The relationship between antenatal provider type and maternal care in rural Ghana: a cross-sectional study


Corresponding Author Sarah J. Atunah-Jay, Warren Alpert School of Medicine of Brown University, University of Ghana/Brown University Academic Partnership to Address HIV/AIDS, The Miriam Hospital, 164 Summit Avenue, Providence, RI 02906, USA. Tel.: +233264661743; Email



To investigate the influence of antenatal provider type on maternity care in rural Ghana.


An analysis of maternal care by antenatal provider type using the 2008 Ghana Demographic and Health Survey. Study population included rural Ghanaian women aged 15–49 years with report of a live birth between 2003 and 2008. Bivariate chi-square analysis was performed to examine differences in maternal report of WHO Maternal Health Interventions. Multivariate linear and logistic regression were performed to assess differences in antenatal care (ANC) scales and maternal care packages.


Thousand and three hundred and sixty-seven rural women reported a live birth. Provider distribution was: doctor, 15.6%; midwife, 70%; community health officer (CHO), 9.1%; no provider, 5.3%. Women from lower socio-demographic categories were more likely to report midwife or CHO. Report of CHO vs. no provider was positively associated with maternal services (P < 0.01). Report of doctor or midwife vs. CHO was significantly associated with maternal services (P < 0.01).


There is a positive association between antenatal provider length of training and maternal specialization and report of maternal services. Community-based providers are associated with markedly increased report of maternal services compared with no provider. Structural factors appear to underlie some differences in service provision.



Etudier l'influence du type de prestataire de soins prénataux sur les soins maternels dans les régions rurales du Ghana.


Analyse des soins maternels selon le type de prestataire de soins prénataux en utilisant l'Enquête Démographique et de Santé du Ghana de 2008. La population d’étude comprenait des femmes de zones rurales ghanéennes âgées de 15 à 49 ans, ayant un report d'une naissance vivante entre 2003–2008. Une analyse bivariée de chi-carré a été réalisée pour examiner les différences dans les reports maternels des interventions de santé maternelle de l’OMS. La régression linéaire et logistique multivariée a été effectuée pour évaluer les différences dans les échelles de soins prénataux et dans les ensembles de soins maternels.


1367 femmes rurales ont déclaré une naissance vivante. La distribution des prestataires était comme suit: médecins, 15,6%; sages-femmes, 70%; agent de santé communautaire (ASC), 9,1%; non prestataires, 5,3%. Les femmes issues des catégories sociodémographiques plus faibles étaient plus susceptibles de déclarer des prestataires de niveau moyen. Le report d’ASC versus celui des non prestataires était positivement associé à des services de maternité (P < 0,01). Le report de médecin ou de sage-femme versus celui d’ASC était significativement associé à des services de maternité (P < 0,01).


Il existe une association positive entre la durée de formation du prestataire de soins prénataux et la spécialisation maternelle avec le report de services maternels. Les prestataires communautaires sont associés avec un report nettement accru de services de santé maternelle comparés au non prestataires. Des facteurs structurels semblent être à l'origine des différences dans la prestation de services.



Investigar la influencia del tipo de proveedor prenatal sobre los cuidados de maternidad en zonas rurales de Ghana.


Análisis de los cuidados maternos según el proveedor prenatal, utilizando datos del 2008 del Censo Demográfico y Sanitario de Ghana. La población de estudio incluyó a mujeres de zonas rurales de Ghana con edades entre los 15–49 años que reportaron haber tenido un parto de feto vivo entre el 2003–2008. Se realizó un análisis bivariado de chi-cuadrado para examinar las diferencias en el informe de la OMS sobre intervenciones en salud materna. Se realizaron regresiones multivariadas y logísticas para evaluar las diferencias en las escalas de cuidados prenatales (CPN) y los paquetes de cuidados maternos.


1,367 mujeres rurales reportaron un parto de feto vivo. La distribución de los proveedores fue la siguiente: Médico, 15.6%; Partera, 70%; Oficial Sanitario Asistencial (OSA), 9.1%; Sin proveedor, 5.3%. Las mujeres de categorías sociodemográficas inferiores tenían una mayor probabilidad de reportar un proveedor de nivel medio. El reportar un CPN recibido de un OSA versus no el tener un proveedor estaba asociado de forma positiva con los servicios maternos (P < 0.01). El reportar un CPN recibido de un médico o partera versus de un OSA estaba significativamente asociado con los servicios maternos (P < 0.01).


Existe una asociación positiva entre el tiempo de formación del proveedor prenatal y la especialización en maternidad y el reportar servicios maternos. Los proveedores basados en la comunidad están asociados con un aumento marcado en el reporte de servicios maternos comparado con una falta de proveedor. Los factores estructurales parecen estar asociados con algunas diferencias en la provisión del servicio. Los proveedores de nivel medio y basados en la comunidad juegan un papel crítico a la hora de proveer servicios maternos a mujeres de zonas rurales.


In 2006, the World Health Organization (WHO) alerted us of a global shortfall of more than four million health workers (2006). Low-income countries disproportionately shoulder the burden of this shortage. One in every ten people in the world lives on the African continent; yet, Africans suffer one-quarter of global disease burden, have fewer than 3% of all health workers and hold only 1% of global financial health resources. The discrepancy in the distribution of health workers is even larger between rural and urban areas. Half of the global population lives in rural areas, while most health workers live and work in cities (WHO 2010). Innovative programming to train health workers, improve health systems management and expanded multilateral agreements to stem health worker emigration are considered key strategies to increasing global workforce capacity and healthcare availability (Hongoro & McPake 2004; WHO 2008; Gupta et al. 2011).

In Ghana, a West African country that faces a critical health worker shortage, pregnancy and childbirth related mortality are the second leading cause of death in women (GSS et al. 2009a). In children under five years of age, more than one-third of deaths occur in the neonatal period; and infant and under-five mortality is nearly 30% higher in rural vs. urban areas (GHS 2008). Antenatal care (ANC) visits are important to identify and treat prenatal complications and are considered a bridge to skilled delivery care, caesarean section and skilled neonatal resuscitation, services key to achieving Millennium Development Goals Four and Five (Abou-Zahr & Wardlaw 2003; WHO 2009; Bhutta et al. 2010). While most women (95%) in Ghana report some ANC from a health professional, there is evidence that provider type and maternal care differ by maternal demographic profile (GSS et al. 2009b). Notably, compared with their urban counterparts, women in rural areas are less likely to report ANC from doctors (16.4% vs. 33.9%) and more likely to report ANC from midwives (69.6% vs. 62.1%), community health officers (7.8% vs. 1.8%), untrained traditional birth attendants (0.1% vs. 0.0%), other providers (1.0% vs. 0.3%) or no providers (4.8% vs. 1.4%).

Midwives in Ghana receive two to three years of post-secondary training and provide antenatal, delivery, newborn and post-partum care, as well as a variety of other maternal and child health services (Prosser et al. 2006; GSS et al. 2009a,b). Ghanaian doctors receive six years of post-secondary training, with options for extended subspecialty training. They provide maternal and newborn services, including surgical services such as caesarean sections and management of high-risk pregnancies. Doctors and midwives are typically facility-based, such as in hospitals, clinics and private maternity homes (GSS et al. 2003). In 1999, Ghana introduced the Community-Based Health Planning Services (CHPS) initiative as a means to increase availability of primary care to geographically remote areas (Nyonator et al. 2005). The programme moved nurses from regional centres into communities that were engaged in building and supporting nurse-staffed health compounds. The nurses, called community health officers (CHO), receive approximately two and a half years of post-secondary training and provide services such as immunizations, family planning, delivery supervision, antenatal and postnatal care, treatment for minor ailments and health education (Acquah et al. 2006). CHPS compounds normally offer ANC services only when a midwife is affiliated with the compound. In 2003, Ghana adopted CHPS as a national programme, but its rollout has been slower than anticipated, reaching only 33% of eligible zones by 2008 (Binka et al. 2009; GMOH 2010).

Prioritizing and strengthening health workforce capacity-building efforts hinges on knowing the location, activities and abilities of different health workers (Dovlo 2004; Mullan & Frehywot 2007; Fulton et al. 2011; Gupta et al. 2011). Data on the quality of maternal care in Ghana are limited. A 2002 national health delivery assessment provided information on facility capacity to provide maternal health services; and the Ghana Demographic and Health Survey (GDHS) provides a cross-sectional snapshot of maternal health services every five years (GSS et al. 2003, 2009b). To date, there is a paucity of data on the relationship between maternal services and health provider type in Ghana.

In this paper, we investigate the influence of antenatal provider type on maternity care in rural Ghana by identifying characteristics of rural women associated with each provider type and by determining report of maternal services by provider type. We hypothesized that women from lower socio-demographic backgrounds would be positively associated with midwives and CHOs and that report of maternal care would be positively associated with increased length of training and maternal specialization of reported provider.


We performed a secondary data analysis of antenatal provider type and maternal care in rural Ghana using cross-sectional data from the 2008 household-based GDHS (GSS et al. 2009b). The GDHS utilizes a representative probability sample of more than 12 000 households selected nationwide to allow for separate estimates of key demographic (including urban vs. rural) and health indicators for each of the 10 regions in Ghana. The two-stage sample design first selects clusters based on 2000 Ghana Population and Housing Census data and then systematically samples 30 households in each cluster. The second stage ensures adequate completed individual interviews to provide acceptable estimates for key indicators and under-five causes of death.

The following three questionnaires were used for the 2008 GDHS the Household Questionnaire, the Women's Questionnaire and the Men's Questionnaire. Each household selected for the GDHS was eligible for interview with the Household Questionnaire. In half of the survey, households all women age 15–49 and all men age 15–59 were eligible to be interviewed if they were either usual residents of the household or visitors present in the household on the night before the survey. The GDHS questionnaires were based on those used for the 2003 GDHS with additional input from the Ghana Ministry of Health (GMOH) and local and international stakeholders. The questionnaires were translated from English into three local languages and pre-tested before finalization. Concurrent data processing occurred during data collection, allowing for feedback to field teams of problems detected during data entry. The University of Minnesota Institutional Review Board approved this secondary analysis.

Study sample

The study sample was restricted to the 1 367 rural women aged 15–49 who lived in a rural area and reported a live birth in the five years preceding the 2008 GDHS. Restriction to the rural population was based on documented differences in health services between rural and urban locales, such as geographic access to and availability of services as well as relative affordability and acceptability of care, as it was believed these factors would skew results against rural providers (Agyepong 1999; GSS et al. 2009b, 2003). Further, the sequestering of CHPS zones to primarily rural locales created a more relevant context in which to compare doctors, midwives and CHOs (GSS et al. 2009b).


The dependent variables were 18 GDHS antenatal, delivery and post-partum measures endorsed by the WHO (2009) (Table 1). General ANC and skilled delivery/post-partum care interventions were asked of the entire study population. Provider-based ANC interventions were asked only of women who reported having a skilled antenatal provider. The GDHS definition of skilled provider was used: doctor, nurse, midwife, auxiliary midwife and CHO (GSS et al. 2009b).

Table 1. Dependent variables
  1. a

    HIV-related questions were asked only of women who reported a live birth in the three years preceding the 2008 GDHS.

General antenatal care
Iron supplementAnswered yes to: ‘During this pregnancy, were you given or did you buy any iron tablets or iron syrup?’
Antiparasitic medicationAnswered yes to: ‘During this pregnancy, did you take any drug for intestinal worms?’
≥2 Tetanus injectionsReported at least two injections in response when asked: ‘During this pregnancy, how many times did you get this tetanus injection?’
Antimalarial medicationAnswered yes to: ‘During this pregnancy, did you take any drugs to keep you from getting malaria?’
SP/Fansidar/MalafanAnswered SP/Fansidar/Malafan to: ‘What [antimalarial] drugs did you take?’ SP/Fansidar/Malafan is an acronym for sulphadoxine-pyrimethamine (SP brand names are Fansidar and Malafan). This was the antimalarial medication recommended by the Ghana Ministry of Health and Ghana National Malaria Control Programme at the time of the 2008 GDHS
Provider-based antenatal care
≥4 antenatal visitsReported at least 4 visits when asked: ‘How many times did you receive antenatal care during this pregnancy?’
Pregnancy complication counselingAnswered yes to: ‘During (any of) your antenatal care visit(s), were you told about the signs of pregnancy complications?’
WeighedAnswered yes to: ‘As part of your antenatal care during this pregnancy, were you weighed at least once?’
Blood pressure measuredAnswered yes to: ‘As part of your antenatal care during this pregnancy, was your blood pressure measured at least once?’
Urine sampleAnswered yes to: ‘As part of your antenatal care during this pregnancy, did you give a urine sample at least once?’
Blood sampleAnswered yes to: ‘As part of your antenatal care during this pregnancy, did you give a blood sample at least once?’
Intermittent preventive therapyReported at least 2 doses when asked: ‘How many times did you take (SP/Fansidar/Malafan) during this pregnancy?’ The Ghana Ministry of Health and Ghana National Malaria Control Programme recommended ≥2 doses of SP/Fansidar/Malafan for maternal antimalarial prophylaxis at the time of the 2008 GDHS
HIV counselingAnswered yes to at least one of the following questions: ‘During any of the antenatal visits for your last birth, did anyone talk to you about: Babies getting the AIDS virus from their mother? Things that you can do to prevent getting the AIDS virus? Getting tested for the AIDS virus?’a
Offered HIV testAnswered yes to: ‘Were you offered a test for the AIDS virus as part of your antenatal care?’a
Tested for HIVAnswered yes to: ‘I don't want to know the results, but were you tested for the AIDS virus as part of your antenatal care?’a
Received results from HIV testAnswered yes to: ‘I don't want to know the results, but did you get the results of the test?’a
Skilled delivery/post-partum care
Skilled delivery assistanceReported a doctor, nurse/midwife, auxiliary midwife or CHO/nurse when asked: ‘Who assisted with the delivery of (NAME)?’
Post-partum checkReported at least one post-partum check within the first 6 weeks post-partum

The independent variables included four ANC provider categories: doctor, midwife, CHO and no provider. Nurse/midwife and auxiliary midwife were grouped into one variable based on previous literature (Anand & Barnighausen 2004; GSS et al. 2009a,b). If more than one source of ANC was reported, one provider was assigned based on longest training and maternal care specialization (i.e. if doctor and midwife were reported, analysis was by doctor; if midwife and CHO were reported, analysis was by midwife). Traditional birth attendants (trained or untrained) and ‘other’ providers were excluded from analysis due to the small sample size and the heterogeneity of the group.

For purpose of this study, we developed two ANC service content scales and two maternal care packages (Table 2). The four-point general ANC scale includes all of the general ANC-dependant variables (with half-weight given to ‘ever taking antimalarial during pregnancy’ and half-weight given to ‘taking correct antimalarial’). This scale was used to examine report of general ANC interventions by CHO vs. no provider.

Table 2. Scale & package items
General antenatal care scale itemsIron supplement, antiparasitic medication, >2 tetanus injections, antimalarial medication (0.5 point), SP/Fansidar/Malafan (0.5 point)
Skilled antenatal care scale itemsIron supplement, antiparasitic medication, >2 tetanus injections, antimalarial medication (0.5 point), SP/Fansidar/Malafan (0.5 point), pregnancy complications counseling, weighed, blood pressure measured, urine sample, blood sample, HIV counseling, offered HIV test
Skilled delivery care & Post-partum check package itemsSkilled delivery assistance, post-partum check
Comprehensive maternal care package items>4 antenatal visits, skilled delivery assistance, post-partum check

The 11-point skilled ANC scale was used to examine general ANC interventions and a representative sample of the provider-based ANC interventions by reported antenatal provider type. This scale adds seven of the provider-based ANC variables to the four-point general ANC scale. Provider-based ANC variables were excluded for the following reasons: ‘≥4 antenatal visits’ is not service content; use of general ANC antimalarial variables in lieu of ‘intermittent preventive therapy (IPT)’ improves cross-provider comparisons (i.e. CHOs not working with midwives are not mandated to provide IPT); and using only two of the four HIV care variables reduces potential for overweighting of HIV services.

The skilled delivery and post-partum check package was used to examine skilled delivery care and post-partum checks by report of CHO vs. report of no provider. The comprehensive maternal care package examined report of at least four antenatal care visits, skilled delivery check and post-partum check by skilled antenatal provider type. The three comprehensive package interventions were based on WHO and Ghana Health Service groupings (World Health Organization (WHO) 2009; GSS et al. 2009a,b). A woman was only considered to have received a package if she reported all interventions in the package.

We limited covariates to standard GDHS demographic characteristics and pregnancy-related factors likely to influence recall: region, education background, wealth quintile, age, insurance status, years since last childbirth, number of children and current pregnancy status.

Statistical analysis

Bivariate analyses using cross-tabulation with chi-square tests determined differences in covariates and interventions by provider type. Multivariate linear and logistic regression determined differences in scales and packages by provider type while controlling for covariates.

Use of unweighted analyses was based on DHS guidelines indicating that sample weights are inappropriate for estimating relationships in this data set (Rutstein & Rojas 2006). Weighted analysis (unpublished) confirmed this design and results were similar to those reported below. All analyses were performed using STATA 9.2 SE to control for the complex cluster sampling design.


Of the 4 916 women interviewed in the DHS, 1 367 met study criteria. Antenatal provider distribution was: doctor, 15.6%; midwife, 70%; CHO, 9.1%; and no provider, 5.3%.

Bivariate analysis

Midwives were reasonably equally distributed across the dichotomous covariates of education, wealth, parity and location of practice (Table 3). Compared with other categories, report of doctor was associated with increased report of some primary school education, being at or above the second wealth quintile, and three or fewer births. Report of CHO, compared with other categories, correlated with increased report of no education, being in the lowest wealth quintile, and four or more births. While most of the women in our study were uninsured (62.5%), there was a positive association between being insured and having an ANC provider. There was no difference between groups in maternal age, years since last childbirth and currently being pregnant.

Table 3. Demographic and pregnancy-related characteristics by antenatal provider type
 TotalDoctorMidwifeCHONo ProviderP-value
1367 %n = 213 %n = 957 %n = 125 %n = 72 %
  1. a

    Defined as region in which at least 15 women in sample reported community health officers (CHO). These regions were Northern, Upper East, Upper West and Volta.

  2. b

    Defined as region in which less than 15 women in sample reported CHOs. These regions were Ashanti, Brong-Ahafo, Central, Eastern, Greater Accra and Western.

  3. c

    N/DHIS indicates membership in Ghana's national or district health insurance scheme, adopted in 2003 (GSS et al. 2009b).

Current age
15–34 years old69.471.870.062.465.30.25
35–49 years old30.628.230.037.634.7
High CHO densitya50.538.547.591.255.6<0.01
Low CHO densityb49.561.552.58.844.4
>Primary school54.272.254.924.044.4
Wealth quintile
Lowest quintile44.726.844.174.454.2<0.01
2nd–5th quintile55.373.255.925.645.8
Currently insured
N/DHISc or Other37.547.437.133.919.4
1–3 children52.760.152.842.448.6<0.05
>4 children47.339.947.257.651.4
Years since last childbirth
Curently pregnant

Report of maternal health interventions varied (Table 4). Women who reported any ANC provider were three to 12 times as likely to report receiving general ANC services as those who did not report a provider. Among skilled providers, doctors and midwives vs. CHOs were associated with increased report of services for all items but report of being weighed, IPT and receipt of HIV test results, for which there was no significant difference by provider type. Report of skilled delivery assistance and post-partum check was highest for women who reported doctors as their ANC provider, followed by midwives, CHOs and no providers.

Table 4. Bivariate analysis of maternal report of WHO-recommended maternal health interventions by antenatal provider
 DoctorMidwifeCHONo ProviderP-value
n = 213(%)n = 957(%)n = 125(%)n = 72(%)
General antenatal care
Iron supplement181 (85.0)831 (87.7)96 (78.1)15 (20.8)<0.01
Antiparasitic medication89 (43.7)374 (41.2)26 (23.6)3 (4.2)<0.01
≥2 tetanus injections114 (53.8)539 (57.0)61 (51.3)5 (6.9)<0.01
Antimalarial medication151 (73.3)561 (60.5)58 (50.4)3 (4.2)<0.01
SP/Fansidar/Malafan132 (63.8)486 (52.4)50 (43.5)0 (0.0)<0.01
Provider-based antenatal care
≥4 antenatal visits169 (80.5)728 (78.0)71 (61.7)<0.01
Pregnancy complication counseling152 (71.4)620 (65.3)58 (47.2)<0.01
Weighed210 (98.6)930 (97.6)120 (96.0)0.27
Blood pressure measured211 (99.1)923 (96.6)116 (92.8)<0.05
Urine sample197 (92.5)813 (85.0)81 (64.8)<0.01
Blood sample199 (93.4)820 (85.8)85 (68.0) <0.01
Intermittent Preventive Therapy108 (82.4)385 (79.7)38 (76.0) 0.65
HIV counseling138 (80.2)541 (67.3)67 (64.4) <0.01
Offered HIV test77 (45.0)265 (33.0)25 (24.0) <0.05
Tested for HIV68 (39.5)218 (27.1)27 (26.0) <0.05
Received results from HIV test53 (81.5)159 (75.4)20 (76.9) 0.66
Delivery/Post-partum Care
Skilled Delivery Assistance119 (55.9)436 (45.6)29 (23.2)8 (11.1)<0.01
Post-partum check166 (77.9)665 (69.6)64 (51.2)32 (44.4)<0.01

Multivariate analysis

Multivariate linear and logistic regression analyses of scales and packages revealed that report of CHO vs. no provider was associated with statistically significant increased report of ANC services and almost 6:1 odds of reporting both skilled delivery care and post-partum check (P < 0.01) (Table 5). Report of doctor or midwife vs. CHO was associated with statistically significant increased report of ANC services and nearly 3:1 odds of reporting the comprehensive maternal care package (P < 0.01).

Table 5. Multivariate regressiona analysis of antenatal care scales and maternal care packages
 General antenatal care scale (0–4)bSkilled delivery care & Post-partum check packageSkilled antenatal care scale (0–11)cComprehensive maternal care package
Coefficient (95% CI)Adjusted coefficient (95% CI)OR (95% CI)Adjusted OR (95% CI)Coefficient (95% CI)Adjusted coefficient (95% CI)OR (95% CI)Adjusted OR (95% CI)
  1. The reference group for doctors and midwives is community health officers (CHOs). The reference group for CHOs is no provider.

  2. a

    Multivariate regression includes maternal education, wealth quintile, region, insured status, parity and age of youngest child.

  3. b

    R2 = 0.47.

  4. c

    R2 = 0.13.

  5. d

    < 0.01.

Doctor1.71d (1.08–2.35)1.07d (0.47–1.67)5.92d (3.12–11.6)2.8d (1.37–5.72)
Midwife1.20d (0.62–1.77)0.79d (0.24–1.33)3.75d (2.01–6.98)2.5d (1.33–4.79)
CHO1.56d (1.29–1.82)1.72d (1.45–1.99)3.03d (1.02–9.05)5 .97d (1.5–23.2)


Consistent with prior studies documenting the influence of maternal demographics on antenatal provider type in Ghana, in this sample of rural women we found the covariates of maternal wealth and education to be positively associated with longer training and increased maternal specialization of antenatal provider (GSS et al. 2009b; Arthur 2012). Wealth and education are well-established global predictors of access to health services. In Ghana's most rural regions, the critical workforce shortage is manifest in a doctor to population ratio ranging from 1:30 000 to 1:90 000 persons (GHS 2008). This shortage is a likely driver of women from higher socio-demographic profiles having relatively better access to doctors than women from lower socio-demographic strata. There was no difference within the dichotomous covariates of maternal wealth and education and maternal report of midwives as antenatal provider, suggesting that midwives are a relatively equitable workforce category across socio-demographic status. The negative association between increased wealth and education and maternal report of CHO antenatal providers supports other studies that have found this decade-old mid-level workforce to be successful in reaching the clientele for which their provider type was designed: poor, rural women (Awoonor-Williams et al. 2004; Binka et al. 2007). Finally, we saw a positive association between health insurance and report of skilled antenatal provider, which is consistent with Mensah et al.'s (2010) findings that Ghana's relatively nascent health insurance scheme appears to be an independent predictor of maternal services.

Maternal care in rural Ghana does vary by antenatal provider type. There is a positive association between antenatal provider length of training and maternal specialization and report of maternal services, taking covariate differences into account. Although less than one-fifth of women in this sample report doctors as antenatal provider, those that do, report more antenatal, skilled delivery and post-partum services than other antenatal providers. Accordingly, women who report midwives as antenatal provider report increased maternal services compared with CHOs. Our findings are consistent with our hypothesis and supported by studies in which quality of health services appears to be higher from doctors compared with other provider categories (Rowe et al. 2007; Brentlinger et al. 2010). The positive relationship between report of doctors and midwives and maternal services is no doubt driven by their professional mandate to provide the full breadth of services included in this analysis, including skilled delivery and post-partum check. CHOs, conversely, are only mandated to provide ANC when working in collaboration with a midwife and are not skilled delivery providers. Therefore, our finding that CHOs are associated with relatively less maternal services is anticipated and may be impacted by their restricted breadth of practice.

Community health officers were associated with markedly increased report of maternal services when compared with no provider. In many countries facing critical health worker shortages, the ‘no provider’ category remains the norm rather than the exception (Kruk et al. 2007; Fulton et al. 2011). While most women in Ghana report some antenatal care, those in rural areas are less likely than their urban counterparts to report an antenatal provider or maternal care services (GSS et al. 2009b). Our findings are supported by studies linking community-based providers in Ghana and other low- and middle-income countries to increased maternal services and reductions in child mortality (Awoonor-Williams et al. 2004; Binka et al. 2007; Lassi et al. 2010; Mukanga et al. 2011; Naariyong et al. 2012). Better accessibility of services and engagement of local stakeholders have been suggested as reasons for the effectiveness of community-based approaches to maternal and child services (Agyepong 1999; WHO 2010).

There was a lack of difference between the skilled provider categories in three maternal services: weight check, IPT and receipt of HIV test results. This lack of difference by provider type supports an emerging body of literature suggesting that mid-level providers can administer equivalent quality of care as doctors for some health services (Lin & Franco 2000; Laurant et al. 2005; Chilopora et al. 2007; Vasan et al. 2009). While weight measurement is a ubiquitous service in Ghanaian pregnancy care, this has not been found for administration of IPT and receipt of HIV test results (Birungi et al. 2006; GSS et al. 2003). The latter services are unique amongst the diagnostic and therapeutic interventions studied as they depend upon a longitudinal provider relationship: both require a minimum of two encounters and neither is dependent on availability of new services (i.e. report of IPT depends on a first dose of sulphadoxine–pyrimethamine being previously available; report of receiving HIV test results depends on previously receiving an HIV test). This relationship suggests that structural factors, such as lack of diagnostics and therapeutics and limitations on practice scope, may be more significant barriers to service provision than length or specialization of training. Structural barriers to care, such as availability, affordability, and accessibility to diagnostics and therapeutics, are fundamental to a health worker's ability to provide quality care (WHO 2010, Awoonor-Williams et al. 2004, Agyepong 1999). Another possible explanation for the lack of difference in maternal report of IPT and receipt of HIV test results may be acceptability of provider type. Other studies have found mid-level providers to be more skilled at patient communication than doctors (Laurant et al. 2005; Lassi et al. 2010).

Finally, we would be remiss not to emphasize that all provider categories were associated with relatively low report of many interventions, including taking recommended antimalarial medication, HIV testing, and skilled-delivery assistance. Gaps in provider services should be considered within the context of the complex administrative, professional and structural systems in which health workers operate. Further work is needed to identify and address barriers to these critical services.

Strengths of this study include that this is the first analysis of maternal services by healthcare provider type in Ghana and that the large sample size gives sufficient statistical power (Long 1997). There are several study limitations. It is cross-sectional; therefore, it is not possible to determine causality and potential mediators of the relationships found without unreasonable assumptions. Report of services received once does not fully reflect the potential quantity and quality of services. We were limited in our ability to adjust for all possible confounders (e.g. high-risk pregnancies are more likely to be followed by doctors) or to distinguish between health workers who are mandated to provide a service vs. those who are not (e.g. only CHOs working with a midwife are mandated to provide ANC). Similarly, the GDHS measures many components of ANC, and we were not able to account for all of them in our analysis (e.g. timing of ANC visits is not included). GDHS data are limited in that there may be reporting bias; provider practice location is not available; unmeasured providers and non-antenatal contacts may influence reported services; and health facility and/or health provider report of service is not measured.

Longitudinal studies would be useful to further investigate the relationship between health workers, services and patient outcomes; cost-effectiveness studies would be useful to inform resource prioritization.


Our findings are consistent with studies that link improved maternal and child health outcomes to increased density of health resources (Anand & Barnighausen 2004; Dovlo 2004; Kruk et al. 2009). It has been suggested that cost-effective care in low-income countries requires a strengthened mid-level workforce (Fulton et al. 2011). The positive relationship we report between lower socio-demographic women and midwives and CHOs reinforces the importance of efforts to strengthen the nurse and community-based workforce. We encourage stakeholders in Ghana to reconsider restrictions on CHO ability to provide critical ANC and skilled delivery services. Practitioners, academics and policy makers should continue to investigate means to optimize Ghana's health professional skill-mix in terms of training costs, practice scope and projected practice location and clientele. Finally, strategies to reduce structural barriers in rural health systems must be aggressively pursued.


Thanks to Linda Bearinger, Chandy John, Shalini Kulasingam, Margherita Ghiselli and Irene A. Agyepong for helpful comments during the development of this research and writing of this article. This work, done during SA's fellowship training at the University of Minnesota, was supported in part through funds from the Physician Faculty Development in Primary Care Clinician Research Fellowship – General Pediatrics grant, University of Minnesota.