Where do women give birth in rural Tanzania ?

Introduction: Skilled birth attendance is one of the key factors in improving maternal and neonatal health but coverage is frequently less than 50% in many African and Asian countries, especially in rural areas. This article reports the findings on skilled birth attendance in a remote area with a large nomadic population in northern Tanzania. Methods: In a secondary analysis of data from a retrospective study on immunisation rates, data were compiled on the rates of skilled birth attendance at 8 mobile reproductive and child health clinics run by a rural first-referral hospital in the Mbulu area, covering the years 1998, 1999, 2006 and 2007. These data were analysed according to tribal affiliation and distance from health institutions with obstetric services. Results: Based on 3851 data sets, average rates of skilled birth attendance were 27%, 24%, 28% and 30% in 1998, 1999, 2006 and 2007, respectively (p = 0.02). At individual clinics, rates could be as low as 5–10%. Only at one clinic, significant improvement occurred over time (p < 0.01). In the univariate analysis, affiliation to the Iraqw tribe was a strong predictor of higher rates of skilled birth attendance in comparison with the nomadic Datoga tribe for all years combined (odds ratio [OR] 2.43 [95% confidence interval {CI} 1.92-3.07]), whereas distance showed only a minor influence (OR 1.02 [95% CI 1.01-1.02]). In the multivariate analysis, only tribal affiliation in 2007 (OR 2.69 [95% CI 1.12-6.46]) and for all years combined (OR 1.65 [95% CI 1.04-2.61]) was a significant factor. Conclusions: This study documented lower than the national average rates of skilled birth attendance in a rural area in Tanzania, especially among the nomadic Datoga tribe, over several years. The effect of distance was not consistent. To increase rates of women giving birth with skilled attendance in rural, remote settings and in populations with large proportions of nomadic people, a


Introduction
Qualified antenatal care, skilled birth attendance, access to emergency obstetric care and neonatal resuscitation skills are vital components to substantially reduce maternal, perinatal and neonatal mortality in developing countries 1,2 .The level of skilled birth attendance varies markedly among and within regions and countries, being well below 50% in many countries in South-East Asia and Sub-Saharan Africa 3 .
Although official nation-wide figures may show high coverage rates, this picture can be misleading.Typically, rates of skilled attendance are lower in rural than in urban areas 4,5 .This situation applies to Tanzania where in 2004/2005 the average rate of skilled attendance was as high as 81% in urban areas and as low as 39% in rural, remote districts 6 .Even within rural regions, marked differences may exist which can be related to cultural norms, socioeconomic circumstances, accessibility of health institutions and service provision 4,5,7 .In the case of nomadic populations, it is even more difficult to provide health services, including obstetric care with skilled birth attendance 8 .
In this context, findings are described from a secondary analysis of data on skilled birth attendance obtained during a large retrospective study on immunisation in a remote, rural area in the Manyara region in northern Tanzania 9 .In this area significant numbers of people belong to a nomadic tribe, and health institutions which provide delivery services are scattered over long distances [10][11][12][13] .The authors were especially interested to analyse whether changes would occur between the two periods covered by this study (1998/1999 and 2006/2007), and whether tribal affiliation and/or distance to health institutions with delivery services would affect rates of skilled birth attendance.

Setting
The original retrospective study on immunisation rates was conducted at the facilities of Haydom Lutheran Hospital (HLH), a rural church hospital in northern Tanzania.The HLH is located at the southern edge of Mbulu District, 380 km south-west of Mount Kilimanjaro.The 400 bed hospital serves a population of more than half a million people, and offers surgical, medical, gynaecological, obstetric, and paediatric services 10,12,13 .During the study period, the numbers of in-patients increased from 11 000 to 16 000, and out-patients from 65 000 to 85 000.Deliveries at the hospital increased from 2000 to 4500 12,14-17 .Mobile reproductive and child health (RCH) clinics located up to 100 km from HLH, serve the population on a monthly basis using a four-wheel-drive vehicle or a light aircraft for more remote locations.During 1998-1999 there were such 20 clinics, while by 2006-2007 there were 27 clinics.Annually, these clinics conduct 25 000 to 35 000 antenatal care examinations and 65 000 to 85 000 examinations of children under 5 years 12,[14][15][16][17] .Eight of these RCH clinics (Endaharghadat, Endanyawish, Endamilay, Getanyamba, Gosamwan, Harbarghet, Labay and Yaeda Chini) were purposively chosen to represent differences in tribal affiliation, geographic location and accessibility for the main parent study 9,[14][15][16][17] .

Population characteristics
The Mbulu area, located in the southern Karatu district and the Mbulu district, is a rural, difficult to reach area with few roads and little transportation infrastructure 10,12,13 .Its population consists of all four African language groups and significantly differing ways of life: the main tribes are the Cushitic Iraqw with a population of approximately 500 000, who are mainly subsistence farmers with some domestic cattle 10,13,18,19 ; and the Nilotic Datoga, who number approximately 100 000 to 200 000 people and are nomadic pastoralists, moving with their livestock over long distances 10,11,13,20,21 .The Khoisan Hadzabe, being true hunters-gatherers with a non-sedentary lifestyle, number only 1000-1500 people 22 .The Iraqw populate the highland plateau of the Mbulu and Karatu districts where the RCH clinics Endaharghadat, Endamilay, Getanyamba, Gosamwan, Harbarghet, and Labay are located 18,19 .The Datoga and Hadzabe reside in the Yaeda Valley between Lake Eyasi and the eastern escarpment (location of the RCH clinics Endanyawish and Yaeda Chini) [20][21][22] , with the Datoga additionally occupying areas south of the HLH catchment area in Hanang district 20,21 .The remaining tribes belong to the Bantu who are typically subsistence farmers or smallscale traders 10,13 .The Datoga and the Hadzabe are difficult to reach with any kind of social services, including healthcare provision 10,13,[20][21][22] .The Iraqw and Bantu communities are more easily accessible with social and health services 10,13,18,19 .Traditional birth attendants are present in the Iraqw and Datoga tribes, but they are not trained in modern obstetric care and co-operate only to a limited degree with the official health system [18][19][20][21] .

Data collection
Data on all infants from the 8 RCH clinics who were registered at these sites in the years 1998, 1999, 2006 and 2007, were included in the retrospective analysis.For the purpose of this secondary analysis, the following variables were included: 3. Status of skilled birth attendance for each infant (at a hospital, health centre, dispensary, or at home) registered at the respective RCH clinic.Births in health institutions (dispensary, health centre, hospital) were deliveries with skilled attendance.
This information was self-reported by the mothers when attending the RCH clinics for the first time with their infants.
Information about the variables tribal affiliation and distance was provided by the RCH staff.The data on skilled attendance were taken from the respective records of each single RCH clinic, which had been compiled by the RCH staff as part of the reporting requirements for the national immunisation program.All data were entered manually in data collection sheets, checked for inconsistencies and then entered into SPSS v15.0 (www.spss.com)for analysis.

Data analysis
Differences in rates of skilled birth attendance at delivery among the RCH clinics and among the 4 years of the study were analysed using the χ 2 test.In order to correlate the dependent variable (rate of skilled birth attendance) at the different clinics and over the time periods with possible underlying factors, both univariate and multivariate logistic regression analyses were performed using the independent variables of major tribal affiliation and distance to health institution.Odds ratios (OR) with 95% confidence intervals (CI) were calculated.The level of significance (two-sided) was defined as p < 0.05.

Ethical approval
Ethical approval for the main study was obtained from the  1, Fig1).The significant difference in the yearly average rates of skilled attendance was attributable to the difference between the years 1999 and 2007.
In 3 clinics (Getanyamba, Endaharghadat, Endamilay), no significant changes occurred during the 4 study years (Table 1, Fig1).In 2 clinics (Harbarghet, Gosamwan), there was a trend for higher rates of skilled birth attendance during the second period but this did not reach significance.At only one site (Labay), attendance rates improved significantly over time which was mainly related to higher numbers of women giving birth in the local Labay Dispensary (Table 1, Fig1).All these clinics were attended predominantly by the Iraqw people, and distance to a hospital was not more than 30 km; in the case of Labay and Gosamwan the distance was less than 5-10 km to the next dispensary with obstetric facilities as an alternative to the hospital.
Skilled birth attendance rates decreased over time at 2 clinics, more markedly at Endanyawish than at Yaeda Chini.This was significant for Yaeda Chini but not for Endanyawish, probably due to low numbers (Table 1, Fig1).
These 2 clinics were visited by the Datoga people and were located more than 50 km from the next health institution with obstetric services.
Differences of skilled attendance rates among the clinics were significant in each year when applying the χ 2 test (Table 2).In univariate logistic regression, affiliation to the Iraqw tribe was a strong predictor for higher rates of skilled birth attendance (OR

Discussion
The retrospective analysis revealed that rates of skilled birth attendance varied markedly among different RCH sites in a rural, remote and difficult-to-reach area and were, with a maximum mean of 30% in 2007, considerably lower than national rural figures from corresponding survey years (1999 and 2004/2005).
Only on the island of Pemba and in northern Zanzibar did less than 30% of women give birth with skilled attendance 6,23 .The national average for rural areas (39%) and for the administrative region of Manyara (36%) was rarely reached 6,23 .Only at one clinic (Labay) was there significant improvement over time, where a marked increase in women giving birth at the local dispensary contributed to this development.These findings are in contrast with another study from the same area which reported a rate of 57% of skilled birth attendance for the year 2000 24 .The data in the latter study were derived from facility records, and the high percentage was due to influx of women giving birth from neighbouring districts in search for quality delivery services.An earlier prospective study, conducted in 1995 and 1996, reported that 43% of women gave birth with skilled attendance; this study used a representative sample of the whole area and not only of more remote locations, thus the average rate was higher 25 .Endamilay.In another study from northern Tanzania which included mainly the Maasai tribe, who are comparable in culture to the Datoga, rates of skilled birth attendance were as low as 7% and similar to our results for this nomadic group 26 .Distance to health facilities exerted only a minor influence in 1999, 2006 and 2007 and for all years combined; in 1998 it was not significant.This latter result could be related to the Iraqw-dominated RCH site Harbarghet.Although the distance to the district hospital at Mbulu was only approximately 20 km, many women may have gone elsewhere to give birth due to concerns with the quality of obstetric care.Thus the effective distance to a wellfunctioning health unit with obstetric services (HLH) was more than 50 km.It has been found that rates of skilled birth attendance may not be a linear function of distance, but most likely also of perceived quality of care at the institution 4,5,7 .This is regularly affected by the state of the health facilities, staff availability, their skills and practices, drug supply, and the level of user fees.If quality of care is regarded as low, women will not attend even geographically close sites to give birth 27,28 .
The In the multivariate analysis, tribal affiliation in 2007 and for all years combined was the only remaining significant predictor.These results may reflect the fact that clinics that were mainly attended by Iraqw people and had low rates of skilled birth attendance for various reasons, were sometimes located close to health institutions with obstetric services (Harbarghet and Endamilay), thus diminishing the influence of tribal affiliation, especially in 1998 and 1999.
Tribal affiliation must be viewed as a proxy for the cultural differences between the Iraqw and Datoga and their approaches and access to health care 5,7,10,[18][19][20][21] .As the Datoga are mainly nomadic pastoralists, their ability and willingness to attend health institutions to give birth is restricted as they move with their cattle from one place to another 10,20,21 .At times official health services may be perceived as intimidating 5,7,27,29 .Furthermore, due to strong cultural traditions, childbirth is seen as a matter for the family, and the presence during childbirth of people outside the Datoga culture may be viewed unfavourably 10,20,21 .It is only during recent years that this cultural view has changed slightly, so it is not surprising that the rates at their sites are among the lowest in this study, and probably nationwide.However, the Iraqw encompass the concept of 'foreign healers' in their approach to seeking health care, so they more easily adjust to the modern approach to obstetric care 10,18,19,30,31 .
Although traditional birth attendants are present in both tribes, one study from Tanzania indicated some level of dissatisfaction with the services provided by traditional birth attendants 32 .To improve this situation it may be more appropriate to train community health workers from within the tribe in delivery skills 8,32 .Another approach could be to offer free or subsidised, planned and/or emergency transportation from home to a facility with obstetric services, in order to increase rates of institutional deliveries with skilled attendance.This approach was chosen by HLH, the major hospital in the area, and the number of deliveries at the hospital has increased considerably since this strategy was introduced 16,17 .In 2009, more than 4500 women gave birth at the hospital 17 .

Limitations of the study
A shortcoming of this study was the retrospective approach which made it impossible to collect information on other relevant factors, such as education, household wealth, health knowledge, and actual or perceived quality of care 5,7 .
Although it is commonly assumed that skilled birth attendance contributes to improved maternal and neonatal outcomes, the evidence for this assumption has to be questioned in the light of some recent studies 33,34 .No qualitative data for the reasons to give birth at home could be collected.Following the 2009 review by Gabrysch and Campbell 7 , it was only possible to investigate some aspects of sociocultural factors (tribal affiliation) and physical accessibility (distance), and aspects of economic accessibility or perceived need could not be analysed.Another disadvantage was that only the data of women registered at the RCH clinics could be collected.This group may not have been representative of the whole population.
Those not registered at the RCH clinics may have had even lower rates of skilled birth attendance at delivery, although the two other studies from the area do not confirm this assumption 24,25 .

Conclusions
This study documented lower than the national average rates of skilled birth attendance over several years in a rural area, especially among the nomadic Datoga tribe.Distance to a functioning health institution with obstetric services did not play a consistent role.To increase rates of skilled birth attendance in rural, remote settings and among populations with large proportions of nomadic people, a multi-facetted approach with education in and sensitisation to pregnancyand delivery-related issues, support for planned and emergency transportation, and improved quality of obstetric and neonatal services needs to be explored 26,[35][36][37] .

1 . 2 .
Major tribal affiliation at each RCH clinic.Data on tribal affiliation were not provided on an individual basis in the records because Tanzanian national policy does not favour collection of individual data with regard to tribal affiliation.Thus, major tribal affiliation was defined as the tribe to which the vast majority of infants seen at the particular RCH clinic belonged.Average distance from the catchment area of the respective RCH clinic to main healthcare provider, typically HLH or another health institution where obstetric services with skilled birth attendance were provided (Mbulu district hospital; Labay dispensary and Maretadu Juu dispensary).
improving figures for 2006 and 2007 at Harbarghet may indicate changes in the obstetric services at the district hospital in Mbulu.The example of Labay demonstrates that provision of a functioning health institution close to the population (Labay Dispensary) can contribute significantly to improved rates 5,7 .Low rates of skilled birth attendance at the Endamilay Clinic, which was predominantly used by the Iraqw people, may have contributed to the low OR during the years 1999, 2006 and 2007.It appeared that rates of skilled birth attendance were only close to national averages for rural areas where the distance to functioning health facilities (hospital or dispensary) was less than 20-25 km.

Table 2 : Tribal affiliation and distance to health institutions compared with rates of skilled birth attendance at all reproductive-and-child-health sites Skilled birth attendance Chi- square (df=7) Tribal affiliation (Iraqw vs Datoga) Distance † (km) Tribal affiliation & distance All clinics, year
AOR, Adjusted odds ratio; CI, confidence interval; df, degree(s) of freedom; HLH, Haydom Lutheran Hospital; OR, odds ratio.*P <0.05; †Distance to main health institution with delivery services.