Non-medical costs can constitute a substantial part of total healthcare costs, especially for older people. While these costs may be deemed irrelevant from a healthcare perspective, they may influence the healthcare provided to older people. Most older people will experience at least one health problem that counts as a disability; many will undergo multiple issues by the time they reach the age of 85.Footnote 1 The associated health effects can be profound. There may be high costs relating to carers, travel, food and accommodation for family members who accompany and care for older people during their medical visits.
Notable differences have been observed in socio-economic characteristics and health outcomes. These differences are likely to become more pronounced with advanced age as it has been observed that disabilities are often more prevalent in older people belonging to lower socio-economic groups.Footnote 2 As the poor tend to incur higher opportunity costs than the rich, the impact of non-medical costs (which are often paid out-of-pocket by these individuals) can significantly influence the material living standards of their households.Footnote 3 Older people from poorer socio-economic groups often have difficulties in meeting these costs, thereby worsening existing health inequalities.
This situation may be more problematic for older people residing in rural China, where health facilities are either scarce or of poor quality.Footnote 4 Rural farmers are more likely to bypass local practitioners to seek care in urban hospitals, which may increase both medical and non-medical costs.Footnote 5 Since differences in wealth can also profoundly influence access to care and affordability, healthcare payment-related problems are particularly acute for rural older people with little wealth who are already struggling to maintain their livelihoods.Footnote 6
This study seeks to examine the effects of non-medical costs on catastrophic health payments and health payment-induced poverty among older people from urban and rural areas in China, where few related empirical studies have been conducted. It first presents a literature review to identify meaningful and applicable ways of measuring non-medical costs for inpatient care. This is followed by another literature review on the non-medical costs of patient care for older people and any related implications on inequity. The data used in this study are derived from the China Health and Retirement Longitudinal Survey (CHARLS) 2015. The findings suggest that policymakers need to modify existing policies, or develop new ones, to facilitate the reimbursement of non-medical costs and improve healthcare systems in general and particularly for the rural population.
Conceptualizing Non-medical Healthcare Costs
There is widespread contention and debate over the definition of non-medical healthcare costs. As this study focuses on patient costs borne by the individual, all costs associated with receiving care were considered. Specifically, three main categories of non-medical costs were examined, as follows.
Category 1: costs of loss of productivity
The value of potentially lost productivity often stems from short or long-term absences from work as a consequence of mortality and morbidity.Footnote 7 Although many health and illness studies have tended to ignore or underestimate the costs associated with productivity losses, some of the highest non-medical costs arise from working time lost as a result of illness. These costs are often significantly higher for older people when compared to the general population.Footnote 8 Methods of accounting for the costs related to productivity losses have traditionally either been derived from the potential value of foregone earnings as a result of illness or by considering the estimated elasticity for labour time versus labour productivity.Footnote 9 The latter method requires information on elasticity, which may vary substantially across occupations and sectors and is often difficult to obtain. This study therefore estimated productivity losses owing to lost earnings, which were evident through the data.
Category 2: carer costs
Carer support accounts for a significant part of the total healthcare provided to older people, especially in situations of chronic illness.Footnote 10 The care given to older people usually involves a mixed support network consisting of hospital assistance in addition to paid and unpaid carers. If care is provided at the expense of paid labour, then any analysis should account for the carer's labour costs. If care is voluntarily delivered by informal unpaid family carers, then such unpaid work imposes additional costs through individual household expenditures, travel and food expenses or lost earnings, but not out-of-pocket payments. Opportunity costs can also be incurred by informal carers. However, these are often difficult to accurately assess and therefore were not included in this analysis.Footnote 11
Category 3: travel, accommodation and food costs
It can be complex and/or costly to travel to healthcare facilities in places where health services are geographically distant or there is poor access to public transportation. This means that there are additional costs associated with travel, accommodation and food for family members who accompany older patients while they visit these facilities. These expenses were also considered in this analysis.
Non-medical Healthcare Costs for Older People and Associated Implications for Equity
Most empirical studies on the costs of health and illness have focused on a narrow range of healthcare services costs. However, a few studies that have focused on older people have recognized the importance of including non-medical or societal costs in their analyses. For instance, Sven Heinrich and colleagues examined primary care service utilization and costs among older people aged 75 and above in Germany and identified that costs associated with carers, assisted living and transportation accounted for approximately 18 per cent of the total patient care costs.Footnote 12 Research has also indicated that the costs associated with nursing, community care and domiciliary care, which involve personal care, significantly contribute to the overall patient costs for older people with cognitive impairments.Footnote 13 For example, informal and social care costs can contribute up to 40 per cent of the total costs for dementia patients in high-income countries, while the direct health costs tend to be much lower. Informal care also accounts for the majority of the total costs in low- and middle-income countries.Footnote 14
Few studies have explicitly measured the extent to which non-medical costs vary according to socio-economic status. A Danish study compared the costs incurred by patients suffering from psoriatic arthritis (PsA) (a common chronic inflammatory skin disorder among older people) with those incurred by the healthy population and found that patients with PsA had higher total health care costs and spent an average of €10,641 per patient on non-medical costs. These patients were also more likely to be unemployed, disabled and experience co-morbidities.Footnote 15 Scholars have also argued that poor older people are exposed to increased cardiovascular risk factors and may incur higher costs compared to those with greater financial means; this situation is worse for those in developing countries.Footnote 16
There has been scant research on patient care costs among older people in China, despite the increasing importance of this topic for policymakers. In China, large geographical variations exist in terms of healthcare provider availability. In the vast rural areas, where the number and quality of healthcare facilities are lower than those in urban areas, many older people may need to travel to a city to see a specialist, which often involves high non-medical costs. While these expenses may be affordable for some, they can be a significant deterrent for those belonging to lower socio-economic groups.Footnote 17 For instance, Zeng Yi, Chen Huashuang and Wang Zhenlian together investigated the utilization and cost of healthcare based on the demographic characteristics of older people in 22 provinces in China.Footnote 18 Results indicated that females who were disabled and without children were more likely to incur high health and social care costs. Scholars have also argued that patient costs usually account for a significant proportion of household expenditures and tend to be higher for older people with chronic illnesses.Footnote 19
Methods
Data source and study sample
The data used in this study were drawn from the 2015 China Health and Retirement Longitudinal Survey (CHARLS). Using a multi-stage sampling process, the CHARLS interviewed households comprising persons aged 45 and above from 28 provinces/autonomous regions. The CHARLS was designed to investigate issues relating to demographic characteristics, socio-economic status and health-related factors. It also contained questions on health service utilization, insurance coverage, health providers and health facilities, in addition to items concerning the service accessibility, travel costs and perceived quality of care.
This study primarily focuses on the impact of non-medical costs for older people. Both urban and rural samples for individuals aged 60 years and above were included for analysis. In comparing the level of catastrophe and health payment-induced poverty for rural and urban populations, all empirical analyses were conducted among the whole (N = 5,329), rural (N = 4,038) and urban (N = 1,291) populations. Cross-sectional individual weights were adjusted according to the individual, and household non-responses were applied to all descriptive analyses.
Variable specifications
Table 1 provides a statistical summary and variable specifications. Two categories of inpatient cost variables were distinguished in this analysis. The first category was inpatient out-of-pocket (OOP) health costs, which were defined as the inpatient costs paid to hospitals after insurance reimbursement. Interviewees were first asked if they had received inpatient care in the 12 months prior to the survey. If answering yes, respondents were then asked to recall their most recent inpatient visit, as follows: “How much did you or will you eventually pay, out of pocket, to the hospital for your hospitalization?” The second cost category included inpatient OOP health costs plus non-medical costs. Respondents were also asked to report on the two following items: “the total costs of hired carers” and “the total costs for transportation, food, and accommodation for yourself and relatives.” After adding the costs associated with productivity losses, which were calculated based on the number of days respondents were absent from work because of inpatient visits multiplied by the individual's daily income, the resulting number constituted non-medical costs. All outliers in the inpatient OOP health costs measurement (i.e. the top and bottom 0.5 per cent of all cases) were mitigated through Winsorization.Footnote 20 Per capita household expenditures were used as a measurement of living standards and were adjusted according to household size using the equivalence scale method.Footnote 21 The statistical summary also indicated that rural respondents were considerably poorer, less educated and less healthy when compared to urban respondents. Table 1 shows variable specifications and descriptive statistics for a set of health needs and socio-economic variables.
Table 1: Summary Statistics
Empirical strategies
(a) Measuring catastrophic health payments. In this study, catastrophe is defined as the catastrophic health payments that occur when health costs exceed 40 per cent of the household per capita expenditures (net of food expenditures). This definition was introduced by the World Health Organization and has been used in many studies.Footnote 22 It is estimated using the following equation:
$$H_k = \displaystyle{{\sum\limits_{i = 1}^N {x_{ik}}} \over N}$$ Where H is catastrophic headcount (the proportion of older people who fall below the catastrophic thresholds), i denotes each observation in the sample (N), x i = 1 if 
$\displaystyle{{Q_{ik}} \over {T_i}} \gt z$ (the ratio of inpatient cost (Q ik) as accounted for in the household per capita expenditure (T i) exceeds the 40 per cent threshold (z)); it is 0 otherwise. Q denotes k different categories of inpatient costs (k = inpatient total health costs, inpatient total health costs plus indirect costs, inpatient OOP health costs, and inpatient OOP health costs plus indirect costs).
The intensity of the payment (overshoot) was measured by the average amount exceeding the catastrophic threshold, as follows:
$$O_k = \displaystyle{{\sum\limits_{i = 1}^N {x_{ik}(\displaystyle{{Q_{ik}} \over {T_i}}-z)}} \over N}$$Mean Positive Overshoot (MPO) is defined as intensity divided by headcount, as follows:
$$MPO_k = \displaystyle{{O_k} \over {H_k}}$$The measures of catastrophic headcount and overshoot required consideration of the distributions of these estimates across income groups. This is because the opportunity costs for the poor are usually greater than those for the rich.Footnote 23 Both measures can be adjusted using the Concentration Indices (CI) method. For catastrophic headcount measures, Guido Erreygers's concentration indices were used because the binary nature of the variable formally called for a non-linear measure.Footnote 24 Catastrophic overshoot was measured using the concentration indices introduced by Owen O'Donnell et al.Footnote 25 These indices indicate the distribution of the catastrophic headcount (Ch) and gap (Co) relative to household income, as follows:
$$H_k^w = H_k(1-C_k^h )$$
$$O_k^w = O_k(1-C_k^o )$$ Where 
$H_k^w $ denotes the weighted headcount for inpatient cost category k, and 
$C_k^h $ denotes Erregyers's CI for the catastrophic headcount, 
$O_k^w $ represents the weighted overshoot and 
$C_k^o $ represents the CI for weighted overshoot. This equation is equivalent to a weighted sum of a catastrophic headcount or an overshoot variable by multiplying weights declining linearly from 2 to 0 as the household ranks from poorest to richest. Here, poor households were likely to receive more weight (i.e. if those exceeding the catastrophic threshold tended to be poor).
(b) Health payment-induced poverty. The extent to which health costs impact household material living standards can also be estimated by examining health payment-induced poverty. This study followed the method introduced by O'Donnell and colleagues, wherein incidence and severity of health payment-induced poverty are compared as the two patient cost variables.Footnote 26 Incidence was measured according to the number of people who fall below the poverty line because of health payments (i.e. headcount), while intensity was measured according to the amount by which a household falls below the poverty line because of health payments (i.e. gap). This study used two poverty thresholds: the international poverty line of US$ 1.9 per person per dayFootnote 27 and the Chinese National Poverty Line (NPL), which, as of 2016, was a net per capita income of 2,300 yuan per year (US$ 0.95 per day).Footnote 28
This study also plotted a revised version of Jan Pen's Parade,Footnote 29 which is defined as “a succession of every person in the economy, with their height proportional to their income, and ordered from lowest to greatest.”Footnote 30 In this case, a parade is plotted using household expenditures per capita gross of any health costs on the y-axis against the cumulative proportion of the population ranked by the expenditures on the x-axis. This study plotted two additional parades by using household expenditures per capita net of inpatient OOP costs and inpatient OOP costs plus indirect costs, respectively, against the cumulative proportion of the population ranked by each of these expenditure variables. The poverty lines were then plotted along the y-axis to show the proportion of households that had been pushed below the poverty line because of inpatient costs.Footnote 31
Robustness checks
Two sets of robustness checks were conducted to determine whether non-medical costs had any significant effects on catastrophic health payments and health payment-induced poverty. Separate analyses were run for the entire sample (i.e. rural and urban populations) in the first set of checks. An additional analysis for older people with chronic illnesses to determine whether non-medical costs resulted in a greater burden for that population was also performed. No significant differences were found when compared to the entire population. The second set of robustness checks can be found in the Appendix.
Results
Table 2 shows average per episode health costs as a share of household per capita expenditures by quintiles. Inpatient costs were measured in two different categories to show the differences that arose when including and excluding indirect costs in the measures. The first category was inpatient OOP costs, while the second was inpatient OOP costs plus non-medical costs (i.e. transportation costs, accommodation costs for family members and carer costs). Results showed that health costs, as a share of household per capita expenditures, were highest for the poorest households. In addition, the share for the poorest households was more than three times that of the richest households. For instance, inpatient OOP costs accounted for 10.82 per cent of household per capita expenditures for the poorest households, but only 3.02 per cent for the richest households. The share of inpatient OOP costs increased to 12.75 per cent for the poorest households and 3.56 per cent for the richest households when non-medical costs were considered. Looking at the rural and urban populations, it is evident that rural populations (especially those from the poorest and second-poorest groups) spend a significantly higher proportion of their total household incomes on inpatient care as compared to urban populations. The relative difference between the richest and poorest households in terms of the share of inpatient costs was approximately three times higher for the rural households, whereas it was less than two times higher for the urban ones.
Table 2: Inpatient Costs as Shares of Household Income (net food expenditures)
Table 3 shows the inpatient costs and related non-medical costs for all samples and samples with inpatient visits. For all samples, the average inpatient OOP cost was 801.27 yuan, whereas non-medical costs for transportation, meals and accommodation for family members amounted to 146.6 yuan. That is, 18 per cent higher than the inpatient OOP costs. These findings also show that older people from urban areas tend to spend more on inpatient care as compared to those from rural areas.
Table 3: Healthcare Cost Categories in Chinese Yuan Units
Table 4 shows the incidence and intensity of catastrophic inpatient health payments according to different cost categories using the threshold level of 40 per cent of per capita household expenditures (net of food expenditures). The incidence and severity of catastrophes increased when non-medical costs were considered. For the whole population (before any non-medical costs were included), 2.58 per cent (N = 150) of older people fell below the threshold level. This percentage increased to 3 per cent (N = 175) if only considering non-medical costs. The difference was 17 per cent and statistically significant at the 0.01 level. Similarly, catastrophic overshoot significantly increased (i.e. to 25.42 per cent (p < 0.01)) when comparing inpatient OOP costs with inpatient OOP costs plus non-medical costs. The same result held for the headcount measure.
Table 4: Catastrophic Health Payments for Different Inpatient Cost Categories
Notes:
*p < 0.1, ** p < 0.05, *** p < 0.01.
Notably, older people from urban areas were significantly less likely to fall below the catastrophic thresholds (i.e. 3.28 per cent of rural populations fell below the catastrophic threshold, whereas the percentage was only 1.95 per cent for urban populations). Similar results were found for overshoot and MPO.
All CIs for headcount and overshoot were pro-poor, meaning that catastrophe was more likely to occur among poor households than rich ones. The CIs for overshoot were high, thus indicating the existence of pronounced pro-poor inequities in the intensity of catastrophic inpatient costs. The MPO indicated the overshoot for those that had inpatient visits. Inpatient OOP health costs were 76.78 per cent higher than the threshold level and were 86.54 per cent more when non-medical costs were included for the entire population.
Table 5 shows the results for health payment-induced poverty using the NPL (2,300 yuan per year) and the World Bank's international poverty line (US$1.9). Using the World Bank poverty line, our results indicated that approximately 14.04 per cent of all older people fell below the poverty line before any inpatient costs were considered. However, the percentages increased to 15.02 per cent and 15.28 per cent when inpatient health costs and non-medical costs were considered, respectively. The poverty gap was 277.60 yuan below the poverty line before any health costs were considered. This gap increased to 400.20 yuan when inpatient health costs were added to the analysis. There was a further increase to 436.90 yuan when both health costs and non-medical costs were added. Similar results were found in the OOP patient costs category and when using the NPL. All differences were statistically significant.
Table 5: Healthcare Payment-induced Poverty Measures (gross and net of different inpatient cost categories)
Notes:
NPL = national poverty line; *p < 0.1, ** p < 0.05, *** p < 0.01.
Results unsurprisingly indicated that rural populations were more likely to fall below the poverty line than urban populations as a result of inpatient OOP costs. However, urban populations were more likely to fall below the poverty line owing to non-medical costs. In particular, 1.22 per cent of all rural populations fell below the World Bank poverty line as a result of non-medical costs (the percentage was 5.8 per cent for urban populations).
Figures 1 and 2 show Pen's Parade results for household expenditures per capita for the poorest 20 per cent using both the World Bank poverty line and the NPL for urban and rural populations, respectively. Figure 1 shows three lines indicating gross and net household expenditures per capita for different inpatient cost categories for rural populations. The X-axis indicates the cumulative percentage of the population ranked by household expenditures per capita gross and net of inpatient cost categories, while the Y-axis indicates household expenditures per capita. The dashed black line indicates household per capita expenditures gross of any health costs. The solid black line indicates household per capita expenditures net of inpatient OOP costs. The dotted line represents household per capita expenditure net of inpatient OOP costs and non-medical costs. As a result of inpatient costs, urban populations were less likely to fall below the poverty line than rural populations (Figures 1 and 2; Table 5). However, there was also a higher probability that urban populations would become poor after incurring non-medical costs.
Figure 1: Effect of Different Inpatient Costs on Pen's Parade of the Household Expenditure Distribution for Rural Populations
Figure 2: Effect of Different Inpatient Costs on Pen's Parade of the Household Expenditure Distribution for Urban Populations
Discussion and Conclusions
Although China has made remarkable achievements in strengthening its healthcare services, this study suggests that increased efforts are needed to reduce the financial burden of health costs for poor older people, many of whom are failing to seek medical treatment or are falling below the poverty line as a result of both medical and non-medical costs. This study is among the first to investigate the effects of inpatient care costs on catastrophic health payments and health payment-induced poverty for older people by incorporating non-medical costs into its analysis. Findings suggest that inpatient costs account for a significant portion of household expenditures among older people. Here, non-medical costs may account for approximately 18 per cent. The share is highest for those in the lowest wealth groups. Results also indicate that rural populations are more likely than urban populations to incur catastrophic health payments and suffer from health payment-induced poverty. Non-medical costs also increase the chances that older people will be affected by both of these problems. These effects are more concentrated among rural and poor populations.
This paper offers three related policy implications. First, the non-medical costs associated with receiving healthcare can present difficulties for many older patients; the government needs to recognize this. For instance, travel costs can be especially expensive when patients are required to complete complex and lengthy journeys from rural villages to tertiary hospitals in urban areas. Further, some older patients may require escorts or carers to accompany them in this context. These individuals should have the option to claim these costs. Some developed countries have already implemented healthcare travel-cost programmes to help vulnerable groups cover these expenses. For instance, the National Health Service in the UK reimburses travel and associated costs for low-income patients if they are referred to hospitals or other health facilities for specialist treatment or diagnostic testing.Footnote 32 Similar measures should be considered by the Chinese government. It is also worth pointing out that poor patients seeking care in urban areas may incur particularly high and burdensome costs related to non-medical items (for example, carers, productivity losses and others).
Second, the government needs to address the issue of high healthcare costs. Here, stress is particularly apparent throughout the nation's rural villages, in which 65 per cent of older people live below the poverty line.Footnote 33 Although the New Cooperative Medical Scheme reaches nearly all of China's rural population, the benefits package is rather basic and only covers a narrow range of conditions.Footnote 34 Thus, more comprehensive coverage with higher reimbursement rates for healthcare services should be introduced so that access to care can be improved for older people living in rural areas.Footnote 35
Finally, there should be special emphasis on the need for long-term care while enhancing coordinated and more continuous care for older patients, who are suffering from chronic diseases at an increasing rate. International precedent indicates that services of this type also reduce unnecessary use of and spending on acute health services while helping families avoid catastrophic health payments.Footnote 36
This study has some limitations. First, the costs associated with health deterioration subsequent to inpatient episodes were not evident through the dataset. These costs were therefore not reflected in the findings. Second, this study mainly relied on self-reported data, which may cause bias in the analysis because of inaccurate recall or misreporting.Footnote 37 However, most studies that use data from individual surveys operate with this limitation. Third, approximately 5.71 per cent of those observed in our sample stated that they had forgone treatments. A higher percentage of people having catastrophic health payments or falling below the poverty line can be anticipated if these respondents are included in the analysis. However, it is difficult to know how the results would have been affected in this scenario. Further studies are thus needed to address the methodological omissions of this analysis.
Acknowledgements
I appreciate the valuable comments of the participants at the Long-term Care workshop in April 2018 and the International Long-term Care Policy Network conference in Vienna in 2018. This work is supported by the Future Research Leadership grant of Economics and Social Research Council, UK.
Biographical note
Wei YANG is an assistant professor in global health at King's College London. She is a health economist with an interest in health and long-term care (LTC) research. Her work focuses on measuring the impacts of health and LTC financing policies on inequities in access to care and fairness in financing, and identifying factors that drive inequities using longitudinal survey datasets. She is also interested in issues relating to efficiency in health and LTC financing in China.
Appendix 1: Catastrophic Health Payments for Different Inpatient Cost Categories for Those with Chronic Diseases
Appendix 2: Heath Payment-induced Poverty Measures, Gross and Net, of Different Inpatient Cost Categories
