Access to maternal health services: geographical inequalities, United Republic of Tanzania

Abstract Objective To determine if improved geographical accessibility led to increased uptake of maternity care in the south of the United Republic of Tanzania. Methods In a household census in 2007 and another large household survey in 2013, we investigated 22 243 and 13 820 women who had had a recent live birth, respectively. The proportions calculated from the 2013 data were weighted to account for the sampling strategy. We examined the association between the straight-line distances to the nearest primary health facility or hospital and uptake of maternity care. Findings The percentages of live births occurring in primary facilities and hospitals rose from 12% (2571/22 243) and 29% (6477/22 243), respectively, in 2007 to weighted values of 39% and 40%, respectively, in 2013. Between the two surveys, women living far from hospitals showed a marked gain in their use of primary facilities, but the proportion giving birth in hospitals remained low (20%). Use of four or more antenatal visits appeared largely unaffected by survey year or the distance to the nearest antenatal clinic. Although the overall percentage of live births delivered by caesarean section increased from 4.1% (913/22 145) in the first survey to a weighted value of 6.5% in the second, the corresponding percentages for women living far from hospital were very low in 2007 (2.8%; 35/1254) and 2013 (3.3%). Conclusion For women living in our study districts who sought maternity care, access to primary facilities appeared to improve between 2007 and 2013, however access to hospital care and caesarean sections remained low.


Introduction
Given that there were an estimated 303 000 maternal deaths and 2.6 million neonatal deaths in 2015, 1,2 we clearly need better ways to reach mothers and their babies with effective interventions. 3 Distance to care is known to influence uptake of health services. 4 As the technology for geospatial measurement becomes more widely available, there has been an increasing number of studies, in low-and middle income countries, in which the association between distance to care and either uptake of care or mortality has been investigated. 5 In general, pregnant women who live far from a health facility are those least likely to have a facility delivery. [6][7][8][9][10] This appears to be the situation in the United Republic of Tanzania, 11,12 where distance to the nearest hospital has also been found to be positively correlated with direct obstetric mortality. 13 Relatively little is known about the association between antenatal care or caesarean section and distance to the nearest facility. [14][15][16] The United Republic of Tanzania has made substantial progress in reducing child mortality, but much more limited improvements in maternal health. 17,18 Much of the success in child health has been due to strong preventive actions that have been mediated by a dense network of primary health facilities 17,19 and supported by policies that, since the 1980s, have focused on rural public health. 20,21 One explicit aim of the country's recent policy on primary care is to increase access to delivery care in primary facilities -mainly by establishing one dispensary, that can provide basic antenatal, delivery, outpatient and postnatal care, for every village. Health centres, which already provide basic laboratory diagnostics and inpatient care, are progressively being upgraded so that they can also provide comprehensive emergency obstetric care. 22 The focused antenatal care programme, which was introduced in 2002, encourages pregnant women without known risk factors to give birth in primary facilities. 23 However, while studies have shown that uptake of intrapartum care is increasing in most parts of the United Republic of Tanzania, 18 it is not known whether the Tanzanian women who live in remote rural areas have benefited from the policy change. We therefore examined whether -and, if so, how -over a six-year period, the relationship between uptake of maternity care and distance to a health facility had changed in five rural districts in the south of the United Republic of Tanzania. In surveys in 2007 and 2013, we quantified the effect of both the distance to the nearest primary facility -i.e. dispensary or health centre -and the distance to the nearest hospital on four key indicators of maternity care: (i) four or more visits for antenatal care; (ii) birth in a primary facility; (iii) birth in a hospital; and (iv) birth by caesarean section. By examining the interaction between distance to facility and survey year, we then examined whether changes over time in uptake of care varied by distance to a facility.

Methods
We used information from two geo-referenced household surveys covering the same five districts in the south of the United Republic of Tanzania: (i) a census of all 243 612 households in 2007 -primarily designed to evaluate the impact of Objective To determine if improved geographical accessibility led to increased uptake of maternity care in the south of the United Republic of Tanzania. Methods In a household census in 2007 and another large household survey in 2013, we investigated 22 243 and 13 820 women who had had a recent live birth, respectively. The proportions calculated from the 2013 data were weighted to account for the sampling strategy. We examined the association between the straight-line distances to the nearest primary health facility or hospital and uptake of maternity care. Findings The percentages of live births occurring in primary facilities and hospitals rose from 12% (2571/22 243) and 29% (6477/22 243), respectively, in 2007 to weighted values of 39% and 40%, respectively, in 2013. Between the two surveys, women living far from hospitals showed a marked gain in their use of primary facilities, but the proportion giving birth in hospitals remained low (20%). Use of four or more antenatal visits appeared largely unaffected by survey year or the distance to the nearest antenatal clinic. Although the overall percentage of live births delivered by caesarean section increased from 4.1% (913/22 145) in the first survey to a weighted value of 6.5% in the second, the corresponding percentages for women living far from hospital were very low in 2007 (2.8%; 35/1254) and 2013 (3.3%). Conclusion For women living in our study districts who sought maternity care, access to primary facilities appeared to improve between 2007 and 2013, however access to hospital care and caesarean sections remained low.

Research
Inequality in Tanzanian maternity care Claudia Hanson et al.
intermittent preventive treatment with antimalarials on infant survival; 24 and (ii) a sample survey in 2013 that assessed the impact of a home-based counselling strategy on neonatal care and survival. 25 In both surveys, the study population comprised women who had had a live birth in the 12 months before the survey and reported on uptake of pregnancy and intrapartum care.
The study area covers three districts of the Lindi region and two districts of the Mtwara region. 26 Most of the residents of these districts are poor and live in mud-walled houses in rural villages. Between 2009 and 2013, two dispensaries in the study area were upgraded to become health centres and 14 new dispensaries were inaugurated. By 2013, the study population was served by 156 dispensaries, 15 health centres and six hospitals within the study area and by another two hospitals just outside the district boundaries. All except four of the 179 health facilities serving the study area in 2013 -i.e. two mission hospitals, one mission dispensary and one private health centre -were public facilities that provided maternal health services free of charge. 27 In both 2007 and 2013, all eight hospitals serving the study area provided caesarean sections on a daily 24-hour basis, three of the hospitals had maternity waiting homes and all of the hospitals and seven of the health centres were equipped with ambulances. Ambulance use -e.g. for hospital referral -was, however, severely constrained by shortages of fuel, human resources and funds for repair. Although all except one of the 179 facilities offered delivery care, basic emergency obstetric care was not consistently available in the study area. [27][28][29]

Data collection
The survey methods are described in detail elsewhere. 24,25 In brief, we used a modular questionnaire, administered in Swahili, to assess coverage of essential interventions during pregnancy and childbirth. Use of personal digital assistants to collect data facilitated the checking of standard ranges, consistency and completeness at the time of data entry. 30 Household wealth was assessed by asking each household head about household assets and housing type. We mapped the study households using a global positioning system. The positions of the relevant health facilities had been recorded in previous surveys. a All the subjects were women of reproductive age who reported giving birth in the 12 months before the survey. All of the percentages for 2013 were computed taking sampling weights into consideration. b For each characteristic, the significance of the between-survey difference was investigated in a χ 2 test. c There are not equal numbers of subjects from each quintile because, in our study population, the mean number of women of reproductive age per household tended to increase with increasing household wealth. d Excluding the mothers whose nearest facility was a hospital. Inequality in Tanzanian maternity care Claudia Hanson et al.
In 2007, we surveyed all 243 612 households in the five study districts. In 2013, however, we sampled 169 324 households, which were selected by following a two-stage sampling survey. 25 Using the results of the national 2012 census, in which 247 350 households were recorded in the study area, we first sampled so-called subvillages. This sampling was proportional to the number of households in each subvillage -typically about 80-100. We included all households in the subvillages with fewer than 130 households, but used segmentation for subvillages with more than 131 households.

Outcomes and explanatory variables
Our main outcomes of interest were uptake of at least four visits for antenatal care, delivery in a health facility and delivery by caesarean section. Using a combination of coordinates and the nearstat command in Stata version 13 (StataCorp. LP, College Station, United States of America), we calculated straight-line distances between each surveyed household and: (i) the nearest antenatal clinic, which could have been in a primary facility or a hospital; (ii) the nearest primary facility offering delivery care; and (iii) the nearest hospital. We did this separately for 2007 and 2013. In the 2007 survey, we attempted to impute the coordinates of households for which no such coordinates were recorded, from the coordinates for neighbouring households. Household wealth quintiles were constructed separately for 2007 and 2013, using principal component analysis. 31

Statistical analysis
All analyses were conducted in Stata version 13. For the 2013 data, we accounted for the different sampling structures of the 2007 and 2013 surveys by weighting subvillages by the inverse chance of being included. The percentages reported for 2013 -but not those reported for 2007 -are therefore weighted values. For both 2007 and 2013, we assessed the effect of: (i) distance to nearest antenatal clinic on uptake of at least four visits for antenatal care; (ii) distance to nearest primary facility on delivery in a primary facility; (iii) distance to nearest hospital on hospital delivery; and (iv) distance to nearest hospital on birth by caesarean section. For the analysis of the effect of distance on delivery in a primary facility,

Lindi Region Mtwara Region
Notes: Distances from home to the nearest facility providing delivery care were evaluated for 3114, 1564, 2437, 2907, and 3609 households in Lindi Rural, Ruangwa, Nachingwea, Newala, and Tandahimba districts, respectively. In these box-and-whisker plots, the horizontal bars, boxes, whiskers and dots indicate medians, interquartile ranges, minimum and maximum values -excluding outliers -and outliers, respectively. Notes: Distances from home to the nearest facility providing delivery care were evaluated for 3114, 1564, 2437, 2907, and 3609 households in Lindi Rural, Ruangwa, Nachingwea, Newala, and Tandahimba districts, respectively In these box-and-whisker plots, the horizontal bars, boxes, whiskers and dots indicate medians, interquartile ranges, minimum and maximum values -excluding outliers -and outliers, respectively.
(continues. . .) we excluded births where a hospital was the nearest facility. We first used generalized linear models to calculate crude prevalence ratios (cPR) with 95% confidence intervals (CI). We compared the prevalence of each indicator by increasing distance to a primary health facility or hospital and then compared the prevalence of each indicator between 2007 and 2013 within each distance group. 32 We adjusted the crude prevalence ratios for potential confounding by the mother's age, parity, district of residence, education, ethnic group and occupation and her household's wealth quintile. Using multilevel logistic regression without weighting, we fitted an interaction term between distance to facility and survey year and used the likelihood ratio test to calculate a corresponding P-value. We also used ArcGIS version 9.2 (ESRI, Redlands, USA) to map the absolute increases in facility delivery and caesarean section by administrative ward -as percentages of the live births -between 2007 and 2013.

Ethics
Ethical clearance was obtained from the institutional review boards of Ifakara   The study population was informed about the surveys by the local government authorities and again, one day prior interview, by a sensitizer who used information sheets in the local language. Written consent to participate was obtained from household heads and the women who answered questions about pregnancy and childbirth.

Results
We conducted interviews with 321 093 consenting females who were aged 13-49 years and considered to be women  (Table 1). In general, compared with those interviewed in the 2013 survey, the women interviewed in the 2007 survey were living in poorer households, less educated and of higher parity and lived further from any health facility providing delivery care in the study area (median: 2.7 km in 2007 vs 2.2 km in 2013; Fig. 1). The median distance to a hospital was 18.0 km in both surveys (Fig. 2).
Coverage with four or more antenatal visits increased only marginally, from 41% (9082/21 959) in 2007 to a weighted value of 45% in 2013 (cPR: 1.1; 95% CI: 1.1-1.1), and there was no association between distance to an antenatal clinic and uptake of such an-tenatal care in 2007 or 2013 (Table 2). Although the interaction between study year and distance to an antenatal clinic was statistically significant (P = 0.011), the between-survey changes seen in uptake of antenatal care in each distance category were very small. Table 2 summarizes the cPRs and adjusted PRs (aPR). After excluding the data for areas where a hospital is the nearest facility, the proportion of births occurring in primary facilities increased from 13% (2546/20 270) in 2007 to a weighted value of 41% in 2013 (aPR: 3.3). The proportion of births occurring in hospitals also increased, from 29% (6475/22 242) in 2007 to a weighted value of 40% in 2013 (aPR: 1.3). In both surveys, the distance to a primary facility was strongly associated with delivery in a primary facility. The between-survey increases in the proportion of births occurring in primary facilities were most pronounced among the women who lived relatively far away from a primary facility (P < 0.001;

Fig. 3. Ward map of the five study districts showing the increases in facility deliveries, United Republic of Tanzania, 2007-2013
Notes: The map, which shows the percentage increase in the proportions of births in the previous 12 months that were facility deliveries, is based on data collected in interviews with 22 243 women in 2007 and 13 833 in 2013. All of the percentages for 2013 were computed taking sampling weights into consideration.

Research
Inequality in Tanzanian maternity care Claudia Hanson et al. Table 2). For example, for those living less than 1.0 km from a primary facility, the proportion of births that occurred in such a facility increased from 22% (1219/5472) in 2007 to a weighted value of 50% in 2013 (aPR: 2.3). The corresponding values for those living at least 7.5 km from a primary facility were 6% (63/1056) and 28%, respectively (aPR: 4.1). In contrast, the between-survey increases in the proportion of births occurring in hospitals were greatest for those living at least 5.0 km, but less than 10.0 km from a hospital -34% (824/2420) versus a weighted value of 57% (aPR: 1.6) -or at least 10.0 km but no more than 15.0 km from a hospital -26% (960/3646) versus a weighted value of 41% (aPR: 1.5) ( Table 2).
Overall, the proportion of women giving birth in any health facility, whether it was a primary facility or a hospital, increased from 41% (9021/22 242) in 2007 to a weighted value of 79% in 2013. The greatest absolute increase was seen in the rural, remote wards that were at least 10.0 km from a hospital (Fig. 3). The share of births occurring in primary facilities increased with the distance to the nearest hospital. In terms of the weighted proportions for 2013, only 4% of the women living very close to a hospital -i.e. at a distance of less than 5.0 km -gave birth in a primary facility.
The corresponding proportions for the women living at least 25.0 km but less than 35.0 km and more than 35.0 km from their nearest hospital were much greater: 49% and 52%, respectively (Fig. 4).
The proportion of births represented by caesarean sections increased from 4.1% (913/22 145) in 2007 to a weighted value of 6.5% in 2013. The level of increase in the frequency of caesarean sections appeared unaffected by the distance to the nearest hospital (P = 0.208; Table 2; Fig. 5) even though, in both surveys, there was a strong negative association between distance to the nearest hospital and delivery by caesarean section. For the women living more than 35.0 km from their nearest hospital, there was no between-survey increase in the proportion of births represented by caesarean sections (aPR: 1.0; Table 2).

Discussion
The data presented here provide evidence of substantial increases, between 2007 and 2013, in the proportion of births in the study area, represented by deliveries in primary facilities and hospitals. The large increase we observed in facility births is consistent with findings from other Tanzanian studies. 18,33,34 The increased uptake of delivery in primary facilities for women who live in the more remote areas, often far from a hospital, is particularly noteworthy. The increase probably indicates that the national policy to improve access to maternity care -by promoting delivery care in primary facilities and further increasing the number of such facilities -is being successful. 23,35 The United Republic of Tanzania's substantial socioeconomic development, 36 including improvements in the road network, may have helped women to travel moderate distances while seeking maternity care. Also, two projects to support birth preparedness, at community level, may also have had a beneficial impact in the study area. 25,37 While we may have seen an important reduction in the inequality of geographical access to primary care for childbirth, there appeared to be little between-survey improvement in access to hospital-based delivery care or caesarean sections. In our study area, the district hospitals are expected to send ambulances to dispensaries, to collect patients who need emergency hospital care. However, such emergency referrals are severely constrained by lack of funds at district level to pay for the fuel, maintenance and repairs needed to keep ambulances on the road. 37 In addition, only three of the hospitals serving the study area had maternity waiting homes. While the optimal caesarean section rate remains a matter of controversy, [38][39][40][41] rates of about 3% -as seen in the more remote settings in our study area -are far too low to meet the needs of pregnant women and their babies.
The persistently low uptake of antenatal care by Tanzanian women has been noted previously. 42 In our study, distance to a facility had no apparent effect on uptake of such care. This observation is in line with findings from Zambia, 15 but conflicts with the results of an earlier study in the United Republic of Tanzania. 35 However, this earlier study did not include dispensaries, which are the main providers of antenatal care in the country. 35 As the World Health Organization has now increased the recommended number of antenatal visits to at least eight, 43 it is, perhaps, even more important to examine the reasons for the suboptimal levels of antenatal care seen in the United Republic of Tanzania. 44 Our study had several strengths, including its reliance on two large representative datasets from, effectively, the same study population and the use of the same questionnaire and a short recall period in both surveys. However, there may be limitations. First, the use of straight-line distance to a facility, to evaluate geographical accessibility, is sometimes regarded as inferior to calculating travel time 45 -although this depends on the setting. 46 The results of a Tanzanian study in which topographic maps were used to estimate travel time 47 indicated that, at least in the United Republic of Tanzania, straight-line distances may correlate fairly well with travel times. Second, our analysis is based on a full census of the study area in 2007 but only a sample survey in 2013.
Despite adjusting our estimates to take account of this difference between the surveys, we still found unanticipated and unexpected differences between the composition of the study population in 2007 and that of the study population in 2013. These differences, however, can probably be attributed to migration and other demographic changes 26 rather than to our sampling procedure. Third, our 2007 data came from women who differed, in terms of three known drivers of the uptake of facility care -i.e. age, education and parity 48 -from the women who provided our data in 2013. We did, however, make adjustments in our data analyses for each of these potential confounders. Lastly, we used prevalence ratios to estimate the strength of the effect of distance to the nearest facility on uptake of care. While this improves the ease of interpretation, it also increases confidence intervals. 32 The increased uptake of facility births in our study area is encouraging. However, our analysis indicates that this  increase did not translate into a substantial concurrent increase in caesarean sections. A plausible explanation is the lack of a functioning link between primary and secondary facilities, especially poor emergency referral from the primary facilities. We believe that our findings -together with existing evidence on deficits in the quality of care in primary facilities and on the high levels of neonatal mortality in our study area and other parts of the United Republic of Tanzania 25,27,28,33 -indicate a need to revisit the policy of providing maternity care in primary facilities that are not linked to hospitals through a functioning referral system. Intrapartum care, in East Africa and elsewhere, needs to be strengthened by improvements in the quality of care and referral systems. 49