Inequity in uptake of hospital-based childbirth care in rural Tanzania: analysis of the 2015–16 Tanzania Demographic and Health Survey

Abstract Proportions of facility births are increasing throughout sub-Saharan Africa, but obstetric services vary within the health system. In Tanzania, advanced management of childbirth complications (comprehensive emergency obstetric care) is offered in hospitals, while in frontline, primary health care (PHC) facilities (health centres and dispensaries) mostly only routine childbirth care is available. With over half (54%) of rural births in facilities, we hypothesized the presence of socio-economic inequity in hospital-based childbirth uptake in rural Tanzania and explored whether this relationship was modified by parity. This inequity may compound the burden of greater mortality among the poorest women and their babies. Records for 4456 rural women from the 2015–16 Tanzania Demographic and Health Survey with a live birth in the preceding 5 years were examined. Proportions of births at each location (home/PHC/hospital) were calculated by demographic and obstetric characteristics. Multinomial logistic regression was used to obtain crude and adjusted odds ratios of home/PHC and hospital/PHC births based on household wealth, including interaction between wealth and parity. Post-estimation margins analysis was applied to estimate childbirth location by wealth and parity. Hospital-based childbirth uptake was inequitable. The gap between poorest and richest was less pronounced at first birth. Hospital-based care utilization was lowest (around 10%) among the poorest multiparous women, with no increase at high parity (≥5) despite higher risk. PHC-based childbirth care was used by a consistent proportion of women after the first birth (range 30–51%). The poorest women utilized it at intermediate parity, but at parity ≥5 mostly gave birth at home. In an effort to provide effective childbirth care to all women, context-specific strategies are required to improve hospital-based care use, and poor, rural, high parity women are a particularly vulnerable group that requires specific attention. Improving childbirth care in PHC and strengthening referral linkages would benefit a considerable proportion of women.


Introduction
Mortality around the time of childbirth is essentially a disease of poverty. An inverse relationship between poverty and maternal health has been known for over a century. Wide inequities in maternal and perinatal mortality exist between nations, with low-and middle-income countries being the most affected . Sub-Saharan Africa (SSA), with only 14% of global population, accounted for 66% (201 000) of maternal deaths, 40% (1 027 000) of newborn deaths and 31% (1 060 000) of stillbirths • The proportion of facility births is increasing in rural Tanzania, but obstetric care at different levels of the health system varies. We hypothesized a differential use of hospitals for childbirth and assessed interaction between wealth and parity. Inequity in hospital-based childbirth care use may contribute to the burden of greater mortality among the poorest women and their babies. • Hospital-based childbirth care use was inequitable among women in rural Tanzania. The gap between the poorest and richest use was less pronounced among women at first parity. Uptake of hospital-based care was lowest (around 10%) among poorest multiparous women, remaining low at high parity (≥5) despite higher risk of complications and death. • Rural women's use of PHC childbirth care after the first birth was noteworthy, ranging between 30 and 51% depending on wealth and parity. • As part of efforts to reach all women with appropriate, timely care, strategies are required to improve uptake of hospital-based care particularly among poor, rural, high parity ones. Improving quality of childbirth care in PHC and referral linkages would benefit a substantial proportion of women.
to services, including intrapartum care, for rural populations. Tanzania, with a population of 59 million (2020) (World Bank, 2020), has been at the forefront of PHC development after independence with its founding principles set out in the Arusha Declaration in 1967 (Bustreo et al., 2019;Dominicus and Akamatsu, 1989). A vision of high-quality PHC for all is expressed in current policy (Vision 2025) (Tanzania URO, 2000). PHC facilities are dispensaries at village level and health centres at ward level (Tanzania MoHSW, 2015). Women can give birth at all levels of the health system, and childbirth in PHC facilities is encouraged for women with no known risk factors at onset of labour (Hanson et al., 2015). In spite of a capillary PHC network, with 85% of the population living within 5 km from a facility, Tanzania's maternal mortality ratio in 2017 remained high at 524 per 100 000 live births (a reduction of 39% from 2000) (WHO, 2015b;2019) and was among 10 countries worldwide with the highest absolute numbers of maternal, newborn deaths and stillbirths (Lawn et al., 2016). Obstetric care at different levels of the health system varies markedly in SSA (Campbell et al., 2016). There is growing evidence that outcomes for mothers and their babies improve when women give birth in units offering high-quality care, not just in any facility Hanson et al., 2015;Lohela et al., 2019). In Tanzania, advanced management of childbirth complications (including surgery and blood transfusions, equivalent to comprehensive emergency obstetric care [EmOC]) is available in hospitals, while lower-level, PHC facilities are generally able to provide routine childbirth care only (Campbell et al., 2016;Kruk et al., 2016). Although dispensaries and health centres differ in size, number of beds and staffing levels, both types of facilities have similar obstetric capacity and often do not reach a description of basic EmOC facility (Campbell et al., 2016;Hanson et al., 2013). Challenges to the provision of high-quality obstetric care in lower-level facilities in Tanzania have been amply described and include insufficient staffing, poor infrastructure and low birth volumes (Hanson et al., 2013;Benova et al., 2014;Kruk et al., 2014;Hanson et al., 2015;Straneo et al., 2014;Baker et al., 2015). Although national efforts are underway to increase obstetric care in health centres up to comprehensive EmOC, at the time of analysis very few had been upgraded. Higher mortality among poorer women and their babies may be compounded by their reduced hospital-based childbirth care uptake, where higher-quality obstetric care is more commonly found.
Over the past decade, a shift from home to facility births has been described in SSA, with increasing proportions of women giving birth in facilities in rural and urban contexts and across wealth groups, and Tanzania is no exception (Montagu et al., 2017;Doctor et al., 2019). However, strong socio-demographic differentials continue to be reported (Kyei-Nimakoh et al., 2017;Moyer and Mustafa, 2013;Gabrysch and Campbell, 2009;Campbell et al., 2016;Virgo et al., 2017;Dunlop et al., 2018). There is limited information on how socio-economic groups uptake obstetric care at different levels of the health system. Within the background of a renewed discussion of the most efficient configuration of childbirth care from an equity, quality and cost-efficiency perspective (Kruk et al., 2016;Hanson and Schellenberg, 2019;Kruk et al., 2018;Roder-DeWan, 2020), we aimed to estimate the levels of use of hospital-based childbirth care in rural Tanzania and its association with women's socio-economic status. Given the association of poverty and high parity, and the latter's implications for obstetric care, we investigated whether the association between wealth and hospital-based childbirth depended on parity.

Study setting
In Tanzania, the most recent (2015-16) Demographic and Health Survey (DHS) estimated that 63% of births in the 5-year period preceding the survey took place in health facilities (54% in rural and 86% in urban areas) (Tanzania MoHCDGEC, 2016). In the same period, there were 6790 facilities (including public, faith-based, parastatal and private), from which 257 (3.8%) were hospitals (Tanzania MoHSW, MoHMZ, National Bureau of Statistics [NBS], Office of the Chief Government Statistician [OCGS], and ICF, 2015). There were an estimated 12 million women of reproductive age and approximately 1.9 million births in 2015.

Data and population
Data from the 2015-16 Tanzania DHS were used. DHS are cross-sectional, nationally representative surveys of households, with women of reproductive age (15-49 years) selfreporting on the use of reproductive and maternal healthcare. Approximately 12 500 households were visited, and 13 000 women interviewed. Records of women living in rural areas in mainland Tanzania were used in this analysis, if they reported a live birth in the 5 years preceding the survey. Classification as 'rural' in DHS was based on census enumeration units (Tanzania NBS and Zanzibar OGCS, 2013).

Definitions
The outcome variable was the location of the most recent live birth, in three categories: home (respondent's home, other home and en route to provider), PHC facility (dispensary, health centre, maternity home and 'other facility') and hospital (district, regional, referral or tertiary/university). All public and private (non-profit/profit) PHC facilities and hospitals were included.
Socio-economic status (SES): In DHS, SES is based on availability of durable household assets (Vyas and Kumaranayake, 2006). A wealth score is generated for each sampled household using principal component analysis and the households are then subdivided into equal-size wealth quintiles. Distribution of wealth is uneven across different contexts, with the highest (wealthiest) SES quintile households underrepresented in rural contexts. In rural Tanzania (DHS 2015-16), there were only 14.6% and 2.5% women in Quintile 4 (richer) and Quintile 5 (richest), respectively. Thus, for the purpose of this analysis, the two highest wealth quintiles were merged, resulting in the creation of four wealth groups (poorest, poorer, middle and wealthiest). The terms richest/wealthiest refer to relative wealth in a poor, rural context, thus indicate women from the least poor households. To analyse the interaction of SES and parity, a binary SES variable was created, by generating two equal groups based on wealth scores (wealth score < median recoded as poorer, wealth score ≥ median coded as richer 50%).
Maternal education was recoded into three categories: no education, completed primary and completed secondary or higher.
Marital status at survey was recoded into currently married/cohabiting and not currently married/cohabiting. Zone of residence: Tanzania is divided into 21 administrative regions, grouped into eight zones (Tanzania MoHCDGEC, MoH [Zanzibar], and ICF, 2016). We used the eight zones to account for sub-national variation in outcomes and service availability (Armstrong et al., 2016). All eight zones include rural areas; the Eastern zone includes the Dar es Salaam urban conglomerate.
Antenatal care (ANC) for the index pregnancy was categorized into no visits, 1-3 visits and ≥4 visits.
Other obstetric characteristics studied were the following: multiple index birth, a previous birth in the recall period by Caeserean section (CS), death of a previous child (born in the recall period) aged 1-12 months, death of a previous child (born in the recall period) aged <1 month, a short preceding birth interval (≤12 months).

Statistical analysis
Analysis was performed using STATA IC 15 software. Complex survey design and non-response (stratification, clustering and survey weights) were accounted for using svyset commands. Characteristics of the sample were analysed with proportions and 95% confidence intervals (CIs) of outcome and exposure variables. There were no missing data for the variables examined. Proportions of subgroups of women at each level of outcome (hospital/PHC/home birth) for each exposure variable were examined using bivariate analysis. As the interaction between SES and parity was of interest, the proportion of women giving birth at each location by combinations of parity levels and a binary SES variable (poorer/richer) were determined. Associations between the outcome variable and dependent variables (demographic, geographical characteristics, SES, ANC care received and available obstetric factors) were assessed in bivariate analysis. Variables which were significant at P <0.05 level in bivariate analysis were included in the final multivariable model. Multinomial logistic regression was used as the outcome variable had three categories, thus allowing to include all births in one model. The baseline outcome was birth in PHC, thus the model produced odds ratios (ORs) of home vs PHC and hospital vs PHC birth. In the final multivariable model, we tested for an interaction between SES and parity group. We calculated the margins to obtain predicted percentages of women giving birth at each of the three locations, depending on their SES and parity group combination. Results were used to calculate the difference or gap in hospital or PHC uptake for birth between the wealthiest and poorest women.

Ethical approval
The DHS receive government permission, use informed consent and assure respondents of confidentiality. Permission to use the dataset for the purpose of this analysis was obtained from the DHS programme.

Population characteristics
Observations of 4456 women living in rural mainland Tanzania and the circumstances of their most recent live birth in the 5 years preceding the survey were included in the analysis. Home birth was reported by 41% of women and a slight majority reported a facility birth (59%): 35% in PHCs and 24% in hospitals. Women from the wealthiest households were under-represented, with only 17% in the highest group compared with 28% in the lowest. Approximately one in five women was nulliparous at index birth (22%), while 25% had parity five or higher. Background characteristics are summarized in Supplementary Table S1.
Results of bivariate analysis are reported in Table 1. The percentage of rural women using hospitals for childbirth increased with higher SES, from 16% in the poorest group to 45% in the wealthiest. PHC births also rose with increasing wealth, although less steeply than hospital births. As parity increased, hospital births reduced sharply, while PHC births had a less clear trend across SES and parity and varied only marginally at around one-third of births. Hospital births increased with higher maternal age, maternal education and number of ANC visits. There was a wide variation in hospital births across the eight zones, ranging from 16% in the Lake Zone to 39% in the Southern Highlands. PHC facilities provided a substantial proportion of childbirth care in rural areas of all zones (range 23-48%, median 38%).
Examining SES and parity together, hospital births were more frequent in women from wealthier households in all parity groups, with percentages reducing as parity increased across all wealth groups. The drop of hospital use for childbirth was seen among poorer women already at Parity 1-2, while among wealthier women, this reduction was seen at Parity 3-4. Despite the decrease, the percentage of births in hospitals remained higher among wealthier than poorer women in all parity groups. The gap between the poorest and wealthiest women in hospital births was greatest at Parity 1-2 (Supplementary Graph 1).

Logistic regression
Results of bivariate and multivariate logistic regression are reported in Table 2.
In adjusted analysis, compared with women from the poorest households' group, all wealthier women were less likely to have given birth at home vs in PHC. The wealthiest were 66% less likely to do so. High-parity women (≥5) had higher odds of home birth (OR 1.54, 95% CI 1.05-2.25) compared with the reference group Parity 1-2, while odds in other parity groups were not significantly different from baseline. Higher odds of a home birth were seen in women with no ANC or 1-3 ANC visits compared with women with ≥4 visits. Compared with women with primary education, those with no education had higher odds of a home birth, while those with secondary or higher education had reduced odds. Women residing in four zones (Southern Highlands; Southern and Southern West Highlands; and Eastern) had reduced odds compared with those residing in the reference Lake Zone.
In adjusted analysis, the wealthiest rural women had higher odds (OR 1.78, 95% CI 1.26-2.50) of a hospital vs a PHC birth compared with the poorest, while other wealth groups were not significantly different from the poorest. Higher odds of hospital vs PHC birth were found in Parity 0 women compared with baseline Parity 1-2 (OR 3.22, 95%CI, 2.34-4.43), while the odds were reduced in higher-parity groups. The effect of maternal age was confounded in crude analysis; in adjusted analysis, the odds of hospital vs PHC birth increased with age. Women with a previous birth by CS had higher odds of a hospital birth compared with those with no previous CS, while women with no education, compared with those with primary education, had reduced odds of hospital vs PHC birth. Higher odds were observed in women residing in two zones (Northern and Central) compared with those residing in the reference Lake Zone.

Interaction between SES and parity
To assess the joint effects of parity and SES, the final adjusted multinomial logistic regression model was run with an interaction term between the two variables. A likelihood ratio test comparing the model with and without interaction indicated better fit of the model with interaction (P = 0.006). The reference group included the poorest women at parity ≥5, as this group had the lowest use of hospital-based childbirth care and was the most numerous wealth/parity subgroup (n = 406). Results are shown in Supplementary Tables S2a  and S2b.
All combinations of SES and parity had lower odds than the baseline category of a home vs a PHC birth, although not all reached statistical significance at P < 0.05. The richest women at high parity (≥5) had the lowest adjusted OR (0.29, 95% CI 0.15-0.57) compared with the reference group. The poorest women at Parity 0 had an OR of 8.03 (95% CI, 4.45-14.46) compared with the baseline of a hospital vs PHC birth, while at other parity levels the ORs were not significantly different. Women at Parity 1-2 from poorer, medium and richest groups had higher odds of hospital vs PHC childbirth compared with the baseline group; the OR was non-significant in the poorest group. In other groups, ORs were not significantly different from the baseline.
We predicted the percentages of childbirth for each combination of SES and parity in each location using margins analysis; results are reported in Table 3 and displayed in Graph 1(A-C). Across all SES groups, hospital-based childbirth (Graph 1A) was highest at first birth, at >40%. Use of hospitals reduced in all SES groups with increasing parity, but the shape of this decline varied. Among the wealthiest women, hospital use decreased gradually, reaching its lowest (around 25%) at parity 3-4. Among the poorest, the decline was abrupt after parity 0, levelling out at 12% at parity 1-2. The effect of wealth on PHC births was more complex (Graph 1B). The predicted percentages at this level were lowest among nulliparous women in all wealth groups. Among the wealthiest, the percentage rose with parity, reaching its maximum (51%) at parity 3-4. Among the poorest women, the predicted utilization reached the highest level at parity 1-2 (39%) and then levelled off at around 30%. In all wealth groups, after parity 0, ≥30% of women were predicted to give birth in a PHC. Median utilization of PHC facilities in parous women was 35% (range 30-51%), while in women at first parity it was 27% (range 20-31%). The percentage of births at home (Graph 1C) increased as parity rose in all wealth groups and was lowest among the wealthiest women in all parity groups.
The profiles of birth location among the two extremes of wealth (poorest and richest women) are compared in Graph 2. Among the richest women, there was a shift in the location of births from mainly hospital (at parity 0) to mainly PHC facilities (at parity ≥5). In the poorest women's group, between parity groups 0 and ≥5, decline in hospital births was accompanied by a sharp rise in home births, with a small increase in PHC births.

Discussion
This study explored rural women's differential use of childbirth care in Tanzania. We report three key findings. First, there was a socio-economic inequity in rural women's use of hospital-based childbirth, which additionally varied with parity. Second, the poorest multiparous women had the lowest use of hospital-based care for childbirth (around 10%), with no increase in uptake at grand multiparity despite increased risk. This group also had lower uptake of PHC care. Third, PHC facilities provided care to a sizeable proportion of women after the first birth, with a median uptake of 35% (range 30-51%) by women after the first birth (compared with a median of 27% by women at first birth).
The poorest women in rural Tanzania were less likely than those from the wealthiest households to give birth in hospitals, where advanced management of childbirth complications was available. The study adds to existing evidence that wealth is not just a determinant for facility birth, but also for uptake of hospital-based childbirth care within the health system. It expands findings of a previous sub-national study (Straneo et al., 2014) and earlier studies (Benova et al., 2014). We found that the gap between the poorest and wealthiest women in use of hospitals was less pronounced among nulliparous Table 2. Crude and adjusted ORs by multinomial logistic regression of home vs PHC birth (left) and hospital vs primary care births (right) in rural women, with a live birth in the last 5 years (Tanzania, DHS 2015-16)

Hospital birth vs PHC birth
Variable Crude OR

P-value
Adjusted OR a

P-value
Adjusted OR a

P-value
Adjusted OR a

P-value
Adjusted OR a

P-value
Zone of residence Western  women. Health policy in Tanzania (Tanzania MoHSW, 2008) recommends that women's first births should take place in hospitals. In spite of the existing recommendations, uptake by rural women at first birth was not universal, as just over half used hospital-based care. We found that hospital-based care use among nulliparous women was very similar across wealth groups. Factors other than wealth are likely to limit hospital use at first birth; amongst these distance to hospital from a woman's residence, which could not be accounted for, stands out, and interaction between distance and wealth has been described (Bai et al., 2002;Hanson et al., 2015;. Our finding is in line with that of other researchers indicating that women and their families recognize the first birth as a higher-risk one (Jahn et al., 1998;Dunlop et al., 2018). Utilization of hospitals decreased at different rates in SES groups with increasing parity and the gap between the poorest and wealthiest was widest at Parity Level 1-2 but persisted across all higher-parity groups. A switch in birth location away from facilities between first-and second-order births was found to be less likely in wealthier households across low-and middle-income countries (Benova et al., 2017). What this study adds is that among the poorest women between first-order and successive births (Parity 1-2), there was a switch within the health system, from hospital-based care to PHC-based care, and to home-based care.
Utilization of hospital-based childbirth is lowest among the poorest, multiparous women. Use, as estimated by margins analysis, dropped to around 10% at all levels of parity after the first-order birth. Despite greater risk in women with ≥5 previous births of adverse pregnancy outcomes, including haemorrhagic complications (Bai et al., 2002;Mgaya et al., 2013;Filippi et al., 2016), there was no increased hospital care uptake among the poorest women. Factors contributing to this may be inadequate counselling during ANC on hospital-based childbirth resulting in low perceived risk (Pembe et al., 2008), childcare duties at home, and greater economic constraints due to larger families. Reducing facility Graph 2. Predicted margins (percentages) of Hospital/PHC/Home births of poorest and wealthiest rural women by parity, Tanzania 2015-16 use with increasing parity is well documented (Kyei-Nimakoh et al., 2017;Moyer and Mustafa, 2013;Gabrysch and Campbell, 2009;Berhan and Berhan, 2014). Results of our study add that, in rural Tanzania, multiparous women will opt for a home birth when economic means are limited but will uptake PHC-based childbirth when resources are available. Poor multiparous women constitute a disadvantaged group, least served by hospitals or indeed by any facility. Qualitative studies in southern Tanzania (Kowalewski et al., 2000) indicated that in the community, these women were perceived as vulnerable due 'to fatigue and (being) overburdened with household duties', and precisely these factors prevented them from accessing health services.
Although our analysis focused on hospital births, rural women's uptake of childbirth care in lower-level facilities is noteworthy. PHC units (health centres and dispensaries) have a critical role in childbirth care in rural Tanzania, as 61% of all facility births took place here, and utilization ranged from 30 to 51% (median 35%) across all SES groups after Parity 0. The greatest use is among the wealthiest, at intermediate and high parity. This has relevance in the current debate on reorganization of maternity care in low-income countries (Campbell et al., 2016;Kruk et al., 2018;Hanson and Schellenberg, 2019). From these data, childbirth care in PHC facilities is used by all wealth groups: among the wealthiest, uptake increases with parity, while among the poorest, utilization is mostly at intermediate levels of parity. Comparing the shift in births between the two extremes of parity in the poorest and richest women suggests that, in this context, a reduction in the availability of facilities providing childbirth care without other measures may result in an increase in the already-high level of home births among the poorest women.
Policy recommendations arising from this study include three main points. Firstly, aiming attention on the poorest women allows identification of health system adjustments to mitigate the effects of poverty on childbirth-related deaths (WHO, 2008). A subsidized voucher scheme has been applied in Kenya (Dennis et al., 2019;. Maternity waiting homes may contribute to facilitating access to hospitals (Virgo et al., 2017), and there is evidence that they are utilized preferentially by poorer women (Fogliati et al., 2017). Secondly, high-parity women's low use of hospital-based childbirth care, particularly among the poorest, requires urgent action. All women should receive appropriate, timely care. National policy should focus attention on grand-multiparous women as a particularly higher-risk group. Guidelines should be in place to prepare these women for hospital-based births. They may include adapted birth preparedness plans and emergency transport during labour to improve geographic accessibility. Thirdly, the current debate on centralization of childbirth care must take into account the sizeable proportions of women using PHC facilities for obstetric care. Care at childbirth is part of essential care, as defined in the Alma Ata declaration of PHC (Beard and Redmond, 1979), which also includes the 'scientifically sound' concept. Effective coverage is increasingly advocated, in place of contacts with care (Campbell et al., 2016;Marsh et al., 2020). In this context, to achieve effective coverage for the large proportion of women who uptake PHC-based care, qualityadjusted coverage (Marsh et al., 2020) must be available at the base of the health system pyramid (Hanson and Schellenberg, 2019;Straneo et al., 2014;Fogliati et al., 2015). Comprehensive EmOC in strategically identified rural health centres is one possible solution (Nyamtema et al., 2016); a locally adapted and community-participated reduction of birth sites may be necessary to balance quality and coverage of care (Fogliati et al., 2015). From a policy perspective, the position of childbirth care in PHC should be reappraised.

Limitations
This analysis is the first, to the best of our knowledge, to examine the use of different levels of the health system for childbirth among rural women in Tanzania and analysed the interaction between wealth and parity. It is based on nationally representative data, from a country that has consistently supported the development of a PHC network (Tanzania MoHSW, 2007;Tanzania MoHSW, MoHMZ, NBS, OCGS, and ICF, 2015), and thus is a model for countries developing rural obstetric care. The Tanzania DHS is unique in allowing identification of facility type (hospital, health centre or dispensary) in both the public and the private sectors, thus providing a more detailed picture of where women report giving birth (Tanzania MoHCDGEC, 2016). The DHS data set was complete, with very little non-response and missing data. Multinomial logistic regression allowed us to include all three locations in one model and thus study factors significant in use of hospital vs PHC facilities, and home vs PHC. Some caution should be applied when interpreting the findings, in terms of the cross-sectional nature of the DHS data and the possible response bias. Since in Swahili all facilities may be referred to as 'hospitali', lower-level public health facilities may be misreported as district hospitals. The DHS interviewers are instructed to circle a type of facility, if known, and if not, to write down the name of the facility, which is later coded as a specific type of facility by the field supervisor. This non-differential misclassification of facilities may bias results and may have led to weaker associations. This DHS collects limited information on obstetric risk factors (Virgo et al., 2017); thus, use of facilities due to risk factors identified during ANC (such as hypertension or maternal infections) cannot be fully captured. Even the risk factors for which some information is available may not fully reflect women's knowledge prior to index birth; e.g., limitations in DHS questionnaire identification of twin pregnancies have been described recently . This is unlikely to modify findings, as in a study on four East African countries no association between obstetric risk and birth location was found (Virgo et al., 2017), while wealth and education were strong determinants. Information on referral was lacking; thus, hospital births may include women who had been referred during labour from a PHC unit. In previous studies we found that intra-partum referral rates were very low (Straneo et al., 2016;Fogliati et al., 2017); thus, this too is unlikely to change the findings. Distance travelled to facilities could not be taken into account. Recent studies indicate that travel time is an important factor for hospital births in Tanzania , and current distribution leaves the rural poor underserved (Wong et al., 2019). Additionally, the study broadly categorized facilities by level (PHC and hospitals) but could not account for variation of quality of care within levels, such as more limited quality at hospital level or more advanced care in health centres (Nyamtema et al., 2016;Tanzania MoHSW, MoHMZ, NBS, OCGS, and ICF, 2015).
In conclusion, the study found that in rural Tanzania the use of hospital-based childbirth was not equitable. Inequity varied with parity level: at first birth, uptake varied only minimally with wealth, while in successive births it was strongly dependent on SES and parity. Uptake was the lowest amongst the poorest, multiparous women, with no increase in uptake at grand multiparity (≥5), in spite of increased risk. PHCbased childbirth accounted for a median of 35% of births after the first in this setting.
To leave no one behind in attaining Sustainable Development Goal 3 on maternal and preventable newborn mortality (Boerma et al., 2018), it is necessary to identify who is underserved at childbirth and make adjustments to improve the use of high-quality care, bearing in mind that from a human rights' perspective health care should contribute to equity (Tudor Hart, 1971). Strategies are needed to improve uptake of hospital-based care among the poorest, rural women, particularly at high parity. A reassessment of the whole district health system, which may involve re-evaluation of childbirth care in PHC and strengthening referral linkages, would benefit a substantial proportion of women.

Supplementary data
Supplementary data are available at Health Policy and Planning online.

Data availability statement
The dataset is available to the public at www.measuredhs. com.