Differences in the determinants of health insurance enrolment among working-age adults in two regions in Ghana

Differences in the determinants of health insurance enrolment among working-age adults in two... Background: Ghana’s National Health Insurance Scheme (NHIS) has achieved varying levels of enrolment within the regions with different rural-urban populations with associated income inequalities. This study sought to investigate the differences in the determinants of enrolment between the Greater Accra (GAR) and Western (WR) regions of Ghana to inform the NHIS reforms. Method: Data from 4214 adults, 18 years and above from a household survey conducted in the two regions was analyzed. Bivariate analysis (t-test for continuous and Pearson chi-square for categorical) was performed to examine differences in respondents characteristics (socio-economic and insurance enrolment) between the two regions for the total, urban and rural samples. Logistic regression estimation was performed to establish differences in determinant of enrolment between the regions. Results: Age, sex, educational level, marital status, health status and travel time to nearest health facility were identified as determinants of enrolment in both regions and among the rural and urban residents within the regions. Although the rich and richest in both regions are more likely to enroll than the poor and poorest, the odds of enrolment for the urban richest in the WR is about twice that of GAR whiles the odds of enrolment for the rural richest in the GAR is also about twice that of the WR. Those who visit public facilities in the GAR are more likely to enroll than those in WR for the total and urban samples. However, those who visit private facilities in rural communities in both regions are more likely to enroll. Conclusion: Differences in the NHIS enrolment between the regions is as a result of differences in socio-economic factors that are intrinsic in the regions and impact on the inhabitants’ ability to afford insurance premium. Policymakers should determine NHIS premium differently at the district level based on socio-economic activities and income levels within the districts. Keywords: Health insurance, Determinants of enrolment, Rural-urban, National Health Insurance Scheme, Ghana Background their entire populations [1, 2]. The 2005 World Health Health insurance has been recognized globally as one of Assembly resolution WHA 58.33 urged members states to the principal methods of financing healthcare to achieve ensure financial protection to all citizens, especially chil- universal coverage, particularly in low and middle income dren and women of reproductive age and “to plan the countries. Many low and middle income countries are transition to universal coverage of their citizens” [2]. currently exploring mechanisms of extending their health Given the high demand for healthcare services of appre- insurance schemes to specific groups to eventually cover ciable quality and the extreme under-utilization of health services in several Sub-Saharan African countries due to financial barriers, health insurance has been recom- Correspondence: stephen.duku@ghsmail.org Department of Epidemiology, Noguchi Memorial Institute for Medical mended as a promising alternative to other criticized fi- Research, University of Ghana, P. O. Box LG 581 Legon, Accra, Ghana nancing systems like cost-recovery and user fees [2]. The Amsterdam Institute for Global health and Development, Amsterdam, The expectation is that health insurance will improve access to Netherlands Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Duku BMC Health Services Research (2018) 18:384 Page 2 of 16 quality healthcare through risk pooling of unforeseeable rural and urban populations [8, 9]. The implementation healthcare cost to fixed premiums [3]. In response to this of the NHIS also saw a steady increase in outpatient call for developing countries to adopt healthcare financing utilization of healthcare services. Outpatients’ utilization mechanisms that remove financial accessibility barriers increased from under 5 million in 2005 to approximately and strive towards the attainment of universal health 24 million in 2012 [9]. coverage, Ghana implemented the National Health Insur- Ghana is considered as a rural country with approxi- ance Scheme (NHIS) in 2004 to replace the “Cash and mately (49%) of the population living in rural areas with Carry” system when patients have to pay out-of-pocket limited socio-economic opportunities [10]. Significant cash before receiving healthcare. inequalities persist between the rural and urban areas in The NHIS was established through an Act of parlia- terms of availability of basic amenities and infrastructure ment, Act 650 in 2003 as part of efforts to make the such as water, sanitation and health facilities [10, 11]. In health goal within the Ghana Poverty Reduction Strategy 2009, only about 10% of the urban population in Ghana (GPRS) achievable and also to accomplish the targets set lack access to portable water as compared to 26% of the in the Health Sector Five-Year Programme of Work, rural population. Whilst 18% of urban population had 2002–2006 [4–6]. The vision of the NHIS was to ensure access to improved sanitation, only 7% of the rural equitable access to acceptable quality package of essen- population had improved sanitation [11]. Large income tial healthcare to all residents of Ghana [4]. Act 650 was inequalities also exist between rural and urban popula- revised in 2012 and replaced with Act 852 to remove ad- tions in Ghana. The overall poverty rate per capita is ministrative bottlenecks, introduce transparency, and re- 39% for rural areas and 10% for urban areas. The rural duce opportunities for corruption and gaming of the areas also have severe poverty rate per capita of 25% as NHIS system [7]. The scheme is financed mainly from compared to 5% in urban areas [11]. Low productivity the National Health Insurance Fund (NHIF). Cash in- and poorly functioning markets for agricultural products flow into the NHIF is from 2.5% of the 17.5% Value have been cited as the main reason for the poverty gap Added Tax (VAT), 2.5% of the 17.5% Social Security and between urban and rural Ghana [11]. The NHIS was National Insurance (SSNIT) contributions made by for- therefore designed as a pro-poor initiative to ensure fi- mal sector employees, member contributions from pre- nancial protection of the vulnerable in society including mium payments and monies that accrue to the fund women and children in rural areas, with a graduation from investments made by the NHIA Council. The Gov- premium based on socio-economic status. However, in ernment of Ghana also allocated funds to the NHIF reality, premiums are generally flat rated at the district through parliament and other donor funds [4–6]. Mem- levels due to the general difficulty in classifying bership of the NHIS involves payment of registration fee subscribers according to their relative socio-economic and insurance premium before an NHIS Identity Card is status [12]. issued. Formal sector employees are exempted from pay- Several studies have examined the determinants of ment of premium and therefore have to pay only the health insurance enrolment and identified economic registration fee to enroll. Informal sector workers have factors, socio-demographic factors, place of residence, to pay the annual premium and registration fee (except behavioral factors and household size as some of the im- those belonging to any of the exemption categories). En- portant determinants. Income, employment, and wealth rolment and financial contribution to the NHIS is legally index are very important economic determinants of mandatory by Act 852, but in practice it is voluntary as health insurance enrolment. Cameron et al. [13], Asenso there are no penalties for non-enrolment [7]. What actu- et al. [14], Sanhueza and Ruiz-Tagle [15] and Ying et al. ally pertains is that the mandatory enrolment and contri- [16], examined the relationship between income and bution only applies to formal sector employees who pay health insurance. They all concluded that income pre- Social Security and National Insurance Trust (SSNIT) dicts health insurance purchase. Economic theory shows contributions. Informal sector workers who form the that income has a fundamental influence on the decision majority of the Ghanaian population have to voluntarily to purchase health insurance as a “normal good”. Thus pay the registration fees and premium to enroll in the higher income increases the affordability of health NHIS. insurance premium because at higher income, the op- After more than a decade of implementation, the portunity cost of insurance purchase reduces [17, 18]. NHIS has made significant progress in extending health Other studies found that education and employment insurance to the people of Ghana. The active member- have direct positive correlation with health insurance en- ship of the scheme increased from 1.3 million in 2005 to rolment [19–21]. 10.15 million in 2013, representing 38% of the Ghanaian Socio-demographic determinants of enrolment such as population [8]. There are however, wide variations in en- age, gender, education and marital status are often used rolment coverage between the 10 regions with different to explain why some individuals may be employed, have Duku BMC Health Services Research (2018) 18:384 Page 3 of 16 high income and in high wealth index but still does not From the evidence available in the literature, it can be to enroll in health insurance. Economic theory indicates deduced that the demographic characteristics of the in- that as individuals’ age, they experience depreciation in habitants and the socio-economic differences that exist their health stock and tend to purchase health insurance within the rural and urban areas and between the differ- as an investment in health to avoid catastrophic health ent regions of Ghana may be responsible for the ob- expenditure in the event of ill health [22]. However, the served differences in health insurance enrolment in the empirical evidence on the effect of age on health insur- regions. Although several studies have been conducted ance enrolment has presented inconsistent results. to identify the determinants of health insurance enrol- While Mwaura [23], Bhat and Jain [24], Kronick and ment in Ghana [12, 14, 20, 21, 27, 28, 34, 36, 44], to the Gilmer [25], Savage and Wright [26] and Ayitey et al. best of my knowledge, no study has specifically exam- [27] found advanced age to increase the likelihood of ined the differences in these determinants between rural health insurance enrolment, Ying et al. [16], Jutting [25] and urban communities and between the different and Brugiavini and Pace [28] found that being young in- regions of the country. This paper seeks to contribute to creases the probability of health insurance enrolment. the broader understanding of determinants of health Other studies also found that higher level of education insurance enrolment in resource constraint setting. The and being married also increases the likelihood of health paper compares the determinants of enrolment in two re- insurance enrolment [14, 16, 23, 27–29]. A study by gions that are geographically similar but socio-economically Muurinen [30] however reported a contrary finding that different. The paper specifically assesses the differences in the highly educated are less likely to purchase health determinants of NHIS enrolment between the two regions insurance and explained that highly educated people are for the total samples, the urban sample and finally the rural likely to be healthier with low probability of risk, hence sample. This study will be relevant to policy makers as it will have lower likelihood of health insurance enrolment. will deepen their understanding of the factors that influence Studies by Butler [31] and Ayitey [14] revealed that the decision to enroll in the NHIS differently in the differ- people who are employed and those on executive ent regions. This knowledge is expected to help policy positions have a higher probability of health insurance makers decide on the best strategies to adopt to increase enrolment. health insurance enrolment in the current era of NHIS The evidence on the effect of gender on health insur- reforms. ance enrolment have reported inconsistent findings. While Asenso et al. [14] and Bourne & Kerr-Campbell Methods [29] found that male headed households and being male Study settings increases the likelihood of health insurance enrolment, This study uses data from the Client-oriented Health Ayitey et al. [27], Jutting [32] and Mwaura [23] found Insurance System in Ghana (COHEiSION) Project base- that female headed households and females are more line survey that was conducted in the Greater Accra likely to enroll in health insurance. The literature is (GAR) and Western (WR) regions of Ghana in April however emphatic that married couples are more likely 2012. These two coastal regions have similarities and to enroll in health insurance [27, 33]. Similarly, larger differences as far as rural and urban populations and household sizes have been reported to significantly in- socio-economic activities are concerned. Table 1 pre- crease the probability of health insurance enrolment [27, sents the differences and similarities between the 34, 35]. Religion has also been reported as a significant Greater Accra and Western regions. These two regions predictor of health insurance enrolment [36]. were purposively selected to provide rural/urban as well Again the empirical evidence on the effect of residen- as socio-economic differences that are of interest to the tial locality on health insurance enrolments has been in- study and allow the assessment of its impact on health consistent. While some studies found little or no insurance enrolment. difference in rural-urban health insurance enrolment [36, 37], others found residential remoteness (rural Data source areas) to be a significant determinant of health insurance This study uses primary data from the baseline survey of enrolment [38–44]. Yet, other studies also found urban COHEiSION project. Data was collected from 1920 ran- locality of residence to be a significant determinant of domly selected households within 10 km radius of se- health insurance enrolment [14, 17, 45–49]. In terms of lected primary healthcare facilities in the two regions. effect of health status and frequency of health facility Respondents were sampled through a multistage sam- visits on health insurance enrolment, the empirical evi- pling strategy. First 16 (8 in each region) districts with dence shows that individuals who are ill are more likely the same or almost same characteristics such as total to enroll in health insurance and so are those who make Population, NHIS enrolment coverage, total number of more health facility visits [1, 27, 50]. accredited health centres/clinics and urban or rural Duku BMC Health Services Research (2018) 18:384 Page 4 of 16 Table 1 Geographical and Socio-economic Comparison of with different implications on wealth. Thus at the time Study Regions of health insurance enrolment decision, individuals are Description Greater Accra Region Western Region uncertain about whether they will be ill or not and also of the financial implications should they become ill. In- Topography Coastal region Coastal region dividuals enroll in health insurance to protect them- Total Population 4,010,054 2,376,021 selves from catastrophic health expenditures in the Percentage rural 9.5% live in rural areas 57.6% live in rural event of ill health. The expectation is that in the event of Population areas ill health, the cost of treatment will be covered by the Percentage urban 90.5% live in urban areas 42.4% live in urban health insurance and in most instances, this cost is more population areas than the health insurance premium paid, representing a Jobs of inhabitants Mainly engaged in white Few are engaged in Urban Areas collar jobs (Government in white collar jobs. gain. The decision to enroll is therefore arrived at by Ministries, Departments Mainly engaged in comparing the expected utility with health insurance to and Agencies; small scale businesses expected utility without health insurance. Risk averse in- manufacturing and large (retailing) scale businesses) dividuals prefer to pay a certain known amount as health insurance premium to uncertain amounts of the same Economic activities Predominantly fishermen, Predominantly in rural areas farmers, small scale salt fishermen, farmers, expected utility in the event of ill health [52–54]. producers and some small small scale salt According to the expected utility theory, the demand for scale miners producers and some health insurance by risk averse individuals to avoid the small scale miners risk of wealth loss should be higher than risk neutral Regional poverty 12% 18% individuals who are indifferent about health insurance incidence enrolment and risk loving individuals who would not NHIS coverage 25.6% 32.2% want to purchase health insurance [55, 56]. This paper Source: 2011 NHIA Annual Report [9]; 2010 Population and Housing Census will rely on the expected utility theory to understand the [10] and Gender Inequalities in Rural Employment in Ghana: An Overview [11] determinants of enrolment in the NHIS in Ghana. categorization were selected for the project. Sixty-four (32 in each region) clusters of NHIS accredited primary Empirical model healthcare facilities were then selected from the selected The focus of this study is to interpret the dependent districts on the basis of their ownership (public/private), variable as a likelihood of enrolling in health insurance location (rural/urban) and NHIS accreditation quality or not given other explanatory variables. The logit model scores. Subsequently, 30 households were randomly is employed in the empirical estimation. This is because sampled from within a 10 km radius of each selected the logit model is able to overcome the problems associ- primary healthcare facility. This sampling process en- ated with the Linear Probability Model (LPM) which sured that a selected facility is the only primary health- allow the use of Ordinary Least Square (OLS) to esti- care facility within the 10 km radius catchment area. It mate the parameters. The LPM is plagued with hetero- also satisfied the randomized controlled trial design of scedasticity, non-normality of the disturbance term, low the COHEiSION project that required equal number of R and non-fulfilment of the 0 ≤ (Y /X ) ≤ 1 restriction i i intervention and control health facilities. Respondents of binary models. The logit model has the advantages of were contacted for the interview in April 2012. A being more robust such that the independent variables semi-structured questionnaire was used to collect infor- don’t have to be normally distributed or have equal vari- mation on respondents’ socio-demographics, social cap- ance in each group, does not assume a linear relation- ital and social schemas, employment status, health status ship between the independent and dependent variables, and healthcare utilization behavior, NHIS enrolment sta- does not assume homogeneity of variance and does not tus, consumption expenditure patterns and dwelling assume normality of error term. The maximum likeli- characteristics. In total, data on 7097 household mem- hood method is used to estimate the parameters [27]. bers was generated from the survey. This paper analysis The functional model for the determinants of health in- data on 4214 individuals who were 18 years and above. surance enrolment can be formulated as follows: Theoretical framework H ¼ fDðÞ ; L; π ; I ð1Þ The Expected Utility Theory of demand for Insurance under conditions of uncertainty and risk aversion by Where, Von Neuman and Morgenstern contends that health in- H = NHIS status of the individual. surance enrolment decision is one of a discrete choice to D = Demographic characteristics of the individual. enroll or not [51]. The theory assumes that individuals L = Health status of the individual. are risk averse and make choices between taking risk π = Probability of illness. Duku BMC Health Services Research (2018) 18:384 Page 5 of 16 I = Individual’s income. Table 2 Measurement of Variables When the logit model is applied to eq. (1), it can be Dependent Variable Operational Measurement expressed as: Health insurance enrolment 0 = Uninsured and 1 = Insured NHIS enrolment 0 = currently uninsured and 1 = currently insured LogitðÞ H ¼ β þ β X þ ε ð2Þ i ij 0 j i; j¼1 Key Independent Variable Operational Measurement Rural-urban location 0 = Urban and 1 = Rural The assumption with the logit mode is that there is a continuous latent variable y* that determines enrolment Type of health insurance 0 = Other insurance and enrollment 1 = Enrolled in NHIS in the NHIS. Thus if y* is positive, then the individual will enrol in the NHIS and the observed binary outcome Age Continuous positive whole numbers in years is one (1), otherwise the outcome is zero (0). The latent Sex 0 = Female and 1 = Male variable y* is therefore modelled by a linear regression function of the individual [27]. The estimable equation Marital Status 1 = Never married, 2 = Married and 3 = Divorced is therefore formulated as: Religion 0 = Other religion and 1 = Christian NHISstatus ¼ β þ β female þ β age þ β nevermarried 0 1 2 i 3 Household size Continuous positive whole numbers of the number of þβ otherreligion þ β Hhsize 4 5 household members þβ noformaledu þ β unemployed 6 7 Educational Level 1 = No education, 2 = Basic education and 3 = Secondary þβ poorhealthstatus þ β poorestwealth 8 i 9 education and above þβ disttofacility þ β public 10 11 Employment Status 0 = Unemployed and 1 = Employed Type of Employment 0 = part-time employment and 1 = Full-time employment Measurement of variables Occupation 1 = Farmer, 2 = Artisan/Trader, Current enrolment in the NHIS is defined as the 3 = Labourer/Casual, Managerial/ dependent variable of interest for this study. It assumes Professional, 5 = Business Owner a value of 1 if the individual is currently enrolled in the Annual Income Amount earned measured in Ghana scheme otherwise 0. The Ghana Statistical Service Cedis (GSS), 2010 Population and Housing Census classifica- Wealth Status 1 = poorest; 2 = poor; 3 = average; tion of rural and urban areas was used to classify 4 = rich and 5 = richest (per capita household food and non-food individuals as living in either rural or urban area [4]. consumption expenditure) According to this classification, five (Ablekuma, Frequency of health facility Continuous positive whole numbers Ayawaso, Tema, Kpeshie and Okaikoi) out of the eight visits of the number of health facility visits selected districts in the Greater Accra Region are urban within the last 6 months prior to the survey districts and the remaining three (Dangme East, Dangme West and Ga West) are rural districts. It also classified Health Status 1 = Bad health, 2 = Fair health and 3 = Good Health five (Bia, Amenfi East, Wassa West, Jomoro, and Ahantaman) out of the eight selected districts in the Time taken from home to Continuous positive whole numbers facility of the time in minutes it takes to move Western Region as rural and three (Amanfiman, Sekondi by road from home to health facility and Takoradi) as urban districts. Ownership of Health Facility 0 = Private and 1 = Public Other demographic, socio-economic, health status characteristics of respondents and characteristics of the health facilities that empirical evidence suggest can in- Data analysis fluence health insurance enrolment were included in the In analyzing the data, 4214 individuals for the age cohort estimation as explanatory variables. The explanatory var- of 18 years and above was used. First descriptive statis- iables included in the estimation are age, gender, reli- tics was used to present respondents proportion/average gion, marital status, household size, educational level, demographic and socio-economic characteristics in the employment status, wealth status (a 5 quintile proxy two regions for the total, urban and rural samples. measure for annual food and non-food household con- Bivariate analysis (t-test for continuous and Pearson sumption expenditure per capita), health status, private/ chi-square for categorical) was performed to examine public health facility and time taken to move from home differences in respondents characteristics (socio-eco- to the nearest health facility. Table 2 summarizes the nomic and insurance enrolment) between the two re- dependent and independent variables used in the ana- gions for the total, urban and rural samples. Finally, lysis and how they were measured. logistic regression estimation was performed to identify Duku BMC Health Services Research (2018) 18:384 Page 6 of 16 the determinants of enrolment in the two regions for the the highest proportion of 23% of respondents are in the total, urban and rural sample. Respondents who were poorest wealth quintile. The proportion of respondents enrolled in insurance schemes other than the NHIS were in the poorest wealth quintile in urban GAR (8%) is far excluded from the regression estimation. smaller than their urban WR (17%) counterparts. Conversely, the proportion of respondents in the richest Results wealth quintile in rural GAR (34%) is much higher than their Characteristics of respondents rural WR (18%) counterparts. Majority of respondents in the Table 3 present respondents’ demographic, socio-economic total sample (87%), GAR (85%) and WR (88%) indicated that and health status characteristics by rural and urban areas in they are of good health. The average annual income from all the total, GAR and WR samples. The average age in the total sourcesfor respondentsinthe totalsampleisGH 2340.76, sample is 38 years. The average age in the GAR is slightly GAR (GH 2470.33) and WR (GH 2371.23). Whilst aver- higher (39 years) than their WR (37 years) counterparts. age annual income level in urban WR (GH 2864.29) is Urban adults in the GAR are older (39 years) than urban higher than urban GAR (GH 2256.41), average annual in- adults in the WR (37 years). Similarly, rural adults in the GAR come in rural GAR (GH 2413.73) is higher than their rural are older (38 years) than rural adults in the WR. There are WR (GH 2072.97) counterparts. The proportion of urban more females (56%) in the total sample than males. Similarly, respondents in WR (92%) with good health status is more there are more females in the GAR (57%) and WR (56%) re- than their urban GAR (85%) counterparts. Conversely, the spectively than males. proportion of respondents in rural GAR (86%) with good There are more married respondents (51%) in the total health status is more than their rural WR (85%) counter- sample, the GAR (49%) and WR (54%) respectively than parts. Most (61%) of health facilities in the total sample, the other categories of marital status. There are however, GAR (62%) and WR (60%) were public owned facilities. more married couples in both urban (47%) and rural Whiles urban GAR has more private (74%) than public (53%) GAR than in urban and rural WR. Majority (90%) (26%) health facilities, urban WR has approximately equal of respondents in the total sample, GAR (89%) and WR (50%) private and public health facilities. On the average, it (90%) are Christians. There are however slightly more takes approximately 12 min to move from home to the near- Christians in urban WR (92%) than urban GAR (90%) est health facility in the total sample, 13 min in GAR and whiles rural GAR (89%) has more Christians than rural 14 min in the WR. It takes less time to get to the health WR (88%). facility in urban GAR (10 min) than urban WR (17 min). There are approximately 5 members per household in However, it takes more time to get to the health facility in the total sample, 4 members in GAR and 5 members in ruralGAR (13min)thanin rural WR (12min). WR respectively. Urban GAR and urban WR have the same household size of 4 members. Similarly, the aver- Regional differences in characteristics by rural and urban age household size in rural GAR (5) is equal to average samples household size in rural WR (5). More than half of re- Table 4 presents the bivariate analysis of health insur- spondents in the total sample (52%), GAR (51%) and ance enrolment and other explanatory variables between WR (53%) have completed basic level of education. The the two regions for the total, urban and rural samples. proportion of adults with basic level of education in The WR have a higher (54%) health insurance coverage urban GAR (46%) is less than their urban WR (50%) than the GAR (45%) in the total sample. Health insur- counterparts. Majority of respondents in the total sam- ance coverage in the urban sample is higher for GAR ple (70%), GAR (68%) and WR (71%) are gainfully (59%) than the WR (41%). employed. Those employed in urban WR (69%) are more However, in the rural sample, health insurance cover- than those in urban GAR (66%). However, both rural age is higher in the WR (69%) than the GAR (31%). GAR and rural WR have equal proportion of 72% These differences in insurance coverages are statistically employed adults. Almost all (more than 98%) of the significant at the 90% confidence interval. In terms of employed adults in the total sample, GAR and WR are type of health insurance, the WR (56%) again have a in full-time employment. Of those employed, 47% in higher NHIS coverage than the GAR (44%) in the total total, 52% in GAR and 42% in WR samples are arti- sample. For the urban sample, the GAR have a higher sans/traders. There are more traders/artisans in urban (57%) NHIS coverage than the WR (43%) whiles in the GAR (52%) than urban WR (44%). Similarly, there are rural sample the WR have a significantly higher (69%) more artisans/traders in rural GAR (52%) than rural NHIS coverage than the GAR (31%). WR (40%). The average age of respondents in the GAR (39 years) The richest wealth quintile had the highest proportion is significantly higher than the WR (37 years) in the total of 24% of respondents in the total sample and 30% in sample. There is however no statistically significant dif- GAR than other wealth quintile. However, in the WR, ference in the average age of respondents in the urban Duku BMC Health Services Research (2018) 18:384 Page 7 of 16 Table 3 Characteristics of Working-Age Adults by Rural and Urban Area in the Two Regions Total Sample (N = 4214) Greater Accra Region (N = 2187) Western Region (N = 2027) Mean (N = 4214) Urban Rural Mean Urban Rural Mean Urban Rural (N = 2171) (N = 2043) (N = 2187) (N = 1407) (N = 780) (N = 2027) (N =764) (N = 1263) Mean Age 37.74 38.61 36.81 38.61 39.26 37.98 36.59 37.43 36.09 Sex (%) Female 56.29 56.06 56.53 56.65 56.15 57.56 55.90 55.89 55.90 Male 43.71 43.94 43.47 43.35 43.85 42.44 44.10 44.11 44.10 Marital Status (%) Never Married 34.77 36.99 32.40 37.27 38.59 34.87 32.07 34.03 30.88 Married 51.50 48.60 54.58 48.98 46.91 52.56 54.27 51.70 55.82 Divorced 13.74 14.42 13.02 13.81 14.50 12.56 13.67 14.27 13.30 Religion (%) Other Religion 10.39 9.44 11.40 10.52 10.31 10.90 10.26 7.85 11.72 Christian 89.61 90.56 88.60 89.48 89.69 89.10 89.74 92.15 88.28 Mean Household Size 4.5 4.4 4.6 4.4 4.3 4.5 4.6 4.4 4.7 Educational Level (%) No Education 13.69 9.95 17.67 10.75 7.18 17.18 16.87 15.05 17.97 Basic Education 52.06 47.31 57.12 50.80 45.70 60.00 53.43 50.26 55.34 Secondary Educ. 34.24 42.75 25.21 38.45 47.12 22.82 29.70 34.69 26.68 and above Employment (%) Not Employed 30.45 33.12 27.61 32.14 34.47 27.95 28.61 30.63 27.40 Employed 69.55 66.88 72.39 67.86 65.53 72.05 71.39 69.37 72.26 Type of Employment Part-time 1.71 2.13 1.28 2.09 2.60 1.25 1.31 1.32 1.31 (%) (N = 2931) Full-Time 98.29 97.87 98.72 97.91 97.40 98.75 98.69 98.68 98.69 Occupation (%) Farmer 18.85 10.22 27.34 8.41 3.44 16.76 29.40 22.01 33.63 (N = 1931) Artisan/Trader 46.85 49.05 44.69 51.95 51.94 51.96 41.70 44.02 40.38 Laborer/Casual 7.83 8.67 7.01 8.07 8.10 8.01 7.59 9.65 6.42 Worker Managerial/ 18.46 22.41 14.57 20.79 24.20 15.08 16.10 19.35 14.27 Professional Business Owner 8.01 9.68 6.28 10.78 12.32 8.19 5.20 5.02 5.31 Wealth Status (%) Poorest 17.56 11.56 23.94 12.53 8.53 1974 22.99 17.15 26.52 Poor 17.13 15.02 19.38 15.96 14.14 19.23 18.40 16.62 19.48 Average 19.06 19.81 18.26 18.98 18.91 19.10 19.14 21.47 17.74 Rich 22.19 24.74 19.48 22.95 23.95 21.15 21.36 26.18 18.45 Richest 24.06 28.88 18.94 29.58 34.47 20.77 18.11 18.59 17.81 Average Annual 2340.76 2470.33 2203.07 2470.33 2256.41 2413.73 2371.23 2864.29 2072.97 Income Duku BMC Health Services Research (2018) 18:384 Page 8 of 16 Table 3 Characteristics of Working-Age Adults by Rural and Urban Area in the Two Regions (Continued) Total Sample (N = 4214) Greater Accra Region (N = 2187) Western Region (N = 2027) Mean (N = 4214) Urban Rural Mean Urban Rural Mean Urban Rural (N = 2171) (N = 2043) (N = 2187) (N = 1407) (N = 780) (N = 2027) (N =764) (N = 1263) Health Status (%) Bad Health 4.77 4.65 4.89 5.21 5.90 3.97 4.29 2.36 5.46 Fair Health 8.66 8.01 9.35 9.33 9.10 9.74 7.94 6.02 9.11 Good Health 86.57 87.33 85.76 85.46 85.00 86.28 87.77 91.62 85.43 Ownership of Private 39.44 20.73 59.32 38.45 18.62 74.23 40.50 24.61 50.12 Facility (%) Public 60.56 79.27 40.68 61.55 81.38 25.77 59.50 75.39 49.88 Time to Health 12.43 12.64 12.19 12.64 10.46 12.63 13.75 16.68 11.92 Facility Source: COHEiSION Project Baseline Survey of 2012 GAR Greater Accra Region, WR Western region Duku BMC Health Services Research (2018) 18:384 Page 9 of 16 Table 4 Bivariate Analysis of Health Insurance, Independent Variables by Regional and Rural/Urban Total Sample (N = 4214) Urban (N = 2171) Rural (N = 2043) GAR (N = 2171) WR (N = 2043) p-value GAR (N = 1407) WR (N = 764) P-value GAR (N = 780) WR (N = 1263) P-value Enrolment in Health Insurance Uninsured (%) 56.51 43.49 0.000*** 68.92 31.08 0.004** 43.04 56.96 0.019* Insured (%) 45.35 54.65 58.82 41.18 31.47 68.53 Type of Health Insurance Other Health Insurance (%) 60.18 39.82 0.032* 81.03 18.97 0.024* 38.18 61.82 0.429 NHIS (%) 44.32 55.68 57.26 42,74 31.01 68.99 Average Age 38.79 36.59 0.002** 39.26 37.43 0.078 27.97 36.09 0.051* Sex Females (%) 52.23 47.77 0.601 64.91 35.09 0.892 38.87 61.13 0.446 Males (%) 51.47 48.53 64.68 35.32 37.27 62.73 Marital Status Never Married (%) 55.63 44.37 0.034* 67.62 32.38 0.319 41.09 58.91 0.366 Married (%) 49.31 50.69 62.56 37.44 36.77 63.23 Divorced (%) 52.16 47.83 65.18 34.82 36.84 63.16 Religion Other Religion (%) 52.51 47.49 0.889 70.73 27.27 0.260 36.48 63.52 0.786 Christian (%) 51.83 48.17 64.81 35.19 38.40 61.60 Average Household Size 4.4 4.6 0.252 4.3 4.4 0.655 4.5 4.7 0.519 Educational Level No Education (%) 40.73 59.27 0.009** 46.76 53.24 0.006** 37.12 62.88 0.408 Basic Education (%) 50.64 49.36 62.61 37.39 40.10 59.90 Secondary Educ. and above (%) 58.28 41.72 71.44 28.56 34.56 65.44 Employment Unemployed (%) 54.79 45.21 0.122 67.45 32.55 0.263 38.65 61.35 0.857 Employed (%) 50.63 49.37 63.50 36.50 38.00 62.00 Health Status Bad Health (%) 56.72 43.28 0.324 82.18 17.82 0.005** 31.00 69.00 0.493 Fair Health (%) 55.89 44.11 73.56 26.44 39.79 60.21 Good Health (%) 51.23 48.77 63.08 36.92 38.41 61.59 Wealth Status Poorest (%) 37.03 62.97 0.000*** 47.81 52.19 0.005** 31.49 68.51 0.482 Poor (%) 48.34 51.66 61.04 38.96 37.88 62.12 Duku BMC Health Services Research (2018) 18:384 Page 10 of 16 Table 4 Bivariate Analysis of Health Insurance, Independent Variables by Regional and Rural/Urban (Continued) Total Sample (N = 4214) Urban (N = 2171) Rural (N = 2043) GAR (N = 2171) WR (N = 2043) p-value GAR (N = 1407) WR (N = 764) P-value GAR (N = 780) WR (N = 1263) P-value Average (%) 51.68 48.32 61.86 38.14 39.95 60.05 Rich (%) 53.69 46.31 62.76 37.24 41.46 58.54 Richest (%) 63.81 36.19 77.35 22.65 41.86 58.14 Ownership of Facility Private (%) 50.60 49.4 0.869 58.22 41.78 0.696 47.77 52.23 0.204 Public (%) 52.74 47.26 66.53 33.47 24.19 75.81 Average Time to Health 11.23 13.75 0.106 10.46 16.68 0.018* 12.63 11.92 0.761 Facility Source: COHEiSION Project Baseline Survey of 2012. For continuous variables, means are reported and significant test is based on t-test whiles for categorical variables percentages are reported and significant test is based on Chi-square test.; Note; *p < 0.05; ** p < 0.01; ***p < 0.001 Duku BMC Health Services Research (2018) 18:384 Page 11 of 16 Table 5 Determinant of Health Insurance Enrolment Adjusted Odds Ratio Total Sample Urban Sample Rural Sample Independent Variables GAR WR GAR WR GAR WR Sex Males 0.66*** (0.53–0.82) 0.52*** (0.44–0.63) 0.74** (0.57–0.97) 0.49*** (0.36–0.67) 0.46*** (0.34–0.64) 0.55*** (0.42–0.71) Age 1.02*** (1.01–1.03) 1.03*** (1.02–1.04) 1.01** (1.01–1.03) 1.04*** (1.02–1.05) 1.01* (0.99–1.04) 1.02*** (1.01–1.04) Marital Status Married 1.28 (0.93–1.76) 1.14 (0.84–1.56) 1.14 (0.82–1.58) 0.86 (0.59–1.26) 1.60 (0.76–3.38) 1.29 (0.77–2.13) Divorced 1.11 (0.71–1.74) 0.66* (0.44–0.99) 0.96 (0.60–1.52) 0.42** (0.21–0.85) 1.22 (0.41–3.62) 0.73 (0.45–1.20) Religion Christian 1.08 (0.67–1.77) 1.10 (0.78–1.56) 1.20 (0.67–2.13) 1.08 (0.70–1.67) 0.68 (0.29–1.58) 1.05 (0.63–1.76) Household Size 1.04 (0.98–1.10) 1.08 (0.98–1.18) 1.03 (0.94–1.13) 1.15* (0.99–1.31) 1.03 (0.93–1.16) 1.04 (0.92–1.18) Educational Level Basic Education 0.89 (0.63–1.28) 1.18 (0.89–1.57) 0.93 (0.50–1.72) 1.87* (1.06–3.31) 0.89 (0.59–1.34) 0.99 (0.75–1.32) Secondary Edu. & 1.49* (1.00–2.21) 1.61* (1.06–2.44) 1.37 (0.75–2.52) 1.95 (0.87–4.39) 2.07*** (1.33–3.22) 1.67* (1.05–2.66) Above Employment Employed 0.82 (0.61–1.11) 0.79 (0.61–1.03) 0.79 (0.55–1.13) 0.84 (0.52–1.39) 0.85 (0.49–1.47) 0.79 (0.55–1.14) Health Status Fair Health 0.84 (0.53–1.34) 0.63 (0.26–1.53) 0.85 (0.45–1.62) 0.08 (0.01–0.72) 1.10 (0.60–2.02) 0.87 (0.35–2.19) Good Health 0.60* (0.39–0.96) 0.54 (0.22–1.34) 0.58 (0.32–1.08) 0.06 (0.01–0.59) 0.82 (0.33–2.03) 0.80 (0.33–1.95) Wealth Status Poor 0.91 (0.49–1.67) 1.47 (0.88–2.46) 0.67 (0.29–1.54) 2.54* (0.99–6.52) 0.98 (0.38–2.55) 1.07 (0.57–2.03) Average 1.11 (0.60–2.04) 1.43 (0.90–2.26) 0.66 (0.28–1.58) 1.49 (0.56–3.97) 1.75 (0.75–4.13) 1.25 (0.72–2.17) Rich 1.35 (0.72–2.50) 1.77* (1.04–3.02) 1.16 (0.47–2.86) 2.68 (0.84–8.49) 1.04 (0.45–2.42) 1.26 (0.68–2.34) Richest 1.73* (0.97–3.10) 2.04** (1.26–3.28) 1.13 (0.49–2.63) 2.59 (0.93–7.22) 2.55** (1.16–5.59) 1.76* (1.04–2.98) Travel time to Facility 1.00 (0.98–1.02) 0.99 (0.98–1.00) 1.00 (0.98–1.03) 0.99 (0.99–1.01 1.01 (0.98–1.04) 0.98** (0.97–0.99) Ownership of Facility Public 1.83*** (1.55–2.16) 0.93 (0.82–1.06) 1.78*** (1.39–2.26) 0.95 (0.82–1.11) 0.48*** (0.31–0.73) 0.31*** (0.26–0.37) Constant 0.14*** (0.06–0.33) 0.34* (0.11–0.04) 0.17*** (0.05–0.59) 1.08 (0.14–8.65) 0.366 (0.09–1.58) 0.42 (0.11–1.59) Pseudo R-squared 0.086 0.079 0.075 0.102 0.135 0.086 No. of Observation 2132 1946 1336 736 735 1167 Source: COHEiSION Project Baseline Survey of 2012. Note: Adjusted odds ratio are reported with reference categories: Age (1 year increase in age); Sex (Female); Marital status (Never Married); Religion (Other Religion); Household size (Increase of 1 person); Educational level (No Formal Education); Employment (Unemployed); Health Status (Poor Health Status); Wealth status (Poorest); Travel time (1 min increase in travel time); Facility ownership (Private); Missing values are excluded from the analyses. Note: *p < 0.05; ** p < 0.01; ***p < 0.001 Duku BMC Health Services Research (2018) 18:384 Page 12 of 16 or rural samples. Similarly, there is no statistically sig- WR for the total, urban and rural samples. For the total nificant difference in the proportion of females and sample, females are significantly more likely to enroll in males in either the total, urban or rural samples. the NHIS than males for both the GAR and WR. Although in the total sample, there are more married re- Similarly, for both the urban and rural samples, fe- spondents in the GAR (56%) than the WR (44%), the dif- males are significantly more likely to enroll in the ferences in the proportion of married respondents NHIS than males in the GAR and WR. These find- between the regions in the urban and rural samples are ings are consistent with findings by Mwaura [23], statistically insignificant as shown in Table 4. The pro- Ayitey et al. [27]and Juttingetal.[32]. Age also im- portion of respondents with basic level education is sig- pacts positively on NHIS enrolments. The results nificantly higher in the GAR (51%) than the WR (49%) show that an increase of 1 year in age significantly in the total sample. Similarly, in the urban sample, the increases the odds of NHIS enrolment in both regions proportion of respondents with basic level education is for the total, rural and urban samples. These findings higher in the GAR (63%) than the WR (37%). There is are consistent with findings by Bhat and Jain [24], however no statistically significantly difference in the Kronick and Gilmer [25], Savage and Wright [26]and proportion of respondents with basic level education be- Ayitey et al. [27]. In terms of marital status, for the tween the GAR and WR in the rural sample. In terms of total sample, whiles divorcees are 1.08 (CI = 0.67– health status, there is no statistically significant differ- 1.77) times more likely to enroll in the NHIS in the ence in the categories of health status between the GAR GAR, their WR counterparts are 0.66 (CI = 0.44–0.99) and WR for the total and rural samples. The GAR how- times significantly less likely to enroll. For the urban ever, have higher proportions the health status categories sample, although divorcees in both regions are less likely than the WR in the urban sample. The proportion of re- to enroll as compared to those married, the results for the spondents in the richest wealth quintile is significantly WR is statistically significant at the 95% confidence inter- higher in the GAR (64%) than the WR (36%) in the total val. However, for the rural sample, whiles divorcees in the sample. Similarly, the proportion of respondents in the GAR are 1.22 (CI = 0.41–3.38) more likely to enroll, their richest wealth quintile is significantly higher in the GAR WR counterparts are 0.73 (CI = 0.45–1.20) less likely to (77%) than the WR (23%). However, there is no statisti- enroll even though this result is statistically insignificant cally significant difference in wealth status categories be- at the 90% confidence interval. This finding is in contrast tween the GAR and WR in the rural sample. with earlier studies by Asenso Okyere et al. [14], Ayitey et There is also no statistically significant differences be- al. [27] and Brugiavini and Pace [28] who found married tween the GAR and WR in terms of religion, employment, individuals to be significantly more likely to enroll in health facility ownership and average time to health facil- health insurance. Although the results further shows that ity either for the total, urban or rural samples. being married increases the odds of NHIS enrolment in both regions for the total, urban and rural samples, these Determinant of health insurance enrolment findings are statistically insignificant at the 90% confi- Table 5 presents the logistic regression estimations of dence interval. Similarly, although the results show that the determinant of NHIS enrolment among working-age being a Christian and a 1 member increase in household adults for the total, urban and rural samples. Estimating sizes increases the odds of NHIS enrolment, these findings individuals’ income in developing countries is difficult are statistically insignificant even at the 90% confidence and unreliable because most people are reluctant to dis- interval. close their true income. A five quintile wealth status was The results also show that an individuals’ level of therefore computed from household food and non-food education influence the odds of NHIS enrolment. For consumption expenditure as a proxy for income levels the total sample, having secondary level education and and used in the regression estimation. Again due to the above significantly increases the likelihood of NHIS en- inability of respondents to accurately determine the dis- rolment by 1.49 (CI = 1.00–1.11) times in the GAR and tance in kilometers from their home to the nearest pri- 1.62 (CI = 1.06–2.44) times in the WR respectively as mary healthcare facility, time taken in minutes to move compared to individuals with no formal education. For from home to the nearest primary healthcare facility was the urban sample, although the results shows that sec- used as a proxy to measure distance from home to ondary level education and above increases the odds of health facility. The differences in the determinants of en- enrolment by 1.37 (CI = 0.75–2.52) times in the GAR rolment between the two regions are presented. and 1.95 (CI = 0.87–4.39) times in the WR, these find- The results show that generally, age, sex, marital status, ings are statistically insignificant at the 90% confidence educational level, health status, wealth status and health interval. However, for the rural sample, individuals with facility ownership are significant determinants of NHIS secondary level education and above are significantly enrolment at varying extents between the GAR and the more likely to enroll in the NHIS by 2.07 (CI = 1.33– Duku BMC Health Services Research (2018) 18:384 Page 13 of 16 3.22) times in the GAR and 1.67 (CI = 1.05–1.14) times in educational level, marital status, health status, wealth sta- the WR. These findings are consistent with findings by tus and health facility ownership are significant determi- Asenso Okyere et al. [14], Ayitey et al. [27] and Brugiavini nants of NHIS enrolment. and Pace [28]. The results show being employed does not Females in both regions are more likely to enroll com- positively influence NHIS enrolment in both regions for pared to males. Similarly, urban and rural females in the total, urban or the rural samples. These findings are both regions are more likely to enroll in the NHIS than however statistically insignificant at the 90% confidence males. This may be as a result of the Free Maternal interval. Health Policy under the NHIS which offer premium ex- Although the finding was statistically insignificant, in- emption to expectant and nursing mothers and therefore dividuals with poor self-assessed health status are more most females in the reproductive age-group might have likely to enroll in the NHIS than those with fair or good enrolled under this exemptions category. The likelihood health status in both regions for the total, urban and of NHIS enrolment was also found to increase with age rural samples. Ayitey et al. [27] reported similar findings for both regions in the total sample and similarly for of the effect of self-assessed health status on NHIS en- both regions in the urban and rural areas. This positive rollment. The results show that wealth status is a signifi- relationship of increasing age with NHIS enrolment can cant determinant of NHIS enrolment at varying extent be attributed to degeneration in health as people age in the two regions depending on the rural/urban locality and the need for increased healthcare utilization. Older of residence of the individual. For the total sample, the people therefore prefer to make investment in their richest in the GAR are 1.73 (CI = 0.97–3.10) times more health through the purchase of health insurance. This likely to enroll in the NHIS than the poorest whiles in findings is consistent with the findings of previous stud- the WR the rich and the richest are 1.77 (CI = 1.04– ies [20, 27, 36]. The increased likelihood of the aged to 3.02) times and 2.04 (CI = 1.26–3.28) times respectively health insurance enrolment can also be attributed to the more likely to enroll in the NHIS than the poorest. exemptions of people aged 70 years and above from pre- However, when the urban and rural samples are consid- mium payment under the NHIS exemption policy. Most ered separately, whiles in the rural sample, individuals in of the aged therefore enrolled under the 70+ age exemp- the richest wealth quintile in the GAR are 2.55 (CI = tion category. Married people are more likely to enroll 1.16–5.59) times and WR are 1.76 (CI = 1.04–2.98) times in the NHIS in both regions for the total sample and significantly more likely than those in the poorest quin- similarly in both regions in the urban and rural areas tile, in the urban sample individuals in the richest wealth than never married people. This may be because married quintile in both regions are more likely to enroll. These couples purchase health insurance to mitigate the finan- finding are statistically insignificant at the 90% confi- cial burden that is likely to accrue from raising children dence interval. These findings are in agreement with after marriage [27, 31, 33]. findings by Ayitey et al. [27] and Fowler et al. [50]. In The findings also point to educational class dimen- terms of travel time from home to the nearest primary sions in NHIS enrolment. The better educated in both healthcare facility, although a 1 min increase in the regions in the total and rural sample are significantly travel time reduces the odds of NHIS enrolment of indi- more likely to enroll than the uneducated. Although the viduals in the WR for the total, urban and rural samples, results from the urban sample is statistically insignifi- these findings were statistically insignificant at the 90% cance, there is still a positive relationship between higher confidence interval. From the results, ownership of the education and NHIS enrolment in both regions. This nearest primary healthcare facilities is another significant positive relationship between higher education and determinant of NHIS enrolment. Individuals who live NHIS is because higher education makes people better around the catchment area of a publicly owned primary understand and appreciate the insurance concept and its health facility in the GAR are 1.83 (CI = 1.55–2.16) times benefits to the household. People with poor self-assessed and 1.78 (CI = 1.39–2.26) times significantly more likely health status in both regions are more likely to enroll in to enroll in the NHIS for the total and urban samples re- the NHIS in the total and urban samples than those with spectively. However, in the rural sample, individuals who fair and good health status. However, in the rural sam- live in the catchment area of private facilities are signifi- ple, people who are of fair health status in the GAR are cantly more likely to enroll in the NHIS than those who more likely to enroll in the NHIS than their WR coun- live around public facilities. terparts. This suggest that people who are of poor health in urban communities on both regions self-select into Discussion the NHIS. Perhaps the availability of both public and The study assessed the differences in determinants of private health facilities in urban communities is such NHIS enrolment among working-aged adults in the GAR that people of poor health get to know of their under- and WR in Ghana. The findings indicates that age, sex, lying propensity to increased healthcare utilization. They Duku BMC Health Services Research (2018) 18:384 Page 14 of 16 therefore enroll in the NHIS to protect themselves from urban and rural samples. It is therefore not surprising catastrophic healthcare expenditure. Theoretical con- that a1minincreaseinthetravel timein theWR cepts such as adverse selection, risk aversion, affordabil- reduces the odds of NHIS enrolment for the total, ity and trust which are beyond the scope of this study urban and rural samples. The ownership of the near- can help explain and put these findings into proper est primary health facility is another important deter- perspective. minant of NHIS enrolment. People who visit public Working-age adults in the rich and richest quintiles in health facilities in the GAR are more likely to enroll both regions in the total sample are significantly more in the NHIS than their WR counterparts for the total likely to enroll in the NHIS than those in the poorest and urban samples. However in the rural areas, this wealth quintile. This suggest that wealth status and is not the case as those who visit private facilities are therefore affordability of insurance premium is an im- significantly more likely to enroll in the NHIS in both portant determinant of NHIS enrolment. In the urban regions. A possible explanation to this could be about sample, the odds of NHIS enrolment for WR individuals perceived poor quality care from these public facilities in the rich and richest quintiles is about twice that of which may influence peoples decision to enroll in the those in the GAR. Conversely, in the rural sample, the NHIS. This is because per the NHIS gate keeping sys- odds of NHIS enrolment for GAR individuals in the tem, card holders are expected to first visit a primary richest quintile are about twice that of those in the WR. health facility when sick. If people perceive the qual- Thus the rich in urban WR enroll more in the NHIS ity of care they receive from their nearest primary than the rich in urban areas in the GAR. However, in health facility to be poor, chances are that it will dis- the rural areas, the rich in rural GAR enroll more in the courage them from enrolling in the NHIS. NHIS than the rich in rural WR. The expectation is that with the pro-poor design of the NHIS and its extensive Conclusion exemption policy that exempt premium payment for in- The study assessed the differences in determinants of digents, the poor in rural areas in both regions will en- NHIS enrolment among working-aged adults in the roll more than the rich. These finding further galvanize GAR and WR in Ghana. The analysis employed logistic findings by earlier studies that poverty is a major barrier regression in the empirical estimation. The study did not to NHIS enrolment and that the NHIS is not pro-poor find any differences in the demographic determinants of as envisaged [27]. The fact that the odds of enrolment NHIS enrolment between the two regions and among among the rural rich in the GAR is higher than rural the rural and urban residents in the two regions. The rich in WR further indicates that levels economic activ- findings indicate that generally, age, sex, educational ities, income levels and affordability of premium in the level, marital status, health status and travel time to rural areas also impact on the decision to enroll in the health facility are significant determinants of NHIS en- NHIS. Although the WR has a high poverty rate of 18% rolment in both regions and similarly in the rural and compared to the GAR 12%, the NHIS enrolment rate in urban communities in the two regions. The study how- the WR (55.7%) is higher than the GAR (44.3%). ever found some differences between the two regions in However, the likelihood of enrolment is significantly terms of wealth status and health facility ownership as higher among the richest in the WR (OR = 2.04) than in determinants of NHIS enrolment. Although the rich and the GAR (OR = 1.73). This can be explained from the fact richest in both regions are more likely to enroll in the that the richest in the WR do not have access to many pri- NHIS than the poor and poorest, the odds of NHIS vate health facility and private health insurance companies enrolment for the richest in urban areas in the WR (OR like the richest in the GAR. So even if the richest in WR = 2.59) is about twice as that of GAR (OR = 1.13) whiles are not satisfied with the quality of care from public health in the rural areas the odds of NHIS enrolment in the GAR facilities which they will attend should they enroll in the (OR = 2.55) is also about twice that of the WR (OR = NHIS, they will still have to attend these same public facil- 1.76). People who visit public health facilities in the GAR ities should they opt out of the NHIS and pay are more likely to enroll in the NHIS than those in WR out-of-pocket or enroll with other private health insur- for the total and urban samples. However in the rural ance. Unlike the GAR where the richest have access to areas, those who visit private facilities are significantly many private insurance companies and private healthcare more likely to enroll in the NHIS in both regions. These facilities, the richest who decide to opt out of the NHIS findings suggest that inequalities still exists in NHIS enrol- for quality issues have readily available alternatives. ment in favor of the wealthy, communities with better Travel time from home to health facility is another im- socio-economic activities and communities with health fa- portant determinant of NHIS enrolment between the re- cilities that are perceived to provide better quality health- gions. The average travel time to the nearest health care. This thus raises concerns as to whether the NHIS is facility is higher in the WR than the GAR for the total, truly pro-poor as envisaged. Duku BMC Health Services Research (2018) 18:384 Page 15 of 16 The study contributes to the literature on the de- Funding This study received financial support from the Government of the terminants of NHIS enrolment by identifying factors Netherlands through the Netherlands Organization for Scientific Research that might be responsible for the observed differ- (NWO/WOTRO) in the form of a research grant (Grant No. W07.45.104.00) for ences in the NHIS regional enrolment coverages. the COHEiSION project. Collaborators of the study include Noguchi Memorial Institute for Medical Research, University of Ghana Legon; Amsterdam The differences in NHIS enrolment coverages be- Institute for Global Health and Development, University of Amsterdam; Vrije tween the regions may be as a result of differences University of Amsterdam and University of Groningen. in socio-economic factors that impact on the ability Availability of data and materials of the inhabitants to afford the insurance premium. All data supporting my findings and conclusions is contained in the This should serve as an important indicator to pol- manuscript. The full data is available in the COHEiSION Project data icymakers on the need to focus on regional specific repository at Noguchi Memorial Institute for Medical Research. There are no restrictions to data sources and details of the full data may be accessed geographic and proxy means targeting strategies through the Co-PI, Dr. Daniel Kojo Arhinful (Noguchi Memorial Institute for aimed at identifying the poor in resource constraints Medical Research, University of Ghana Legon. P. O. Box, LG 581, Legon. E- communities for premium exemptions. The study mail: DArhinful@noguchi.ug.edu.gh). recommends that policymakers should use innovative Author’s contributions approaches to determine NHIS premium at the dis- SKOD initiated the conceptualization, collected the data, conducted data trict level based on socio-economic activities and in- analysis and prepared the manuscript. The author read and approved the final manuscript. come levels within the district. They should also consider introducing quality healthcare dimension Ethics approval and consent to participate into provider payment mechanisms to rewards pro- Ethical clearance for the study was obtained from the Ghana Health Service (GHS) Ethical Review Committee (ERC clearance number: GHS-ERC 08/5/11). viders who meet quality criteria as expressed by Informed consent was also obtained from respondents. All literate respondents their clients. This will serve as an incentive for both signed written informed consent while illiterate respondents had the informed public and private accredited healthcare providers to consent form read to them in a local language that they understand before thumb-printing the form to participate in the study. provide quality healthcare that attract individuals to enroll in the NHIS. Competing interests Finally, the findings of this study is subject to the fol- The author declares he has no competing interests. lowing limitations. The analysis relied on self-reported measures such as health insurance enrolment status, Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in patterns of enrolment and self-assessed health status. published maps and institutional affiliations. Therefore any systematic differences due to respondents reporting bias could affect the precision of the reported Author details Department of Epidemiology, Noguchi Memorial Institute for Medical estimates. The study did not include individuals insured Research, University of Ghana, P. O. Box LG 581 Legon, Accra, Ghana. with other private insurance companies other than the 2 Amsterdam Institute for Global health and Development, Amsterdam, The NHIS due to the small sample size in the data. These Netherlands. Vrije Universiteit Amsterdam, Amsterdam, The Netherlands. limitation however does not invalidate the entirety of Received: 12 September 2017 Accepted: 8 May 2018 the self-reported measures and the analysis of these measure which have been well-documented. The find- References ings of the determinants of NHIS enrolment reported in 1. Carrin G, James C. Social health insurance: key factors affecting the this study are limited to the Greater Accra and Western transition towards universal coverage. Int Soc Secur Rev. 2005;58:1–15. regions of Ghana and therefore caution should be taken 2. WHO. Sustainable health financing, universal coverage and social health insurance. In World Health Assembly Resolution: WHA58.33. Geneva: WHO; when generalizing the findings to other populations. 3. Griffin C. Health care in Asia: a comparative study of cost and financing. Abbreviations Washington: World Bank Regional and Sectoral Studies; 1992. COHEiSION Project: Client-Oriented Health Insurance System in Ghana 4. GOG. The National Health Insurance Act: Act 650. Accra: Parliament House; Project; GAR: Greater Accra Region; GSS: Ghana Statistical Service; NHIA: National Health Insurance Authority; NHIF: National Health Insurance 5. Arhinful DK. The solidarity of self-interest: social and cultural feasibility Fund; NHIS: National Health Insurance Scheme; NOW: Netherlands of rural health insurance in Ghana. Leiden: African studies canter Organization for Scientific Research; SSNIT: Social Security and National research report; 2003. Available: https://openaccess.leidenuniv.nl/ Insurance Trust; WHA: World Health Assembly; WOTRO: Science for Global bitstream/handle/1887/12919/ASC-075287668-077-01.pdf?sequence=2. Development; WR: Western Region Accessed 20 August 2011. 6. Addo JS. Consultants & tri Star actuarial and management consultant. A Acknowledgements feasibility study for the establishment of a national health insurance scheme I acknowledge the support of the Government of the Netherlands through in Ghana. Accra: Ministry of Health; 1995. the Netherlands Organization for Scientific Research (NWO/WOTRO) in the 7. GOG. The national health insurance act: Act 852. Accra: Parliament House; form of a research grant (Grant No. W07.45.104.00) for the COHEiSION 2012. project. I am most grateful to staff of the Epidemiology Department, 8. NHIA. National health insurance authority 2013 annual report. Accra, Ghana: Noguchi Memorial Institute for Medical Research, University of Ghana and NHIA; 2014. staff of the Faculty of Economics and Business Administration, Vrije 9. NHIA. National health insurance authority 2011 annual report. Accra, Ghana: Universiteit Amsterdam for their support in the project design. NHIA; 2012. Duku BMC Health Services Research (2018) 18:384 Page 16 of 16 10. GSS. 2010 population and housing census. Accra-Ghana: Ghana Statistical 36. Nketiah-Amponsah E. Demand for health insurance among women in Service; 2012. Ghana: cross-sectional evidence. International Research Journal of Finance 11. FAO. Gender inequalities in rural employment in Ghana: An overview. Accra- and Economics. 2009;33:179–91. Ghana; 2012. http://www.statsghana.gov.gh/docfiles/Gender/FAO_GHANA_ 37. Duncan RP, Seccombe K, Amey C. Changes in health insurance coverage COUNTRY_PROFILE_FINAL2012.pdf. Accessed 20 Sept 2012. within rural and urban environments. J Rural Health. 1995;11:160–76. 12. Akazili J, Welaga P, Bawah A, Achana FS, Oduro A, et al. Is Ghana’s pro-poor 38. Stock R. Distance and utilization of health facilities in rural Nigeria. Soc Sci health insurance scheme really for the poor? Evidence from northern Med. 1983;17:63–570. Ghana. BMC Health Serv Res. 2014;14:637. 39. Jong KE, Smith DP, Yu XQ, O’Connell L, Goldstein D, Armstrong BK. Remoteness of residence and survival from cancer in new South Wales. 13. Cameron AC, Trivedi PK, Milne F, Piggott J. A micro-econometric model of Med J Aust. 2004;180:618–22. the demand for health care and health insurance in Australia. Rev Econ 40. Criel B, Van der Stuyft P, Van Lerberghe W. The Bwamanda hospital Stud. 1988;55(1):85–106. insurance scheme: effective for whom? A study of its impact on hospital 14. Asenso-Okyere W, Osei-Akoto I, Anum A, Appiah E. Willingness to pay for utilization patterns. Soc Sci Med. 1999;48:897–911. health insurance in a developing economy. A pilot study of the informal sector 41. Nemet GF, Bailey AJ. Distance and healthcare utilization among the rural of Ghana using contingency valuation. Health Policy. 1997;42(3):223–37. elderly. Soc Sci Med. 2000;50:1197–208. 15. Sanhueza R, Ruiz-Tagle J. Choosing health insurance in a dual health care 42. Frenzen PD. Health insurance coverage in U.S. urban and rural areas. J Rural SYSTEM: the Chilean case. J Appl Econ. 5(1):157–84. Health. 1993;9:204–14. 16. Ying XH, Hu TW, Ren J, Chen W, Xu K, Huang JH. Demand for private health 43. Taylor AK, Cohen JW, Machlin SR. Being uninsured in 1996 compared to insurance in Chinese urban areas. Health Econ. 2007;16(10):1041–50. 1987: how has the experience of the uninsured changed over time? Health 17. Chankova S, Sulzbach S, Diop F. Impact of mutual health organizations: Serv Res. 2001;36:16–31. evidence from West Africa. Health Policy Plan. 2008;23:264–76. 44. Jehu-Appiah C, Aryeetey G, Spaan E, De Hoop T, Agyepong I, Baltussen R. 18. Dong H, Gbangou A, De Allegri M, Pokhrel S & Sauerborn R. The differences Equity aspects of the national health insurance scheme in Ghana: who is in characteristics between health-care users and non-users: implication for enrolling, who is not and why? Soc Sci Med. 2011;72:157–65. introducing community-based health insurance in Burkina Faso. European 45. Coburn AF, Mcbridge TD, Ziller EC. Patterns of health insurance coverage Journal of Health Economics, 2008; 9(1): 41–50. among rural and urban children. Med Care Res Rev. 2002;59:272–92. 19. Gius MP. An analysis of the health insurance coverage of young adults. Int J 46. Holmes M, Ricketts T. In: North Carolina Rural Health Research and Policy Appl Econ. 2010;7(1):1–17. Analysis Centre, editor. Rural-urban differences in the rate of health 20. Sarpong N, Loag W, Fobil J, Meyer CG, Adu-Sarkodie Y, May J, et al. National insurance coverage. North Carolina; 2008. http://www.shepscenter.unc.edu/ health insurance coverage and socio-economic status in a rural district of wp-content/uploads/2013/07/ruralgap_finalbrief.pdf. Accessed 21 Ghana. Trop Med Int Health. 2010;15(2):191–7. September 2012. 21. Dixon J, Tenkorang EY, Luginaah I. Ghana’s national health insurance 47. Hu HM, Duncan RP, Radcliff TA, Porter CK, Hall AG. Variations in health scheme: helping the poor or leaving them behind? Environment and insurance coverage for rural and urban nonelderly adult residents of Florida, Planning C: Politics and Space. 2011;29(6):1102–15. Indiana, and Kansas. J Rural Health. 2006;22:147–50. 22. Grossman M. On the concept of health capital and the demand for health. 48. Comer J, Mueller K. Access to health care: urban-rural comparisons from a J Polit Econ. 1972;80:223–55. Midwestern agricultural state. J Rural Health. 1995;11:128–36. 23. Mwaura JW, Pongpanich S. Access to health care: the role of community 49. Ketsche P. Employment-based health insurance: analysis of rural-urban based health insurance in Kenya. Pan African Medical Journal. 2012;12:35. differences in one state. Med Care Res Rev. 2005;62:458–78. 24. Bhat R, Jain N. Factors affecting the demand for health insurance in a micro 50. Fowler R, Noyahr L, Thornton J, Pinto R, Kahn J. An official thoracic society insurance scheme. Indian Institute of Management, Working Paper 07-02. systematic review: the association between health insurance status and Ahmedabad: Research and Publication Department; 2006. access, care delivery and outcomes for patients who are critically ill. 25. Kronick R, Gilmer T. Explaining the decline in health insurance coverage, American Journal of Critical Care Medicine. 2010;181:1003–11. 1979 – 1995. Health Aff. 1999;18(2):30–47. 51. Von Neumann J, Morgenstern O. Theory of games and economic 26. Savage E, Wright D. Health Insurance and health care utilization: theory and behaviour. Princeton, New Jersey, USA: Princeton University Press; 1994. evidence from Australia 1989–1990. CHERE Discussion Paper 44. Sydney: 52. Propper C. The demand for private health care in the UK. J Health Econ. University of Technology; 2001. 2000;19(6):855–76. 27. Ayitey AM, Nketiah-Amponsah E, Barimah A. Determinants of insurance 53. Schneider P. Why should the poor insure? Theories of decision-making in enrolment among Ghanaian adults: the case of the national health insurance the context of health insurance. Health Policy Plan. 2004;19(6):347–55. scheme. Economics Management and Financial Markets. 2013;8(3):37. 54. Fin C, Harmon C. A dynamic model of demand for private health insurance 28. Brugiavini A, Pace M. Extending health insurance: effects of the national in Ireland. IZA Discussion Paper No. 2472. Bonn: IZA; 2006. http://ftp.iza.org/ health insurance scheme in Ghana. Italy: European report on development, dp2472.pdf. European University Institute, Badia Fiesolana; 2011. http://cadmus.eui.eu/ 55. Arrow K. Uncertainty and the welfare economics of medical care. Am Econ bitstream/handle/1814/17221/RSCAS_2011_27.pdf?sequence=1&isAllowed= Rev. 1963;53(5):941–73. y. Accessed 20 Sept 2012. 56. Feldstein MS. The welfare loss of excess health insurance. J Polit Econ. 1973; 29. Bourne PA, Kerr-Campbell MD. Determinants of self-rated private health 81(2, part 1):251–80. insurance coverage in Jamaica. Health. 2010;2(06):541. 30. Muurinen JM. Demand for health: a generalized Grossman model. J Health Econ. 1982;1(1):5–28. 31. Butler J. Estimating elasticity of demand for private health insurance in Australia. National center for epidemiology and population health. Canberra:ANU;1999. 32. Jutting J. Health Insurance for the Rural Poor? Community financing scheme in Senegal to protect against illness. Development and Cooperation. 2001;6:4–5. 4 33. Cameron AC, McCallum. Private health insurance choices in Australia: The role of long-term utilization of health services, in Economics and health, 1995. Sydney: School of Health Services Management, UNSW; 1996. p. 143–157. 34. Cameron AC, Trivedi PK. The role of income and health risk in the choice of health insurance: evidence from Ghana. J Public Econ. 1991;45(1):1–28. 35. Kirigia JM, Sambo LG, Nganda B, Mwabu GM, Chatora R, Mwase T. Determinants of health insurance ownership among south African women. BMC Health Serv Res. 2005;5:17. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Health Services Research Springer Journals

Differences in the determinants of health insurance enrolment among working-age adults in two regions in Ghana

Free
16 pages
Loading next page...
 
/lp/springer_journal/differences-in-the-determinants-of-health-insurance-enrolment-among-tSVgmrDPa9
Publisher
BioMed Central
Copyright
Copyright © 2018 by The Author(s).
Subject
Medicine & Public Health; Public Health; Health Administration; Health Informatics; Nursing Research
eISSN
1472-6963
D.O.I.
10.1186/s12913-018-3192-9
Publisher site
See Article on Publisher Site

Abstract

Background: Ghana’s National Health Insurance Scheme (NHIS) has achieved varying levels of enrolment within the regions with different rural-urban populations with associated income inequalities. This study sought to investigate the differences in the determinants of enrolment between the Greater Accra (GAR) and Western (WR) regions of Ghana to inform the NHIS reforms. Method: Data from 4214 adults, 18 years and above from a household survey conducted in the two regions was analyzed. Bivariate analysis (t-test for continuous and Pearson chi-square for categorical) was performed to examine differences in respondents characteristics (socio-economic and insurance enrolment) between the two regions for the total, urban and rural samples. Logistic regression estimation was performed to establish differences in determinant of enrolment between the regions. Results: Age, sex, educational level, marital status, health status and travel time to nearest health facility were identified as determinants of enrolment in both regions and among the rural and urban residents within the regions. Although the rich and richest in both regions are more likely to enroll than the poor and poorest, the odds of enrolment for the urban richest in the WR is about twice that of GAR whiles the odds of enrolment for the rural richest in the GAR is also about twice that of the WR. Those who visit public facilities in the GAR are more likely to enroll than those in WR for the total and urban samples. However, those who visit private facilities in rural communities in both regions are more likely to enroll. Conclusion: Differences in the NHIS enrolment between the regions is as a result of differences in socio-economic factors that are intrinsic in the regions and impact on the inhabitants’ ability to afford insurance premium. Policymakers should determine NHIS premium differently at the district level based on socio-economic activities and income levels within the districts. Keywords: Health insurance, Determinants of enrolment, Rural-urban, National Health Insurance Scheme, Ghana Background their entire populations [1, 2]. The 2005 World Health Health insurance has been recognized globally as one of Assembly resolution WHA 58.33 urged members states to the principal methods of financing healthcare to achieve ensure financial protection to all citizens, especially chil- universal coverage, particularly in low and middle income dren and women of reproductive age and “to plan the countries. Many low and middle income countries are transition to universal coverage of their citizens” [2]. currently exploring mechanisms of extending their health Given the high demand for healthcare services of appre- insurance schemes to specific groups to eventually cover ciable quality and the extreme under-utilization of health services in several Sub-Saharan African countries due to financial barriers, health insurance has been recom- Correspondence: stephen.duku@ghsmail.org Department of Epidemiology, Noguchi Memorial Institute for Medical mended as a promising alternative to other criticized fi- Research, University of Ghana, P. O. Box LG 581 Legon, Accra, Ghana nancing systems like cost-recovery and user fees [2]. The Amsterdam Institute for Global health and Development, Amsterdam, The expectation is that health insurance will improve access to Netherlands Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Duku BMC Health Services Research (2018) 18:384 Page 2 of 16 quality healthcare through risk pooling of unforeseeable rural and urban populations [8, 9]. The implementation healthcare cost to fixed premiums [3]. In response to this of the NHIS also saw a steady increase in outpatient call for developing countries to adopt healthcare financing utilization of healthcare services. Outpatients’ utilization mechanisms that remove financial accessibility barriers increased from under 5 million in 2005 to approximately and strive towards the attainment of universal health 24 million in 2012 [9]. coverage, Ghana implemented the National Health Insur- Ghana is considered as a rural country with approxi- ance Scheme (NHIS) in 2004 to replace the “Cash and mately (49%) of the population living in rural areas with Carry” system when patients have to pay out-of-pocket limited socio-economic opportunities [10]. Significant cash before receiving healthcare. inequalities persist between the rural and urban areas in The NHIS was established through an Act of parlia- terms of availability of basic amenities and infrastructure ment, Act 650 in 2003 as part of efforts to make the such as water, sanitation and health facilities [10, 11]. In health goal within the Ghana Poverty Reduction Strategy 2009, only about 10% of the urban population in Ghana (GPRS) achievable and also to accomplish the targets set lack access to portable water as compared to 26% of the in the Health Sector Five-Year Programme of Work, rural population. Whilst 18% of urban population had 2002–2006 [4–6]. The vision of the NHIS was to ensure access to improved sanitation, only 7% of the rural equitable access to acceptable quality package of essen- population had improved sanitation [11]. Large income tial healthcare to all residents of Ghana [4]. Act 650 was inequalities also exist between rural and urban popula- revised in 2012 and replaced with Act 852 to remove ad- tions in Ghana. The overall poverty rate per capita is ministrative bottlenecks, introduce transparency, and re- 39% for rural areas and 10% for urban areas. The rural duce opportunities for corruption and gaming of the areas also have severe poverty rate per capita of 25% as NHIS system [7]. The scheme is financed mainly from compared to 5% in urban areas [11]. Low productivity the National Health Insurance Fund (NHIF). Cash in- and poorly functioning markets for agricultural products flow into the NHIF is from 2.5% of the 17.5% Value have been cited as the main reason for the poverty gap Added Tax (VAT), 2.5% of the 17.5% Social Security and between urban and rural Ghana [11]. The NHIS was National Insurance (SSNIT) contributions made by for- therefore designed as a pro-poor initiative to ensure fi- mal sector employees, member contributions from pre- nancial protection of the vulnerable in society including mium payments and monies that accrue to the fund women and children in rural areas, with a graduation from investments made by the NHIA Council. The Gov- premium based on socio-economic status. However, in ernment of Ghana also allocated funds to the NHIF reality, premiums are generally flat rated at the district through parliament and other donor funds [4–6]. Mem- levels due to the general difficulty in classifying bership of the NHIS involves payment of registration fee subscribers according to their relative socio-economic and insurance premium before an NHIS Identity Card is status [12]. issued. Formal sector employees are exempted from pay- Several studies have examined the determinants of ment of premium and therefore have to pay only the health insurance enrolment and identified economic registration fee to enroll. Informal sector workers have factors, socio-demographic factors, place of residence, to pay the annual premium and registration fee (except behavioral factors and household size as some of the im- those belonging to any of the exemption categories). En- portant determinants. Income, employment, and wealth rolment and financial contribution to the NHIS is legally index are very important economic determinants of mandatory by Act 852, but in practice it is voluntary as health insurance enrolment. Cameron et al. [13], Asenso there are no penalties for non-enrolment [7]. What actu- et al. [14], Sanhueza and Ruiz-Tagle [15] and Ying et al. ally pertains is that the mandatory enrolment and contri- [16], examined the relationship between income and bution only applies to formal sector employees who pay health insurance. They all concluded that income pre- Social Security and National Insurance Trust (SSNIT) dicts health insurance purchase. Economic theory shows contributions. Informal sector workers who form the that income has a fundamental influence on the decision majority of the Ghanaian population have to voluntarily to purchase health insurance as a “normal good”. Thus pay the registration fees and premium to enroll in the higher income increases the affordability of health NHIS. insurance premium because at higher income, the op- After more than a decade of implementation, the portunity cost of insurance purchase reduces [17, 18]. NHIS has made significant progress in extending health Other studies found that education and employment insurance to the people of Ghana. The active member- have direct positive correlation with health insurance en- ship of the scheme increased from 1.3 million in 2005 to rolment [19–21]. 10.15 million in 2013, representing 38% of the Ghanaian Socio-demographic determinants of enrolment such as population [8]. There are however, wide variations in en- age, gender, education and marital status are often used rolment coverage between the 10 regions with different to explain why some individuals may be employed, have Duku BMC Health Services Research (2018) 18:384 Page 3 of 16 high income and in high wealth index but still does not From the evidence available in the literature, it can be to enroll in health insurance. Economic theory indicates deduced that the demographic characteristics of the in- that as individuals’ age, they experience depreciation in habitants and the socio-economic differences that exist their health stock and tend to purchase health insurance within the rural and urban areas and between the differ- as an investment in health to avoid catastrophic health ent regions of Ghana may be responsible for the ob- expenditure in the event of ill health [22]. However, the served differences in health insurance enrolment in the empirical evidence on the effect of age on health insur- regions. Although several studies have been conducted ance enrolment has presented inconsistent results. to identify the determinants of health insurance enrol- While Mwaura [23], Bhat and Jain [24], Kronick and ment in Ghana [12, 14, 20, 21, 27, 28, 34, 36, 44], to the Gilmer [25], Savage and Wright [26] and Ayitey et al. best of my knowledge, no study has specifically exam- [27] found advanced age to increase the likelihood of ined the differences in these determinants between rural health insurance enrolment, Ying et al. [16], Jutting [25] and urban communities and between the different and Brugiavini and Pace [28] found that being young in- regions of the country. This paper seeks to contribute to creases the probability of health insurance enrolment. the broader understanding of determinants of health Other studies also found that higher level of education insurance enrolment in resource constraint setting. The and being married also increases the likelihood of health paper compares the determinants of enrolment in two re- insurance enrolment [14, 16, 23, 27–29]. A study by gions that are geographically similar but socio-economically Muurinen [30] however reported a contrary finding that different. The paper specifically assesses the differences in the highly educated are less likely to purchase health determinants of NHIS enrolment between the two regions insurance and explained that highly educated people are for the total samples, the urban sample and finally the rural likely to be healthier with low probability of risk, hence sample. This study will be relevant to policy makers as it will have lower likelihood of health insurance enrolment. will deepen their understanding of the factors that influence Studies by Butler [31] and Ayitey [14] revealed that the decision to enroll in the NHIS differently in the differ- people who are employed and those on executive ent regions. This knowledge is expected to help policy positions have a higher probability of health insurance makers decide on the best strategies to adopt to increase enrolment. health insurance enrolment in the current era of NHIS The evidence on the effect of gender on health insur- reforms. ance enrolment have reported inconsistent findings. While Asenso et al. [14] and Bourne & Kerr-Campbell Methods [29] found that male headed households and being male Study settings increases the likelihood of health insurance enrolment, This study uses data from the Client-oriented Health Ayitey et al. [27], Jutting [32] and Mwaura [23] found Insurance System in Ghana (COHEiSION) Project base- that female headed households and females are more line survey that was conducted in the Greater Accra likely to enroll in health insurance. The literature is (GAR) and Western (WR) regions of Ghana in April however emphatic that married couples are more likely 2012. These two coastal regions have similarities and to enroll in health insurance [27, 33]. Similarly, larger differences as far as rural and urban populations and household sizes have been reported to significantly in- socio-economic activities are concerned. Table 1 pre- crease the probability of health insurance enrolment [27, sents the differences and similarities between the 34, 35]. Religion has also been reported as a significant Greater Accra and Western regions. These two regions predictor of health insurance enrolment [36]. were purposively selected to provide rural/urban as well Again the empirical evidence on the effect of residen- as socio-economic differences that are of interest to the tial locality on health insurance enrolments has been in- study and allow the assessment of its impact on health consistent. While some studies found little or no insurance enrolment. difference in rural-urban health insurance enrolment [36, 37], others found residential remoteness (rural Data source areas) to be a significant determinant of health insurance This study uses primary data from the baseline survey of enrolment [38–44]. Yet, other studies also found urban COHEiSION project. Data was collected from 1920 ran- locality of residence to be a significant determinant of domly selected households within 10 km radius of se- health insurance enrolment [14, 17, 45–49]. In terms of lected primary healthcare facilities in the two regions. effect of health status and frequency of health facility Respondents were sampled through a multistage sam- visits on health insurance enrolment, the empirical evi- pling strategy. First 16 (8 in each region) districts with dence shows that individuals who are ill are more likely the same or almost same characteristics such as total to enroll in health insurance and so are those who make Population, NHIS enrolment coverage, total number of more health facility visits [1, 27, 50]. accredited health centres/clinics and urban or rural Duku BMC Health Services Research (2018) 18:384 Page 4 of 16 Table 1 Geographical and Socio-economic Comparison of with different implications on wealth. Thus at the time Study Regions of health insurance enrolment decision, individuals are Description Greater Accra Region Western Region uncertain about whether they will be ill or not and also of the financial implications should they become ill. In- Topography Coastal region Coastal region dividuals enroll in health insurance to protect them- Total Population 4,010,054 2,376,021 selves from catastrophic health expenditures in the Percentage rural 9.5% live in rural areas 57.6% live in rural event of ill health. The expectation is that in the event of Population areas ill health, the cost of treatment will be covered by the Percentage urban 90.5% live in urban areas 42.4% live in urban health insurance and in most instances, this cost is more population areas than the health insurance premium paid, representing a Jobs of inhabitants Mainly engaged in white Few are engaged in Urban Areas collar jobs (Government in white collar jobs. gain. The decision to enroll is therefore arrived at by Ministries, Departments Mainly engaged in comparing the expected utility with health insurance to and Agencies; small scale businesses expected utility without health insurance. Risk averse in- manufacturing and large (retailing) scale businesses) dividuals prefer to pay a certain known amount as health insurance premium to uncertain amounts of the same Economic activities Predominantly fishermen, Predominantly in rural areas farmers, small scale salt fishermen, farmers, expected utility in the event of ill health [52–54]. producers and some small small scale salt According to the expected utility theory, the demand for scale miners producers and some health insurance by risk averse individuals to avoid the small scale miners risk of wealth loss should be higher than risk neutral Regional poverty 12% 18% individuals who are indifferent about health insurance incidence enrolment and risk loving individuals who would not NHIS coverage 25.6% 32.2% want to purchase health insurance [55, 56]. This paper Source: 2011 NHIA Annual Report [9]; 2010 Population and Housing Census will rely on the expected utility theory to understand the [10] and Gender Inequalities in Rural Employment in Ghana: An Overview [11] determinants of enrolment in the NHIS in Ghana. categorization were selected for the project. Sixty-four (32 in each region) clusters of NHIS accredited primary Empirical model healthcare facilities were then selected from the selected The focus of this study is to interpret the dependent districts on the basis of their ownership (public/private), variable as a likelihood of enrolling in health insurance location (rural/urban) and NHIS accreditation quality or not given other explanatory variables. The logit model scores. Subsequently, 30 households were randomly is employed in the empirical estimation. This is because sampled from within a 10 km radius of each selected the logit model is able to overcome the problems associ- primary healthcare facility. This sampling process en- ated with the Linear Probability Model (LPM) which sured that a selected facility is the only primary health- allow the use of Ordinary Least Square (OLS) to esti- care facility within the 10 km radius catchment area. It mate the parameters. The LPM is plagued with hetero- also satisfied the randomized controlled trial design of scedasticity, non-normality of the disturbance term, low the COHEiSION project that required equal number of R and non-fulfilment of the 0 ≤ (Y /X ) ≤ 1 restriction i i intervention and control health facilities. Respondents of binary models. The logit model has the advantages of were contacted for the interview in April 2012. A being more robust such that the independent variables semi-structured questionnaire was used to collect infor- don’t have to be normally distributed or have equal vari- mation on respondents’ socio-demographics, social cap- ance in each group, does not assume a linear relation- ital and social schemas, employment status, health status ship between the independent and dependent variables, and healthcare utilization behavior, NHIS enrolment sta- does not assume homogeneity of variance and does not tus, consumption expenditure patterns and dwelling assume normality of error term. The maximum likeli- characteristics. In total, data on 7097 household mem- hood method is used to estimate the parameters [27]. bers was generated from the survey. This paper analysis The functional model for the determinants of health in- data on 4214 individuals who were 18 years and above. surance enrolment can be formulated as follows: Theoretical framework H ¼ fDðÞ ; L; π ; I ð1Þ The Expected Utility Theory of demand for Insurance under conditions of uncertainty and risk aversion by Where, Von Neuman and Morgenstern contends that health in- H = NHIS status of the individual. surance enrolment decision is one of a discrete choice to D = Demographic characteristics of the individual. enroll or not [51]. The theory assumes that individuals L = Health status of the individual. are risk averse and make choices between taking risk π = Probability of illness. Duku BMC Health Services Research (2018) 18:384 Page 5 of 16 I = Individual’s income. Table 2 Measurement of Variables When the logit model is applied to eq. (1), it can be Dependent Variable Operational Measurement expressed as: Health insurance enrolment 0 = Uninsured and 1 = Insured NHIS enrolment 0 = currently uninsured and 1 = currently insured LogitðÞ H ¼ β þ β X þ ε ð2Þ i ij 0 j i; j¼1 Key Independent Variable Operational Measurement Rural-urban location 0 = Urban and 1 = Rural The assumption with the logit mode is that there is a continuous latent variable y* that determines enrolment Type of health insurance 0 = Other insurance and enrollment 1 = Enrolled in NHIS in the NHIS. Thus if y* is positive, then the individual will enrol in the NHIS and the observed binary outcome Age Continuous positive whole numbers in years is one (1), otherwise the outcome is zero (0). The latent Sex 0 = Female and 1 = Male variable y* is therefore modelled by a linear regression function of the individual [27]. The estimable equation Marital Status 1 = Never married, 2 = Married and 3 = Divorced is therefore formulated as: Religion 0 = Other religion and 1 = Christian NHISstatus ¼ β þ β female þ β age þ β nevermarried 0 1 2 i 3 Household size Continuous positive whole numbers of the number of þβ otherreligion þ β Hhsize 4 5 household members þβ noformaledu þ β unemployed 6 7 Educational Level 1 = No education, 2 = Basic education and 3 = Secondary þβ poorhealthstatus þ β poorestwealth 8 i 9 education and above þβ disttofacility þ β public 10 11 Employment Status 0 = Unemployed and 1 = Employed Type of Employment 0 = part-time employment and 1 = Full-time employment Measurement of variables Occupation 1 = Farmer, 2 = Artisan/Trader, Current enrolment in the NHIS is defined as the 3 = Labourer/Casual, Managerial/ dependent variable of interest for this study. It assumes Professional, 5 = Business Owner a value of 1 if the individual is currently enrolled in the Annual Income Amount earned measured in Ghana scheme otherwise 0. The Ghana Statistical Service Cedis (GSS), 2010 Population and Housing Census classifica- Wealth Status 1 = poorest; 2 = poor; 3 = average; tion of rural and urban areas was used to classify 4 = rich and 5 = richest (per capita household food and non-food individuals as living in either rural or urban area [4]. consumption expenditure) According to this classification, five (Ablekuma, Frequency of health facility Continuous positive whole numbers Ayawaso, Tema, Kpeshie and Okaikoi) out of the eight visits of the number of health facility visits selected districts in the Greater Accra Region are urban within the last 6 months prior to the survey districts and the remaining three (Dangme East, Dangme West and Ga West) are rural districts. It also classified Health Status 1 = Bad health, 2 = Fair health and 3 = Good Health five (Bia, Amenfi East, Wassa West, Jomoro, and Ahantaman) out of the eight selected districts in the Time taken from home to Continuous positive whole numbers facility of the time in minutes it takes to move Western Region as rural and three (Amanfiman, Sekondi by road from home to health facility and Takoradi) as urban districts. Ownership of Health Facility 0 = Private and 1 = Public Other demographic, socio-economic, health status characteristics of respondents and characteristics of the health facilities that empirical evidence suggest can in- Data analysis fluence health insurance enrolment were included in the In analyzing the data, 4214 individuals for the age cohort estimation as explanatory variables. The explanatory var- of 18 years and above was used. First descriptive statis- iables included in the estimation are age, gender, reli- tics was used to present respondents proportion/average gion, marital status, household size, educational level, demographic and socio-economic characteristics in the employment status, wealth status (a 5 quintile proxy two regions for the total, urban and rural samples. measure for annual food and non-food household con- Bivariate analysis (t-test for continuous and Pearson sumption expenditure per capita), health status, private/ chi-square for categorical) was performed to examine public health facility and time taken to move from home differences in respondents characteristics (socio-eco- to the nearest health facility. Table 2 summarizes the nomic and insurance enrolment) between the two re- dependent and independent variables used in the ana- gions for the total, urban and rural samples. Finally, lysis and how they were measured. logistic regression estimation was performed to identify Duku BMC Health Services Research (2018) 18:384 Page 6 of 16 the determinants of enrolment in the two regions for the the highest proportion of 23% of respondents are in the total, urban and rural sample. Respondents who were poorest wealth quintile. The proportion of respondents enrolled in insurance schemes other than the NHIS were in the poorest wealth quintile in urban GAR (8%) is far excluded from the regression estimation. smaller than their urban WR (17%) counterparts. Conversely, the proportion of respondents in the richest Results wealth quintile in rural GAR (34%) is much higher than their Characteristics of respondents rural WR (18%) counterparts. Majority of respondents in the Table 3 present respondents’ demographic, socio-economic total sample (87%), GAR (85%) and WR (88%) indicated that and health status characteristics by rural and urban areas in they are of good health. The average annual income from all the total, GAR and WR samples. The average age in the total sourcesfor respondentsinthe totalsampleisGH 2340.76, sample is 38 years. The average age in the GAR is slightly GAR (GH 2470.33) and WR (GH 2371.23). Whilst aver- higher (39 years) than their WR (37 years) counterparts. age annual income level in urban WR (GH 2864.29) is Urban adults in the GAR are older (39 years) than urban higher than urban GAR (GH 2256.41), average annual in- adults in the WR (37 years). Similarly, rural adults in the GAR come in rural GAR (GH 2413.73) is higher than their rural are older (38 years) than rural adults in the WR. There are WR (GH 2072.97) counterparts. The proportion of urban more females (56%) in the total sample than males. Similarly, respondents in WR (92%) with good health status is more there are more females in the GAR (57%) and WR (56%) re- than their urban GAR (85%) counterparts. Conversely, the spectively than males. proportion of respondents in rural GAR (86%) with good There are more married respondents (51%) in the total health status is more than their rural WR (85%) counter- sample, the GAR (49%) and WR (54%) respectively than parts. Most (61%) of health facilities in the total sample, the other categories of marital status. There are however, GAR (62%) and WR (60%) were public owned facilities. more married couples in both urban (47%) and rural Whiles urban GAR has more private (74%) than public (53%) GAR than in urban and rural WR. Majority (90%) (26%) health facilities, urban WR has approximately equal of respondents in the total sample, GAR (89%) and WR (50%) private and public health facilities. On the average, it (90%) are Christians. There are however slightly more takes approximately 12 min to move from home to the near- Christians in urban WR (92%) than urban GAR (90%) est health facility in the total sample, 13 min in GAR and whiles rural GAR (89%) has more Christians than rural 14 min in the WR. It takes less time to get to the health WR (88%). facility in urban GAR (10 min) than urban WR (17 min). There are approximately 5 members per household in However, it takes more time to get to the health facility in the total sample, 4 members in GAR and 5 members in ruralGAR (13min)thanin rural WR (12min). WR respectively. Urban GAR and urban WR have the same household size of 4 members. Similarly, the aver- Regional differences in characteristics by rural and urban age household size in rural GAR (5) is equal to average samples household size in rural WR (5). More than half of re- Table 4 presents the bivariate analysis of health insur- spondents in the total sample (52%), GAR (51%) and ance enrolment and other explanatory variables between WR (53%) have completed basic level of education. The the two regions for the total, urban and rural samples. proportion of adults with basic level of education in The WR have a higher (54%) health insurance coverage urban GAR (46%) is less than their urban WR (50%) than the GAR (45%) in the total sample. Health insur- counterparts. Majority of respondents in the total sam- ance coverage in the urban sample is higher for GAR ple (70%), GAR (68%) and WR (71%) are gainfully (59%) than the WR (41%). employed. Those employed in urban WR (69%) are more However, in the rural sample, health insurance cover- than those in urban GAR (66%). However, both rural age is higher in the WR (69%) than the GAR (31%). GAR and rural WR have equal proportion of 72% These differences in insurance coverages are statistically employed adults. Almost all (more than 98%) of the significant at the 90% confidence interval. In terms of employed adults in the total sample, GAR and WR are type of health insurance, the WR (56%) again have a in full-time employment. Of those employed, 47% in higher NHIS coverage than the GAR (44%) in the total total, 52% in GAR and 42% in WR samples are arti- sample. For the urban sample, the GAR have a higher sans/traders. There are more traders/artisans in urban (57%) NHIS coverage than the WR (43%) whiles in the GAR (52%) than urban WR (44%). Similarly, there are rural sample the WR have a significantly higher (69%) more artisans/traders in rural GAR (52%) than rural NHIS coverage than the GAR (31%). WR (40%). The average age of respondents in the GAR (39 years) The richest wealth quintile had the highest proportion is significantly higher than the WR (37 years) in the total of 24% of respondents in the total sample and 30% in sample. There is however no statistically significant dif- GAR than other wealth quintile. However, in the WR, ference in the average age of respondents in the urban Duku BMC Health Services Research (2018) 18:384 Page 7 of 16 Table 3 Characteristics of Working-Age Adults by Rural and Urban Area in the Two Regions Total Sample (N = 4214) Greater Accra Region (N = 2187) Western Region (N = 2027) Mean (N = 4214) Urban Rural Mean Urban Rural Mean Urban Rural (N = 2171) (N = 2043) (N = 2187) (N = 1407) (N = 780) (N = 2027) (N =764) (N = 1263) Mean Age 37.74 38.61 36.81 38.61 39.26 37.98 36.59 37.43 36.09 Sex (%) Female 56.29 56.06 56.53 56.65 56.15 57.56 55.90 55.89 55.90 Male 43.71 43.94 43.47 43.35 43.85 42.44 44.10 44.11 44.10 Marital Status (%) Never Married 34.77 36.99 32.40 37.27 38.59 34.87 32.07 34.03 30.88 Married 51.50 48.60 54.58 48.98 46.91 52.56 54.27 51.70 55.82 Divorced 13.74 14.42 13.02 13.81 14.50 12.56 13.67 14.27 13.30 Religion (%) Other Religion 10.39 9.44 11.40 10.52 10.31 10.90 10.26 7.85 11.72 Christian 89.61 90.56 88.60 89.48 89.69 89.10 89.74 92.15 88.28 Mean Household Size 4.5 4.4 4.6 4.4 4.3 4.5 4.6 4.4 4.7 Educational Level (%) No Education 13.69 9.95 17.67 10.75 7.18 17.18 16.87 15.05 17.97 Basic Education 52.06 47.31 57.12 50.80 45.70 60.00 53.43 50.26 55.34 Secondary Educ. 34.24 42.75 25.21 38.45 47.12 22.82 29.70 34.69 26.68 and above Employment (%) Not Employed 30.45 33.12 27.61 32.14 34.47 27.95 28.61 30.63 27.40 Employed 69.55 66.88 72.39 67.86 65.53 72.05 71.39 69.37 72.26 Type of Employment Part-time 1.71 2.13 1.28 2.09 2.60 1.25 1.31 1.32 1.31 (%) (N = 2931) Full-Time 98.29 97.87 98.72 97.91 97.40 98.75 98.69 98.68 98.69 Occupation (%) Farmer 18.85 10.22 27.34 8.41 3.44 16.76 29.40 22.01 33.63 (N = 1931) Artisan/Trader 46.85 49.05 44.69 51.95 51.94 51.96 41.70 44.02 40.38 Laborer/Casual 7.83 8.67 7.01 8.07 8.10 8.01 7.59 9.65 6.42 Worker Managerial/ 18.46 22.41 14.57 20.79 24.20 15.08 16.10 19.35 14.27 Professional Business Owner 8.01 9.68 6.28 10.78 12.32 8.19 5.20 5.02 5.31 Wealth Status (%) Poorest 17.56 11.56 23.94 12.53 8.53 1974 22.99 17.15 26.52 Poor 17.13 15.02 19.38 15.96 14.14 19.23 18.40 16.62 19.48 Average 19.06 19.81 18.26 18.98 18.91 19.10 19.14 21.47 17.74 Rich 22.19 24.74 19.48 22.95 23.95 21.15 21.36 26.18 18.45 Richest 24.06 28.88 18.94 29.58 34.47 20.77 18.11 18.59 17.81 Average Annual 2340.76 2470.33 2203.07 2470.33 2256.41 2413.73 2371.23 2864.29 2072.97 Income Duku BMC Health Services Research (2018) 18:384 Page 8 of 16 Table 3 Characteristics of Working-Age Adults by Rural and Urban Area in the Two Regions (Continued) Total Sample (N = 4214) Greater Accra Region (N = 2187) Western Region (N = 2027) Mean (N = 4214) Urban Rural Mean Urban Rural Mean Urban Rural (N = 2171) (N = 2043) (N = 2187) (N = 1407) (N = 780) (N = 2027) (N =764) (N = 1263) Health Status (%) Bad Health 4.77 4.65 4.89 5.21 5.90 3.97 4.29 2.36 5.46 Fair Health 8.66 8.01 9.35 9.33 9.10 9.74 7.94 6.02 9.11 Good Health 86.57 87.33 85.76 85.46 85.00 86.28 87.77 91.62 85.43 Ownership of Private 39.44 20.73 59.32 38.45 18.62 74.23 40.50 24.61 50.12 Facility (%) Public 60.56 79.27 40.68 61.55 81.38 25.77 59.50 75.39 49.88 Time to Health 12.43 12.64 12.19 12.64 10.46 12.63 13.75 16.68 11.92 Facility Source: COHEiSION Project Baseline Survey of 2012 GAR Greater Accra Region, WR Western region Duku BMC Health Services Research (2018) 18:384 Page 9 of 16 Table 4 Bivariate Analysis of Health Insurance, Independent Variables by Regional and Rural/Urban Total Sample (N = 4214) Urban (N = 2171) Rural (N = 2043) GAR (N = 2171) WR (N = 2043) p-value GAR (N = 1407) WR (N = 764) P-value GAR (N = 780) WR (N = 1263) P-value Enrolment in Health Insurance Uninsured (%) 56.51 43.49 0.000*** 68.92 31.08 0.004** 43.04 56.96 0.019* Insured (%) 45.35 54.65 58.82 41.18 31.47 68.53 Type of Health Insurance Other Health Insurance (%) 60.18 39.82 0.032* 81.03 18.97 0.024* 38.18 61.82 0.429 NHIS (%) 44.32 55.68 57.26 42,74 31.01 68.99 Average Age 38.79 36.59 0.002** 39.26 37.43 0.078 27.97 36.09 0.051* Sex Females (%) 52.23 47.77 0.601 64.91 35.09 0.892 38.87 61.13 0.446 Males (%) 51.47 48.53 64.68 35.32 37.27 62.73 Marital Status Never Married (%) 55.63 44.37 0.034* 67.62 32.38 0.319 41.09 58.91 0.366 Married (%) 49.31 50.69 62.56 37.44 36.77 63.23 Divorced (%) 52.16 47.83 65.18 34.82 36.84 63.16 Religion Other Religion (%) 52.51 47.49 0.889 70.73 27.27 0.260 36.48 63.52 0.786 Christian (%) 51.83 48.17 64.81 35.19 38.40 61.60 Average Household Size 4.4 4.6 0.252 4.3 4.4 0.655 4.5 4.7 0.519 Educational Level No Education (%) 40.73 59.27 0.009** 46.76 53.24 0.006** 37.12 62.88 0.408 Basic Education (%) 50.64 49.36 62.61 37.39 40.10 59.90 Secondary Educ. and above (%) 58.28 41.72 71.44 28.56 34.56 65.44 Employment Unemployed (%) 54.79 45.21 0.122 67.45 32.55 0.263 38.65 61.35 0.857 Employed (%) 50.63 49.37 63.50 36.50 38.00 62.00 Health Status Bad Health (%) 56.72 43.28 0.324 82.18 17.82 0.005** 31.00 69.00 0.493 Fair Health (%) 55.89 44.11 73.56 26.44 39.79 60.21 Good Health (%) 51.23 48.77 63.08 36.92 38.41 61.59 Wealth Status Poorest (%) 37.03 62.97 0.000*** 47.81 52.19 0.005** 31.49 68.51 0.482 Poor (%) 48.34 51.66 61.04 38.96 37.88 62.12 Duku BMC Health Services Research (2018) 18:384 Page 10 of 16 Table 4 Bivariate Analysis of Health Insurance, Independent Variables by Regional and Rural/Urban (Continued) Total Sample (N = 4214) Urban (N = 2171) Rural (N = 2043) GAR (N = 2171) WR (N = 2043) p-value GAR (N = 1407) WR (N = 764) P-value GAR (N = 780) WR (N = 1263) P-value Average (%) 51.68 48.32 61.86 38.14 39.95 60.05 Rich (%) 53.69 46.31 62.76 37.24 41.46 58.54 Richest (%) 63.81 36.19 77.35 22.65 41.86 58.14 Ownership of Facility Private (%) 50.60 49.4 0.869 58.22 41.78 0.696 47.77 52.23 0.204 Public (%) 52.74 47.26 66.53 33.47 24.19 75.81 Average Time to Health 11.23 13.75 0.106 10.46 16.68 0.018* 12.63 11.92 0.761 Facility Source: COHEiSION Project Baseline Survey of 2012. For continuous variables, means are reported and significant test is based on t-test whiles for categorical variables percentages are reported and significant test is based on Chi-square test.; Note; *p < 0.05; ** p < 0.01; ***p < 0.001 Duku BMC Health Services Research (2018) 18:384 Page 11 of 16 Table 5 Determinant of Health Insurance Enrolment Adjusted Odds Ratio Total Sample Urban Sample Rural Sample Independent Variables GAR WR GAR WR GAR WR Sex Males 0.66*** (0.53–0.82) 0.52*** (0.44–0.63) 0.74** (0.57–0.97) 0.49*** (0.36–0.67) 0.46*** (0.34–0.64) 0.55*** (0.42–0.71) Age 1.02*** (1.01–1.03) 1.03*** (1.02–1.04) 1.01** (1.01–1.03) 1.04*** (1.02–1.05) 1.01* (0.99–1.04) 1.02*** (1.01–1.04) Marital Status Married 1.28 (0.93–1.76) 1.14 (0.84–1.56) 1.14 (0.82–1.58) 0.86 (0.59–1.26) 1.60 (0.76–3.38) 1.29 (0.77–2.13) Divorced 1.11 (0.71–1.74) 0.66* (0.44–0.99) 0.96 (0.60–1.52) 0.42** (0.21–0.85) 1.22 (0.41–3.62) 0.73 (0.45–1.20) Religion Christian 1.08 (0.67–1.77) 1.10 (0.78–1.56) 1.20 (0.67–2.13) 1.08 (0.70–1.67) 0.68 (0.29–1.58) 1.05 (0.63–1.76) Household Size 1.04 (0.98–1.10) 1.08 (0.98–1.18) 1.03 (0.94–1.13) 1.15* (0.99–1.31) 1.03 (0.93–1.16) 1.04 (0.92–1.18) Educational Level Basic Education 0.89 (0.63–1.28) 1.18 (0.89–1.57) 0.93 (0.50–1.72) 1.87* (1.06–3.31) 0.89 (0.59–1.34) 0.99 (0.75–1.32) Secondary Edu. & 1.49* (1.00–2.21) 1.61* (1.06–2.44) 1.37 (0.75–2.52) 1.95 (0.87–4.39) 2.07*** (1.33–3.22) 1.67* (1.05–2.66) Above Employment Employed 0.82 (0.61–1.11) 0.79 (0.61–1.03) 0.79 (0.55–1.13) 0.84 (0.52–1.39) 0.85 (0.49–1.47) 0.79 (0.55–1.14) Health Status Fair Health 0.84 (0.53–1.34) 0.63 (0.26–1.53) 0.85 (0.45–1.62) 0.08 (0.01–0.72) 1.10 (0.60–2.02) 0.87 (0.35–2.19) Good Health 0.60* (0.39–0.96) 0.54 (0.22–1.34) 0.58 (0.32–1.08) 0.06 (0.01–0.59) 0.82 (0.33–2.03) 0.80 (0.33–1.95) Wealth Status Poor 0.91 (0.49–1.67) 1.47 (0.88–2.46) 0.67 (0.29–1.54) 2.54* (0.99–6.52) 0.98 (0.38–2.55) 1.07 (0.57–2.03) Average 1.11 (0.60–2.04) 1.43 (0.90–2.26) 0.66 (0.28–1.58) 1.49 (0.56–3.97) 1.75 (0.75–4.13) 1.25 (0.72–2.17) Rich 1.35 (0.72–2.50) 1.77* (1.04–3.02) 1.16 (0.47–2.86) 2.68 (0.84–8.49) 1.04 (0.45–2.42) 1.26 (0.68–2.34) Richest 1.73* (0.97–3.10) 2.04** (1.26–3.28) 1.13 (0.49–2.63) 2.59 (0.93–7.22) 2.55** (1.16–5.59) 1.76* (1.04–2.98) Travel time to Facility 1.00 (0.98–1.02) 0.99 (0.98–1.00) 1.00 (0.98–1.03) 0.99 (0.99–1.01 1.01 (0.98–1.04) 0.98** (0.97–0.99) Ownership of Facility Public 1.83*** (1.55–2.16) 0.93 (0.82–1.06) 1.78*** (1.39–2.26) 0.95 (0.82–1.11) 0.48*** (0.31–0.73) 0.31*** (0.26–0.37) Constant 0.14*** (0.06–0.33) 0.34* (0.11–0.04) 0.17*** (0.05–0.59) 1.08 (0.14–8.65) 0.366 (0.09–1.58) 0.42 (0.11–1.59) Pseudo R-squared 0.086 0.079 0.075 0.102 0.135 0.086 No. of Observation 2132 1946 1336 736 735 1167 Source: COHEiSION Project Baseline Survey of 2012. Note: Adjusted odds ratio are reported with reference categories: Age (1 year increase in age); Sex (Female); Marital status (Never Married); Religion (Other Religion); Household size (Increase of 1 person); Educational level (No Formal Education); Employment (Unemployed); Health Status (Poor Health Status); Wealth status (Poorest); Travel time (1 min increase in travel time); Facility ownership (Private); Missing values are excluded from the analyses. Note: *p < 0.05; ** p < 0.01; ***p < 0.001 Duku BMC Health Services Research (2018) 18:384 Page 12 of 16 or rural samples. Similarly, there is no statistically sig- WR for the total, urban and rural samples. For the total nificant difference in the proportion of females and sample, females are significantly more likely to enroll in males in either the total, urban or rural samples. the NHIS than males for both the GAR and WR. Although in the total sample, there are more married re- Similarly, for both the urban and rural samples, fe- spondents in the GAR (56%) than the WR (44%), the dif- males are significantly more likely to enroll in the ferences in the proportion of married respondents NHIS than males in the GAR and WR. These find- between the regions in the urban and rural samples are ings are consistent with findings by Mwaura [23], statistically insignificant as shown in Table 4. The pro- Ayitey et al. [27]and Juttingetal.[32]. Age also im- portion of respondents with basic level education is sig- pacts positively on NHIS enrolments. The results nificantly higher in the GAR (51%) than the WR (49%) show that an increase of 1 year in age significantly in the total sample. Similarly, in the urban sample, the increases the odds of NHIS enrolment in both regions proportion of respondents with basic level education is for the total, rural and urban samples. These findings higher in the GAR (63%) than the WR (37%). There is are consistent with findings by Bhat and Jain [24], however no statistically significantly difference in the Kronick and Gilmer [25], Savage and Wright [26]and proportion of respondents with basic level education be- Ayitey et al. [27]. In terms of marital status, for the tween the GAR and WR in the rural sample. In terms of total sample, whiles divorcees are 1.08 (CI = 0.67– health status, there is no statistically significant differ- 1.77) times more likely to enroll in the NHIS in the ence in the categories of health status between the GAR GAR, their WR counterparts are 0.66 (CI = 0.44–0.99) and WR for the total and rural samples. The GAR how- times significantly less likely to enroll. For the urban ever, have higher proportions the health status categories sample, although divorcees in both regions are less likely than the WR in the urban sample. The proportion of re- to enroll as compared to those married, the results for the spondents in the richest wealth quintile is significantly WR is statistically significant at the 95% confidence inter- higher in the GAR (64%) than the WR (36%) in the total val. However, for the rural sample, whiles divorcees in the sample. Similarly, the proportion of respondents in the GAR are 1.22 (CI = 0.41–3.38) more likely to enroll, their richest wealth quintile is significantly higher in the GAR WR counterparts are 0.73 (CI = 0.45–1.20) less likely to (77%) than the WR (23%). However, there is no statisti- enroll even though this result is statistically insignificant cally significant difference in wealth status categories be- at the 90% confidence interval. This finding is in contrast tween the GAR and WR in the rural sample. with earlier studies by Asenso Okyere et al. [14], Ayitey et There is also no statistically significant differences be- al. [27] and Brugiavini and Pace [28] who found married tween the GAR and WR in terms of religion, employment, individuals to be significantly more likely to enroll in health facility ownership and average time to health facil- health insurance. Although the results further shows that ity either for the total, urban or rural samples. being married increases the odds of NHIS enrolment in both regions for the total, urban and rural samples, these Determinant of health insurance enrolment findings are statistically insignificant at the 90% confi- Table 5 presents the logistic regression estimations of dence interval. Similarly, although the results show that the determinant of NHIS enrolment among working-age being a Christian and a 1 member increase in household adults for the total, urban and rural samples. Estimating sizes increases the odds of NHIS enrolment, these findings individuals’ income in developing countries is difficult are statistically insignificant even at the 90% confidence and unreliable because most people are reluctant to dis- interval. close their true income. A five quintile wealth status was The results also show that an individuals’ level of therefore computed from household food and non-food education influence the odds of NHIS enrolment. For consumption expenditure as a proxy for income levels the total sample, having secondary level education and and used in the regression estimation. Again due to the above significantly increases the likelihood of NHIS en- inability of respondents to accurately determine the dis- rolment by 1.49 (CI = 1.00–1.11) times in the GAR and tance in kilometers from their home to the nearest pri- 1.62 (CI = 1.06–2.44) times in the WR respectively as mary healthcare facility, time taken in minutes to move compared to individuals with no formal education. For from home to the nearest primary healthcare facility was the urban sample, although the results shows that sec- used as a proxy to measure distance from home to ondary level education and above increases the odds of health facility. The differences in the determinants of en- enrolment by 1.37 (CI = 0.75–2.52) times in the GAR rolment between the two regions are presented. and 1.95 (CI = 0.87–4.39) times in the WR, these find- The results show that generally, age, sex, marital status, ings are statistically insignificant at the 90% confidence educational level, health status, wealth status and health interval. However, for the rural sample, individuals with facility ownership are significant determinants of NHIS secondary level education and above are significantly enrolment at varying extents between the GAR and the more likely to enroll in the NHIS by 2.07 (CI = 1.33– Duku BMC Health Services Research (2018) 18:384 Page 13 of 16 3.22) times in the GAR and 1.67 (CI = 1.05–1.14) times in educational level, marital status, health status, wealth sta- the WR. These findings are consistent with findings by tus and health facility ownership are significant determi- Asenso Okyere et al. [14], Ayitey et al. [27] and Brugiavini nants of NHIS enrolment. and Pace [28]. The results show being employed does not Females in both regions are more likely to enroll com- positively influence NHIS enrolment in both regions for pared to males. Similarly, urban and rural females in the total, urban or the rural samples. These findings are both regions are more likely to enroll in the NHIS than however statistically insignificant at the 90% confidence males. This may be as a result of the Free Maternal interval. Health Policy under the NHIS which offer premium ex- Although the finding was statistically insignificant, in- emption to expectant and nursing mothers and therefore dividuals with poor self-assessed health status are more most females in the reproductive age-group might have likely to enroll in the NHIS than those with fair or good enrolled under this exemptions category. The likelihood health status in both regions for the total, urban and of NHIS enrolment was also found to increase with age rural samples. Ayitey et al. [27] reported similar findings for both regions in the total sample and similarly for of the effect of self-assessed health status on NHIS en- both regions in the urban and rural areas. This positive rollment. The results show that wealth status is a signifi- relationship of increasing age with NHIS enrolment can cant determinant of NHIS enrolment at varying extent be attributed to degeneration in health as people age in the two regions depending on the rural/urban locality and the need for increased healthcare utilization. Older of residence of the individual. For the total sample, the people therefore prefer to make investment in their richest in the GAR are 1.73 (CI = 0.97–3.10) times more health through the purchase of health insurance. This likely to enroll in the NHIS than the poorest whiles in findings is consistent with the findings of previous stud- the WR the rich and the richest are 1.77 (CI = 1.04– ies [20, 27, 36]. The increased likelihood of the aged to 3.02) times and 2.04 (CI = 1.26–3.28) times respectively health insurance enrolment can also be attributed to the more likely to enroll in the NHIS than the poorest. exemptions of people aged 70 years and above from pre- However, when the urban and rural samples are consid- mium payment under the NHIS exemption policy. Most ered separately, whiles in the rural sample, individuals in of the aged therefore enrolled under the 70+ age exemp- the richest wealth quintile in the GAR are 2.55 (CI = tion category. Married people are more likely to enroll 1.16–5.59) times and WR are 1.76 (CI = 1.04–2.98) times in the NHIS in both regions for the total sample and significantly more likely than those in the poorest quin- similarly in both regions in the urban and rural areas tile, in the urban sample individuals in the richest wealth than never married people. This may be because married quintile in both regions are more likely to enroll. These couples purchase health insurance to mitigate the finan- finding are statistically insignificant at the 90% confi- cial burden that is likely to accrue from raising children dence interval. These findings are in agreement with after marriage [27, 31, 33]. findings by Ayitey et al. [27] and Fowler et al. [50]. In The findings also point to educational class dimen- terms of travel time from home to the nearest primary sions in NHIS enrolment. The better educated in both healthcare facility, although a 1 min increase in the regions in the total and rural sample are significantly travel time reduces the odds of NHIS enrolment of indi- more likely to enroll than the uneducated. Although the viduals in the WR for the total, urban and rural samples, results from the urban sample is statistically insignifi- these findings were statistically insignificant at the 90% cance, there is still a positive relationship between higher confidence interval. From the results, ownership of the education and NHIS enrolment in both regions. This nearest primary healthcare facilities is another significant positive relationship between higher education and determinant of NHIS enrolment. Individuals who live NHIS is because higher education makes people better around the catchment area of a publicly owned primary understand and appreciate the insurance concept and its health facility in the GAR are 1.83 (CI = 1.55–2.16) times benefits to the household. People with poor self-assessed and 1.78 (CI = 1.39–2.26) times significantly more likely health status in both regions are more likely to enroll in to enroll in the NHIS for the total and urban samples re- the NHIS in the total and urban samples than those with spectively. However, in the rural sample, individuals who fair and good health status. However, in the rural sam- live in the catchment area of private facilities are signifi- ple, people who are of fair health status in the GAR are cantly more likely to enroll in the NHIS than those who more likely to enroll in the NHIS than their WR coun- live around public facilities. terparts. This suggest that people who are of poor health in urban communities on both regions self-select into Discussion the NHIS. Perhaps the availability of both public and The study assessed the differences in determinants of private health facilities in urban communities is such NHIS enrolment among working-aged adults in the GAR that people of poor health get to know of their under- and WR in Ghana. The findings indicates that age, sex, lying propensity to increased healthcare utilization. They Duku BMC Health Services Research (2018) 18:384 Page 14 of 16 therefore enroll in the NHIS to protect themselves from urban and rural samples. It is therefore not surprising catastrophic healthcare expenditure. Theoretical con- that a1minincreaseinthetravel timein theWR cepts such as adverse selection, risk aversion, affordabil- reduces the odds of NHIS enrolment for the total, ity and trust which are beyond the scope of this study urban and rural samples. The ownership of the near- can help explain and put these findings into proper est primary health facility is another important deter- perspective. minant of NHIS enrolment. People who visit public Working-age adults in the rich and richest quintiles in health facilities in the GAR are more likely to enroll both regions in the total sample are significantly more in the NHIS than their WR counterparts for the total likely to enroll in the NHIS than those in the poorest and urban samples. However in the rural areas, this wealth quintile. This suggest that wealth status and is not the case as those who visit private facilities are therefore affordability of insurance premium is an im- significantly more likely to enroll in the NHIS in both portant determinant of NHIS enrolment. In the urban regions. A possible explanation to this could be about sample, the odds of NHIS enrolment for WR individuals perceived poor quality care from these public facilities in the rich and richest quintiles is about twice that of which may influence peoples decision to enroll in the those in the GAR. Conversely, in the rural sample, the NHIS. This is because per the NHIS gate keeping sys- odds of NHIS enrolment for GAR individuals in the tem, card holders are expected to first visit a primary richest quintile are about twice that of those in the WR. health facility when sick. If people perceive the qual- Thus the rich in urban WR enroll more in the NHIS ity of care they receive from their nearest primary than the rich in urban areas in the GAR. However, in health facility to be poor, chances are that it will dis- the rural areas, the rich in rural GAR enroll more in the courage them from enrolling in the NHIS. NHIS than the rich in rural WR. The expectation is that with the pro-poor design of the NHIS and its extensive Conclusion exemption policy that exempt premium payment for in- The study assessed the differences in determinants of digents, the poor in rural areas in both regions will en- NHIS enrolment among working-aged adults in the roll more than the rich. These finding further galvanize GAR and WR in Ghana. The analysis employed logistic findings by earlier studies that poverty is a major barrier regression in the empirical estimation. The study did not to NHIS enrolment and that the NHIS is not pro-poor find any differences in the demographic determinants of as envisaged [27]. The fact that the odds of enrolment NHIS enrolment between the two regions and among among the rural rich in the GAR is higher than rural the rural and urban residents in the two regions. The rich in WR further indicates that levels economic activ- findings indicate that generally, age, sex, educational ities, income levels and affordability of premium in the level, marital status, health status and travel time to rural areas also impact on the decision to enroll in the health facility are significant determinants of NHIS en- NHIS. Although the WR has a high poverty rate of 18% rolment in both regions and similarly in the rural and compared to the GAR 12%, the NHIS enrolment rate in urban communities in the two regions. The study how- the WR (55.7%) is higher than the GAR (44.3%). ever found some differences between the two regions in However, the likelihood of enrolment is significantly terms of wealth status and health facility ownership as higher among the richest in the WR (OR = 2.04) than in determinants of NHIS enrolment. Although the rich and the GAR (OR = 1.73). This can be explained from the fact richest in both regions are more likely to enroll in the that the richest in the WR do not have access to many pri- NHIS than the poor and poorest, the odds of NHIS vate health facility and private health insurance companies enrolment for the richest in urban areas in the WR (OR like the richest in the GAR. So even if the richest in WR = 2.59) is about twice as that of GAR (OR = 1.13) whiles are not satisfied with the quality of care from public health in the rural areas the odds of NHIS enrolment in the GAR facilities which they will attend should they enroll in the (OR = 2.55) is also about twice that of the WR (OR = NHIS, they will still have to attend these same public facil- 1.76). People who visit public health facilities in the GAR ities should they opt out of the NHIS and pay are more likely to enroll in the NHIS than those in WR out-of-pocket or enroll with other private health insur- for the total and urban samples. However in the rural ance. Unlike the GAR where the richest have access to areas, those who visit private facilities are significantly many private insurance companies and private healthcare more likely to enroll in the NHIS in both regions. These facilities, the richest who decide to opt out of the NHIS findings suggest that inequalities still exists in NHIS enrol- for quality issues have readily available alternatives. ment in favor of the wealthy, communities with better Travel time from home to health facility is another im- socio-economic activities and communities with health fa- portant determinant of NHIS enrolment between the re- cilities that are perceived to provide better quality health- gions. The average travel time to the nearest health care. This thus raises concerns as to whether the NHIS is facility is higher in the WR than the GAR for the total, truly pro-poor as envisaged. Duku BMC Health Services Research (2018) 18:384 Page 15 of 16 The study contributes to the literature on the de- Funding This study received financial support from the Government of the terminants of NHIS enrolment by identifying factors Netherlands through the Netherlands Organization for Scientific Research that might be responsible for the observed differ- (NWO/WOTRO) in the form of a research grant (Grant No. W07.45.104.00) for ences in the NHIS regional enrolment coverages. the COHEiSION project. Collaborators of the study include Noguchi Memorial Institute for Medical Research, University of Ghana Legon; Amsterdam The differences in NHIS enrolment coverages be- Institute for Global Health and Development, University of Amsterdam; Vrije tween the regions may be as a result of differences University of Amsterdam and University of Groningen. in socio-economic factors that impact on the ability Availability of data and materials of the inhabitants to afford the insurance premium. All data supporting my findings and conclusions is contained in the This should serve as an important indicator to pol- manuscript. The full data is available in the COHEiSION Project data icymakers on the need to focus on regional specific repository at Noguchi Memorial Institute for Medical Research. There are no restrictions to data sources and details of the full data may be accessed geographic and proxy means targeting strategies through the Co-PI, Dr. Daniel Kojo Arhinful (Noguchi Memorial Institute for aimed at identifying the poor in resource constraints Medical Research, University of Ghana Legon. P. O. Box, LG 581, Legon. E- communities for premium exemptions. The study mail: DArhinful@noguchi.ug.edu.gh). recommends that policymakers should use innovative Author’s contributions approaches to determine NHIS premium at the dis- SKOD initiated the conceptualization, collected the data, conducted data trict level based on socio-economic activities and in- analysis and prepared the manuscript. The author read and approved the final manuscript. come levels within the district. They should also consider introducing quality healthcare dimension Ethics approval and consent to participate into provider payment mechanisms to rewards pro- Ethical clearance for the study was obtained from the Ghana Health Service (GHS) Ethical Review Committee (ERC clearance number: GHS-ERC 08/5/11). viders who meet quality criteria as expressed by Informed consent was also obtained from respondents. All literate respondents their clients. This will serve as an incentive for both signed written informed consent while illiterate respondents had the informed public and private accredited healthcare providers to consent form read to them in a local language that they understand before thumb-printing the form to participate in the study. provide quality healthcare that attract individuals to enroll in the NHIS. Competing interests Finally, the findings of this study is subject to the fol- The author declares he has no competing interests. lowing limitations. The analysis relied on self-reported measures such as health insurance enrolment status, Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in patterns of enrolment and self-assessed health status. published maps and institutional affiliations. Therefore any systematic differences due to respondents reporting bias could affect the precision of the reported Author details Department of Epidemiology, Noguchi Memorial Institute for Medical estimates. The study did not include individuals insured Research, University of Ghana, P. O. Box LG 581 Legon, Accra, Ghana. with other private insurance companies other than the 2 Amsterdam Institute for Global health and Development, Amsterdam, The NHIS due to the small sample size in the data. These Netherlands. Vrije Universiteit Amsterdam, Amsterdam, The Netherlands. limitation however does not invalidate the entirety of Received: 12 September 2017 Accepted: 8 May 2018 the self-reported measures and the analysis of these measure which have been well-documented. The find- References ings of the determinants of NHIS enrolment reported in 1. Carrin G, James C. Social health insurance: key factors affecting the this study are limited to the Greater Accra and Western transition towards universal coverage. Int Soc Secur Rev. 2005;58:1–15. regions of Ghana and therefore caution should be taken 2. WHO. Sustainable health financing, universal coverage and social health insurance. In World Health Assembly Resolution: WHA58.33. Geneva: WHO; when generalizing the findings to other populations. 3. Griffin C. Health care in Asia: a comparative study of cost and financing. Abbreviations Washington: World Bank Regional and Sectoral Studies; 1992. COHEiSION Project: Client-Oriented Health Insurance System in Ghana 4. GOG. The National Health Insurance Act: Act 650. Accra: Parliament House; Project; GAR: Greater Accra Region; GSS: Ghana Statistical Service; NHIA: National Health Insurance Authority; NHIF: National Health Insurance 5. Arhinful DK. The solidarity of self-interest: social and cultural feasibility Fund; NHIS: National Health Insurance Scheme; NOW: Netherlands of rural health insurance in Ghana. Leiden: African studies canter Organization for Scientific Research; SSNIT: Social Security and National research report; 2003. Available: https://openaccess.leidenuniv.nl/ Insurance Trust; WHA: World Health Assembly; WOTRO: Science for Global bitstream/handle/1887/12919/ASC-075287668-077-01.pdf?sequence=2. Development; WR: Western Region Accessed 20 August 2011. 6. Addo JS. Consultants & tri Star actuarial and management consultant. A Acknowledgements feasibility study for the establishment of a national health insurance scheme I acknowledge the support of the Government of the Netherlands through in Ghana. Accra: Ministry of Health; 1995. the Netherlands Organization for Scientific Research (NWO/WOTRO) in the 7. GOG. The national health insurance act: Act 852. Accra: Parliament House; form of a research grant (Grant No. W07.45.104.00) for the COHEiSION 2012. project. I am most grateful to staff of the Epidemiology Department, 8. NHIA. National health insurance authority 2013 annual report. Accra, Ghana: Noguchi Memorial Institute for Medical Research, University of Ghana and NHIA; 2014. staff of the Faculty of Economics and Business Administration, Vrije 9. NHIA. National health insurance authority 2011 annual report. Accra, Ghana: Universiteit Amsterdam for their support in the project design. NHIA; 2012. Duku BMC Health Services Research (2018) 18:384 Page 16 of 16 10. GSS. 2010 population and housing census. Accra-Ghana: Ghana Statistical 36. Nketiah-Amponsah E. Demand for health insurance among women in Service; 2012. Ghana: cross-sectional evidence. International Research Journal of Finance 11. FAO. Gender inequalities in rural employment in Ghana: An overview. Accra- and Economics. 2009;33:179–91. Ghana; 2012. http://www.statsghana.gov.gh/docfiles/Gender/FAO_GHANA_ 37. Duncan RP, Seccombe K, Amey C. Changes in health insurance coverage COUNTRY_PROFILE_FINAL2012.pdf. Accessed 20 Sept 2012. within rural and urban environments. J Rural Health. 1995;11:160–76. 12. Akazili J, Welaga P, Bawah A, Achana FS, Oduro A, et al. Is Ghana’s pro-poor 38. Stock R. Distance and utilization of health facilities in rural Nigeria. Soc Sci health insurance scheme really for the poor? Evidence from northern Med. 1983;17:63–570. Ghana. BMC Health Serv Res. 2014;14:637. 39. Jong KE, Smith DP, Yu XQ, O’Connell L, Goldstein D, Armstrong BK. Remoteness of residence and survival from cancer in new South Wales. 13. Cameron AC, Trivedi PK, Milne F, Piggott J. A micro-econometric model of Med J Aust. 2004;180:618–22. the demand for health care and health insurance in Australia. Rev Econ 40. Criel B, Van der Stuyft P, Van Lerberghe W. The Bwamanda hospital Stud. 1988;55(1):85–106. insurance scheme: effective for whom? A study of its impact on hospital 14. Asenso-Okyere W, Osei-Akoto I, Anum A, Appiah E. Willingness to pay for utilization patterns. Soc Sci Med. 1999;48:897–911. health insurance in a developing economy. A pilot study of the informal sector 41. Nemet GF, Bailey AJ. Distance and healthcare utilization among the rural of Ghana using contingency valuation. Health Policy. 1997;42(3):223–37. elderly. Soc Sci Med. 2000;50:1197–208. 15. Sanhueza R, Ruiz-Tagle J. Choosing health insurance in a dual health care 42. Frenzen PD. Health insurance coverage in U.S. urban and rural areas. J Rural SYSTEM: the Chilean case. J Appl Econ. 5(1):157–84. Health. 1993;9:204–14. 16. Ying XH, Hu TW, Ren J, Chen W, Xu K, Huang JH. Demand for private health 43. Taylor AK, Cohen JW, Machlin SR. Being uninsured in 1996 compared to insurance in Chinese urban areas. Health Econ. 2007;16(10):1041–50. 1987: how has the experience of the uninsured changed over time? Health 17. Chankova S, Sulzbach S, Diop F. Impact of mutual health organizations: Serv Res. 2001;36:16–31. evidence from West Africa. Health Policy Plan. 2008;23:264–76. 44. Jehu-Appiah C, Aryeetey G, Spaan E, De Hoop T, Agyepong I, Baltussen R. 18. Dong H, Gbangou A, De Allegri M, Pokhrel S & Sauerborn R. The differences Equity aspects of the national health insurance scheme in Ghana: who is in characteristics between health-care users and non-users: implication for enrolling, who is not and why? Soc Sci Med. 2011;72:157–65. introducing community-based health insurance in Burkina Faso. European 45. Coburn AF, Mcbridge TD, Ziller EC. Patterns of health insurance coverage Journal of Health Economics, 2008; 9(1): 41–50. among rural and urban children. Med Care Res Rev. 2002;59:272–92. 19. Gius MP. An analysis of the health insurance coverage of young adults. Int J 46. Holmes M, Ricketts T. In: North Carolina Rural Health Research and Policy Appl Econ. 2010;7(1):1–17. Analysis Centre, editor. Rural-urban differences in the rate of health 20. Sarpong N, Loag W, Fobil J, Meyer CG, Adu-Sarkodie Y, May J, et al. National insurance coverage. North Carolina; 2008. http://www.shepscenter.unc.edu/ health insurance coverage and socio-economic status in a rural district of wp-content/uploads/2013/07/ruralgap_finalbrief.pdf. Accessed 21 Ghana. Trop Med Int Health. 2010;15(2):191–7. September 2012. 21. Dixon J, Tenkorang EY, Luginaah I. Ghana’s national health insurance 47. Hu HM, Duncan RP, Radcliff TA, Porter CK, Hall AG. Variations in health scheme: helping the poor or leaving them behind? Environment and insurance coverage for rural and urban nonelderly adult residents of Florida, Planning C: Politics and Space. 2011;29(6):1102–15. Indiana, and Kansas. J Rural Health. 2006;22:147–50. 22. Grossman M. On the concept of health capital and the demand for health. 48. Comer J, Mueller K. Access to health care: urban-rural comparisons from a J Polit Econ. 1972;80:223–55. Midwestern agricultural state. J Rural Health. 1995;11:128–36. 23. Mwaura JW, Pongpanich S. Access to health care: the role of community 49. Ketsche P. Employment-based health insurance: analysis of rural-urban based health insurance in Kenya. Pan African Medical Journal. 2012;12:35. differences in one state. Med Care Res Rev. 2005;62:458–78. 24. Bhat R, Jain N. Factors affecting the demand for health insurance in a micro 50. Fowler R, Noyahr L, Thornton J, Pinto R, Kahn J. An official thoracic society insurance scheme. Indian Institute of Management, Working Paper 07-02. systematic review: the association between health insurance status and Ahmedabad: Research and Publication Department; 2006. access, care delivery and outcomes for patients who are critically ill. 25. Kronick R, Gilmer T. Explaining the decline in health insurance coverage, American Journal of Critical Care Medicine. 2010;181:1003–11. 1979 – 1995. Health Aff. 1999;18(2):30–47. 51. Von Neumann J, Morgenstern O. Theory of games and economic 26. Savage E, Wright D. Health Insurance and health care utilization: theory and behaviour. Princeton, New Jersey, USA: Princeton University Press; 1994. evidence from Australia 1989–1990. CHERE Discussion Paper 44. Sydney: 52. Propper C. The demand for private health care in the UK. J Health Econ. University of Technology; 2001. 2000;19(6):855–76. 27. Ayitey AM, Nketiah-Amponsah E, Barimah A. Determinants of insurance 53. Schneider P. Why should the poor insure? Theories of decision-making in enrolment among Ghanaian adults: the case of the national health insurance the context of health insurance. Health Policy Plan. 2004;19(6):347–55. scheme. Economics Management and Financial Markets. 2013;8(3):37. 54. Fin C, Harmon C. A dynamic model of demand for private health insurance 28. Brugiavini A, Pace M. Extending health insurance: effects of the national in Ireland. IZA Discussion Paper No. 2472. Bonn: IZA; 2006. http://ftp.iza.org/ health insurance scheme in Ghana. Italy: European report on development, dp2472.pdf. European University Institute, Badia Fiesolana; 2011. http://cadmus.eui.eu/ 55. Arrow K. Uncertainty and the welfare economics of medical care. Am Econ bitstream/handle/1814/17221/RSCAS_2011_27.pdf?sequence=1&isAllowed= Rev. 1963;53(5):941–73. y. Accessed 20 Sept 2012. 56. Feldstein MS. The welfare loss of excess health insurance. J Polit Econ. 1973; 29. Bourne PA, Kerr-Campbell MD. Determinants of self-rated private health 81(2, part 1):251–80. insurance coverage in Jamaica. Health. 2010;2(06):541. 30. Muurinen JM. Demand for health: a generalized Grossman model. J Health Econ. 1982;1(1):5–28. 31. Butler J. Estimating elasticity of demand for private health insurance in Australia. National center for epidemiology and population health. Canberra:ANU;1999. 32. Jutting J. Health Insurance for the Rural Poor? Community financing scheme in Senegal to protect against illness. Development and Cooperation. 2001;6:4–5. 4 33. Cameron AC, McCallum. Private health insurance choices in Australia: The role of long-term utilization of health services, in Economics and health, 1995. Sydney: School of Health Services Management, UNSW; 1996. p. 143–157. 34. Cameron AC, Trivedi PK. The role of income and health risk in the choice of health insurance: evidence from Ghana. J Public Econ. 1991;45(1):1–28. 35. Kirigia JM, Sambo LG, Nganda B, Mwabu GM, Chatora R, Mwase T. Determinants of health insurance ownership among south African women. BMC Health Serv Res. 2005;5:17.

Journal

BMC Health Services ResearchSpringer Journals

Published: May 29, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off