% CI: 0.12 to 0.30; p<0.001) more likely than the poorest to experience CHE, while the average and the richer groups were 0.45 (95% CI: 0.33 to 0.61; p<0.001) and 0.33 (95% CI: 0.23 to 0.47; p<0.001) times more likely, respectively. The poor group was 0.58 (95% CI: 0.45 to 0.75; p<0.001) times more likely than the poorest group to experience CHE.
In column (3), we use the 2015 CHARLS cross-sectional data to further examine the impact of zone and area on CHE. Here, we select 40% of non-food expenditure as the standard for CHE. The results reveal that households in the middle and western zones face a 1.36 (95% CI: 1.08 to 1.71; p=0.008) and 1.32 (95% CI: 1.05 to 1.65; p=0.017) times higher prevalence of CHE, respectively, than households in the eastern zone. Households in rural areas face a lower prevalence of CHE, approximately 0.83 (95% CI: 0.67 to 1.02; p=0.078) times that of households in urban areas. The results for the other variables were very similar.