Why not? Understanding the spatial clustering of private facility-based delivery and financial reasons for homebirths in Nigeria

Why not? Understanding the spatial clustering of private facility-based delivery and financial... Background: In Nigeria, the provision of public and private healthcare vary geographically, contributing to variations in one’s healthcare surroundings across space. Facility-based delivery (FBD) is also spatially heterogeneous. Levels of FBD and private FBD are significantly lower for women in certain south-eastern and northern regions. The potential influence of childbirth services frequented by the community on individual’s barriers to healthcare utilization is under-studied, possibly due to the lack of suitable data. Using individual-level data, we present a novel analytical approach to examine the relationship between women’s reasons for homebirth and community-level, health-seeking surroundings. We aim to assess the extent to which cost or finance acts as a barrier for FBD across geographic areas with varying levels of private FBD in Nigeria. Method: The most recent live births of 20,467 women were georeferenced to 889 locations in the 2013 Nigeria Demographic and Health Survey. Using these locations as the analytical unit, spatial clusters of high/low private FBD were detected with Kulldorff statistics in the SatScan software package. We then obtained the predicted percentages of women who self-reported financial reasons for homebirth from an adjusted generalized linear model for these clusters. Results: Overall private FBD was 13.6% (95%CI = 11.9,15.5). We found ten clusters of low private FBD (average level: 0.8, 95%CI = 0.8,0.8) and seven clusters of high private FBD (average level: 37.9, 95%CI = 37.6,38.2). Clusters of low private FBD were primarily located in the north, and the Bayelsa and Cross River States. Financial barrier was associated with high private FBD at the cluster level – 10% increase in private FBD was associated with + 1.94% (95%CI = 1.69,2.18) in nonusers citing cost as a reason for homebirth. Conclusions: In communities where private FBD is common, women who stay home for childbirth might have mild increased difficulties in gaining effective access to public care, or face an overriding preference to use private services, among other potential factors. The analytical approach presented in this study enables further research of the differentials in individuals’ reasons for service non-uptake across varying contexts of healthcare surroundings. This will help better devise context-specific strategies to improve health service utilization in resource-scarce settings. Keywords: Spatial epidemiology, Clustering, Facility childbirth delivery, Maternal health service utilization, Financial barrier, Private health services * Correspondence: kerry.wong@lshtm.ac.uk Department of Infectious Disease Epidemiology, Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Wong et al. BMC Health Services Research (2018) 18:397 Page 2 of 12 Background dominations of lower-level and primary care and private Despite ongoing efforts by the Nigerian health system to health services in some areas but not others [13]. In increase maternal health service utilization, including addition, despite the Nigerian government’s aspiration to midwives service schemes, removal of user fees and in- provide free/subsidized maternity care in the public sec- creasing the involvement of the private sector [1, 2], tor, some women who stay home for childbirth reported population usage of many life-saving obstetric interven- cost or finance as a barrier to using maternity care, tions remains suboptimal. National statistics for 2009– among other factors [14, 15]. This raises questions re- 2013 show that, for instance, 22.6% of all births occurred garding current understanding of the factors for service in a public health facility and 13.2% in the private sector uptake vs. non-uptake in relation to one’s healthcare sur- – leaving approximately two thirds of childbirths based roundings. In some settings, e.g. where public maternity outside of a health facility [3]. At the subnational scale, care is free of charge, it is likely that some of those who likelihoods for facility-based delivery (FBD) and private stay home for childbirth for financial reasons only con- FBD vary considerably, and were significantly lower for sidered using private services, the alternative being women residing in parts of the South South zone, and homebirths (over public care). This speculation might be majority of the Northern zone [3]. more pertinent where private FBD is common, due to Both in Nigeria and other low- and middle-income the potential impact that one’s peers and healthcare sur- countries (LMICs), having a FBD is a practical way to roundings have on their reasons for service non-uptake. ensure assistance by a skilled birth attendant and ac- The aim of this study is to assess the extent to which cost cess to life-saving interventions for mothers and new- or finance is a barrier for FBD across geographic areas with born [4]. Previous reviews addressing factors related varying levels of private FBD in Nigeria. To overcome the to FBD in sub-Saharan Africa and other LMICs have limitation of community-level data availability, we present identified an array of determinants [4–7]. Moyer and an innovative approach applying geographic information Mustafa’s literature review, published in 2013, system (GIS) tools to examine the clustering of maternity highlighted an overwhelming reliance on population/ care utilization using individual-level survey data. This survey data with which maternal sociodemographic study will help motivate and enable further investigation of factors were well-represented [4]. The limited body of the way in which childbirth services frequented by the literature around community-level factors of FBD in community influences community members to deliver in LMICs emphasizes community socio-demographic or outside a health facility, adding contribution to the characteristics, community views on skilled and trad- current effort to support maternity care utilization for itional births [8, 9], service accessibility such as dis- groups and individuals most “left behind”. tance to care and community uptake of antenatal care [4]. Communities likely have other unique characteris- Methods tics that influence demand for and supply of health- Data and study sample care [10], many of which are overlooked. This analysis was based on data from the 2013 Nigeria Unlike other health service seeking, childbirth can Demographic and Health Survey (NDHS). The data is happen unexpectedly throughout the day and the representative at the national level, of the six geopolitical woman may need to reach a nearby care provider at zones and of the 36 states and the Federal Capital Terri- relatively short notice. The types of childbirth delivery tory (FCT-Abuja). The survey sample was selected using services most accessible to, or most accessed by, the a stratified multi-stage cluster probability sampling de- community directly relate to an individual’s perception sign with census ward as the primary sampling unit. As of, wishes for, and actual uptake of services. Women also part of the DHS sampling procedure, all households in exchange information and experience surrounding child- each sampled ward was enlisted, which was then used as birth in social settings, and one’s planning for future de- the sampling frame for household selection [16]. Eligible livery may be conditioned by assessing factors important individuals aged 15–49 in selected households were to their peers, culture and community [11, 12]. A better interviewed with a standardized questionnaire. The final understanding of one’s healthcare surroundings is im- sample of the 2013 NDHS consisted 896 census wards perative to developing effective strategies to increase and 39,902 eligible women; 98% (38,948) were success- healthcare utilization among groups currently “left be- fully interviewed. Women with a live birth in the five hind”. Part of the dearth of research in this area might years before the interview were asked to self-report the be due to the lack of suitable data, especially at the na- care received during pregnancy and delivery. The sample tional scale. of the current analysis was restricted to the circum- In a study of the characteristics of health facilities stances of 20,467 women’s most recent live birth during across Nigeria, Nwakeze and Kandala found vast geo- the five-year survey recall period as some of the required graphic disparities in the country, including greater data was only collected for this subsample. Wong et al. BMC Health Services Research (2018) 18:397 Page 3 of 12 Geography and administration of Nigeria or governmental health facilities (HFs), private or Nigeria is divided into six geopolitical zones (Fig. 1): non-governmental HFs, as well as all other unspecified North Central, North East, North West, South West, locations [3]. FBD was obtained by coding responses as South South and South West; and within these zones, “any HF” and “not in a HF”. For the analysis of private into 36 states and the FCT-Abuja. For administrative FBD, all births were categorized as “any private HF” purposes, the states are subdivided into 774 local gov- and “not in a private HF”. We note that the 2013 ernments areas [3], each made up of approximately 10– NDHS had conflated all non-governmental, for-profit 15 wards [17]. and not-for-profit providers as one category of “private” provider. The outcome of interest was financial barrier for Measurement FBD. Women who did not deliver in a HF indicated Population centroids of wards, recorded as latitude and the reasons that applied to them from a list of po- longitude, were obtained by DHS enumerators using tential barriers, including “cost too much”. Other co- Global Position System (GPS) receivers [18]. All individ- variates and demographic information considered as uals residing in a ward therefore have the same georefer- potential confounders were: wealth quintile, maternal ence. For privacy considerations, the coordinates were education, maternal age and parity at the time of the randomly displaced by up to 2 km in urban areas and up most recent birth, and whether the woman had to 5 km in rural areas by the NDHS. An additional 1% health insurance coverage. Household wealth quintile of rural wards were displaced by 10 km. was derived from the wealth index – constructed by Delivery location was based on women’sanswerto: the DHS using household asset data via a principal “Where did you give birth to [name of child]?” on the component analysis [19]. The sampled households Women’s Questionnaire. The major categories of were than ranked and divided into five quintiles. response were domestic environments (home of re- Each woman is assigned her household’swealth spondent or of traditional birth assistant (TBA)), public quintile. Fig. 1 Map of Nigeria showing boundaries of six geopolitical zones, 36 states and Federal Capital Territory (FCT-Abuja). Shapefile is obtained from gadm.org. The 2018 GADM license allows data re-use for academic and other non-commercial purposes (https://gadm.org/license.html, last accessed: 14th May 2018) Wong et al. BMC Health Services Research (2018) 18:397 Page 4 of 12 Spatial scan statistics of private FBD outcome) was related to the percentage of births occur- To identify geographic clusters of high and low private ring in private facilities. The SatScan spatial clusters FBD, the number of most recent births and those based were weighted by the number of most recent births cir- in a private HF were aggregated at the ward level, with cled within. To account for a proportion as outcome adjustment of survey sampling weighting. Together with (bounded between 0 and 100%), we adopted a general- ward latitude and longitude as inputs, each ward was ized linear model, specifying a logit link and the bino- treated as an analytical unit to test whether private FBDs mial family [27–29]: were distributed randomly in space or not. logitðÞ p ¼ xβ þ ε where Y  BinðÞ N ; p ; ε At the ward level, the observed numbers of private i i i i i i FBD varied from zero to the total number of eligible  N0; σ births. To detect clustering of private FBDs, we chose a Poisson distribution to represent the expected distribu- We denoted y = number of private FBD in SaTScan tion of this count over space. Under the null hypothesis, spatial cluster i, N = number of most recent births in i the expected number of private FBDs in each area is and p = probability of having a private FBD. We also proportional to its population size (approximated using specified the Huber-White (i.e. robust) estimators of the sample size) [20, 21]. Spatial scan statistics was per- standard errors in case of heteroskedasticity arising from formed using the SaTScan™ software (version 9.4) [20, potential misspecification in the distribution family [30]. 21]. Spatial clusters were identified by taking into con- The z test was used for significance testing of model co- sideration the rates of nearby wards [22, 23]. The spatial efficients. We generated predictions from both the bi- scan method used circular windows of various sizes that variate and multivariate fits and back-transformed them move across the map to find clusters of wards with ei- as the percentages of women with a non-facility birth ther higher and lower than expected rates under the null who cited cost was a barrier at 5%-intervals of hypothesis of uniform spatial distribution [24, 25]. The community-level private FBD. radius of the circle varies continuously from zero to a predefined value that specified the percentage of the Missing data maximum total population at risk within the scanning We found missing data in geographic coordinates in window [21]. The recommended maximum size is 50%; seven wards, containing < 1% of the respondents from we conducted additional scans at the 10 and 5% levels to the study sample. These were removed from analyses account for independent smaller clusters that may be where location data was required. We also found 0.4% of contained in a large cluster. The alternative hypothesis is missing data for health insurance coverage and coded that there is a reduced/elevated rate within the scanning these as uninsured. There was no missing data in the window as compared to outside. The test of significance, other variables in the model. based on likelihood ratio and the null distribution, was obtained from Monte Carlo Simulation [26]. The num- Results ber of permutations was set at 999 and the significance Facility-based delivery level was set at 0.05 [21]. Identified clusters are ordered Of the 20,467 births in our sample, 7649 (37.4, 95%CI = based on their likelihood ratio test values. 34.7,40.2) occurred in health facilities: 23.8% (95%CI = Geographic locations of, and wards contained in each, 22.0,25.5) in public and 13.6% (95%CI = 11.9,15.5) in pri- identified spatial cluster were merged back to the 2013 vate facilities. More of those who were rural residents, NDHS women’s data. We considered women living in from the poorest wealth quintile, without any education, the same SaTScan spatial cluster to be in the same uninsured and having a second or higher order birth de- “community”. Estimates on private service use as a per- livered outside of a health facility (Table 1). Geographic centage of all most recent births, financial barriers re- variations of FBD were observed – highest in the South ported among women who did not deliver in a HF, as East zone (78.5, 95%CI = 73.2,83.0) and lowest in the well as other covariates were recalculated for each SaTS- North West zone (12.8, 95%CI = 10.2,15.9). can spatial cluster to generate community-level data. This was done in Stata SE version 14 (StataCorp LP, Col- Sub-national private facility-based delivery lege Station, TX, USA), adjusted for survey-specific Regional averages of private FBD varied between 0.5% weighting and stratified, cluster sampling design. (95%CI = 0.3,1.1) in North West zone to 44.8% (95%CI = 38.4,51.4) in the South East zone (Table 1). Using SaTS- Relating community -level private facility use to nonusers’ can analysis, ten spatial clusters of low level and seven self-reporting of financial barrier spatial clusters of high private FBD were identified Using SaTScan spatial cluster as the analytical unit, the (Table 2). The number of wards contained in these geo- percentage of nonusers reporting financial barrier (main graphic clusters ranged from five to 88; the number of Wong et al. BMC Health Services Research (2018) 18:397 Page 5 of 12 Table 1 Percentage distribution and 95% confidence intervals of sample sociodemographic characteristics by place of delivery Number of most Place of delivery recent births Outside of a health facility Public health facility Private Health facility N proportion 20,467 12,818 5100 2620 (100) 62.6 (59.8,65.3) 23.8 (22.0,25.6) 13.6 (11.9,15.5) Area of residence Urban 6790 36.8 (32.6,41.2) 36.3 (33.5,36.3) 26.9 (23.2,30.8) Rural 13,402 76.9 (74.2,79.4) 16.8 (15.0,18.8) 6.3 (5.2,7.7) Wealth quintile Poorest 4379 93.8 (92.4,95.0) 5.0 (4.1,6.1) 1.2 (0.8,1.9) Poorer 4603 81.8 (79.2,84.1) 13.4 (11.7,15.3) 4.8 (3.7,6.2) Middle 4069 62.2 (58.7,65.5) 26.6 (24.0,29.3) 11.3 (9.4,13.4) Richer 3798 42.5 (38.9,46.1) 39.4 (36.5,42.3) 18.2 (15.7,20.9) Richest 3343 18.6 (16.2,21.3) 42.5 (38.7,46.3) 38.9 (34.3,43.7) Maternal education No education 9171 88.0 (86.4,89.5) 10.2 (8.9,11.6) 1.8 (1.4,2.3) Primary 4113 57.1 (54.0,60.3) 27.5 (25.2,29.9) 15.4 (13.4,17.6) Secondary 5565 33.8 (31.1,36.5) 39.0 (36.4,41.6) 27.2 (24.0,30.7) Higher 1343 8.3 (8.3,10.7) 51.3 (46.7,55.9) 40.4 (35.5,45.5) Health insurance Yes 363 14.7 (10.4,20.4) 50.6 (43.8,57.4) 34.7 (27.8,42.3) No 19,829 63.4 (60.6,66.1) 23.3 (21.6,25.1) 13.3 (11.6,15.1) Parity First birth 3624 51.7 (48.3,55.0) 31.4 (28.9,34.0) 16.9 (14.6,19.5) Higher order birth(s) 16,568 65.0 (62.2,67.7) 22.1 (20.4,23.9) 12.9 (11.3,14.7) Geopolitical zones North Central 3095 53.0 (47.4,58.5) 31.3 (27.7,35.2) 15.7 (12.7,19.3) North East 4001 79.5 (74.9,83.4) 19.2 (15.5,23.5) 1.3 (0.8,2.1) North West 6206 87.2 (84.1,89.8) 12.3 (9.8,15.2) 0.5 (0.3,1.1) South East 1724 21.5 (17.0,26.8) 33.7 (28.9,38.9) 44.8 (38.4,51.4) South South 2500 49.2 (44.1,54.4) 36.6 (32.6,40.7) 14.2 (10.8,18.3) South West 2666 23.8 (19.3,29.0) 23.8 (31.9,40.6) 40.1 (35.2,45.1) Age at birth Mean (interquartile range) 29.42 (29.2,29.6) Self-reported financial barrier to deliver in a health facility 9.1 (8.9,10.5) Not applicable Not applicable most recent births circled within a geographic cluster 0.8,0.8), respectively. Average public FBD among all births ranged between 63 and 1201, and spatial cluster radii was 37.2% (95%CI = 37.1,37.4) in high private FBD clus- varied between 21.2 and 208.5 km. Altogether, 648 wards ters. On the other hand, 14.8% of all births were public and 14,434 births occurred in these 17 clusters. facility-based in the ten spatial clusters of low private The location and size of these geographic clusters were FBD. Substantial differences in sociodemographic charac- drawn in Fig. 2. Clusters of low private FBD were pri- teristics of women living in the two groups of spatial clus- marily located in the North West and North East zones, ters were also seen (Fig. 2). Women in low private FBD with an exception near Jos North in Plateau State, where clusters were more rural, poorer and less educated com- one spatial cluster of high private FBD (50.5, 95%CI = pared to women in high private FBD clusters. 35.5,65.5) was identified. In addition, southern Cross We performed additional cluster detections setting River state and central and southern Bayelsa state (in the maximum cluster size to 10 and 5% of the survey sam- South South zone) also showed spatial clustering of low ple. The first yielded the same set of results. The details private FBD: 2.9% (95%CI = 1.0,4.7) and 2.1% (95%CI = of the 19 SaTScan spatial clusters returned from using 0.0,5.5), respectively. Communities of high private FBD the 5% limit is given in Additional file 1: Figure S1. No were identified around the Lagos and Ogun States (52.8, substantial differences to the model with 17 SatScan 95%CI = 47.7,57.9), Edo State (32.9, 95%CI = 24.7,41.1) as clusters were observed. well as large parts of the South-East zone (e.g., Imo and Abia States) and the North Central zone. Reporting cost as a barrier for facility-based delivery Mean percentages of private FBD in high and low clus- Across the seven spatial clusters of high private FBD, ters were 37.9% (95%CI = 37.6,38.2) and 0.8% (95%CI = 24.5% (95%CI = 21.1,27.8) of women delivered at home Wong et al. BMC Health Services Research (2018) 18:397 Page 6 of 12 Table 2 Seventeen significantly higher and lower than expected proportions of FBD spatial clusters ID Cluster location No. of No. of most Observed Observed Expected number Relative p-value wards circled recent births number % private of private FBD risk Latitude Longitude Radius (km) of private FBD FBD under H High 1 6.7 3.7 97.2 75 1182 605 52.8 154 4.82 < 0.001 2 9.9 8.9 21.2 5 63 33 50.5 8 4.06 < 0.001 3 5.8 7.2 85.2 88 1199 569 48.9 156 4.38 < 0.001 4 6.7 5.4 78.7 28 457 146 32.9 60 2.54 < 0.001 5 8.5 4.6 132.5 67 1198 338 28.0 156 2.34 < 0.001 6 7.3 9.0 79.4 12 249 73 27.4 32 2.29 < 0.001 7 7.9 7.1 148.3 78 1200 295 25.3 156 2.00 < 0.001 Low 8 9.8 9.7 64.0 8 233 8 3.1 30 0.26 0.014 9 5.3 8.6 75.9 17 280 8 2.9 37 0.22 < 0.001 10 4.6 5.7 88.6 24 558 5 2.1 73 0.07 < 0.001 11 8.7 11.6 178.0 41 1187 22 1.8 155 0.14 < 0.001 12 10.8 7.3 155.6 38 1174 14 1.4 153 0.09 < 0.001 13 10.8 3.9 184.8 24 770 7 0.5 100 0.07 < 0.001 14 13.3 8.0 144.9 31 1201 1 0.1 156 0.01 < 0.001 15 12.0 12.5 208.5 46 1197 2 0.1 156 0.01 < 0.001 16 11.7 9.3 84.7 30 1085 1 0.1 141 0.01 < 0.001 17 13.2 5.5 145.6 36 1201 0 0.0 156 0.00 < 0.001 FBD = facility based delivery; H = null hypothesis of spatial randomness The likelihood ratio test is used for testing cluster significance Cluster 1 is the most likely cluster; all other clusters are non-overlapping secondary clusters Relative risk of private FBD within cluster compared to the risk in all other areas and 14.9% (95%CI = 14.7,15.1) reported cost as a barrier showed that the factors associated with self-reported fi- (Fig. 2). In contrast, 85.7% (95%CI = 83.2,88.2) of women nancial barrier for FBD at the spatial cluster unit level living in the 10 clusters with low FBD delivered at home, included living in Cross River and Bayelsa States, the and 8.8% (95%CI = 8.6,8.9) cited cost as a barrier. Fig- percentage of public facility utilization, rural setting, ure 3 illustrates that in contrast to other spatial clusters wealth, the level of maternal education, and the percent- of low private FBD, exceptionally high proportions of age of women covered by health insurance (Table 3). All nonusers living in Cross River (32.7, 95%CI = 26.3,39.1) of these were significant at the p < 0.001 level. and Bayelsa State (25.2, 95%CI = 19.1,31.4) said cost was In multivariate analysis, all predictors remained signifi- a reason to deliver outside a facility. Unadjusted analysis cantly associated with the proportion of women Fig. 2 Seventeen SaTScan spatial clusters (drawn proportionate to cluster radii) of higher (red) and lower (blue) than expected proportions of private facility birth among all most recent births. The DHS wards contained in each spatial clusters are also shown Wong et al. BMC Health Services Research (2018) 18:397 Page 7 of 12 Fig. 3 Proportions of women delivering outside a health facility who self-reported financial barrier as a reason for homebirth in 17 spatial clusters of high and low private facility births. Predicted percentages and confidence intervals at various levels of private facility birth from an adjusted generalized linear model weighted by numbers of most recent births in spatial clusters are also shown (represented by size of bubbles) reporting financial barrier in the community (Table 3). and Abia States had particularly high levels of private After controlling for the proportion of birth occurring in FBD. Using a novel approach, we examined the associ- public HFs, rurality, wealth, maternal education, health ation between private healthcare utilization contexts and insurance and residency in Cross River and Bayelsa financial barriers for FBD. We found cost was more States, a 10% point increase in private facility use for likely to be cited as a barrier to FBD in settings where childbirth was associated with an average 1.94% point private FBD was high. We found exceptions, however, increase (95%CI = 1.69,2.18) in nonusers citing cost as a for southern Cross River and Bayelsa States, where a barrier for FBD. The adjusted predicted percentages of large proportion of nonusers reported cost as a barrier self-reported financial barrier across varying levels of and overall facility delivery (in both the public and pri- private service use were also computed based on the ad- vate sectors) very low. justed regression model. Table 3 and Fig. 3 illustrate a steady rise in the extent to which financial consideration Limitations was a barrier as community-level private FBD increased. Our findings have important implications, but they should be understood with certain limitations. Firstly, Discussion the 2013 NDHS response option for private delivery in- To our knowledge, this is the first study to examine na- cluded both for-profit and not-for-profit establishments tional geographic disparities in private facility use for operating under different financial motives and poten- childbirth in a sub-Saharan African country at a small tially charging widely varying fees for childbirth care. geographic scale. We found substantial spatial variation However, we still believe that our assumption that pri- in the utilization of private facilities for delivery care vate sector childbirth costs more than public sector is across Nigeria. The level of private FBD was very low in valid. Self-reported reasons to deliver in non-healthcare the northern part of the country except for Jos in Plat- settings might also be subject to accuracy and reliability eau State. Private FBD was medium to high in North issues [31]. In addition, women could list more than one Central zone and the highest in the South West and barrier of FBD – approximately 50% of women who South East zones. Certain areas in Lagos, Imo, Ogun cited cost as a barrier also listed one or more other Wong et al. BMC Health Services Research (2018) 18:397 Page 8 of 12 Table 3 Effect sizes of predictor variables and estimates of proportion citing financial barriers Community-level factors Unadjusted estimates Adjusted estimates Average change in proportion of nonusers citing financial barriers with 95%CI and p-value Private facility delivery (every + 10%) 1.82 (1.79,1.86) < 0.001 1.94 (1.69,2.18) < 0.001 Public facility delivery (every + 10%) 1.17 (1.14,1.20) < 0.001 −1.64 (−1.88,-1.41) < 0.001 Rural sample (every + 10%) −1.20 (−1.24,-1.16) < 0.001 0.66 (0.54,0.78) < 0.001 Wealth: Q1 sample (every + 10%) −2.32 (2.37,2.27) < 0.001 0.50 (0.31,0.70) < 0.001 No to primary education (every + 10%) −1.83 (−1.86,-1.80) < 0.001 − 1.61 (− 1.78,-1.43) < 0.001 Health insurance (every + 10%) 21.8 (21.0,22.7) < 0.001 3.58 (2.11,5.05) < 0.001 Geographic location Others Reference Reference Cross River and Bayelsa 17.61 (17.34,17.87) < 0.001 17.34 (15.74,18.94) < 0.001 % of births in private facility Predicted percentage of nonusers citing financial barriers as reason for homebirth with 95%CI * 0 8.23 (8.08,8.37) 7.48 (7.20,7.76) 5 8.96 (8.82,9.10) 8.22 (8.00,8.43) 10 9.75 (9.62,9.89) 9.02 (8.88,9.16) 15 10.60 (10.48,10.73) 9.90 (9.81,9.99) 20 11.52 (11.40,11.64) 10.85 (10.69,11.01) 25 12.51 (12.40,12.62) 11.88 (11.59,12.17) 30 13.57 (13.46,13.68) 12.99 (12.55,13.44) 35 14.70 (14.60,14.81) 14.19 (13.56,14.83) 40 15.92 (15.80,16.02) 15.49 (14.64,16.33) 45 17.21 (17.08,17.33) 16.87 (15.79,17.96) 50 18.58 (18.42,18.73) 18.36 (17.00,19.71) 55 20.03 (19.85,20.22) 19.94 (18.29,21.59) 60 21.57 (21.35,21.80) 21.62 (19.66,23.59) ^ Unadjusted and adjusted effects were back-transformed from parameter estimates obtained using a logit link transformation. The z test was used for significance testing of model coefficients + Adjusted estimates describe the adjusted curve drawn in Fig. 4 *Adjusted predicted percentage of nonusers citing financial barriers were obtained with all other coverages fixed their mean values reasons (data not shown) – and the relative importance geographic coordinates of individuals and those at of cost compared to other reasons is not known. Contri- aggregated levels [32, 33]. butions of other potential factors – including, but not limited to individuals’ perceptions towards the care re- Giving birth in the private sector ceived and healthcare professionals – warrants further In Nigeria and other LMICs, pregnant women who opt investigation. The analytical approach presented in this for private FBD have a similar sociodemographic profile study offers a novel method for such future research – higher SES, higher education and, in some contexts, with available, secondary data. certain ethnicity or religious affiliations [34–37]. A The SaTScan spatial clusters identified were rela- search of peer-reviewed articles and the grey literature tively large in geographical size (even with a smaller returned little information on the cost of private FBD in maximum allowable limit), and there might be sub- Nigeria. However, a study showed 1.8 times more spend- stantial heterogeneity in the characteristics of the ing in private hospitals than public hospitals by users women living in the same spatial cluster. Some of residing in urban south-eastern Nigeria [38]. Despite this heterogeneity, including parity, pregnancy com- higher cost, for-profit healthcare care may have more plication and marital status, may confound our pri- appeal due to a wide range of reasons, such as privacy, mary association of interest at the individual-level, shorter waiting times, higher perceived quality of care, butwereomittedastheir relevance atthe commu- empathy and respectful approach, availability of doctors nity level is likely low. Lastly, some loss of power in and as a status symbol [39, 40]. For users of private ser- cluster detection might have occurred through a vices, cost or affordability might be a relatively weaker degradation of spatial information between the exact determinant of where to seek care. Wong et al. BMC Health Services Research (2018) 18:397 Page 9 of 12 Community-level private service use and self-reported highlights the importance of contextualizing personal financial barriers for facility-based delivery factors alongside other community- or macro-level Our findings extend the current knowledge about prefer- factors. Bayelsa State is primarily covered by marshlands ence towards private HFs for their users. We found that and waterways; it is also an important gas- and in contexts with relatively high private FBD, a greater petrol-producing region in Nigeria that has generated proportion of facility non-users reported financial bar- interest among prospective companies [46, 47]. However, riers for any care, including both private care and the most Bayelsans remain poor, and the state’s public infra- relatively more affordable public care. In Edo, Ogun and structure development insufficient [47]. Lack of trans- Abia States, for instance, the majority of health facilities portation and the riverine setting pose tremendous are privately owned [13]. Our results may indicate that impediments to overcoming physical barriers to reaching facility nonusers living in high FBD contexts are unable health services [46, 48]. In a study looking at barriers to to gain effective access to any healthcare due to personal utilization of maternal health services in Bayelsa State, a financial barrier (for private care) and insufficient majority of respondents reported infrastructure-related provision of public services in their lived environment. barriers to access (availability of facilities/equipment, In other places of high private FBD where such practice schedule of maternal health clinic, accessibility and so may have become normalized, women who lack ad- on); and much lower percentages of women reported de- equate funds for private providers might perceive deliv- terrents such as cultural acceptance and language prob- ering at home or a TBA’s home as their best alternative lems [49]. Compared to the rest of the country, special due to social pressure and low acceptability of publicly economic and environmental contexts and the additional provided services. The observed preference for home- resources required to overcome physical accessibility births is in line with qualitative findings from various barriers may have caused financial considerations to op- states including FCT-Abuja and Lagos, where women erate differently among people living in Bayelsa and who do not deliver in a health facility had poor confi- Cross River States. The role of financial barriers, separat- dence in the public health sector and strong desires to ing direct payment for delivery from other expenses and deliver with a TBA [41–43]. According to these studies, trade-offs, including cost of transport, as well as time women perceive home delivery with a TBA, and espe- and financial lost from other daily/productive activities, cially with family members present, to be personal and warrants further research. supporting [41]. Some TBAs often allow for flexible fi- nance options, such as payment in kind or in instal- A note on using DHS data to study healthcare utilization ments, making it easier for families to pay [42]. surroundings On the other hand, in settings where private facility de- Various studies have looked at the service provision en- livery use is relatively low, and especially where overall vironment as a determinant of FBD. A common ap- FBD utilization is also low, such as most of North West proach consists of conducting interviews with women zone and North East zone, women’s reasons to not give about the availability of maternity care in their commu- birth in a HF were less connected to cost. In these set- nity as a measure of service provision [50–54]. Alterna- tings, other cited barriers included service availability, dis- tively, geocoded master facility list (MFL) data or the tance or physical accessibility, social norms and lack of like, with which the entire health infrastructure of a perceived need [43]. In a study set in the Jigawa State, ap- spatial area is mapped out, are geographically linked to proximately 25% of nonusers claimed they did not attend population data in a GIS to facilitate calculation of mea- facilities for childbirth because they did not think it was sures of people’s healthcare availability [55–57]. The necessary [44]. In addition, household decision-making present study used available secondary data on dynamics also varies across this large multi-ethnic coun- individual-level service utilization and women’s location try; Abuja city/FCT-Abuja, for instance, is generally asso- of residence to construct the geographic patterning of ciated with greater gender equality when compared to healthcare surroundings across Nigeria. Our variable of other southern and northern cities [45]. Especially in the interest was community-level utilization surrounding the north, women’s relative lack of participation in individuals, which is somewhat conditioned on health- intra-household decision making and access to money care provision environment, but is also a consequence of have been associated with very low FBD rates [45]. other cultural, contextual and individual-level determi- Exceptions to the inverse relationship found between nants. Nwakeze and Kandala examined the spatial distri- financial barriers and private FBD were noted in south- bution of health establishments using data collected by ern Cross River State and Bayelsa State, where overall the National Bureau of Statistics of Nigeria, and found percentages of FBD were midrange, private FBD very moderate to low numbers of private health establish- low, and a relatively large proportion of nonusers re- ments in the Benue, Nasarawa and Kogi States, com- ported financial barriers to delivering in a HF. This pared to the number of public health facilities [13]. In Wong et al. BMC Health Services Research (2018) 18:397 Page 10 of 12 the present analysis, however, parts of these places research is needed to help inform policies and health sys- showed high level of private FBD. Our findings therefore tem responses to provide adequate health services that also tangentially shed light on people’s decision-making people will utilize. of the services to use from the options that are available to them. Such knowledge is useful for the formulation of Additional file appropriate interventions to concurrently address provision of and demand for services [58, 59]. In the Additional file 1: Figure S1. Nineteen SaTScan spatial clusters (drawn case of these states, additional provision of public health proportionate to cluster radii) of higher and lower than expected services might not be as effective a strategy to boost proportions of private facility birth among all most recent births. The DHS wards contained in each spatial clusters are also shown. (DOCX 130 kb) FBD as trying to strengthen the quality and acceptability of existing public services. Funding Conclusion This research is supported by funding from MSD, through its MSD for Mothers programme. MSD has no role in the design, collection, analysis, and In this study, we found an inverse relationship between interpretation of data, in the writing of manuscripts, or in decisions to community private care-seeking for childbirth and submit manuscripts for publication. The content of all publications is solely self-reported financial reasons of service non-uptake. the responsibility of the authors and does not represent the official views of MSD. MSD for Mothers is an initiative of Merck & Co., Inc., Kenilworth, N.J., This extends current understanding of the influence of U.S.A. OJB is supported by a Sir Henry Wellcome Fellowship funded by the financial barriers for maternity care. We argue that fur- Wellcome Trust (grant number 206471/Z/17/Z). ther investigation of determinants of maternal health-seeking, and potentially other health-seeking, Availability of data and materials should look beyond individual-level barriers to consider The dataset is available to the public freely at dhsprogram.com. Questionnaires used for the survey are attached to the final report published, community-level factors. Many LMICs continue to be which can be found at https://dhsprogram.com/pubs/pdf/FR293/FR293.pdf challenged by poor maternal health outcomes driven to (last accessed: 12th May 2018). some extent by wide subnational disparities in maternal healthcare provision, utilization and care quality. The Authors’ contributions lack of research and attention in the existing literature KW and OMRC conceptualised the study. KW conducted the analysis, to study community-level factor is possibly due to the developed the statistical methodology and approach, and prepared the first draft of the manuscript. ER contributed to drafts of the paper, interpretation lack of suitable data, especially since studies of determi- of the findings and revising of the paper. OO contributed to interpretation of nants of FBD are mostly based on individual- and the findings and revising of the paper. OB contributed to developing the household-level data. Working with geographic data and statistical methods, interpretation of the findings and revising of the paper. CL contributed to interpretation of the findings. LB contributed to GIS tools, including mapping techniques and spatial developing the statistical methods, drafts of the paper, interpretation of cluster detection, we developed a novel way to bridge the findings and revising of the paper. All authors read and approved this persistent knowledge gap. Our approach offers new the final manuscript. approaches to examine the way in which childbirth ser- vices frequented by the community influences commu- Ethics approval and consent to participate The DHS receive government permission and obtain informed consent from nity members to deliver in or outside a health facility. all participants. The Research Ethics Committee of the London School of Hygiene The method presented can be extended to other re- and Tropical Medicine approved our secondary analysis of anonymised data. search questions related to barriers and different health service characteristics, such as service acceptability and Consent for publication the level/standard of care most frequently sought, as well The consent to publish is not applicable for the current analysis as individual as perceived need, cultural drivers and social norms data is not reported. against overall utilization rate. Our approach also pre- serves spatial patterns in the data, a component that is Competing interests The authors declare that they have no competing interests. often neglected but requires specific analytical consider- ations and carries contextual significance, including pol- icy implications. Publisher’sNote Overall, we suggest that the approach presented to be Springer Nature remains neutral with regard to jurisdictional claims in published best for 1) illustrating the service utilization environment maps and institutional affiliations. in the population and 2) examining associations between Author details individual-level and community-level factors. The com- 1 Department of Infectious Disease Epidemiology, Faculty of Epidemiology plex reasons behind underutilization of delivery care ser- and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Guttmacher Institute, 125 Maiden vices indicates the need for a multi-focus approach that Lane 7th Floor, New York, NY 10038, USA. Centre for Mathematical addresses service provision and usage suited for the local Modelling for Infectious Diseases, London School of Hygiene and Tropical context of healthcare uptake and non-uptake. Further Medicine, Keppel Street, London WC1E 7HT, UK. Wong et al. BMC Health Services Research (2018) 18:397 Page 11 of 12 Received: 5 February 2018 Accepted: 22 May 2018 24. Kulldorff M, Nagarwalla N. Spatial disease clusters: detection and inference. Stat Med. 1995;14:799–810. 25. Kulldorff M. Geographic information systems (GIS) and community health: some statistical issues. J Public Health Manag Pract. 1999;5:100–6. 26. Kulldorff, M., Feuer, E. Miller, B. Breast cancer clusters in the northeast United References States: a geographic analysis. Am. J. (1997);146(2):161-170. at <http://aje. 1. Wekesah, F. M., Adedini, S. A. Osotimehin, B. Chimaraoke O. Izugbara. 2016 oxfordjournals.org/content/146/2/161.short> at <http://aphrc.org/wp-content/uploads/2016/05/Maternal-Health-in- 27. de Smith MJ, College London U. Statistical Analysis Handbook A Nigeria_Final-Report.pdf> comprehensive handbook of statistical concepts, techniques and software 2. Kana MA, Doctor HV, Peleteiro B, Lunet N, Barros H. Maternal and child tools. In: Edinburgh: The Winchelsea Press, Drumlin Security Ltd. 2018 at health interventions in Nigeria: a systematic review of published studies <http://www.statsref.com/StatsRefSample.pdf>. from 1990 to 2014. BMC Public Health. 2015;15:334. 28. Jackman, S. Models for Binary Outcomes and Proportions. (2007). at <http:// 3. NPC/Nigeria, N. P. C.- & International, I. Nigeria Demographic and Health citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.505.5574&rep= Survey 2013. (2014). at <http://dhsprogram.com/publications/publication- rep1&type=pdf> fr293-dhs-final-reports.cfm> 29. Crawley, M. J. The R book. (2012). at <https://www.wiley.com/en-us/The+R 4. Moyer CA, Mustafa A. Drivers and deterrents of facility delivery in sub- +Book%2C+2nd+Edition-p-9780470973929> Saharan Africa: a systematic review. Reprod Health. 2013;10:40. 30. Huber, P. in Berkeley Symposium on Mathematical Statistics and Probability 5. Kiwanuka SN, et al. Access to and utilisation of health services for the poor 221–223 (1967). in Uganda: a systematic review of available evidence. Trans R Soc Trop Med 31. Eberth JM, Vernon SW, White A, Abotchie PN, Coan SP. Accuracy of self- Hyg. 2008;102(11):1067–74. reported reason for colorectal Cancer testing. Cancer Epidemiol Biomark 6. Ikeako LC, et al. Influence of formal maternal education on the use of Prev. 2010;19:196–200. maternityservices in Enugu, Nigeria. J Obstet Gynaecol (Lahore). 2006;26(1):30–4. 32. Higgs BW, Mohtashemi M, Grinsdale J, Kawamura LM. Early detection of 7. Berhan Y, Berhan A. A meta-analysis of socio-demographic factors tuberculosis outbreaks among the San Francisco homeless: trade-offs predicting birth in health facility. Ethiop J Health Sci. 2014;24 Suppl:81–92. between spatial resolution and temporal scale. PLoS One. 2007;2:e1284. 8. Mills S, Williams JE, Adjuik M, Hodgson A. Use of health professionals for 33. Ozonoff A, Jeffery C, Manjourides J, Forsberg White L, Pagano M. Effect of delivery following the availability of free obstetric Care in Northern Ghana. spatial resolution on cluster detection: a simulation study. Int J Health Matern Child Health J. 2008;12:509–18. Geogr. 2007;6:52. 9. Kruk ME, Rockers PC, Mbaruku G, Paczkowski MM, Galea S. Community and 34. DO M. Utilization of skilled birth attendants in pubilc and private sectors in health system factors associated with facility delivery in rural Tanzania: a Vietnam. J Biosoc Sci. 2009;41:289. multilevel analysis. Health Policy (New York). 2010;97:209–16. 35. Barua A, et al. Implementing reproductive and child health Services in 10. Musoke D, Boynton P, Butler C, Musoke MB. Health seeking behaviour and Rural Maharashtra, India: a pragmatic approach. Reprod Health Matters. challenges in utilising health facilities in Wakiso district, Uganda. Afr Health 2003;11:140–9. Sci. 2014;14:1046–55. 36. Olusanya BO, Roberts AA, Olufunlayo TF, Inem VA. Preference for private 11. Onyeneho NG, Amazigo UV, Njepuome NA, Nwaorgu OC, Okeibunor JC. hospital-based maternity services in inner-city Lagos, Nigeria: An Perception and utilization of public health services in Southeast Nigeria: observational study. Health Policy (New York). 2010;96(3):210–6. implication for health care in communities with different degrees of 37. Benova L, Campbell OM, Ploubidis GB. A mediation approach to urbanization. Int J Equity Health. 2016;15:12. understanding socio-economic inequalities in maternal health-seeking 12. Edmonds JK, Hruschka D, Bernard HR, Sibley L. Women’s social networks behaviours in Egypt. BMC Health Serv Res. 2015;15:1. and birth attendant decisions: application of the network-episode model. 38. Onwujekwe O, Hanson K, Uzochukwu B. Examining inequities in incidence Soc Sci Med. 2012;74:452–9. of catastrophic health expenditures on different healthcare services and 13. Nwakeze NM, Kandala N-B. The spatial distribution of health health facilities in Nigeria. PLoS One. 2012;7(7):e40811. establishmentsin Nigeria. African Popul Stud. 2011;680–96. 14. Ajayi AI, Akpan W. Who benefits from free institutional delivery? Evidence 39. Onah HE, Ikeako LC, Iloabachie GC. Factors associated with the use of from a cross sectional survey of north central and southwestern Nigeria. maternity services in Enugu, southeastern Nigeria. Soc Sci Med. 2006;63: BMC Health Serv Res. 2017;17:620. 1870–8. 40. Ogunbekun I, Ogunbekun A, Orobaton N. Private health care in Nigeria: 15. Edu BC, Agan TU, Monjok E, Makowiecka K. Effect of free maternal health walking the tightrope. Health Policy Plan. 1999;14:174–81. care program on health-seeking behaviour of women during pregnancy, 41. Bohren MA, et al. Mistreatment of women during childbirth in Abuja, Intra-partum and Postpartum Periods in Cross River State of Nigeria: A Nigeria: a qualitative study on perceptions and experiences of women and Mixed Method Study. Open access Maced J Med Sci. 2017;5:370–82. healthcare providers. Reprod Health. 2017;14(1):9. 16. Abuja & Nigeria. NIGERIA DEMOGRAPHIC AND HEALTH SURVEY 2013 National 42. Akeju DO, et al. Determinants of health care seeking behaviour during Population Commission Federal Republic of Nigeria. (2014). at <https:// pregnancy in Ogun state, Nigeria. Reprod Health. 2016;13(Suppl 1):3. https:// dhsprogram.com/pubs/pdf/FR293/FR293.pdf> doi.org/10.1186/s12978-016-0139-7. 17. INEC Distribution of Senatorial Districts, Federal and State Constituencies, Electoral Wards, Polling States - NigerianMuse NigerianMuse 2016 https:// 43. Weis, J. Bedford, J. Quality care for mothers and babies at the time of birth www.nigerianmuse.com/20070414084834zg/sections/important-documents/ and immediate postpartum period national and state level barriers to inec-distribution-of-senatorial-districtsfederal-and-state-constituencies- scaling maternal and newborn care interventions in Nigeria. (2013). electoral. 44. Findley, S. et al. Changes in Maternal and Child Health Care Behaviors: Early 18. Burgert, C. R., Colston, J., Roy, T. Zachary, B. Geographic displacement evidence of the impact of community-based programs. (2012). at <http:// procedure and georeferenced data release policy for the Demographic and www.prrinn-mnch.org/documents/ Health Surveys. (2013). at <https://dhsprogram.com/publications/ ChangesinMaternalChildHealthCareBehaviors.pdf> publication-SAR7-Spatial-Analysis-Reports.cfm> 45. Corroon M, et al. The role of gender empowerment on reproductive health 19. Rutstein, S. O. Johnson, K. The DHS wealth index. (2004). at <http:// outcomes in urban Nigeria. Matern Child Health J. 2014;18:307–15. dhsprogram.com/publications/publication-CR6-Comparative-Reports.cfm> 46. Onokerhoraye, A. G. Access and Utilization of Modern Health Care Facilities in the Petroleum-producing Region of Nigeria: The Case of Bayelsa State. 20. Kulldorff M. A spatial scan statistic. Commun Stat - Theory Methods. 1997; 26:1481–96. (1999). at <https://cdn1.sph.harvard.edu/wp-content/uploads/sites/114/ 2012/10/rp162.pdf> 21. Kulldorff, M. SaTScan TM User Guide. (2015). at <http://www.satscan.org/> 22. The DHS Program User Forum: Sampling » What is the DHS position on the 47. Chiegeonu, A. S. Nigeria: State by State. (2006). at <https://books.google.co.uk/ use of cluster-level data? at 2016 <http://userforum.dhsprogram.com/index. books?id=IcmwBgAAQBAJ&printsec=frontcover&dq=nigeria+state+by+state&hl= php?t=msg&goto=9496&S=Google> en&sa=X&redir_esc=y#v=onepage&q=nigeria state by state&f=false>. 23. Kravdal Ø. A simulation-based assessment of the bias produced when using 48. Bayelsa State of Nigeria . Nigeria Information & Guide. (2017). at https:// averages from small DHS clusters as contextual variables in multilevel www.nigeriagalleria.com/Nigeria/States_Nigeria/Bayelsa/Bayelsa_State. models. Demogr Res. 2006;15:1–20. html Wong et al. BMC Health Services Research (2018) 18:397 Page 12 of 12 49. Olayinka OA, Achi OT, Amos AO, Chiedu EM. International journal of nursing and midwifery awareness and barriers to utilization of maternal health care services among reproductive women in Amassoma community. Bayelsa State. 2014;6:10–5. 50. Nakua EK, et al. Home birth without skilled attendants despite millennium villages project intervention in Ghana: insight from a survey of women’s perceptions of skilled obstetric care. BMC Pregnancy Childbirth. 2015;15:243. 51. Hailu D, Berhe H. Determinants of institutional childbirth service utilisation among women of childbearing age in urban and rural areas of Tsegedie district, Ethiopia. Midwifery. 2014;30:1109–17. 52. De Allegri M, et al. Determinants of utilisation of maternal care services after the reduction of user fees: a case study from rural Burkina Faso. Health Policy (New. York). 2011;99:210–8. 53. Wilunda C, et al. Determinants of utilisation of antenatal care and skilled birth attendant at delivery in south west Shoa zone, Ethiopia: a cross sectional study. Reprod Health. 2015;12:74. 54. Gitimu A, et al. Determinants of use of skilled birth attendant at delivery in Makueni, Kenya: a cross sectional study. BMC Pregnancy Childbirth. 2015;15:9. 55. McLaren ZM, Ardington C, Leibbrandt M. Distance decay and persistent health care disparities in South Africa. BMC Health Serv Res. 2014;14:541. 56. Lohela TJ, Campbell OMR, Gabrysch S, Mbewe R, Campbell O. Distance to care, facility delivery and early neonatal mortality in Malawi and Zambia. PLoS One. 2012;7:e52110. 57. Johnson FA, et al. Evaluating the impact of the community-based health planning and services initiative on uptake of skilled birth Care in Ghana. PLoS One. 2015;10:e0120556. 58. De Brouwere V, Richard F, Witter S. Access to maternal and perinatal health services: lessons from successful and less successful examples of improving access to safe delivery and care of the newborn. Trop Med Int Heal. 2010; 15:901–9. 59. O’Donnell O. Access to health care in developing countries: breaking down demand side barriers. Cad Saude Publica. 2007;23:2820–34. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Health Services Research Springer Journals

Why not? Understanding the spatial clustering of private facility-based delivery and financial reasons for homebirths in Nigeria

Free
12 pages

Loading next page...
 
/lp/springer_journal/why-not-understanding-the-spatial-clustering-of-private-facility-based-n0cSOosObd
Publisher
BioMed Central
Copyright
Copyright © 2018 by The Author(s).
Subject
Medicine & Public Health; Public Health; Health Administration; Health Informatics; Nursing Research
eISSN
1472-6963
D.O.I.
10.1186/s12913-018-3225-4
Publisher site
See Article on Publisher Site

Abstract

Background: In Nigeria, the provision of public and private healthcare vary geographically, contributing to variations in one’s healthcare surroundings across space. Facility-based delivery (FBD) is also spatially heterogeneous. Levels of FBD and private FBD are significantly lower for women in certain south-eastern and northern regions. The potential influence of childbirth services frequented by the community on individual’s barriers to healthcare utilization is under-studied, possibly due to the lack of suitable data. Using individual-level data, we present a novel analytical approach to examine the relationship between women’s reasons for homebirth and community-level, health-seeking surroundings. We aim to assess the extent to which cost or finance acts as a barrier for FBD across geographic areas with varying levels of private FBD in Nigeria. Method: The most recent live births of 20,467 women were georeferenced to 889 locations in the 2013 Nigeria Demographic and Health Survey. Using these locations as the analytical unit, spatial clusters of high/low private FBD were detected with Kulldorff statistics in the SatScan software package. We then obtained the predicted percentages of women who self-reported financial reasons for homebirth from an adjusted generalized linear model for these clusters. Results: Overall private FBD was 13.6% (95%CI = 11.9,15.5). We found ten clusters of low private FBD (average level: 0.8, 95%CI = 0.8,0.8) and seven clusters of high private FBD (average level: 37.9, 95%CI = 37.6,38.2). Clusters of low private FBD were primarily located in the north, and the Bayelsa and Cross River States. Financial barrier was associated with high private FBD at the cluster level – 10% increase in private FBD was associated with + 1.94% (95%CI = 1.69,2.18) in nonusers citing cost as a reason for homebirth. Conclusions: In communities where private FBD is common, women who stay home for childbirth might have mild increased difficulties in gaining effective access to public care, or face an overriding preference to use private services, among other potential factors. The analytical approach presented in this study enables further research of the differentials in individuals’ reasons for service non-uptake across varying contexts of healthcare surroundings. This will help better devise context-specific strategies to improve health service utilization in resource-scarce settings. Keywords: Spatial epidemiology, Clustering, Facility childbirth delivery, Maternal health service utilization, Financial barrier, Private health services * Correspondence: kerry.wong@lshtm.ac.uk Department of Infectious Disease Epidemiology, Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Wong et al. BMC Health Services Research (2018) 18:397 Page 2 of 12 Background dominations of lower-level and primary care and private Despite ongoing efforts by the Nigerian health system to health services in some areas but not others [13]. In increase maternal health service utilization, including addition, despite the Nigerian government’s aspiration to midwives service schemes, removal of user fees and in- provide free/subsidized maternity care in the public sec- creasing the involvement of the private sector [1, 2], tor, some women who stay home for childbirth reported population usage of many life-saving obstetric interven- cost or finance as a barrier to using maternity care, tions remains suboptimal. National statistics for 2009– among other factors [14, 15]. This raises questions re- 2013 show that, for instance, 22.6% of all births occurred garding current understanding of the factors for service in a public health facility and 13.2% in the private sector uptake vs. non-uptake in relation to one’s healthcare sur- – leaving approximately two thirds of childbirths based roundings. In some settings, e.g. where public maternity outside of a health facility [3]. At the subnational scale, care is free of charge, it is likely that some of those who likelihoods for facility-based delivery (FBD) and private stay home for childbirth for financial reasons only con- FBD vary considerably, and were significantly lower for sidered using private services, the alternative being women residing in parts of the South South zone, and homebirths (over public care). This speculation might be majority of the Northern zone [3]. more pertinent where private FBD is common, due to Both in Nigeria and other low- and middle-income the potential impact that one’s peers and healthcare sur- countries (LMICs), having a FBD is a practical way to roundings have on their reasons for service non-uptake. ensure assistance by a skilled birth attendant and ac- The aim of this study is to assess the extent to which cost cess to life-saving interventions for mothers and new- or finance is a barrier for FBD across geographic areas with born [4]. Previous reviews addressing factors related varying levels of private FBD in Nigeria. To overcome the to FBD in sub-Saharan Africa and other LMICs have limitation of community-level data availability, we present identified an array of determinants [4–7]. Moyer and an innovative approach applying geographic information Mustafa’s literature review, published in 2013, system (GIS) tools to examine the clustering of maternity highlighted an overwhelming reliance on population/ care utilization using individual-level survey data. This survey data with which maternal sociodemographic study will help motivate and enable further investigation of factors were well-represented [4]. The limited body of the way in which childbirth services frequented by the literature around community-level factors of FBD in community influences community members to deliver in LMICs emphasizes community socio-demographic or outside a health facility, adding contribution to the characteristics, community views on skilled and trad- current effort to support maternity care utilization for itional births [8, 9], service accessibility such as dis- groups and individuals most “left behind”. tance to care and community uptake of antenatal care [4]. Communities likely have other unique characteris- Methods tics that influence demand for and supply of health- Data and study sample care [10], many of which are overlooked. This analysis was based on data from the 2013 Nigeria Unlike other health service seeking, childbirth can Demographic and Health Survey (NDHS). The data is happen unexpectedly throughout the day and the representative at the national level, of the six geopolitical woman may need to reach a nearby care provider at zones and of the 36 states and the Federal Capital Terri- relatively short notice. The types of childbirth delivery tory (FCT-Abuja). The survey sample was selected using services most accessible to, or most accessed by, the a stratified multi-stage cluster probability sampling de- community directly relate to an individual’s perception sign with census ward as the primary sampling unit. As of, wishes for, and actual uptake of services. Women also part of the DHS sampling procedure, all households in exchange information and experience surrounding child- each sampled ward was enlisted, which was then used as birth in social settings, and one’s planning for future de- the sampling frame for household selection [16]. Eligible livery may be conditioned by assessing factors important individuals aged 15–49 in selected households were to their peers, culture and community [11, 12]. A better interviewed with a standardized questionnaire. The final understanding of one’s healthcare surroundings is im- sample of the 2013 NDHS consisted 896 census wards perative to developing effective strategies to increase and 39,902 eligible women; 98% (38,948) were success- healthcare utilization among groups currently “left be- fully interviewed. Women with a live birth in the five hind”. Part of the dearth of research in this area might years before the interview were asked to self-report the be due to the lack of suitable data, especially at the na- care received during pregnancy and delivery. The sample tional scale. of the current analysis was restricted to the circum- In a study of the characteristics of health facilities stances of 20,467 women’s most recent live birth during across Nigeria, Nwakeze and Kandala found vast geo- the five-year survey recall period as some of the required graphic disparities in the country, including greater data was only collected for this subsample. Wong et al. BMC Health Services Research (2018) 18:397 Page 3 of 12 Geography and administration of Nigeria or governmental health facilities (HFs), private or Nigeria is divided into six geopolitical zones (Fig. 1): non-governmental HFs, as well as all other unspecified North Central, North East, North West, South West, locations [3]. FBD was obtained by coding responses as South South and South West; and within these zones, “any HF” and “not in a HF”. For the analysis of private into 36 states and the FCT-Abuja. For administrative FBD, all births were categorized as “any private HF” purposes, the states are subdivided into 774 local gov- and “not in a private HF”. We note that the 2013 ernments areas [3], each made up of approximately 10– NDHS had conflated all non-governmental, for-profit 15 wards [17]. and not-for-profit providers as one category of “private” provider. The outcome of interest was financial barrier for Measurement FBD. Women who did not deliver in a HF indicated Population centroids of wards, recorded as latitude and the reasons that applied to them from a list of po- longitude, were obtained by DHS enumerators using tential barriers, including “cost too much”. Other co- Global Position System (GPS) receivers [18]. All individ- variates and demographic information considered as uals residing in a ward therefore have the same georefer- potential confounders were: wealth quintile, maternal ence. For privacy considerations, the coordinates were education, maternal age and parity at the time of the randomly displaced by up to 2 km in urban areas and up most recent birth, and whether the woman had to 5 km in rural areas by the NDHS. An additional 1% health insurance coverage. Household wealth quintile of rural wards were displaced by 10 km. was derived from the wealth index – constructed by Delivery location was based on women’sanswerto: the DHS using household asset data via a principal “Where did you give birth to [name of child]?” on the component analysis [19]. The sampled households Women’s Questionnaire. The major categories of were than ranked and divided into five quintiles. response were domestic environments (home of re- Each woman is assigned her household’swealth spondent or of traditional birth assistant (TBA)), public quintile. Fig. 1 Map of Nigeria showing boundaries of six geopolitical zones, 36 states and Federal Capital Territory (FCT-Abuja). Shapefile is obtained from gadm.org. The 2018 GADM license allows data re-use for academic and other non-commercial purposes (https://gadm.org/license.html, last accessed: 14th May 2018) Wong et al. BMC Health Services Research (2018) 18:397 Page 4 of 12 Spatial scan statistics of private FBD outcome) was related to the percentage of births occur- To identify geographic clusters of high and low private ring in private facilities. The SatScan spatial clusters FBD, the number of most recent births and those based were weighted by the number of most recent births cir- in a private HF were aggregated at the ward level, with cled within. To account for a proportion as outcome adjustment of survey sampling weighting. Together with (bounded between 0 and 100%), we adopted a general- ward latitude and longitude as inputs, each ward was ized linear model, specifying a logit link and the bino- treated as an analytical unit to test whether private FBDs mial family [27–29]: were distributed randomly in space or not. logitðÞ p ¼ xβ þ ε where Y  BinðÞ N ; p ; ε At the ward level, the observed numbers of private i i i i i i FBD varied from zero to the total number of eligible  N0; σ births. To detect clustering of private FBDs, we chose a Poisson distribution to represent the expected distribu- We denoted y = number of private FBD in SaTScan tion of this count over space. Under the null hypothesis, spatial cluster i, N = number of most recent births in i the expected number of private FBDs in each area is and p = probability of having a private FBD. We also proportional to its population size (approximated using specified the Huber-White (i.e. robust) estimators of the sample size) [20, 21]. Spatial scan statistics was per- standard errors in case of heteroskedasticity arising from formed using the SaTScan™ software (version 9.4) [20, potential misspecification in the distribution family [30]. 21]. Spatial clusters were identified by taking into con- The z test was used for significance testing of model co- sideration the rates of nearby wards [22, 23]. The spatial efficients. We generated predictions from both the bi- scan method used circular windows of various sizes that variate and multivariate fits and back-transformed them move across the map to find clusters of wards with ei- as the percentages of women with a non-facility birth ther higher and lower than expected rates under the null who cited cost was a barrier at 5%-intervals of hypothesis of uniform spatial distribution [24, 25]. The community-level private FBD. radius of the circle varies continuously from zero to a predefined value that specified the percentage of the Missing data maximum total population at risk within the scanning We found missing data in geographic coordinates in window [21]. The recommended maximum size is 50%; seven wards, containing < 1% of the respondents from we conducted additional scans at the 10 and 5% levels to the study sample. These were removed from analyses account for independent smaller clusters that may be where location data was required. We also found 0.4% of contained in a large cluster. The alternative hypothesis is missing data for health insurance coverage and coded that there is a reduced/elevated rate within the scanning these as uninsured. There was no missing data in the window as compared to outside. The test of significance, other variables in the model. based on likelihood ratio and the null distribution, was obtained from Monte Carlo Simulation [26]. The num- Results ber of permutations was set at 999 and the significance Facility-based delivery level was set at 0.05 [21]. Identified clusters are ordered Of the 20,467 births in our sample, 7649 (37.4, 95%CI = based on their likelihood ratio test values. 34.7,40.2) occurred in health facilities: 23.8% (95%CI = Geographic locations of, and wards contained in each, 22.0,25.5) in public and 13.6% (95%CI = 11.9,15.5) in pri- identified spatial cluster were merged back to the 2013 vate facilities. More of those who were rural residents, NDHS women’s data. We considered women living in from the poorest wealth quintile, without any education, the same SaTScan spatial cluster to be in the same uninsured and having a second or higher order birth de- “community”. Estimates on private service use as a per- livered outside of a health facility (Table 1). Geographic centage of all most recent births, financial barriers re- variations of FBD were observed – highest in the South ported among women who did not deliver in a HF, as East zone (78.5, 95%CI = 73.2,83.0) and lowest in the well as other covariates were recalculated for each SaTS- North West zone (12.8, 95%CI = 10.2,15.9). can spatial cluster to generate community-level data. This was done in Stata SE version 14 (StataCorp LP, Col- Sub-national private facility-based delivery lege Station, TX, USA), adjusted for survey-specific Regional averages of private FBD varied between 0.5% weighting and stratified, cluster sampling design. (95%CI = 0.3,1.1) in North West zone to 44.8% (95%CI = 38.4,51.4) in the South East zone (Table 1). Using SaTS- Relating community -level private facility use to nonusers’ can analysis, ten spatial clusters of low level and seven self-reporting of financial barrier spatial clusters of high private FBD were identified Using SaTScan spatial cluster as the analytical unit, the (Table 2). The number of wards contained in these geo- percentage of nonusers reporting financial barrier (main graphic clusters ranged from five to 88; the number of Wong et al. BMC Health Services Research (2018) 18:397 Page 5 of 12 Table 1 Percentage distribution and 95% confidence intervals of sample sociodemographic characteristics by place of delivery Number of most Place of delivery recent births Outside of a health facility Public health facility Private Health facility N proportion 20,467 12,818 5100 2620 (100) 62.6 (59.8,65.3) 23.8 (22.0,25.6) 13.6 (11.9,15.5) Area of residence Urban 6790 36.8 (32.6,41.2) 36.3 (33.5,36.3) 26.9 (23.2,30.8) Rural 13,402 76.9 (74.2,79.4) 16.8 (15.0,18.8) 6.3 (5.2,7.7) Wealth quintile Poorest 4379 93.8 (92.4,95.0) 5.0 (4.1,6.1) 1.2 (0.8,1.9) Poorer 4603 81.8 (79.2,84.1) 13.4 (11.7,15.3) 4.8 (3.7,6.2) Middle 4069 62.2 (58.7,65.5) 26.6 (24.0,29.3) 11.3 (9.4,13.4) Richer 3798 42.5 (38.9,46.1) 39.4 (36.5,42.3) 18.2 (15.7,20.9) Richest 3343 18.6 (16.2,21.3) 42.5 (38.7,46.3) 38.9 (34.3,43.7) Maternal education No education 9171 88.0 (86.4,89.5) 10.2 (8.9,11.6) 1.8 (1.4,2.3) Primary 4113 57.1 (54.0,60.3) 27.5 (25.2,29.9) 15.4 (13.4,17.6) Secondary 5565 33.8 (31.1,36.5) 39.0 (36.4,41.6) 27.2 (24.0,30.7) Higher 1343 8.3 (8.3,10.7) 51.3 (46.7,55.9) 40.4 (35.5,45.5) Health insurance Yes 363 14.7 (10.4,20.4) 50.6 (43.8,57.4) 34.7 (27.8,42.3) No 19,829 63.4 (60.6,66.1) 23.3 (21.6,25.1) 13.3 (11.6,15.1) Parity First birth 3624 51.7 (48.3,55.0) 31.4 (28.9,34.0) 16.9 (14.6,19.5) Higher order birth(s) 16,568 65.0 (62.2,67.7) 22.1 (20.4,23.9) 12.9 (11.3,14.7) Geopolitical zones North Central 3095 53.0 (47.4,58.5) 31.3 (27.7,35.2) 15.7 (12.7,19.3) North East 4001 79.5 (74.9,83.4) 19.2 (15.5,23.5) 1.3 (0.8,2.1) North West 6206 87.2 (84.1,89.8) 12.3 (9.8,15.2) 0.5 (0.3,1.1) South East 1724 21.5 (17.0,26.8) 33.7 (28.9,38.9) 44.8 (38.4,51.4) South South 2500 49.2 (44.1,54.4) 36.6 (32.6,40.7) 14.2 (10.8,18.3) South West 2666 23.8 (19.3,29.0) 23.8 (31.9,40.6) 40.1 (35.2,45.1) Age at birth Mean (interquartile range) 29.42 (29.2,29.6) Self-reported financial barrier to deliver in a health facility 9.1 (8.9,10.5) Not applicable Not applicable most recent births circled within a geographic cluster 0.8,0.8), respectively. Average public FBD among all births ranged between 63 and 1201, and spatial cluster radii was 37.2% (95%CI = 37.1,37.4) in high private FBD clus- varied between 21.2 and 208.5 km. Altogether, 648 wards ters. On the other hand, 14.8% of all births were public and 14,434 births occurred in these 17 clusters. facility-based in the ten spatial clusters of low private The location and size of these geographic clusters were FBD. Substantial differences in sociodemographic charac- drawn in Fig. 2. Clusters of low private FBD were pri- teristics of women living in the two groups of spatial clus- marily located in the North West and North East zones, ters were also seen (Fig. 2). Women in low private FBD with an exception near Jos North in Plateau State, where clusters were more rural, poorer and less educated com- one spatial cluster of high private FBD (50.5, 95%CI = pared to women in high private FBD clusters. 35.5,65.5) was identified. In addition, southern Cross We performed additional cluster detections setting River state and central and southern Bayelsa state (in the maximum cluster size to 10 and 5% of the survey sam- South South zone) also showed spatial clustering of low ple. The first yielded the same set of results. The details private FBD: 2.9% (95%CI = 1.0,4.7) and 2.1% (95%CI = of the 19 SaTScan spatial clusters returned from using 0.0,5.5), respectively. Communities of high private FBD the 5% limit is given in Additional file 1: Figure S1. No were identified around the Lagos and Ogun States (52.8, substantial differences to the model with 17 SatScan 95%CI = 47.7,57.9), Edo State (32.9, 95%CI = 24.7,41.1) as clusters were observed. well as large parts of the South-East zone (e.g., Imo and Abia States) and the North Central zone. Reporting cost as a barrier for facility-based delivery Mean percentages of private FBD in high and low clus- Across the seven spatial clusters of high private FBD, ters were 37.9% (95%CI = 37.6,38.2) and 0.8% (95%CI = 24.5% (95%CI = 21.1,27.8) of women delivered at home Wong et al. BMC Health Services Research (2018) 18:397 Page 6 of 12 Table 2 Seventeen significantly higher and lower than expected proportions of FBD spatial clusters ID Cluster location No. of No. of most Observed Observed Expected number Relative p-value wards circled recent births number % private of private FBD risk Latitude Longitude Radius (km) of private FBD FBD under H High 1 6.7 3.7 97.2 75 1182 605 52.8 154 4.82 < 0.001 2 9.9 8.9 21.2 5 63 33 50.5 8 4.06 < 0.001 3 5.8 7.2 85.2 88 1199 569 48.9 156 4.38 < 0.001 4 6.7 5.4 78.7 28 457 146 32.9 60 2.54 < 0.001 5 8.5 4.6 132.5 67 1198 338 28.0 156 2.34 < 0.001 6 7.3 9.0 79.4 12 249 73 27.4 32 2.29 < 0.001 7 7.9 7.1 148.3 78 1200 295 25.3 156 2.00 < 0.001 Low 8 9.8 9.7 64.0 8 233 8 3.1 30 0.26 0.014 9 5.3 8.6 75.9 17 280 8 2.9 37 0.22 < 0.001 10 4.6 5.7 88.6 24 558 5 2.1 73 0.07 < 0.001 11 8.7 11.6 178.0 41 1187 22 1.8 155 0.14 < 0.001 12 10.8 7.3 155.6 38 1174 14 1.4 153 0.09 < 0.001 13 10.8 3.9 184.8 24 770 7 0.5 100 0.07 < 0.001 14 13.3 8.0 144.9 31 1201 1 0.1 156 0.01 < 0.001 15 12.0 12.5 208.5 46 1197 2 0.1 156 0.01 < 0.001 16 11.7 9.3 84.7 30 1085 1 0.1 141 0.01 < 0.001 17 13.2 5.5 145.6 36 1201 0 0.0 156 0.00 < 0.001 FBD = facility based delivery; H = null hypothesis of spatial randomness The likelihood ratio test is used for testing cluster significance Cluster 1 is the most likely cluster; all other clusters are non-overlapping secondary clusters Relative risk of private FBD within cluster compared to the risk in all other areas and 14.9% (95%CI = 14.7,15.1) reported cost as a barrier showed that the factors associated with self-reported fi- (Fig. 2). In contrast, 85.7% (95%CI = 83.2,88.2) of women nancial barrier for FBD at the spatial cluster unit level living in the 10 clusters with low FBD delivered at home, included living in Cross River and Bayelsa States, the and 8.8% (95%CI = 8.6,8.9) cited cost as a barrier. Fig- percentage of public facility utilization, rural setting, ure 3 illustrates that in contrast to other spatial clusters wealth, the level of maternal education, and the percent- of low private FBD, exceptionally high proportions of age of women covered by health insurance (Table 3). All nonusers living in Cross River (32.7, 95%CI = 26.3,39.1) of these were significant at the p < 0.001 level. and Bayelsa State (25.2, 95%CI = 19.1,31.4) said cost was In multivariate analysis, all predictors remained signifi- a reason to deliver outside a facility. Unadjusted analysis cantly associated with the proportion of women Fig. 2 Seventeen SaTScan spatial clusters (drawn proportionate to cluster radii) of higher (red) and lower (blue) than expected proportions of private facility birth among all most recent births. The DHS wards contained in each spatial clusters are also shown Wong et al. BMC Health Services Research (2018) 18:397 Page 7 of 12 Fig. 3 Proportions of women delivering outside a health facility who self-reported financial barrier as a reason for homebirth in 17 spatial clusters of high and low private facility births. Predicted percentages and confidence intervals at various levels of private facility birth from an adjusted generalized linear model weighted by numbers of most recent births in spatial clusters are also shown (represented by size of bubbles) reporting financial barrier in the community (Table 3). and Abia States had particularly high levels of private After controlling for the proportion of birth occurring in FBD. Using a novel approach, we examined the associ- public HFs, rurality, wealth, maternal education, health ation between private healthcare utilization contexts and insurance and residency in Cross River and Bayelsa financial barriers for FBD. We found cost was more States, a 10% point increase in private facility use for likely to be cited as a barrier to FBD in settings where childbirth was associated with an average 1.94% point private FBD was high. We found exceptions, however, increase (95%CI = 1.69,2.18) in nonusers citing cost as a for southern Cross River and Bayelsa States, where a barrier for FBD. The adjusted predicted percentages of large proportion of nonusers reported cost as a barrier self-reported financial barrier across varying levels of and overall facility delivery (in both the public and pri- private service use were also computed based on the ad- vate sectors) very low. justed regression model. Table 3 and Fig. 3 illustrate a steady rise in the extent to which financial consideration Limitations was a barrier as community-level private FBD increased. Our findings have important implications, but they should be understood with certain limitations. Firstly, Discussion the 2013 NDHS response option for private delivery in- To our knowledge, this is the first study to examine na- cluded both for-profit and not-for-profit establishments tional geographic disparities in private facility use for operating under different financial motives and poten- childbirth in a sub-Saharan African country at a small tially charging widely varying fees for childbirth care. geographic scale. We found substantial spatial variation However, we still believe that our assumption that pri- in the utilization of private facilities for delivery care vate sector childbirth costs more than public sector is across Nigeria. The level of private FBD was very low in valid. Self-reported reasons to deliver in non-healthcare the northern part of the country except for Jos in Plat- settings might also be subject to accuracy and reliability eau State. Private FBD was medium to high in North issues [31]. In addition, women could list more than one Central zone and the highest in the South West and barrier of FBD – approximately 50% of women who South East zones. Certain areas in Lagos, Imo, Ogun cited cost as a barrier also listed one or more other Wong et al. BMC Health Services Research (2018) 18:397 Page 8 of 12 Table 3 Effect sizes of predictor variables and estimates of proportion citing financial barriers Community-level factors Unadjusted estimates Adjusted estimates Average change in proportion of nonusers citing financial barriers with 95%CI and p-value Private facility delivery (every + 10%) 1.82 (1.79,1.86) < 0.001 1.94 (1.69,2.18) < 0.001 Public facility delivery (every + 10%) 1.17 (1.14,1.20) < 0.001 −1.64 (−1.88,-1.41) < 0.001 Rural sample (every + 10%) −1.20 (−1.24,-1.16) < 0.001 0.66 (0.54,0.78) < 0.001 Wealth: Q1 sample (every + 10%) −2.32 (2.37,2.27) < 0.001 0.50 (0.31,0.70) < 0.001 No to primary education (every + 10%) −1.83 (−1.86,-1.80) < 0.001 − 1.61 (− 1.78,-1.43) < 0.001 Health insurance (every + 10%) 21.8 (21.0,22.7) < 0.001 3.58 (2.11,5.05) < 0.001 Geographic location Others Reference Reference Cross River and Bayelsa 17.61 (17.34,17.87) < 0.001 17.34 (15.74,18.94) < 0.001 % of births in private facility Predicted percentage of nonusers citing financial barriers as reason for homebirth with 95%CI * 0 8.23 (8.08,8.37) 7.48 (7.20,7.76) 5 8.96 (8.82,9.10) 8.22 (8.00,8.43) 10 9.75 (9.62,9.89) 9.02 (8.88,9.16) 15 10.60 (10.48,10.73) 9.90 (9.81,9.99) 20 11.52 (11.40,11.64) 10.85 (10.69,11.01) 25 12.51 (12.40,12.62) 11.88 (11.59,12.17) 30 13.57 (13.46,13.68) 12.99 (12.55,13.44) 35 14.70 (14.60,14.81) 14.19 (13.56,14.83) 40 15.92 (15.80,16.02) 15.49 (14.64,16.33) 45 17.21 (17.08,17.33) 16.87 (15.79,17.96) 50 18.58 (18.42,18.73) 18.36 (17.00,19.71) 55 20.03 (19.85,20.22) 19.94 (18.29,21.59) 60 21.57 (21.35,21.80) 21.62 (19.66,23.59) ^ Unadjusted and adjusted effects were back-transformed from parameter estimates obtained using a logit link transformation. The z test was used for significance testing of model coefficients + Adjusted estimates describe the adjusted curve drawn in Fig. 4 *Adjusted predicted percentage of nonusers citing financial barriers were obtained with all other coverages fixed their mean values reasons (data not shown) – and the relative importance geographic coordinates of individuals and those at of cost compared to other reasons is not known. Contri- aggregated levels [32, 33]. butions of other potential factors – including, but not limited to individuals’ perceptions towards the care re- Giving birth in the private sector ceived and healthcare professionals – warrants further In Nigeria and other LMICs, pregnant women who opt investigation. The analytical approach presented in this for private FBD have a similar sociodemographic profile study offers a novel method for such future research – higher SES, higher education and, in some contexts, with available, secondary data. certain ethnicity or religious affiliations [34–37]. A The SaTScan spatial clusters identified were rela- search of peer-reviewed articles and the grey literature tively large in geographical size (even with a smaller returned little information on the cost of private FBD in maximum allowable limit), and there might be sub- Nigeria. However, a study showed 1.8 times more spend- stantial heterogeneity in the characteristics of the ing in private hospitals than public hospitals by users women living in the same spatial cluster. Some of residing in urban south-eastern Nigeria [38]. Despite this heterogeneity, including parity, pregnancy com- higher cost, for-profit healthcare care may have more plication and marital status, may confound our pri- appeal due to a wide range of reasons, such as privacy, mary association of interest at the individual-level, shorter waiting times, higher perceived quality of care, butwereomittedastheir relevance atthe commu- empathy and respectful approach, availability of doctors nity level is likely low. Lastly, some loss of power in and as a status symbol [39, 40]. For users of private ser- cluster detection might have occurred through a vices, cost or affordability might be a relatively weaker degradation of spatial information between the exact determinant of where to seek care. Wong et al. BMC Health Services Research (2018) 18:397 Page 9 of 12 Community-level private service use and self-reported highlights the importance of contextualizing personal financial barriers for facility-based delivery factors alongside other community- or macro-level Our findings extend the current knowledge about prefer- factors. Bayelsa State is primarily covered by marshlands ence towards private HFs for their users. We found that and waterways; it is also an important gas- and in contexts with relatively high private FBD, a greater petrol-producing region in Nigeria that has generated proportion of facility non-users reported financial bar- interest among prospective companies [46, 47]. However, riers for any care, including both private care and the most Bayelsans remain poor, and the state’s public infra- relatively more affordable public care. In Edo, Ogun and structure development insufficient [47]. Lack of trans- Abia States, for instance, the majority of health facilities portation and the riverine setting pose tremendous are privately owned [13]. Our results may indicate that impediments to overcoming physical barriers to reaching facility nonusers living in high FBD contexts are unable health services [46, 48]. In a study looking at barriers to to gain effective access to any healthcare due to personal utilization of maternal health services in Bayelsa State, a financial barrier (for private care) and insufficient majority of respondents reported infrastructure-related provision of public services in their lived environment. barriers to access (availability of facilities/equipment, In other places of high private FBD where such practice schedule of maternal health clinic, accessibility and so may have become normalized, women who lack ad- on); and much lower percentages of women reported de- equate funds for private providers might perceive deliv- terrents such as cultural acceptance and language prob- ering at home or a TBA’s home as their best alternative lems [49]. Compared to the rest of the country, special due to social pressure and low acceptability of publicly economic and environmental contexts and the additional provided services. The observed preference for home- resources required to overcome physical accessibility births is in line with qualitative findings from various barriers may have caused financial considerations to op- states including FCT-Abuja and Lagos, where women erate differently among people living in Bayelsa and who do not deliver in a health facility had poor confi- Cross River States. The role of financial barriers, separat- dence in the public health sector and strong desires to ing direct payment for delivery from other expenses and deliver with a TBA [41–43]. According to these studies, trade-offs, including cost of transport, as well as time women perceive home delivery with a TBA, and espe- and financial lost from other daily/productive activities, cially with family members present, to be personal and warrants further research. supporting [41]. Some TBAs often allow for flexible fi- nance options, such as payment in kind or in instal- A note on using DHS data to study healthcare utilization ments, making it easier for families to pay [42]. surroundings On the other hand, in settings where private facility de- Various studies have looked at the service provision en- livery use is relatively low, and especially where overall vironment as a determinant of FBD. A common ap- FBD utilization is also low, such as most of North West proach consists of conducting interviews with women zone and North East zone, women’s reasons to not give about the availability of maternity care in their commu- birth in a HF were less connected to cost. In these set- nity as a measure of service provision [50–54]. Alterna- tings, other cited barriers included service availability, dis- tively, geocoded master facility list (MFL) data or the tance or physical accessibility, social norms and lack of like, with which the entire health infrastructure of a perceived need [43]. In a study set in the Jigawa State, ap- spatial area is mapped out, are geographically linked to proximately 25% of nonusers claimed they did not attend population data in a GIS to facilitate calculation of mea- facilities for childbirth because they did not think it was sures of people’s healthcare availability [55–57]. The necessary [44]. In addition, household decision-making present study used available secondary data on dynamics also varies across this large multi-ethnic coun- individual-level service utilization and women’s location try; Abuja city/FCT-Abuja, for instance, is generally asso- of residence to construct the geographic patterning of ciated with greater gender equality when compared to healthcare surroundings across Nigeria. Our variable of other southern and northern cities [45]. Especially in the interest was community-level utilization surrounding the north, women’s relative lack of participation in individuals, which is somewhat conditioned on health- intra-household decision making and access to money care provision environment, but is also a consequence of have been associated with very low FBD rates [45]. other cultural, contextual and individual-level determi- Exceptions to the inverse relationship found between nants. Nwakeze and Kandala examined the spatial distri- financial barriers and private FBD were noted in south- bution of health establishments using data collected by ern Cross River State and Bayelsa State, where overall the National Bureau of Statistics of Nigeria, and found percentages of FBD were midrange, private FBD very moderate to low numbers of private health establish- low, and a relatively large proportion of nonusers re- ments in the Benue, Nasarawa and Kogi States, com- ported financial barriers to delivering in a HF. This pared to the number of public health facilities [13]. In Wong et al. BMC Health Services Research (2018) 18:397 Page 10 of 12 the present analysis, however, parts of these places research is needed to help inform policies and health sys- showed high level of private FBD. Our findings therefore tem responses to provide adequate health services that also tangentially shed light on people’s decision-making people will utilize. of the services to use from the options that are available to them. Such knowledge is useful for the formulation of Additional file appropriate interventions to concurrently address provision of and demand for services [58, 59]. In the Additional file 1: Figure S1. Nineteen SaTScan spatial clusters (drawn case of these states, additional provision of public health proportionate to cluster radii) of higher and lower than expected services might not be as effective a strategy to boost proportions of private facility birth among all most recent births. The DHS wards contained in each spatial clusters are also shown. (DOCX 130 kb) FBD as trying to strengthen the quality and acceptability of existing public services. Funding Conclusion This research is supported by funding from MSD, through its MSD for Mothers programme. MSD has no role in the design, collection, analysis, and In this study, we found an inverse relationship between interpretation of data, in the writing of manuscripts, or in decisions to community private care-seeking for childbirth and submit manuscripts for publication. The content of all publications is solely self-reported financial reasons of service non-uptake. the responsibility of the authors and does not represent the official views of MSD. MSD for Mothers is an initiative of Merck & Co., Inc., Kenilworth, N.J., This extends current understanding of the influence of U.S.A. OJB is supported by a Sir Henry Wellcome Fellowship funded by the financial barriers for maternity care. We argue that fur- Wellcome Trust (grant number 206471/Z/17/Z). ther investigation of determinants of maternal health-seeking, and potentially other health-seeking, Availability of data and materials should look beyond individual-level barriers to consider The dataset is available to the public freely at dhsprogram.com. Questionnaires used for the survey are attached to the final report published, community-level factors. Many LMICs continue to be which can be found at https://dhsprogram.com/pubs/pdf/FR293/FR293.pdf challenged by poor maternal health outcomes driven to (last accessed: 12th May 2018). some extent by wide subnational disparities in maternal healthcare provision, utilization and care quality. The Authors’ contributions lack of research and attention in the existing literature KW and OMRC conceptualised the study. KW conducted the analysis, to study community-level factor is possibly due to the developed the statistical methodology and approach, and prepared the first draft of the manuscript. ER contributed to drafts of the paper, interpretation lack of suitable data, especially since studies of determi- of the findings and revising of the paper. OO contributed to interpretation of nants of FBD are mostly based on individual- and the findings and revising of the paper. OB contributed to developing the household-level data. Working with geographic data and statistical methods, interpretation of the findings and revising of the paper. CL contributed to interpretation of the findings. LB contributed to GIS tools, including mapping techniques and spatial developing the statistical methods, drafts of the paper, interpretation of cluster detection, we developed a novel way to bridge the findings and revising of the paper. All authors read and approved this persistent knowledge gap. Our approach offers new the final manuscript. approaches to examine the way in which childbirth ser- vices frequented by the community influences commu- Ethics approval and consent to participate The DHS receive government permission and obtain informed consent from nity members to deliver in or outside a health facility. all participants. The Research Ethics Committee of the London School of Hygiene The method presented can be extended to other re- and Tropical Medicine approved our secondary analysis of anonymised data. search questions related to barriers and different health service characteristics, such as service acceptability and Consent for publication the level/standard of care most frequently sought, as well The consent to publish is not applicable for the current analysis as individual as perceived need, cultural drivers and social norms data is not reported. against overall utilization rate. Our approach also pre- serves spatial patterns in the data, a component that is Competing interests The authors declare that they have no competing interests. often neglected but requires specific analytical consider- ations and carries contextual significance, including pol- icy implications. Publisher’sNote Overall, we suggest that the approach presented to be Springer Nature remains neutral with regard to jurisdictional claims in published best for 1) illustrating the service utilization environment maps and institutional affiliations. in the population and 2) examining associations between Author details individual-level and community-level factors. The com- 1 Department of Infectious Disease Epidemiology, Faculty of Epidemiology plex reasons behind underutilization of delivery care ser- and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Guttmacher Institute, 125 Maiden vices indicates the need for a multi-focus approach that Lane 7th Floor, New York, NY 10038, USA. Centre for Mathematical addresses service provision and usage suited for the local Modelling for Infectious Diseases, London School of Hygiene and Tropical context of healthcare uptake and non-uptake. Further Medicine, Keppel Street, London WC1E 7HT, UK. Wong et al. BMC Health Services Research (2018) 18:397 Page 11 of 12 Received: 5 February 2018 Accepted: 22 May 2018 24. Kulldorff M, Nagarwalla N. Spatial disease clusters: detection and inference. Stat Med. 1995;14:799–810. 25. Kulldorff M. Geographic information systems (GIS) and community health: some statistical issues. J Public Health Manag Pract. 1999;5:100–6. 26. Kulldorff, M., Feuer, E. Miller, B. Breast cancer clusters in the northeast United References States: a geographic analysis. Am. J. (1997);146(2):161-170. at <http://aje. 1. Wekesah, F. M., Adedini, S. A. Osotimehin, B. Chimaraoke O. Izugbara. 2016 oxfordjournals.org/content/146/2/161.short> at <http://aphrc.org/wp-content/uploads/2016/05/Maternal-Health-in- 27. de Smith MJ, College London U. Statistical Analysis Handbook A Nigeria_Final-Report.pdf> comprehensive handbook of statistical concepts, techniques and software 2. Kana MA, Doctor HV, Peleteiro B, Lunet N, Barros H. Maternal and child tools. In: Edinburgh: The Winchelsea Press, Drumlin Security Ltd. 2018 at health interventions in Nigeria: a systematic review of published studies <http://www.statsref.com/StatsRefSample.pdf>. from 1990 to 2014. BMC Public Health. 2015;15:334. 28. Jackman, S. Models for Binary Outcomes and Proportions. (2007). at <http:// 3. NPC/Nigeria, N. P. C.- & International, I. Nigeria Demographic and Health citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.505.5574&rep= Survey 2013. (2014). at <http://dhsprogram.com/publications/publication- rep1&type=pdf> fr293-dhs-final-reports.cfm> 29. Crawley, M. J. The R book. (2012). at <https://www.wiley.com/en-us/The+R 4. Moyer CA, Mustafa A. Drivers and deterrents of facility delivery in sub- +Book%2C+2nd+Edition-p-9780470973929> Saharan Africa: a systematic review. Reprod Health. 2013;10:40. 30. Huber, P. in Berkeley Symposium on Mathematical Statistics and Probability 5. Kiwanuka SN, et al. Access to and utilisation of health services for the poor 221–223 (1967). in Uganda: a systematic review of available evidence. Trans R Soc Trop Med 31. Eberth JM, Vernon SW, White A, Abotchie PN, Coan SP. Accuracy of self- Hyg. 2008;102(11):1067–74. reported reason for colorectal Cancer testing. Cancer Epidemiol Biomark 6. Ikeako LC, et al. Influence of formal maternal education on the use of Prev. 2010;19:196–200. maternityservices in Enugu, Nigeria. J Obstet Gynaecol (Lahore). 2006;26(1):30–4. 32. Higgs BW, Mohtashemi M, Grinsdale J, Kawamura LM. Early detection of 7. Berhan Y, Berhan A. A meta-analysis of socio-demographic factors tuberculosis outbreaks among the San Francisco homeless: trade-offs predicting birth in health facility. Ethiop J Health Sci. 2014;24 Suppl:81–92. between spatial resolution and temporal scale. PLoS One. 2007;2:e1284. 8. Mills S, Williams JE, Adjuik M, Hodgson A. Use of health professionals for 33. Ozonoff A, Jeffery C, Manjourides J, Forsberg White L, Pagano M. Effect of delivery following the availability of free obstetric Care in Northern Ghana. spatial resolution on cluster detection: a simulation study. Int J Health Matern Child Health J. 2008;12:509–18. Geogr. 2007;6:52. 9. Kruk ME, Rockers PC, Mbaruku G, Paczkowski MM, Galea S. Community and 34. DO M. Utilization of skilled birth attendants in pubilc and private sectors in health system factors associated with facility delivery in rural Tanzania: a Vietnam. J Biosoc Sci. 2009;41:289. multilevel analysis. Health Policy (New York). 2010;97:209–16. 35. Barua A, et al. Implementing reproductive and child health Services in 10. Musoke D, Boynton P, Butler C, Musoke MB. Health seeking behaviour and Rural Maharashtra, India: a pragmatic approach. Reprod Health Matters. challenges in utilising health facilities in Wakiso district, Uganda. Afr Health 2003;11:140–9. Sci. 2014;14:1046–55. 36. Olusanya BO, Roberts AA, Olufunlayo TF, Inem VA. Preference for private 11. Onyeneho NG, Amazigo UV, Njepuome NA, Nwaorgu OC, Okeibunor JC. hospital-based maternity services in inner-city Lagos, Nigeria: An Perception and utilization of public health services in Southeast Nigeria: observational study. Health Policy (New York). 2010;96(3):210–6. implication for health care in communities with different degrees of 37. Benova L, Campbell OM, Ploubidis GB. A mediation approach to urbanization. Int J Equity Health. 2016;15:12. understanding socio-economic inequalities in maternal health-seeking 12. Edmonds JK, Hruschka D, Bernard HR, Sibley L. Women’s social networks behaviours in Egypt. BMC Health Serv Res. 2015;15:1. and birth attendant decisions: application of the network-episode model. 38. Onwujekwe O, Hanson K, Uzochukwu B. Examining inequities in incidence Soc Sci Med. 2012;74:452–9. of catastrophic health expenditures on different healthcare services and 13. Nwakeze NM, Kandala N-B. The spatial distribution of health health facilities in Nigeria. PLoS One. 2012;7(7):e40811. establishmentsin Nigeria. African Popul Stud. 2011;680–96. 14. Ajayi AI, Akpan W. Who benefits from free institutional delivery? Evidence 39. Onah HE, Ikeako LC, Iloabachie GC. Factors associated with the use of from a cross sectional survey of north central and southwestern Nigeria. maternity services in Enugu, southeastern Nigeria. Soc Sci Med. 2006;63: BMC Health Serv Res. 2017;17:620. 1870–8. 40. Ogunbekun I, Ogunbekun A, Orobaton N. Private health care in Nigeria: 15. Edu BC, Agan TU, Monjok E, Makowiecka K. Effect of free maternal health walking the tightrope. Health Policy Plan. 1999;14:174–81. care program on health-seeking behaviour of women during pregnancy, 41. Bohren MA, et al. Mistreatment of women during childbirth in Abuja, Intra-partum and Postpartum Periods in Cross River State of Nigeria: A Nigeria: a qualitative study on perceptions and experiences of women and Mixed Method Study. Open access Maced J Med Sci. 2017;5:370–82. healthcare providers. Reprod Health. 2017;14(1):9. 16. Abuja & Nigeria. NIGERIA DEMOGRAPHIC AND HEALTH SURVEY 2013 National 42. Akeju DO, et al. Determinants of health care seeking behaviour during Population Commission Federal Republic of Nigeria. (2014). at <https:// pregnancy in Ogun state, Nigeria. Reprod Health. 2016;13(Suppl 1):3. https:// dhsprogram.com/pubs/pdf/FR293/FR293.pdf> doi.org/10.1186/s12978-016-0139-7. 17. INEC Distribution of Senatorial Districts, Federal and State Constituencies, Electoral Wards, Polling States - NigerianMuse NigerianMuse 2016 https:// 43. Weis, J. Bedford, J. Quality care for mothers and babies at the time of birth www.nigerianmuse.com/20070414084834zg/sections/important-documents/ and immediate postpartum period national and state level barriers to inec-distribution-of-senatorial-districtsfederal-and-state-constituencies- scaling maternal and newborn care interventions in Nigeria. (2013). electoral. 44. Findley, S. et al. Changes in Maternal and Child Health Care Behaviors: Early 18. Burgert, C. R., Colston, J., Roy, T. Zachary, B. Geographic displacement evidence of the impact of community-based programs. (2012). at <http:// procedure and georeferenced data release policy for the Demographic and www.prrinn-mnch.org/documents/ Health Surveys. (2013). at <https://dhsprogram.com/publications/ ChangesinMaternalChildHealthCareBehaviors.pdf> publication-SAR7-Spatial-Analysis-Reports.cfm> 45. Corroon M, et al. The role of gender empowerment on reproductive health 19. Rutstein, S. O. Johnson, K. The DHS wealth index. (2004). at <http:// outcomes in urban Nigeria. Matern Child Health J. 2014;18:307–15. dhsprogram.com/publications/publication-CR6-Comparative-Reports.cfm> 46. Onokerhoraye, A. G. Access and Utilization of Modern Health Care Facilities in the Petroleum-producing Region of Nigeria: The Case of Bayelsa State. 20. Kulldorff M. A spatial scan statistic. Commun Stat - Theory Methods. 1997; 26:1481–96. (1999). at <https://cdn1.sph.harvard.edu/wp-content/uploads/sites/114/ 2012/10/rp162.pdf> 21. Kulldorff, M. SaTScan TM User Guide. (2015). at <http://www.satscan.org/> 22. The DHS Program User Forum: Sampling » What is the DHS position on the 47. Chiegeonu, A. S. Nigeria: State by State. (2006). at <https://books.google.co.uk/ use of cluster-level data? at 2016 <http://userforum.dhsprogram.com/index. books?id=IcmwBgAAQBAJ&printsec=frontcover&dq=nigeria+state+by+state&hl= php?t=msg&goto=9496&S=Google> en&sa=X&redir_esc=y#v=onepage&q=nigeria state by state&f=false>. 23. Kravdal Ø. A simulation-based assessment of the bias produced when using 48. Bayelsa State of Nigeria . Nigeria Information & Guide. (2017). at https:// averages from small DHS clusters as contextual variables in multilevel www.nigeriagalleria.com/Nigeria/States_Nigeria/Bayelsa/Bayelsa_State. models. Demogr Res. 2006;15:1–20. html Wong et al. BMC Health Services Research (2018) 18:397 Page 12 of 12 49. Olayinka OA, Achi OT, Amos AO, Chiedu EM. International journal of nursing and midwifery awareness and barriers to utilization of maternal health care services among reproductive women in Amassoma community. Bayelsa State. 2014;6:10–5. 50. Nakua EK, et al. Home birth without skilled attendants despite millennium villages project intervention in Ghana: insight from a survey of women’s perceptions of skilled obstetric care. BMC Pregnancy Childbirth. 2015;15:243. 51. Hailu D, Berhe H. Determinants of institutional childbirth service utilisation among women of childbearing age in urban and rural areas of Tsegedie district, Ethiopia. Midwifery. 2014;30:1109–17. 52. De Allegri M, et al. Determinants of utilisation of maternal care services after the reduction of user fees: a case study from rural Burkina Faso. Health Policy (New. York). 2011;99:210–8. 53. Wilunda C, et al. Determinants of utilisation of antenatal care and skilled birth attendant at delivery in south west Shoa zone, Ethiopia: a cross sectional study. Reprod Health. 2015;12:74. 54. Gitimu A, et al. Determinants of use of skilled birth attendant at delivery in Makueni, Kenya: a cross sectional study. BMC Pregnancy Childbirth. 2015;15:9. 55. McLaren ZM, Ardington C, Leibbrandt M. Distance decay and persistent health care disparities in South Africa. BMC Health Serv Res. 2014;14:541. 56. Lohela TJ, Campbell OMR, Gabrysch S, Mbewe R, Campbell O. Distance to care, facility delivery and early neonatal mortality in Malawi and Zambia. PLoS One. 2012;7:e52110. 57. Johnson FA, et al. Evaluating the impact of the community-based health planning and services initiative on uptake of skilled birth Care in Ghana. PLoS One. 2015;10:e0120556. 58. De Brouwere V, Richard F, Witter S. Access to maternal and perinatal health services: lessons from successful and less successful examples of improving access to safe delivery and care of the newborn. Trop Med Int Heal. 2010; 15:901–9. 59. O’Donnell O. Access to health care in developing countries: breaking down demand side barriers. Cad Saude Publica. 2007;23:2820–34.

Journal

BMC Health Services ResearchSpringer Journals

Published: Jun 1, 2018

References

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


DeepDyve is your
personal research library

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

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

All for just $49/month

Explore the DeepDyve Library

Search

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

Organize

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

Access

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

Your journals are on DeepDyve

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

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off