Household Expenditures And Direct Medical Costs Among Older Adults In Ghana: Evidence From WHO-SAGE Wave 2

Background: The Ghanaian population aged 60 years and older will almost double to reach 14% of the total population by 2050. Ascertaining the pattern of health expenditures among this growing population group is important to inform policy makers about the targets for nancial risk protection as part of achieving Universal Health Coverage (UHC) by 2030. This study aimed to estimate household expenditures among older adults and determine their direct medical costs. Methods: The World Health Organization's Study on global AGEing and adult health (SAGE) Wave 2 was conducted in China, Ghana, India, Mexico, Russian Federation, and South Africa between 2014 and 2015, as a follow-up to Wave 0 in 2003 to 2004 and Wave 1 in 2007 to 2010. Survey questions explored sources of income and total directhousehold expenditures in the year preceding the interview. SAGE Ghana was implemented using face-to-face interviews in a nationally representative sample of persons aged 50+ years with a comparison sample of younger adults aged 18–49 years. Findings: Analyses included a total of 4,735 participants, with 1,948 (43.8%) males and 2,787 (56.2%) females, with a mean age of 57.6 (±16.7). The average annual household expenditure was US$ 2,458 (±7,374). Older adults had higher expenditure levels at US$ 2,501 (±8,307), as compared to younger adults, US$ 2,309 (±3,056). Direct health-related and food costs accounted for 10% and 35% respectively of the total household expenditure. The prevalence of catastrophic health expenditure in households of older adults in Ghana was 7.5% (95% CI 6.0% - 9.3%). Conclusion: These updated estimates on household expenditure among older adults provide needed evidence to support the inclusion of social protection mechanisms in the national ageing agenda. The National Health Insurance reduced the burden of direct health expenditure for households with older adults.


Background
Like many other countries, Ghana is experiencing a demographic transition charcterised by rapid aging. The population aged 60 years and older in Ghana is currently 5.3% and is projected to reach 14.1% of the total population by 2050 (1,2). The ageing population typically modi es disease burden patterns in countries, with increasing impact of age-related health conditions, while emerging economies like Ghana, in addition, continue to cope with the burden of infectious diseases, injuries and non-communicable diseases (3). It seems however, that many health services may not adequately be prepared for services tailored to the peculiar needs of the elderly (2). Furthermore, older adults in Ghana could often be at higher nancial risk of impoverishing health expenditures as a result of often having greater health and long-term care needs compared to younger people.
The introduction of the National Health Insurance Scheme in 2003 has contributed to Ghana's targets to achieve Universal Health Coverage (UHC), although inequalities remain in health service coverage and nancial risk protection (4). A National Ageing Policy for Ghana (5) was adopted in 2010, seeking to achieve holistic social, economic and cultural re-integration of older people into mainstream society. It also sought to equip them to participate fully in the national development process. Enforcing social protection systems for this population group would be a crucial component of achieving the National Ageing Policy objectives, and would support older Ghanaians to get the health care services they need in the pursuit of Universal Health Coverage (UHC). Despite some comprehensive policies, not much has materialized towards implementing the National Ageing Policy and Implementation Action Plan (7,8) The aging population in uences many aspects of society, including economy, labour markets, pensions, taxation, housing demand, and healthcare services (1,3). They are therefore a critical component of the household, which largely remains the unit of production and utilization, especially in emerging economies. Household expenditure analysis is used widely for monitoring general household living standards, wellbeing, and consumption patterns, giving indicators for measures to alleviate severe nancial hardship or impoverishment, such as those for healthcare (catastrophic health expenditure) (8). While continuous evaluation of the household expenditure patterns is relevant for policy planning and implementation, there seems to be insu cient data in this regard in Ghana.
Ascertaining the pattern of expenditure among this unique group does not only inform policy makers about heterogeneity in demand among older people, but gives an idea of how demand and costs are evolving in this population (9). A degree of ambiguity remains in the relationship between ageing populations and health care expenditure (10). The objective of this study is therefore, to explore the levels and patterns of household expenditures in a nationally representative sample of households of the elderly in Ghana, with a focus on determining direct medical costs within overall expenditures.

Data collection
Data was obtained from the 2014/15 World Health Organisation (WHO) Study on global AGEing and adult health (SAGE) Ghana Wave 2. This included a follow-up sample from Wave 1, that consisted of nationally representative cohorts of adults aged 50 years and older, and smaller comparative samples of people aged 18-49 years (younger adults). Further information on SAGE, with country-speci c household-level and person-level analysis weights, was made available by WHO (11).

Participants, sampling and variables
Recruitment of participants and implementation of the study was coordinated by the SAGE Ghana Team. SAGE Ghana employed a strati ed random sampling strategy with households as the nal sampling units (12). More details about the sampling strategies are available elsewhere (13). The sociodemographic characteristics information collected included age, sex, educational level, nancial assistance, marital status and health insurance. Participants also self-assessed their perception of health as being 'Very Good', 'Good', 'Moderate', 'Bad', or 'Very Bad'. Outcome variables were total household expenditures and total direct medical costs over the year preceeding the interview. For each participant, the overall consumption and health expenditures of their households were determined from sub-variables obtained by interviewing an identi ed household informant. The total household expenditure and direct medical costs were de ned as the sum of obtained sub-variables listed in Table 1. The catastrophic health expenditure was also determined for participants. In this study, expenditure was de ned as being catastrophic if a household's direct medical cost was greater than or equal to 40% of total expenditure of the household, after subsistence needs (food) have been met (13,14).

Data analysis
Total direct medical costs and total household expenditures were estimated by summing the costs of the individual categories (Table 1). Variables which were ascertained for a 30-day period were multiplied by 12 to obtain estimates of annual costs. All estimated costs were converted into US$ using an exchange rate of GHS 3.21 (the average exchange rate over the period this study was conducted).
All variables were analyzed using Stata 14 software (StataCorp. College Station, TX). SAGE wave 2 used adjusted weights due to the complex nature of the study design. In determining estimations, adjusting for the participants' weighting, strati cation and clustering in a complex survey dataset was a key issue considered during data analysis. SAGE wave 2 adopted a multistage cluster survey design, and adjusted for the primary sampling units, strati cation, and the sampling weights to reduce bias and to improve data analysis in all estimates.
Descriptive statistics were used to describe participant characteristics and mean values of cost estimates between groups compared. Proportions of participants with catastrophic health expenditure were reported, while Chi-square test was used to test association between categorical variables, at 95% con dence, assuming an alpha level of 0.05. Most of the participants with no formal education were older adults (49.9%), compared to the 18.3% of the younger adult population (P<0.001). Older adults were signi cantly more likely than younger adults to perceive their health to be either 'bad' or 'very bad' (OR=3.9, P<0.001). The marital status and health insurance coverage distribution among older adult age groups did not vary signi cantly. About 40.3% of households had received nancial assistance within the past year ( Figure 1). Of these, 32.2% were in the older adult group, compared to 8.0% of the younger adult population (P=0.073). Also, 14.6% were in the rural areas, compared with 25.6% in the urban areas (P<0.001).
The average annual household expenditures among adults in Ghana varied considerably among different subgroups ( Table 2). The average household expenditure per annum for the entire population was US$ 2,459±7,375, with a US$ 246±1,922 average annual household expenditure on health. Direct medical costs averagely accounted for 10% of the total household expenditures with levels different for younger and older adults (12.3% and 9.5% respectively). The proportions of the household expenditure components among older adults are shown in Figure 2. The prevalence of catastrophic health expenditure generally among older people in Ghana was 7.5% (95% CI 6.0% -9.3%). That of the entire population was however 7.8% (95% CI 6.5% to 9.4%). The household catastrophic health expenditure did not vary signi cantly across any of the variables studied.
Comparing the pattern of direct medical costs and total household expenditure across the sociodemographic characteristics, the average direct medical cost differed signi cantly across the health insurance status and the residential status. There was however no signi cant difference in the average total household expenditure across the different characteristics (Table 2). Out-of-pocket expenditures on transportation and traditional or alternative health therapies were higher among households of older adults, compared to the younger adult population. Though not signi cant, the inverse was observed for expenditurs on medications, dental care, food, housing and utilities (Table 3).

Discussion
This study set out to determine the pattern of household expenditure among older adults in Ghana, while ascertaining their direct medical costs. The results from this study comes as pertinent for several reasons, especially in a low-middle income country such as Ghana and similar countries in the subregion.
Ghana has shown a gradual change in its demographic characteristics due to population ageing (16). The somewhat gradual pace provides valuable lead time for strategic planning and policy development to ensure systems are prepared for the coming changes. Estimates of household expenditure, such as those from this study, could therefore be used to inform policies, with regards to prioritization, health insurance, reimbursements, social services, housing, transportation and agriculture among older persons.
It is important to establish that though the pattern of household expenditure holds great predictive value in planning, it is affected by many individual preferences which may be di cult to account for (behavioural heterogeneity) (9). There are also assertions of age-associated health conditions which could increase the household expenditures among older adults.
This study found the average household expenditure among the study participants to be US$ 2,459. These low estimates, though excluding taxes, are considered characteristic of low income countries which Ghana was, at the time of the study. Furthermore, the average direct medical cost, being US$ 246, accounted for 10% of the average total expenditure. It has been suggested that at low income levels, spending diversity is low, as food and healthcare might be expected to dominate spending (9). In corroboration with this, food and healthcare took a proportionate 35% and 10% of the total household expenditure in this study. In comparison to these, average annual health expenditure was estimated to be US$7,439 among older populations in Australia at about the time of this study (10). Also, in this study, 5% of household expenditure was accrued by medications, while 3% was spent on herbal/alternative medicines. This further substantiates the high prevalence of alternative/traditional medicine usage among older Ghanaian adults (17). Compared to younger adults, older adults spent more on transportation and traditional or alternative health therapies. While the elderly are likely to spend more on conveyance, especially for those with mobility related illnesses, the cultural context in uences ageing and health practices. Traditional and alternative medicine use is inadmisible to the cultural setting of many people in sub-Saharan Africa, possibly accounting for its signi cant presence in the household expenditure of older adults in this study (18).
The population characteristics of this study is similar to ndings of Issahaku and Neysmith (2013), who also reported that a higher proportion of the elderly in West Africa is female, with many having little or no formal education, and found more in rural areas (19). Comparatively, Ogura and Jakovljevic (2018) also describe the elderly in higher income countries as being mostly female, poor, and living alone (3).
Aside being a LMIC, the pattern of expenditure among older Ghanaians generally differs from that of Japan (20), UK (9,21) , and the USA (22), due to the entire prevailing social, political and cultural systems. For instance, health nancing in Ghana is mainly publicly funded through the National Health Insurance scheme (NHIS), which, as of the commencement of this study in 2014, covered some 10.4 million Ghanaians, approximating 40% of the country's population (23). The fairly large coverage, which is free for adults aged 70-plus years, and the 'core poor' de ned as the unemployed with no visible source of income, no xed residence, and not living with someone employed and with a xed residence (24), could signi cantly have contributed to the comparatively lower out-of-pocket expenditure on health in this study. Also, the 27% expenditure on household and utilities could be a manifestation of the deep-rooted housing crises in Ghana, which has seen several attempts at reforms over the past few decades (25).
With rising cost of rent, cheaper alternatives and settlements, especially in rural areas could be the preference among older adults. The 17% 'Other' category ( Figure 1) could be a re ection of the heterogeneity of the study population. Included in this category are taxes, gifts, recreation, among others.
Safeguarding people from catastrophic payments, that is, nancial hardship caused by health payments, has come to be a widely accepted index, and a desirable objective in health policy. This study found the prevalence of catastrophic health expenditure among older adults in Ghana to be 7.8%, using a 40% threshold. The catastrophic household health expenditure was reported to be 1.6% from a survey in 1998 (8). As catastrophic health spending is generally thought to be rising, our gure comparatively relates to the 7% prevalence found in India (15). It was also reported to be 22.2% in Iran (26), 6-15% in Burkina Faso (27), and 8-14% in Thailand (28). Nearly 40% of the households had obtained nancial assistance, while many (30%) had salaries being the major source of nancing. A signi cant 34% of older population had to borrow money or sell their property to nance their expenditure. Nearly half of the older population had also not had any formal education, an observation which could inform the policy process and its implementation. These ndings necessitates the need for a broader discussion on health nancing, as the transition to a more aged society brings considerable nancial challenges to any economy (3).
Beyond the effective and more tailored implementation of the NHIS, some authors have suggested for the provision of basic income for persons aged 60 years and above (19), while free or subsidized transportation could be considered of the very elderly in the society.
This study is limited by the few variables explored. Also, being cross-sectional, this study provides only point estimates, for which care must be taken in the interpretation of its results. Recall bias, as in many costing studies could have in uenced households' report of expenditures.

Conclusion
This study in many ways highlights the unmet need for social support for older Ghanaians. In providing estimates and pattern of the expenditure of this group, detailed planning could be incorporated into social interventions. Findings from this study could also be used in hypotheses generation for future studies to explore how the older population in Ghana is evolving.  Proportions of cost areas among the elderly in Ghana