Background
Although substantial progress has been made in reducing maternal mortality, the World Health Organization (WHO) estimated that in 2015, 303,000 women died from potentially avoidable problems in pregnancy or childbirth worldwide. Africa accounted for more than half of the global burden of maternal deaths, with women in the region having a 1 in 37 chance of dying in pregnancy/child birth compared to a 1 in 3400 chance in Europe – the largest difference between poor and rich countries on any health indicator. Nigeria, which constituted less than 1 % of the world’s population, accounted for 19 % of global maternal deaths and had an estimated maternal mortality ratio of 814 maternal deaths per 100,000 live births in 2015 [
1]. Uptake of maternity care is low in Nigeria, with only 36 % of births occurring in a health facility (HF) and 38 % being assisted by a skilled provider [
2].
Access to good quality obstetric care is critical for reducing maternal mortality. Globally, 15 % of pregnant women will experience obstetric complications, many of which are unpredictable, making the availability and quality of basic emergency obstetric and neonatal care (BEmONC) a public health imperative. While multiple factors may account for a high number of maternal deaths, empirical studies have suggested that poor quality services may lead to low coverage of maternal health (MH) care and non-effective and non-timely management of life-threatening complications of pregnancy/childbirth [
3‐
7]. Existing evidence indicates that MH services in many parts of sub-Saharan Africa are deficient in terms of providing basic emergency obstetric care (BEmOC) [
8,
9]. Some settings have been noted to lack essential drugs, reagents and equipment for MH care, high quality patient education, timely referral and transportation services, and an adequate number and mix of skilled providers [
10,
11].
Studies of supply-side factors affecting the utilization of MH care services in Nigeria have revealed wide gaps in HF and provider performance. The health care system has been found to be suboptimal in terms of the availability of human resources and commodities at service delivery points [
12]. However, evidence suggests that privately-owned facilities are better prepared to provide basic and comprehensive emergency obstetric care than public HFs [
13]. Available evidence showed that the level of provision of almost all signal functions for basic and comprehensive emergency obstetric care was below the UN-recommended level while the direct obstetric case fatality rate, which is reflective of the quality of care, exceeded the United Nations-recommended maximum [
14]. Of primary health care (PHC) facilities enrolled in the Midwives Service Scheme, a government program designed to address the national shortage of skilled birth attendants, 44 % did not provide all the components of ANC [
15]. At the sub-national level, the availability and quality of life-saving obstetric services have been demonstrated to be poorer in northern compared to southern states, and in rural than in urban areas, a reflection of broader health system constraints [
14].
Qualitative studies have revealed that barriers to HF delivery include cost, women’s concern that a companion will not be allowed to stay with them during labor, unfriendly attitude of health care providers, and concern about not being able to deliver in the preferred position [
16]. Some studies have uncovered neglectful, abusive and disrespectful treatment of women during childbirth by healthcare providers, which tended to deter HF delivery [
17‐
19]. Other factors identified by community members as affecting the quality of services delivered were inadequate trained health workers, shortage of essential drugs, and long distances to HFs [
20].
Research has shown that at the individual-level, older, more educated, wealthier, urban and working Nigerian women are more like to deliver in a HF than their counterparts [
21,
22], while those residing in the northern states are significantly less likely to deliver in a HF than those in southern states [
23]. HF delivery has also been associated with Nigerian women’s participation in household decision making and attitudes regarding a wife’s ability to refuse sex [
22,
24]; timing of initiation of antenatal care (ANC) and number of ANC visits at the primary health care center [
21] and household levels [
25]; and enrolment in a health insurance scheme [
25]. Community-level determinants have included residence in areas with a high proportion of women who had secondary education and ethnic diversity, with the former factor being positively and the latter, negatively associated with HF delivery in Nigeria [
23]. However, there is a lack of studies that have examined the extent to which structural aspects of health care are predictive of HF delivery or MH care utilization, more broadly, after controlling for individual-level factors.
This study contributed to filling this gap in the literature by examining the association between HF readiness to provide MH care and HF delivery in five states of Nigeria. The study is especially valuable since Nigeria’s contribution to the global burden of maternal deaths is one of the highest and since the country faces many health-system challenges, which are also found in other sub-Saharan African settings. Specific study objectives were to: (1) assess changes in HF readiness to deliver MH services over time; (2) examine the association with HF delivery of area-level indices of readiness to provide MH services and quality management practices at HFs; and (3) assess whether these associations changed significantly over time and were different in rural and urban areas. The results of the analysis could help improve understanding of the supply and demand nexus for HF delivery, which is critical for reducing deaths arising from complications of pregnancy.
Methods
Data were drawn from the 2005 and 2009 surveys of the Community Participation for Action in the Social Sectors (COMPASS) project in Nigeria. The COMPASS Project was launched in 2004 with the aim of expanding participation, ownership and use of healthcare and education sector services at the community level in 51 local government areas (LGAs) across four states (Bauchi, Lagos, Kano and Nasarawa) and the Federal Capital Territory (FCT) of Nigeria over a period of 5 years. The project was designed with the aim of stimulating and promoting the integration of education and health development across all project activities at all levels.
The study was conducted in the 51 COMPASS LGAs and comprised three surveys: household, HF, and school. The household survey used a multi-stage stratified sampling design and collected information on reproductive and MH, child health, and HIV/AIDS-related knowledge and behaviors among women aged 15–49 and men aged 15–64 years. At the first stage of sampling, enumeration areas (EAs) were selected within each state, with probability proportional to the number of LGAs per state as follows: 1:1:2:2:1 for Bauchi, FCT, Kano, Lagos, and Nasarawa, respectively. At the second stage, 25 households were selected within each sample EA using systematic random sampling. Fieldwork was conducted from July-August 2005 and from mid-June to August, 2009.
The HF surveys (including comprehensive health care centers; public PHC centers; health, maternity, private, or uniformed services clinics; health posts; and dispensaries and patent medicine vendors (PMVs)) were implemented at the same time as the household surveys. The HF survey was census of all public and primary HFs and PMVs serving the households surveyed. Consequently, the HF survey included some service delivery points that were located outside of the sample EAs and LGAs selected for the household survey. A total of 233 and 286 HFs were surveyed and facility inventories and provider interviews conducted in 2005 and 2009, respectively. The LGA was used to link the HF and household data.
The outcome was binary and indicated whether the most recent birth in the past 5 years was delivered in a medical institution. The first two indices of service readiness were constructed based on World Health Organization guidelines on tracer elements for assessment of general service readiness [
26]. The third index measured management practices supportive of quality MH services. Each component of the index was binary unless otherwise indicated.
(1)
Adapted Index of basic amenities for the provision of MH services:
This was a 7-item additive index measuring the presence of the following resources in HFs: power (a grid or a functional generator and fuel for it); a protected water source; communication equipment (a working phone or shortwave radio); access to an incinerator for disposal of potentially contaminated waste and items that are not reused such as bandages and syringes; HF assessed to be clean; public transportation within 1 kilometer; and beds for overnight stay. Unfortunately, the HF questionnaire excluded four recommended components – access to computer with email/internet, access to adequate sanitation facilities, availability of emergency transportation, and availability of a room with auditory and visual privacy for patient consultations. The resulting HF index ranged from 0 to 7 and had a Cronbach’s alpha of 0.612 in 2005 and 0.793 in 2009. The LGA-based measure was the mean adapted index of basic amenities per HF surveyed in the LGA.
(2)
Adapted Index of readiness to deliver BEmONC:
Components of the index were based on WHO (2010) recommendations and covered staff training, equipment and medicines/commodities, and included: availability of guidelines for delivery; staff trained; emergency transportation not considered problematic; examination light; suction apparatus/mucous extractor; vacuum aspirator or dilation and curettage kit; newborn bag and mask; partograph; clean gloves; injectable uterotonic; injectable antibiotic; and intravenous solution with infusion set. Data were not collected on three recommended components: manual vacuum extractor, antibiotic eye ointment for the newborn, and magnesium sulphate. HFs that did not provide delivery/newborn care were assigned the value “0” on this indicator. The resulting 22-item additive index represented the cumulative availability of components required to provide BEmONC, had a Cronbach’s alpha of 0.925 in 2005 and 0.939 in 2009, and ranged from 0 to 21 for HFs in the sample. The LGA-based measure was the unweighted average number of items present in the HFs that provided delivery and newborn care in the LGA.
Index of management practices supportive of quality MH services
This index measured the routine use of quality assurance methods by the HF; the occurrence and content of supervisory visits in the past six months; the availability of systems for client feedback; the presence of up-to-date client and birth registers; and the availability of a skilled provider. Questions on use of quality assurance methods asked non-PMVs whether any of the following methods of quality assurance were routinely used by the facility: (a) supervisory checklist for health system components (e.g., service-specific equipment, medications and records) based on standard and protocol; (b) supervisory checklist for health service provision (e.g., observation checklist) based on standards and protocol; (c) system for identifying and addressing quality of care that is implemented by staff or specific service level; (d) facility-wide review of mortality; (e) periodic audit of medical records or service registers; (f) quality assurance committee/team; (g) regional/district health management teams; and (h) other method. Components pertaining to supervision were asked separately for ANC/postpartum care and delivery/newborn care and measured (h) number of times in the last six months the provider’s delivery/newborn care was supervised and for the most recent supervisory visit, whether the supervisor (i) checked the provider’s records/reports; (j) observed his/her work; (k) provided feedback on his/her performance; (l) provided updates on administrative or technical issues related to his/her work; (m) discussed problems the provider had encountered; (n) discussed job expectations; and (n) anything else. Components pertaining to the availability of systems for client feedback asked whether the HF had the following systems for determining client opinion about the HF or its services: (o) suggestion box; (p) client survey form; (q) client interview; (r) other system. The variable measuring up-to-date birth registers consisted of four categories: no register, register not seen, register seen – last entry more than 7 days ago, and register seen – entry in past 7 days. One component of the index measured whether a skilled birth attendant (doctor, nurse or midwife) was present at the facility or on call 24 h a day, including weekends to provide delivery care and their actual involvement in conducting deliveries. This variable was coded as follows: 4 if a skilled attendant was present and always conducted deliveries; 3 if a skilled attendant was present but deliveries were sometimes conducted by primary- or auxiliary-level staff; 2 if a skilled attendant was on call and always conducted deliveries; 1 if a skilled attendant was on call but deliveries were sometimes conducted by primary or auxiliary level staff; and 0 if a skilled attendant was not present or on call 24 h a day, including weekends, to provide delivery care. The resulting 28-item HF index ranged from 0 to 31, with a Cronbach’s alpha of 0.8834 in 2005 and 0.895 in 2009. We calculated the LGA-level mean index based on the scores of all non-PMVs that provided MH services in the LGA.
The analysis controlled for the following individual-level variables: year of survey (2009 versus 2005); duration of residence in the area (years); age as reported; number of children ever born; education (none, primary, secondary/higher); marital status (married, living with a partner, not in union); type of place of residence (urban, semi-urban, or rural); state (Bauchi, Kano, FCT, Lagos, Nasarawa); counseling about pregnancy complications (no ANC from a health professional, not counseled, counseled about pregnancy complications and where to go); and household wealth (low, medium, high). Household wealth represented by tertile of an index constructed from the household ownership of the following amenities/items, using principal components analysis: refrigerator, electricity, piped water, flush toilet, bicycle, motorcycle, car, television, radio, and telephone/cellular phone). The index was based on the first component, which explained 44.2 % of the common variances of all ten components. Scree plot inspection revealed a distinct one-factor solution. The Kaiser-Meyer Olkin measure of sampling adequacy was 0.867.
The analysis was based on women whose most recent birth occurred in the five years preceding the survey. Descriptive statistics were calculated for all variables of interest. We computed
F-tests to investigate the association between the HF delivery rate and LGA measures of service readiness, taking into consideration the multi-stage sampling design. Two-level random-intercept logistic regression models that offered simultaneous consideration of
i women (Level 1) nested in
j LGAs (Level 2) were estimated to take into consideration the hierarchical clustered structure of the data, which if ignored, could generate improper standard errors, and to incorporate random effects at the LGA and individual levels to account for unobserved factors. Adjusted odds ratios (AORs) and 95 % confidence intervals (CIs) were estimated from regression statistics using the generalized latent and mixed model command in Stata 12.1.0.
15 Variance inflation factors (VIFs) suggested that multicollinearity was not a major concern: the mean VIF was 1.82 and the highest was 2.80. Intra-class correlation coefficients (ICC) were used to evaluate how the odds of HF delivery varied between LGAs and were calculated as:
$$ \rho = \left({\upsigma^2}_{\upmu}/\ \left({\upsigma^2}_{\upmu} + {\uppi}^2/3\right)\right) $$
where σ2μ is the intercept variance and π2/3 = 3.29 and represents the level-1 residual variance for a logit model. The analytical sample consisted of 51 LGAs and 2710 mothers whose last birth occurred in the past five years and who had no missing data on variables of interest. No significant differences between missing and non-missing cases were detected.
Discussion
The purpose of this study was to examine how readiness to deliver MH services was associated with HF delivery, while taking into consideration underlying socio-economic determinants of health service use, such as education and household wealth. The study also tried to detect changes in measures of health service readiness over time and modelled some of the dynamic interactions that potentially existed between those aspects of service delivery and type of place of residence.
The analysis uncovered critical gaps in service readiness and HF capacity to provide BEmONC and corroborated national and subnational studies of HFs in Nigeria which showed major deficiencies in the health care system. One national study had found that most PHC facilities had no functional equipment for maternal and child health services while another had showed that 44 % of PHCs enrolled in the Midwives Service Scheme, an initiative implemented by the Federal Government of Nigeria to address the shortage of skilled birth attendants in rural areas, did not provide all components of MH care [
15]. In a LGA of Southwest Nigeria, none of the facilities met the criteria for a BEmOC facility, 46 % were unmanned by unskilled health attendants, and none of the health workers in the LGA had ever been trained in lifesaving skills. In addition, there was a widespread lack of BEmONC equipment and supplies [
27]. Similarly a rapid assessment of 121 PHCs revealed that most were unable to provide all BEmOC services and generally lacked clinical staff needed to dispense maternal and neonatal care services, ambulances, and uninterrupted electricity supply whenever there were obstetric emergencies. Although secondary HFs scored higher on these services, like PHCs, they tended to lack infrastructure for neonatal care [
28]. Similarly, Abegunde et al. found that in Bauchi state, which was also included in the present study, the availability, utilization, and quality of EmOC services were suboptimal [
14].
Although the rate of HF delivery increased significantly from 2005 to 2009, more than half of women in the sample delivered outside of HFs. The regression analysis revealed that the adapted index of basic amenities for the provision of ANC was more positively associated with the odds of HF delivery in 2009 than in 2005, and in rural than urban areas. While assessing and defining the quality of care can be difficult, research suggests that for poor and vulnerable clients, the most important dimensions of quality tend to include facility amenities, the others being technical competence, interpersonal relations, and accessibility [
20]. Good infrastructure and hygiene, which were captured by our adapted index of basic amenities, may have been related to greater client satisfaction with the physical environment of the HF, and served as a greater catalyst for delivery care utilization among rural as compared to urban women.
We also found that the index of management practices supportive of quality MH care was associated with significantly lower odds of HF delivery among rural women compared to their urban counterparts. This index reflected quality assurance and supportive management practices, which together with infrastructural decay, have been described as “the bane of efficient PHC delivery” in Nigeria [
29]. Given the insufficient quantity of care (even of lower-level HFs) in rural areas and rural women’s’ limited choice of HFs, these results are to be expected.
Surprisingly, readiness to deliver BEmONC was unrelated to the odds of HF delivery, after controlling for other factors, and called attention to sociocultural and other explanations for the utilization of MH services, which were not captured by our model. These explanations include transportation difficulties; attitudes of health workers; affordability, especially if supply constraints lead to women being required to bring their own supplies such as gloves; lack of privacy; women’s perception of being in good health; gender norms that constrain women’s mobility; spouses and relative’s disapproval of institutional delivery; and traditional beliefs and practices [
16,
17,
22,
28,
30].
The limited improvements in components of health service readiness may be due to several factors, such as lack of political will and inadequate resource allocation to health system strengthening at the national and local levels over the years. However, the Federal Government of Nigeria is implementing several initiatives to improve the availability of, access to, and quality of MH services and address rural–urban inequities in service provision. The maternal and child care component of the Federal Government of Nigeria’s Subsidy Reinvestment Program has identified 125 general hospitals across 36 states and 500 out of the 23,000 frontline PHC facilities for refurbishment, upgrading, equipping, supply of drugs, and the employment and deployment of skilled health workers. The implementation of the Midwives Service Scheme (MSS) is addressing the shortage of skilled birth attendants at the primary healthcare (PHC) level, particularly in rural areas. Conditional cash transfers are being provided to pregnant women to address the indirect costs of care seeking, which partially contribute to the low demand for ANC and delivery services [
31]. Such programs must be scaled up and accompanied by regular supportive supervision in order to improve the utilization of MH services and save lives.
The dearth of research on supply-side determinants of MH care utilization precluded the specification of hypotheses about the importance of service readiness as a predictor of HF and a nuanced comparison of our findings with those of other studies. Although the evidence is inconsistent, the effects often small in magnitude, and causality a concern, a number of studies have found a significant association between elements of HF readiness and family planning (FP) outcomes, even after controlling for individual-level factors. Using a composite measure of infrastructure and facility readiness to provide FP services, Hong, Montana and Mishra found that measures related to counseling and the examination room had significant positive effects on IUD use in Egypt [
32]. Similarly, using an index score of the service delivery infrastructure, medical equipment, essential medicines, number of contraceptive methods available on the day of the visit and the number of staff trained in FP, Do and Keonig found that residence in communes with higher quality health centers was associated with significantly lower risk of method discontinuation [
33]. However, other studies have found a weak or non-significant link between HF readiness and client contraceptive behavior [
34,
35].
The analysis presented here focused on only one of the WHO-recommended five domains of general service readiness – basic amenities – with some adaptation. At earlier stages of the analysis, we examined two other domains, notably basic equipment and standard precautions for prevention of infection, in the regressions but these domains were unrelated to the odds of institution delivery, and their associations did not vary significantly over time or by type of place of residence. Consequently, they were omitted from the final regression models. No data were collected on laboratory tests (hemoglobin, blood glucose, HIV rapid diagnostic tests (RDT), syphilis RDT, malaria RDT or smear, TB microscopy, general microscopy, urine pregnancy test, urine dipstick), precluding us from including the laboratory domain of general service readiness in our analysis. Due to survey’s focus on maternal and child health and HIV/AIDS, data were not collected on medicines for non-communicable diseases (Salbutamol, Glibenclamide, Atenolol, Captopril, Simvasatin, Amitriptyline, Diazepam, Omeprazole), on one medicine for infectious diseases (Ceftriazone), and on one pain medication (Diclofenac). As data were available for only four of the 14 essential medicines (Ciprofloxacin, Co‐trimoxazole, Amoxicillin, and Paracetamol), a decision was made to exclude the essential-medicines index from the analysis.
Due to lack of data, the adapted index of basic amenities could not include four recommended components: access to a computer with email/internet; access to adequate sanitation facilities; availability of emergency transportation; and availability of a room with auditory and visual privacy for patient consultations. The addition of those components could have increased the reliability of the index given that the more items there are in an index that is designed to measure a particular concept, the more reliable will be the index. It is to be noted that Cronbach’s alpha for the adapted index of basic amenities was much lower in 2005 (0.612) than in 2009 (0.793), which signified poorer consistency of the item responses in the earlier survey. One factor that contributes to the consistency of an index is stable characteristics of the attribute being measured but between 2005 and 2009, several interventions implemented by development partners such as COMPASS and PATH targeted the provision and improvement of basic amenities in HFs, among other things. While the adapted index of basic amenities enabled us to have one statistical measure with which to gauge the availability of basic amenities in HFs serving COMPASS LGAs, the low reliability of the adapted index of basic amenities in 2005 increased the risk that our analysis underestimated or failed to detect the true association between the adapted index and the odds of HF delivery in 2005. This limitation should be borne in mind in the interpretation of the findings.
The study raised important methodological issues regarding how best to link population and HF data. Although the survey identified the universe of HFs actually used by households, the administrative boundaries of the LGA were artificially imposed on the data in order to identify level-two units for the multilevel regression and administrative areas for a potential programmatic response. The question arose as to whether the sampled HFs that fell within a given LGA characterized well where residents were obtaining MH care. There were two caveats. First, many individuals did not use HFs for MH care, which likely introduced biases in the sample. Second, our methodological approach combined two concepts: “Where LGA residents could go” and “where LGA residents did go” for health services. In the survey, GPS coordinates were obtained for HFs but not for households or EAs, an unfortunate omission. In addition, ensuring that there were enough facilities surveyed and service providers interviewed in areas that were closer approximations of neighborhoods than the LGA was a challenge and influenced how level-two units were defined.
Other limitations of the data stemmed from their cross-sectional nature, making it difficult to establish causality. Endogeneity was a concern as MH services may have been placed in areas with higher demand and fertility levels, potentially leading to an overstatement of the results. Temporal ordering was also of concern as measures of the MH service delivery environment may not have preceded delivery. Other limitations were that the data were not representative of the states surveyed and that we were unable to measure the quality of the services provided. Whether a provider carries out the right actions and the extent to which this translates into the right actions by the patient are best captured through exit interviews, provider-patient observations, mystery client studies, and follow-up studies of patients. These methods of data collection were not included in the HF survey. Finally, the odds of HF delivery may be determined by unobserved factors such as cultural beliefs surrounding delivery, transportation networks, financial costs of care, and provider shortage and absenteeism. Future research should explore these issues.
Acknowledgements
We are grateful to the team of experts led by Dr. Alfred Adewuyi from the Center for Research, Evaluation, and Resource Development (CRERD) for their leadership during all stages of survey implementation. We also owe a debt of gratitude to the supervisors and field-workers who participated in the 2005 and 2009 surveys.