National health insurance scheme enrolment and antenatal care among women in Ghana: is there any relationship?

Authors


Abstract

Objectives

The objective of this study was to examine whether enrolment in the National Health Insurance Scheme (NHIS) affects the likelihood and timing of utilising antenatal care among women in Ghana.

Methods

Data were drawn from the Ghana Demographic and Health Survey, a nationally representative survey collected in 2008. The study used a cross-sectional design to examine the independent effects of NHIS enrolment on two dependent variables (frequency and timing of antenatal visits) among 1610 Ghanaian women. Negative binomial and logit models were fitted given that count and categorical variables were employed as outcome measures, respectively.

Results

Regardless of socio-economic and demographic factors, women enrolled in the NHIS make more antenatal visits compared with those not enrolled; however, there was no statistical association with the timing of the crucial first visit. Women who are educated, living in urban areas and are wealthy were more likely to attend antenatal care than those living in rural areas, uneducated and from poorer households.

Conclusion

The NHIS should be strengthened and resourced as it may act as an important tool for increasing antenatal care attendance among women in Ghana.

Abstract

Objectifs

Evaluer si l'inscription au programme nationale d'assurance santé (NHIS) influe sur la probabilité et le moment de l'utilisation des soins prénataux chez les femmes au Ghana.

Méthodes

Les données ont été tirées de l’Enquête Démographique et de Santé du Ghana, une enquête nationale représentative réalisée en 2008. L’étude a utilisé une conception transversale pour examiner les effets indépendants de l'inscription au programme NHIS sur deux variables dépendantes (la fréquence et le calendrier des visites prénatales) chez 1.610 femmes ghanéennes. Des modèles binomiaux et logit négatifs ont été appliqués, étant donné que les variables de comptage et catégoriques ont été respectivement utilisées comme indicateurs de résultats.

Résultats

Indépendamment des facteurs socioéconomiques et démographiques, les femmes inscrites dans le programme NHIS effectuent plus de visites prénatales que les non inscrites; mais il n'y avait pas d'association statistique avec le moment de la première visite cruciale. Les femmes instruites, vivant dans les zones urbaines et riches, étaient plus susceptibles d'assister à des soins prénataux que celles vivant dans les zones rurales, sans instruction et dans des ménages plus pauvres.

Conclusion

Le programme NHIS devrait être renforcé et doté de ressources car il peut agir comme un outil important pour accroître la fréquentation des consultations prénatales chez les femmes au Ghana.

Abstract

Objetivos

Examinar si la participación dentro del Esquema Nacional de Seguros Sanitarios (ENSS) afecta la probabilidad y el momento de utilizar los servicios prenatales entre mujeres de Ghana.

Métodos

Se utilizaron datos del Censo Demográfico y Sanitario de Ghana, una encuesta nacional representativa recogida en el 2008. El estudio utilizó un diseño croseccional para examinar los efectos independientes de la participación en el ENSS sobre dos variables dependientes (frecuencia y momento de las visitas prenatales) entre 1.610 mujeres de Ghana. Los modelos de distribución binomial negativo y logit se ajustaron, ya que se utilizaron variables de conteo y categóricas como resultados, respectivamente.

Resultados

Independientemente de los factores socioeconómicos y demográficos, las mujeres que tomaban parte en el ENSS tenían más visitas prenatales comparadas con aquellas que no lo hacían. Sin embargo, no existía una asociación estadística con el momento de la tan crucial primera visita. Las mujeres más educadas, viviendo en áreas urbanas o con mayor riqueza tenían una mayor probabilidad de ir a visitas prenatales que aquellas viviendo en áreas rurales, con menor educación o pertenecientes a hogares más pobres.

Conclusión

El ENSS debería fortalecerse y se le deberían asignar recursos puesto que puede actuar como una herramienta importante para aumentar la asistencia a los servicios prenatales de las mujeres de Ghana.

Introduction

Ghana's National Health Insurance Scheme (NHIS), created to abolish user fees and ensure more equitable access to health care (Agyepong & Adjei 2008), is one of the very few sub-Saharan African examples of a nationally organised and financed healthcare scheme, thus making it a trail blazer for the universal healthcare movement on the continent. Since its assentation into law in 2003, studies have focused predominantly on enrolment rates and questions of access (Sarpong et al. 2010; Dixon et al. 2011; Jehu-Appiah et al. 2011). Very few studies have examined the impact of the recently introduced NHIS on health service utilisation and specifically on antenatal care (ANC) attendance. This article seeks to fill this void by examining the effects of the NHIS on both the frequency and timing of ANC attendance among Ghanaian women. Given that maternal and child health issues have recently dominated international and domestic policy initiatives (WHO 2005), ANC usage provides a logical point to test the NHIS' effectiveness.

Antenatal care – contact with a health provider during the time of pregnancy – is an important part of the maternal care continuum. ANC is beneficial to the health of both mother and child including prevention, detection and treatment of anaemia; detection and treatment of hypertension; treatment of eclampsia (convulsions); prevention of obstructed labour; and screening for and prevention of infection (Carroli et al. 2001); immunizations; treatment for malaria (WHO 2003); and reduction in low birthweights (Raatikainen et al. 2007). ANC can facilitate women coming in contact with health workers for the first time and creating a sense of comfort with the staff and environment of the health centre; this in turn may increase the likelihood of delivery and post-natal care at the health facility (Sugathan et al. 2001; Mrisho et al. 2009). It also provides time and space for health education and promotion, which may benefit the health of mother and child down the line (WHO 2003), and reduces the need for more expensive healthcare outputs in the future (Jowett 2000).

Under the previous maternal care programme in Ghana between 2005 and early 2008, ANC was only available with a charge, although delivery was free in public facilities. However, parallel to this, women who had enrolled in the newly implemented NHIS would receive a complete benefit package. Recognising the inequities and inefficacies of these separate health policies, and in an attempt to better address its commitment to Millennium Development Goals (MDGs) 4 and 5 aimed at reducing child mortality and improving maternal health, the Ghana government in 2008 tucked maternal care under the umbrella of the NHIS by creating an exemption for pregnant women from having to pay premium or renewal fees. The benefit package includes ANC, delivery services, post-natal care of the mother and free neonate coverage on the mother's card for 3 months after delivery (Dzakpasu et al. 2012). However, women must prove their pregnancy and go through the process of NHIS registration before they can take advantage of this offer for free maternal health care (Witter et al. 2013).

Even with fee exemptions included with maternal care policies (both historical and current), limited financial resources may still influence access and utilisation of ANC for women in Ghana. Many cannot afford the cost of transport to the health facility, the loss of earnings from time taken away from work and having to pay for medicines that are not covered under exemption policies (Addai 2000; Finlayson & Downe 2013). Studies also suggest that socio-economic status plays a large role in access and utilisation: the poor generally have far less access (Ghana Health Service 2008; Simkhada et al. 2008; Doku et al. 2012), as do those in rural areas (Arthur 2012; Atunah-Jay et al. 2013). Some studies demonstrate a positive relationship between a woman's age and ANC usage (Addai 2000; Doku et al. 2012; Owoo & Lambon-Quayefio 2013). However, there is reason to believe the NHIS may offset inequalities in access to health care. Witter and Garshong (2009) found that the introduction of the NHIS increased access to health care by taking away the out-of-pocket costs. There has been some preliminary evidence to suggest that members of the scheme are more likely to utilise health care (Apoya & Marriott 2011; Blanchet et al. 2012; Sekyi & Domanban 2012; Aboagye & Agyemang 2013) and in a more timely manner (Sulzbach et al. 2005; Health Systems 20/20 2009).

Mensah et al. (2010) used propensity score matching to assess women's access to health care within urban and rural districts in two of Ghana's regions and found that the NHIS improved access to ANC as well as increased the number of deliveries at hospital facilities. Dzakpasu et al. (2012), analysing trends in deliveries in seven districts in the Brong Ahafo Region also suggest that free care via the NHIS could increase the use of maternal health care in Ghana. However, both Mensah et al.'s (2010) and Dzakpasu et al.'s (2012) regional limitations mean that generalizations are difficult; there is a clear need for a national-level investigation. Using a nationally representative survey, Ghana Demographic and Health Survey (GDHS), we contribute to the extant literature that explores linkages between national health programmes and policies, and utilisation of maternal health care among women in Ghana.

We focus on testing women's compliance to the World Health Organization's (WHO) ANC model, operationalized into two aspects. WHO's standard for the provision of effective ANC states: ‘All pregnant women should have at least four [ANC] assessments by or under the supervision of a skilled attendant. These should as a minimum, include all the interventions outlined in the new WHO [ANC] model and be spaced at regular intervals through pregnancy, commencing as early as possible’ (WHO 2007): 49. Taking cues from this, we asked these specific research questions: First, do NHIS-enrolled pregnant women engage in more regular antenatal visits than non-enrolled pregnant women? and Second, do NHIS-enrolled pregnant women access ANC faster or earlier than the non-enrolled after the onset of pregnancy?

Methods

We use data from the 2008 Ghana Demographic and Health Survey (GDHS). The GDHS is a nationally representative data set administered by the Ghana Statistical Service (GSS) and Macro International, and the fifth in such surveys of the Global Demographic and Health Surveys Program. GDHS aims at monitoring the population and social well-being of Ghanaians (Ghana Statistical Service 2009). The 2008 GDHS employed a two-staged stratified sample frame where systematic sampling with probability proportional to size was used to identify enumeration areas from which households were selected. The GDHS identified about 5096 eligible women aged 15–49 years from 11 778 households, of which 4916 were interviewed yielding a response rate of 97%. Data were collected by a team of trained interviewers under the supervision of senior staff from the Ghana Statistical Service (GSS) with technical input from ICF Macro International, Inc. (Ghana Statistical Service 2009).

As the NHIS only began slowly coming into operation after 2005, the 2008 GDHS was the first year of the survey that asked participants about their enrolment in the health insurance scheme. Given this, we were only able to draw on one round of the GDHS and we are not able to make comparisons over time. The analysis was restricted to women whose last birth occurred between 2005 (when NHIS became available) and the time of interview.

Measures

Two major dependent variables employed for this study were ‘number of ANC visits during pregnancy’ and ‘timing of first ANC check’. While number of ANC visits depends on the medical needs and risk status of the specific woman in question, WHO recommends a minimum of four ANC visits, and an early first visit, mostly before the end of the first trimester, for screening and identification of infections during this period.

Our focal independent variable, enrolment in Ghana's NHIS, was measured with the question: ‘What type of health insurance do you have?’ Respondents who answered they had health insurance and then specified the type as ‘National/District Health Insurance’ were considered enrolled in the NHIS. Thus, the variable was dichotomized into the categories ‘yes’ or ‘no’ for enrolment in the NHIS. We controlled for socio-economic and demographic variables. Socio-economic variables used include educational background of respondents coded (no education = 0; primary education = 1; secondary education = 2; higher education = 3), employment status of respondents coded (not employed = 0; employed = 1) and wealth status, a composite index based on the household's ownership of a number of consumer items including television and a car, flooring material, drinking water, toilet facilities, etc., coded (poorest = 0; poorer = 1; middle = 2; richer = 3; richest = 4). Demographic variables included in the analysis are age of respondents as of their last birthday, marital status coded (never married = 0; currently married = 1; formerly married = 2), rural/urban residence (urban = 0; rural = 1), region of residence (Greater Accra= 0; Central = 1; Western = 2; Volta = 3; Eastern = 4; Ashanti = 5; Brong Ahafo = 6; Northern = 7; Upper East = 8; Upper West = 9), ethnicity (Akan = 0; Ga/Adangbe = 1; Ewe = 2; Northern languages = 3; other languages = 4) and the religious denomination of respondents (Christians = 0; Muslims = 1; Traditional = 2; No religion = 3). A latent variable measuring barriers to access was derived using principal component analysis (PCA) from questions that asked women some of the many factors that prevent them from accessing medical advice or treatment. Thus, women were asked to identify whether it was a big problem: getting permission to go the health facility, getting money needed for treatment, distance to the health facility, having to take transport and going to the health facility alone. All these were loaded on the same latent construct with factor loading ranging from 0.4 to 0.9. Variables were coded as ‘0 = no problem’, ‘1 = not a big problem’ and ‘2 = a big problem’. Higher values on the scale indicate more problems faced by women in accessing medical treatment, while lower values indicate fewer problems. Reliability coefficient (Cronbach's alpha) was estimated as 0.836.

Data analysis

We used Poisson regression to analyse the outcome variable ‘number of antenatal visits during pregnancy’, given that this technique is more appropriate for count outcomes (Mullahy 1986; Lawless 1987; Dean & Lawless 1989). Counts are usually positive integers that assume a Poisson rather than a normal distribution. A key assumption underlying the Poisson regression model is that the estimated conditional variance should equal the conditional mean for the data points under observation (Lawless 1987). Preliminary diagnostics indicate, however, that the variance was greater than the mean, a condition described as overdispersion, which if not corrected may bias standard errors and parameter estimates. Thus, to deal with this, we fit a negative binomial model, a specific type of Poisson regression that accounts for the problem of overdispersion (Allison & Waterman 2002). Positive coefficients for a categorical independent variable would mean the expected log count of the number of antenatal visits is higher for a category in question, compared with the reference category. A negative coefficient, on the other hand, means the expected log count of the number of antenatal visits is lower, compared with the reference category. Logit models were also used to examine the timing of first antenatal visit, which was divided into early antenatal visit, mostly within the first 3 month of gestation (0–3 months), and visits beyond the first trimester (4 months and beyond). Both Poisson and logit models are built under the assumption of independence of subjects, but the GDHS has a hierarchical structure with respondents nested within survey clusters, which could potentially bias the standard errors. STATA 12 (StataCorp, College Station, TX, USA). SE, which provides an outlet for handling this problem, is used by imposing on our models a ‘cluster’ variable, usually the identification numbers of respondents at the cluster level. This in turn adjusts the standard errors producing statistically robust parameter estimates (Cleves et al. 2004; Tenkorang & Owusu 2010).

Results

The average age was estimated as 29 years for this sample of women, that is, those who gave birth between 2005 and 2008. The majority of women were married, lived in the rural areas, were employed and identified as Christians (Table 1). Respondents are evenly spread across various wealth quintiles, and quite a substantial proportion reported having secondary/higher education. Results indicate that Ghanaian women average six visits for ANC and that the majority of Ghanaian women (57%) make such visits in the first trimester. While 41.6% of the selected sample indicated that they had enrolled in the NHIS, 56.8% had not. The mean score for barriers to access indicates that women consider such barriers as rather problematic.

Table 1. Univariate analysis of selected dependent and independent variables among women in Ghana, for births between 2005–2008
Variables% or Mean (N = 1610)
Average number of antenatal visits5.7
Timing of antenatal visits
0–3 months56.8
4 months and beyond43.2
Health Insurance (NHIS) 
No58.4
Yes41.6
Wealth quintile
Poorest29.9
Poorer22.7
Middle17.4
Richer17.8
Richest12.2
Education
No education35.2
Primary education24.7
Secondary/Higher education40.1
Employment
Unemployed13.3
Employed86.7
Marital status
Currently married89.7
Never married5.8
Formerly married4.5
Mean age of respondents29.2
Religious affiliation
Christians67.8
Muslims20.5
Traditionalists6.8
No religion4.9
Ethnicity
Akans39.1
Ga Dangbe4.5
Ewe13.1
Northern40.0
Other3.3
Rural–urban residence
Urban34.2
Rural65.8
Region of residence
Greater Accra9.0
Central8.2
Western9.3
Volta8.3
Eastern8.8
Ashanti14.7
Brong Ahafo9.3
Northern15.2
Upper East6.8
Upper West10.6
Means score for barriers to access0.0357

In Table 2, we present the zero-order negative binomial and logit coefficients for dependent and independent variables. Results indicate that the expected log count of the number of ANC visits (b = 0.231) is higher for women enrolled in NHIS than for those not enrolled. Although marginally significant, we find that compared with those not enrolled, women enrolled on NHIS are more likely to make their first antenatal visit earlier (b = 0.189), in the first trimester than beyond. The expected log count of the number of ANC visits is higher for educated and wealthy women than for poorer and uneducated women. Also, educated, wealthy and employed women have a higher likelihood of making their first ANC visit in the first trimester of pregnancy than beyond. Compared with Christians, Muslims and Traditionalists not only made fewer ANC visits, but were also less likely to make such visits in the first trimester. Rural dwellers showed a reduced log count of the number of visits for ANC (b = −0.333) and are less likely to make such visits in the first trimester (b = −0.246). A regional comparison showed that except for women in the Greater Accra Region, those in the other 9 regions had lower/reduced log count of the number of visits for ANC. Compared with those in the Greater Accra Region, women in the Eastern (b = −0.605) and Northern (b = −1.48) regions were less likely to make such visits in the first trimester than beyond this period. Respondents who indicated barriers to access as problematic made fewer antenatal visits and were less likely to have made their first ANC visit in the first trimester.

Table 2. Bivariate analysis of selected dependent and independent variables among women in Ghana, for births between 2005–2008
Independent variables NHISZero-order Negative Binomial and Logit coefficients
Number of ANC visitsTiming of ANC Visits
  1. *P < 0.1; **P < 0.05; ***P < 0.01.

  2. Coefficients are adjusted for clustering and robust standard errors are presented in brackets.

No00
Yes0.231 (0.031)***0.189 (0.107)*
Wealth quintile
Poorest (ref)00
Poorer0.130 (0.046)***0.287 (0.157)*
Middle0.228 (0.045)***0.323 (0.166)**
Richer0.420 (0.044)***0.533 (0.165)***
Richest0.590 (0.043)***1.28 (0.205)***
Education
No Education00
Primary Education0.077 (0.040)**0.246 (0.145)*
Secondary/Higher education0.309 (0.037)***0.451 (0.137)***
Employment
Unemployed00
Employed−0.019 (0.044)0.369 (0.152)***
Marital Status
Currently married00
Never married−0.034 (0.059)−0.155 (0.231)
Formerly married0.004 (0.083)−0.170 (0.253)
Age of Respondents0.004 (0.002)**−0.005 (0.007)
Religious affiliation
Christians00
Muslims−0.035 (0.038)−0.408 (0.161)***
Traditionalists−0.496 (0.082)***−0.988 (0.261)***
No religion−0.338 (0.081)***−0.229 (0.249)
Ethnicity
Akans00
Ga Dangbe−0.105 (0.070)−0.379 (0.251)
Ewe−0.136 (0.046)***−0.034 (0.168)
Northern−0.195 (0.039)***−0.263 (0.140)*
Other−0.125 (0.087)−0.084 (0.302)
Rural–urban residence
Urban00
Rural−0.333 (0.035)***−0.246 (0.126)**
Region of residence
Greater Accra00
Central−0.266 (0.076)***−0.095 (0.262)
Western−0.190 (0.080)***−0.156 (0.254)
Volta−0.377 (0.071)***−0.288 (0.227)
Eastern−0.332 (0.069)***−0.605 (0.269)**
Ashanti−0.166 (0.064)***−0.244 (0.222)
Brong Ahafo−0.175 (0.085)**−0.417 (0.247)*
Northern−0.473 (0.066)***−1.48 (0.226)***
Upper East−0.306 (0.074)***−0.239 (0.255)
Upper West−0.280 (0.057)***−0.018 (0.258)
Barriers to access−0.106 (0.016)***−0.123 (0.063)***

In Table 3, we present four multivariate models: two each for both ‘number of visits for ANC’ and ‘timing of ANC’. Model 1 for each outcome variable has our focal independent variable, ‘enrolment on NHIS’ controlling for socio-economic predictors, and in the final model (model 2), demographic and other socio-cultural variables are included. Table 3 indicates that even after controlling for both socio-economic and demographic variables, enrolling on NHIS was statistically associated with the number of times Ghanaian women attend ANC care. Specifically, women enrolled in NHIS have a higher log count of the number of antenatal visits, compared with those not enrolled on the scheme (b = 0.095). Regarding the timing of antenatal visits, however, we did not find any statistically significant difference between enrolled and non-enrolled women. Compared with poorer women, wealthier women have a higher log count of the number of antenatal visits (b = 0.217 and b = 0.324 for the richer and richest quintiles, respectively), and are more likely to make such visits earlier, in the first trimester of pregnancy (b = 0.595 for the richer and b = 1.43 for the richest quintiles). While employment status had no statistically significant association with the number of times women attended ANC, it was associated with the timing of antenatal visits. Employed women were more likely to attend ANC in the first trimester of pregnancy than the unemployed (b = 0.432). Compared with Christians, Traditionalists and respondents without any religious affiliation had lower log counts for the number of antenatal visits (b = −0.226 and b = −0.213, respectively). Similar results were observed for ‘other’ ethnic groups (b = −0.159) compared with Akans. Consistent with bivariate results, the multivariate findings indicated that rural dwellers have lower log counts of the number of antenatal visits than urban dwellers (b = −0.092). Women who reported barriers to access as problematic had lower log counts for the number of antenatal visits (b = −0.040). The relationship between barriers to access and timing of antenatal care was spurious, as these effects were mediated by socio-economic variables such as education and wealth.

Table 3. Multivariate analyses of ‘number of times attended antenatal care (ANC)’ and ‘timing of ANC’ among women in Ghana, for births between 2005-2008
Independent variablesCoefficients for Negative Binomial modelsCoefficients for Logit models
Model 1Model 2Model 1Model 2
  1. *P < 0.1; **P < 0.05; ***P < 0.01.

  2. Coefficients are adjusted for clustering, and robust standard errors are presented in brackets.

NHIS
No0000
Yes0.125 (0.029)***0.095 (0.029)***0.010 (0.110)−0.050 (0.119)
Wealth quintile
Poorest (ref)0000
Poorer0.095 (0.044)**0.044 (0.041)0.279 (0.160)*0.185 (0.176)
Middle0.169 (0.044)***0.085 (0.046)*0.317 (0.171)*0.369 (0.202)*
Richer0.338 (0.046)***0.217 (0.050)***0.521 (0.176)***0.595 (0.218)***
Richest0.484 (0.048)***0.324 (0.055)***1.27 (0.225)***1.43 (0.292)***
Education
No Education0000
Primary Education−0.004 (0.037)−0.025 (0.038)0.107 (0.150)−0.118 (0.167)
Secondary/Higher education0.092 (0.034)***0.067 (0.040)*0.097 (0.147)−0.095 (0.174)
Employment
Unemployed0000
Employed0.025 (0.038)0.027 (0.039)0.467 (0.155)***0.432 (0.169)***
Marital status
Never married 0 0
Currently married −0.007 (0.057) −0.192 (0.244)
Formerly married 0.012 (0.073) −0.139 (0.269)
Age of respondents 0.004 (0.002) −0.010 (0.008)
Religious affiliation
Christians 0 0
Muslims 0.062 (0.042) −0.188 (0.188)
Traditionalists −0.226 (0.081)*** −0.409 (0.264)
No religion −0.213 (0.079)*** −0.126 (0.259)
Ethnicity
Akans 0 0
Ga Dangbe −0.066 (0.059) −0.380 (0.320)
Ewe −0.022 (0.049) −0.038 (0.211)
Northern −0.087 (0.051) 0.421 (0.215)
Other −0.159 (0.077)** 0.222 (0.340)
Rural–urban residence
Urban 0 0
Rural −0.092 (0.038)*** 0.275 (0.162)*
Region of residence
Greater Accra 0 0
Central −0.066 (0.075) 0.119 (0.307)
Western −0.033 (0.066) 0.075 (0.296)
Volta −0.115 (0.075) 0.060 (0.294)
Eastern −0.148 (0.069)** −0.228 (0.303)
Ashanti −0.034 (0.061) −0.051 (0.271)
Brong Ahafo 0.030 (0.078) −0.133 (0.315)
Northern −0.090 (0.073) −1.21 (0.331)***
Upper East 0.040 (0.082) −0.242 (0.389)
Upper West 0.069 (0.078) 0.170 (0.357)
Barriers to access −0.040 (0.015)*** −0.047 (0.056)
Log Pseudo-likelihood−3873.193−3823.445−1016.936−977.041
Model signficance (Wald chi-square)259.74 (8)***349.44 (29)***49.80 (8)***145.96 (29)***
Sample size1610161016101610

Discussion

The most important result to emerge from this study is that regardless of socio-economic and demographic factors, women enrolled in the NHIS make more ANC visits compared with those not enrolled. The implication of this finding is that enrolment in the NHIS can help reduce inequalities in access that have previously influenced ANC usage (Addai 2000; Arthur 2012; Finlayson & Downe 2013). The positive relationship between NHIS enrolment and frequency of ANC attendance is consistent with evidence from Mensah et al. (2010) and Dzakpasu et al. (2012), who argued that the NHIS increased maternal care usage for some regions in Ghana.

The underlying assumption behind the various iterations of Ghana's maternal healthcare policy is that by removing the financial barriers to maternal care, more women will partake. While there are a multitude of factors in developing countries that affect the utilisation of maternal health care (Addai 2000), and ANC in particular (Simkhada et al. 2008), direct financial barriers are often found to be a prime reason why women do not access care. Our findings support the theory that removing direct financial barriers (in this case, through the NHIS) may increase the use of maternal health care in Ghana. Beyond just maternal care issues, this result also suggests that the NHIS may indeed work as a pro-poor equalizing agent in health financing by increasing access to health care.

It is interesting, however, that despite the association between the NHIS and the total number of ANC visits, our study also found that the NHIS did not have a significant influence on the timing of the first ANC visit. Although 57% of Ghanaian women received ANC within the first 3 months of pregnancy as evinced in Table 1, it was those with high socio-economic characteristics, including the wealthy and employed, who benefited. The finding, however, corroborates the works of Arthur (2012) and Doku et al. (2012) who also argued that women of higher SES are more likely to seek ANC earlier in their pregnancies.

We propose two different explanations as to why the NHIS could be associated with increased likelihood of the number of ANC visits and yet does not influence the timing of the first ANC visit: first that the NHIS can only impact the timing of ANC if women are aware they are pregnant, and many women in SSA do not even realise they are pregnant until the second trimester, making early ANC impossible (Myer & Harrison 2003; Gross et al. 2012; Pell et al. 2013).

Second, and perhaps most importantly, we already know that the poor are less likely to be enrolled in the NHIS (Dixon et al. 2011), and thus, women of higher socio-economic status are more likely to already be enrolled when they discover they are pregnant, making early ANC more practical. Although Ghana's new maternal health policy exempts pregnant women from premium payments, there are multiple points in the process where the ‘free’ enrolment may be deterred or delayed by other obstacles (Pell et al. 2013). For instance, women must provide a laboratory-certified pregnancy test to the NHIS agent, who can then enrol them in the NHIS for free. By the time a woman has realised she is pregnant, travels to a clinic, completes a pregnancy test, travels to the NHIS office to initiate and complete paper work with the agent which sometimes involves multiple visits because NHIS officers may not be present, and then finally receives her temporary enrolment card, she may very well be beyond her first trimester. Thus, the lack of statistical relationship between the timing of the first ANC visit and NHIS enrolment may be an artefact of the already established inequalities between wealth and enrolment, which are delaying the poor in getting ANC.

Our study also revealed that even when controlling for NHIS enrolment, women in rural areas were less likely to make first trimester ANC visits, and generally made fewer total ANC visits, suggesting that availability of health facilities (which are much fewer and harder to travel to in rural areas) remains a factor in ANC access. This finding corroborates previous work in Ghana that women in rural locations were far less likely to access ANC (Overbosch et al. 2004). Despite recent improvements, there are still marked differences between availability of health services in rural and urban areas (Ghana Health Service 2008), and while the NHIS could help to address some financial barriers to care, it does not account for other factors such as cost of transportation, physical difficulty for pregnant women to travel or the availability of health professionals and NHIS agents. The lack of services in rural areas is especially problematic considering that these rural areas have both higher fertility rates and higher proportions of the poor (GSS et al 2009).

Unsurprisingly, women identifying barriers to accessing health facilities as problematic made fewer ANC visits. As with the rural–urban dichotomy, the NHIS is only set up to address a very specific financial barrier and does not address pre-existing socio-cultural or transport barriers. However, unlike previous work that found a woman's age to be a predictor of ANC utilisation (Addai 2000; Doku et al. 2012; Owoo & Lambon-Quayefio 2013), our study found no relationship. This may be attributed to the spurious impact of the NHIS and socio-economic variables. Further research elucidating the role of women's autonomy and healthcare usage with the NHIS will prove useful.

A limitation to this study is that the data collection for the GDHS took place shortly after the implementation of the NHIS maternal exemption policy was created. Thus, caution should be taken when interpreting results and conclusions for their application to the present day. It will be interesting to see whether data from the next round of the GDHS show any improvement in the NHIS' influence on maternal healthcare usage for Ghanaian women, in particular, the timing of ANC, after the policy has had a few years to properly establish itself in the public's awareness; things may have shifted in terms of ANC attendance and timing since 2008. Also, as the data used are largely cross-sectional, we refrain from making ‘causal’ connections between independent variables and outcome variables.

Nevertheless, this study adds to the fledgling literature drawing linkages between the NHIS and the utilisation of health care. As it stands now, while it appears the NHIS may be connected to the number of total ANC visits, the influence of socio-economic and demographic elements on both the number of visits and the timing of visits suggests that there are still serious inequalities existent in Ghana's maternal care. For the NHIS to reach its potential as a social equalizer, these problems will need to be addressed. The NHIS should be strengthened and resourced as it may act as an important tool for increasing ANC attendance among women in Ghana. Future research using longitudinal data with proper time order are required to clearly disentangle the causal connections between NHIS enrolment and ANC attendance. A national-level study investigating the influence of the NHIS on other types of health outcomes will also be helpful in understanding the full impact of the NHIS and improving Ghana's health policies.

Ancillary