The National Health Insurance Act (Act 650) of 2003 permits the establishment and operation of three types of health insurance schemes in Ghana. They include the district mutual health insurance schemes, private commercial health insurance schemes and private mutual health insurance schemes. The private commercial schemes operate as limited liability companies, while the private mutual schemes are organised by individual groups of persons for their own benefit. Both types do not receive any financial support from the government. By Act 650, districts are to establish their own district mutual health insurance schemes (DMHIS). The DMHIS are autonomous from each other but all operate under the National Health Insurance Authority (NHIA). The new NHI Act of 2012 (Act 852) merges all the DMHIS to form a nationwide National Health Insurance Scheme (NHIS) which every resident of Ghana shall belong to 37 . The NHIA provides subsidy from the National Health Insurance Fund (NHIF) to the DMHIS for their operations. The state sponsored DMHIS are the dominant health insurance schemes across the country and operational in all the districts of Ghana. A member of the NHIS may also belong to a private health insurance scheme as stipulated by Act 852. This study focuses on the NHIS comprising of the DMHIS because that is public and of national interest.

The NHIA has five main sources of funds for its operations. The NHI Levy which is a 2.5% value added tax (VAT) on selected goods and services which accounts for about 60% of the total revenue making it the major contributor to the NHIF. The other sources of revenue to the fund include a mandatory 2.5% deduction from formal sector workers’ social security contribution managed by the Social Security and National Insurance Trust (SSNIT), sector budgetary support allocated by the Parliament of Ghana, income accruing to the NHIF from investments made by the NHI Council and grants, fees, donations, gifts and voluntary contributions made to the Fund 37 . The premiums paid by members is another source of funds but this accounts for less than 5% of the total inflows to the NHIS 38 .

Membership in the NHIS is at the individual level and supposed to be mandatory by law for all residents of the country. According to Act 852, all employers are also obliged to ensure that all their employees are registered under the NHIS. Individual adults aged 18–69 years in the informal sector pay annual premiums (i.e. direct premium-paying adults) as determined by the DMHIS and approved by the NHIA. The premium ranges between 7.2 Ghana Cedis (GhȻ) (US$4.8) to GhȻ48.0 (US$32) depending on the socio-economic status (SES) of the individual. However due to the difficulty in determining SES of people in the informal sector, the premium is in practice set at a flat rate and varies from district to district. Formal sector workers whose premiums are deducted from their social security contributions (i.e. SSNIT contributors) are exempted from direct premium payments to the scheme to become members. They are however not automatic members of the NHIS unless they decide to enrol with a DMHIS of their choice by paying a registration fee. In 2011, SSNIT contributors constituted only 4.3% of the total active membership of the NHIS 38 .

The NHI Act 852 exempts children under 18 years from paying the premium if at least one parent or the guardian is a valid card holder of the NHIS (this clause in the LI 1809 has been scrapped by the new Act 852 of 2012 though the legislative instrument for its implementation is not yet out). The registration of children under five years has however been decoupled from that of their parents since 2010 and therefore they can be registered even if their parents are not registered. The elderly (≥70 years), SSNIT pensioners, core poor indigents identified by communities and pregnant women (for ante-natal, delivery and post-natal healthcare services) are also exempted from paying premiums to become members of the scheme. Apart from the indigents and pregnant women, all exempt populations are required to pay a registration fee of GhȻ4.0 (US$2.7) but the fee could vary depending on the district scheme.

The NHIS has a comprehensive benefit package which covers over 95% of disease conditions in Ghana 38 . The benefit package however excludes treatment for cancers apart from breast and cervical cancers, HIV retroviral drugs, dialysis for chronic renal failure, hormone and organ replacement therapy and few others.

The study used data from a representative cross-sectional household survey in three districts. For financial reasons we selected only one district each to represent the three main ecological zones of Ghana (i.e. the southern, middle and northern zones). Three districts were Kwaebibirem in the southern zone, Asutifi in the middle zone and Savelugu-Nanton in the northern zone. In 2010, Kwaebibirem had a total population of 200,000 as against 114,029 in Asutifi and 139,283 in Savelugu-Nanton. The three districts were generally rural with agriculture as the main economic activity of the population. Kwaebibirem district is noted for industrial and small-scale mining of diamond and gold while Asutifi is noted for gold mining. While Kwaebibirem and Asutifi have similar socio-economic characteristics, Savelugu-Nanton is relatively poor. All the districts had well established DMHISs with over 70% of the population reported to have ever registered with the NHIS since its inception.

The key indicator of the survey was active membership in the NHIS in 2010 and this informed the determination of the sample size using the formula for sample size for estimation of a single proportion 39 . Assuming that 50% of households in each district were active members of the NHIS and with a confidence level of 95%, 5% margin of error and an estimated design effect of 2.0, a minimum sample size of 768 households per district was required for the survey. Since it was not possible to identify insured and uninsured households from the district schemes, it was decided to select a representative sample in each district in order to guarantee sufficient numbers of insured and uninsured households. In each district, EAs were stratified into rural and urban after which 27 representative EAs were sampled based on the Ghana Statistical Service’s classification and demarcations. All households in each EA were listed to obtain a sampling frame from which 30 households were randomly sampled. Selecting 30 households from each EA resulted in a total of 810 households per district (27 EAs × 30 households) which was closer to the estimated minimum required size of 768. In all, a total of 2,430 households were surveyed in 81 EAs in the three districts. Field interviewers had up to three visits per households if the household head was not available during the first visit. The response rate was 99.5% resulting in a total of 2, 418 households for the analysis. Data collection took place between February and April, 2011 by ten trained interviewers and two supervisors.

A structured household questionnaire was administered to the household head in a face-to-face interview by ten trained field assistants. The questionnaire had modules on demographic composition and socio-economic characteristics of the household, health status (e.g. illness in the last 6 months, presence of chronic illness), health insurance status of household members, reasons for non-enrolment and non-renewal, access to health services, household dwelling characteristics as well as assets ownership. In addition, data on annual household consumption expenditure consisting of expenditures on food and non-food items including health care and imputed values for home produced items such as food and housing as used in the Ghana Living Standard Surveys were collected 40 . The survey tool was developed specifically for this study. Field assistants were trained for seven days after which the survey tool was pre-tested. Based on the results of the pre-testing, final corrections were made to questionnaire before the main survey.

The household was the unit of analysis for the study. A household is defined as a person or a group of persons, who live together in the same dwelling, sharing the same house-keeping arrangements and are catered for as one unit 40 . Households are not required to enrol as a single unit in the NHIS but the decision as to who gets to enrol in the household to a large extent is influenced by the household head who is a prominent decision maker regarding access to health and intra-household resource allocation 13 , 23 , 34 . Studies on household decision making have also observed that joint decision-making is quite common among married couples in Ghana on issues relating to healthcare, use of a partner’s earnings and major household purchases 41 , 42 . Three types of households were identified from the survey with respect to their health insurance status in the NHIS through the district schemes. A household is defined as uninsured if none of its members was insured with the NHIS, partially insured if at least one member was insured or fully insured if all the members were insured at the time of the survey.

The affordability analysis was based on the ‘normative definition’ of affordability which relates the total price of health insurance for all household members to the overall income of the household. Health insurance is defined as affordable to a household if purchasing it leaves enough income for it to meet its other socially defined minimum necessities required for living. This means that the total payment for health insurance plus the socially defined minimum level of spending on other socially required goods should be less than household’s total annual income 27 , 28 , 43 . Thus, using the ‘budget-based approach to affordability’ by Gruber and Seif 28 , health insurance is affordable if a + b < c , where a is household’s minimum level of spending on necessities, b is the household total annual health insurance contributions and c is the household’s total annual consumption expenditure. In the absence of reliable household income data, we used total annual household consumption expenditure data from our survey as a proxy to household available resources 44 . The household total annual consumption expenditure was the sum of the monetary value of all items purchased by the household and home produced items meant for household consumption during the reference year for the survey as reported by the household.

The national upper poverty line of 370 Ghana cedis (GhȻ) (US$246.6 in March 2011) measured in 2006 45 could not be used to represent the socially defined minimum level of spending on necessities because we considered it low and far below the World Bank’s poverty line of US$1.25 per day (US$456.25 per year). Instead, this study adopted the methodology by Xu 46 which uses a food-share based poverty line to estimate household subsistence spending. The method involves the estimation of household’s food expenditure share by dividing the household food expenditure by its total consumption expenditure. It also involves the estimation of equivalised food expenditures for each household to make it easy to compare welfare across households with difference size and demographic composition 47 . This adjusts for the fact that large households need larger food expenditures while taking into account that food consumption needs varies between adult and children and that the marginal cost of feeding one additional person is diminishing, i.e. economies of scale apply. To obtain the subsistence expenditure per adult equivalent, the total household consumption expenditure is adjusted for household size and composition using the formula: AE = (A + αK) ^θ where AE is the number of adult equivalents, A is the number of adults in the household and K is the number of children. The parameter α is the cost of a child relative to that of an adult while θ captures the effect of economies of scale 44 , 47 . Currently, there is no Ghana-specific adult equivalent scale, indicating parameter values. In a recent study on the progressivity of health care financing and incidence of service benefits in Ghana, Akazili et al. 48 used α = 0.5 for children between 0–14 years (≥15 years can be legally employed) and θ = 0.75 and this was adopted for our study. The food-share poverty line is then defined as the average food expenditures of households whose food expenditure share of total household consumption expenditure is within the 45th and 55th percentile of the total sample. By multiplying the subsistence expenditure per capita (i.e. the estimated poverty line) by the adjusted household size (AE), we estimated the total subsistence spending for each household. By these calculations, we derived a poverty line of GhȻ798.42 (US$532.17). A household was then identified as poor if the total household expenditure was smaller than the estimated poverty line. We also created socio-economic quintiles from the equivalised household consumption expenditure to represent the socio-economic status of the households.

Further, we computed the household total annual health insurance contributions (i.e. premiums and registration fees) required to be fully insured. The expected total annual health insurance contributions for each household depended on the total number of premium-paying adults in the household, the number of premium-exempt individuals who pay only the registration fee as well as the premium and registration fee for the district of residence. Data from the three districts showed that in 2010/2011, the registration fee per registrant were GhȻ3.0 in Asutifi, GhȻ4.0 in Kwaebibirem and GhȻ5.0 in Savelugu-Nanton. The premiums were GhȻ13.0 per person in Asutifi, GhȻ14.0 in Kwaebibirem and GhȻ12.0 in Savelugu-Nanton. This means the total contribution per a premium-paying adult in the informal sector was GhȻ16.0 in Asutifi, GhȻ17.0 in Savelugu-Nanton and GhȻ18.0 in Kwaebibirem. We multiplied the individuals in each household (i.e. premium-paying and premium-exempt individuals but registration fee paying) by their respective NHIS contributions as determined by the district schemes. The sum of these contributions for each household constituted its expected total annual NHIS contributions. The affordability analysis was then performed to identify all households which could not afford the expected NHIS contributions for all their members. These households were termed the ‘unafforders’. The unafforders were then categorised into ‘uninsured unafforders’ (i.e. uninsured households which would not be able to afford full insurance for their members) and ‘insured unafforders’ (i.e. insured households which were identified as unable to afford full insurance but have managed to have it) 27 . Significance tests were conducted to check for differences among uninsured, partially insured and fully insurance households. The choice of the statistical test depended on whether the variable was continuous (F-test of means because there were more than two groups) or categorical (Pearson’s chi-square test). All statistical analyses were performed in Stata 11.2.

Ethical approval for the survey was obtained from the Institutional Review Board (IRB) of the Noguchi Memorial Institute for Medical Research (NMIMR) of the University of Ghana with a certified protocol number of 069/11-12. Informed consent was also sought from all participating households.

A total of 11,089 household members were recorded in 2,418 households surveyed. About 28% of the households were categorised as fully insured and accounted for 23% of the household population. About 26% of the households were partially insured households and also accounted for 30% of the household members. The remaining 46% of the households were uninsured and had 47% of the household members. Table 1 presents descriptive statistics of the households comparing uninsured, partially insured and fully insured households. The results show that the households differed significantly in all the characteristics presented. The fully insured households were older (31.7 years), relatively had a higher number of household members with at least secondary education and had more formal sector workers though only 11% of the surveyed households had at least one formal sector worker. The partially insured households on the other hand had higher numbers of children, elderly and premium-paying adults. The uninsured households had fewer numbers of elderly, fewer formal sector workers and members with secondary or higher education. The results also show that the fully insured households reported higher illness per capita in the last six months prior to the survey and had a higher proportion (14.7%) of members with chronic illnesses. The uninsured reported the lowest illness per capita (0.28) and members with chronic illnesses (4.7%). A higher proportion of the fully insured households (66.2%) lived in urban areas compared to the partially insured (60.5%) and uninsured households (52.2%) (p < 0.01). The uninsured had a higher proportion (81.6%) of male heads than the partially insured (69.8%) and the fully insured (66.7%) households (p < 0.01). About 76% of all the household heads were married or had partners. About 22% of the heads of the fully insured households and about 20% of the partially insured households were divorced or widowed compared to only 14.7% among the uninsured. The heads of the uninsured households were the least educated while the fully insured relatively had better educated heads (p < 0.01). Most of the household heads (84.3%) were employed in the informal sector. Regarding socio-economic status (SES) of households, a higher proportion of the fully insured (52.51%) were in the top two socio-economic quintiles compared to only about 35% of the partially and uninsured households (p < 0.01). Significant proportions of the fully insured (72%) and partially insured (67%) households lived less than two kilometres from the nearest NHIS accredited health facilities compared to 55.5% of the uninsured. Nearly 22% of the uninsured were more than five kilometres from the nearest NHIS accredited health facilities. Finally, the Savelugu-Nanton district accounted for about 41% of the uninsured households compared to 32% in Kwaebibirem and 27% in Asutifi.

Source: Household survey, 2011.

Notes: ^aF-test for continuous variables (test of means) and Pearson’s chi-square (χ²) for categorical variables. Test of statistical significance: ***p<0.01.

^bStandard deviations of continuous variables in parentheses (…).

The results show that about 64% (634) of the uninsured and 47% (128) of the partially insured households reported that their members were not insured with the NHIS because they considered the contributions to be expensive. This was followed by 22.6% (223) of the uninsured and 37% (101) of the partially insured households who attributed their lack of insurance to the fact that their members did not fall sick to require a health insurance. About 14% (175) of the households also mentioned others reasons including perceived poor quality of care to the insured, registration difficulties, lack of trust in NHIS officials and inadequate benefit package, among others (Table 2). Among uninsured and partially insured households who found the NHIS contributions to be expensive, about 55% of them belonged to the two lowest socio-economic quintiles. About a quarter of the uninsured and partially insured households were also in the fourth and fifth socio-economic quintiles. For households who thought they did not need health insurance because their members do not fall sick, only about 10% and 14.6% of the uninsured and partially insured respectively were in the first socio-economic quintile.

Source: Household survey, 2011.

^aPercentages are based on row totals.

^b*Significant at 10%.

Table 3 presents summary statistics of the mean expected annual NHIS contributions if households were to become fully insured (full insurance). It also shows the estimated burden of the expected NHIS contributions on household expenditures. The mean expected annual NHIS contributions significantly differ among the districts (p < 0.01) and by the household’s health insurance status (p < 0.01). Savelugu-Nanton had expected NHIS contribution of GhȻ60.85 (US$40.56) compared to about GhȻ39.00 (UD$26.0) in the other two districts. This is because Savelugu-Nanton district had the highest registration fee (GhȻ5.0 compared to GhȻ3.0 in Asutifi and GhȻ4.0 in Kwaebibirem) and largest mean household size (5.7 compared to 4.1 in Asutifi and 3.5 in Kwaebibirem).

Source: Household survey, 2011.

Notes:

^aIn Ghana cedis. US$1 = 1.5003 in March, 2011.

^bTest of means. ***Significant at 1%.

^cExempted from paying the premium but they must pay the registration fee to be members of NHIS.

The results further show that the partially insured had the highest expected annual NHIS contribution of GhȻ53.69 (US$35.79), followed by the uninsured (GhȻ 47.73; US$31.81) with the fully insured households having the lowest (GhȻ38.01; US$25.33). As shown in Table 3, the partially insured household had the largest household size (5.3) while the fully insured households had the smallest household size (3.8). For all households, the expected annual NHIS contribution as a proportion of household non-food expenditure was 5.9% on the average and 2.0% of household total expenditure. The proportions were highest among the uninsured (6.7% and 2.2%) compared to the fully insured households (4.2% and 1.5%).

The expected annual NHIS contributions were significantly higher among households in the lower socio-economic quintiles (p < 0.01). For instance, households in the first socio-economic quintile had the highest expected annual NHIS contribution of GhȻ56.69 (US$37.79) compared to GhȻ33.14 (US$22.10) for households in the fifth quintile. For households in the first socio-economic quintile, the expected annual NHIS contribution was equivalent to 11.4% of the household non-food expenditure compared to just 2.0% for those in the fifth socio-economic quintile. The results further show that households which perceived the NHIS contributions to be expensive had a significantly higher expected annual contributions (GhȻ52.47; US$34.97) than those who did not complain about it (GhȻ47.81; US$31.87) (p < 0.01). Finally, households which thought their members were healthy and did not need health insurance had the lowest expected annual contributions which was significantly lower (GhȻ 44.37; US$29.57) than the other households (GhȻ51.11; US$34.07).

By the normative definition of affordability used for our analysis, 29% (702) of all the households surveyed were identified as unable to afford the expected annual NHIS contributions for full insurance (“unafforders”) (Table 4). Regarding their health insurance status, 34.5% (385) of the uninsured households were classified as uninsured unafforders while 30.2% (187) of the partially insured households were classified as unable to afford to be fully insured. Nearly 19% (130) of the fully insured households were also identified as insured unafforders.

Source: Household survey, 2011.

Notes: *Significant at 10%; **Significant at 5%; ***Significant at 1%.

Table 4 shows the characteristics of the unafforders by their health insurance status. Savelugu-Nanton district accounted for 58.4% of the unafforders followed by Asutifi (25.5%) and Kwaebibirem (16.0%). This is not too surprising since the Savelugu-Nanton district is located in one of the poorest regions of Ghana. In terms of socio-economic status, all the unafforders were in the first and second socio-economic quintiles with almost 69% in the first quintile alone. Though more than half (53.0%) of the unafforders lived in urban communities, a higher proportion of the uninsured unafforders (52.5%) lived in rural communities. There was a higher proportion of male household heads (85.2%) among the uninsured unafforders compared to the other households. The results further show that the expected annual NHIS contribution of the uninsured unafforders was about 11% of their non-food expenditure compared to 10.1% for the partially insured and 8.6% for the insured unafforders. The unafforders also had a mean household size of 5.5, larger than the mean household size of 4.4 for all the households surveyed.

The results have shown that the partially insured households had the largest household size, the highest numbers of children and elderly as well as the highest expected annual NHIS contributions. About 43% of them were in the first and second socio-economic quintiles. The study sought to find out individuals who were more likely to get insured in the NHIS in partially insured households. Females were more likely than males to insure (56.5% vs. 47.0%, p < 0.01) (Table 5). With the premium waiver for children and the elderly in place, a significant proportion of children and the elderly were insured (p < 0.01). Household members whose health status was perceived to be fair (59.6%) or poor (73.9%) were likely to be insured (p < 0.01). Though only 3.4% (113) of the household members had been diagnosed with chronic health conditions, about 83% of them were insured compared to 51% of those without chronic illnesses (p < 0.01). No significant difference was observed in the insurance status of informal and formal sector workers in partially insured households though formal sector workers were more likely to get insured.

Source: Household survey, 2011.

Notes: ^aPercentages are based on row totals.

***Significant at 1%.

This study has some limitations which could affect our results. First of all, the estimation of the expected annual NHIS contributions for households did not take account of indigents (i.e. destitutes) and pregnant women who are exempted from paying the premium and registration fee to enrol in the NHIS. Such persons were not captured by the household survey but having many of them in a household could reduce the cost of full insurance. Not all pregnant women in the households may however benefit from this exemption package either because they refuse to register for it or had enrolled prior to the pregnancy and therefore do not need a new one. We did not have such data.

It is also important to acknowledge that the normative definition of affordability used for the study is subjective and may have its own methodological and theoretical challenges as well as a lack of common acceptance of what constitute affordable premium 28 , 29 . For instance, some insured households categorised as “insured unafforders” could have borrowed to finance their enrolment which they may not be able to repay or could become a heavy burden on future household resources 28 . The consumption expenditure data used for the various measurements were self-reported by households and could suffer from recall errors which could bias the estimates, though quality control measures were adhered to during the data collection exercise. In addition to the use of the quantitative method, the issues of affordability could have been explored further by the use of qualitative methods. This would have made it possible to have a deeper understanding of the enrolment behavour of households. Finally, the lack of a more current national poverty line did not make it possible for us to assess how reliable our estimated poverty line is. Notwithstanding these limitations, the findings from the study are believed to reflect the general situation in the ecological zones the districts represented.