Abstract

In an effort to reduce maternal mortality, developing countries have been investing in village-level primary care facilities to bring skilled delivery services closer to women. We explored the extent to which women in rural western Tanzania bypass their nearest primary care facilities to deliver at more distant health facilities, using a population-representative survey of households (N = 1204). Using a standardized instrument, we asked women who had a delivery within 5 years about the place of their most recent delivery. Information on all functioning health facilities in the area were obtained from the district health office. Women who delivered in a health facility that was not the nearest available facility were considered bypassers. Forty-four per cent (186/423) of women who delivered in a health facility bypassed their nearest facility. In adjusted analysis, women who bypassed were more likely than women who did not bypass to be 35 or older (OR 2.5, P ≤ 0.01), to have one or no living children (OR 2.2, P = 0.03), to have stayed in a maternity waiting home prior to delivery (OR 4.3, P ≤ 0.01), to choose a facility on the basis of quality or experience (OR 2.1, P ≤ 0.01), to have a high level of trust in health workers at the delivery facility (OR 2.7, P ≤ 0.01), and to perceive the nearest facility to be of low quality (OR 3.1, P ≤ 0.01). Bypassing for facility delivery is frequent among women in rural Tanzania. In addition to obstetric risk factors, a major reason for this appears to be a concern about the quality of care at government dispensaries and health centres. Investing in improved quality of care in primary care facilities may reduce bypassing and improve the efficiency and effectiveness of the health system in providing coverage for facility delivery in rural Africa.

KEY MESSAGES

  • Research has found that a high proportion of people bypass their nearest primary care facilities to seek care in higher level government facilities or private facilities. Bypassing is costly and inefficient for the individuals and the health system.

  • More than 40% of women who chose to deliver in health facilities in a poor, rural district of Tanzania bypassed their nearest health facility, choosing to deliver at the government hospital or mission facilities.

  • Perceived poor quality of care at nearby primary care facilities, as well as older age, fewer children, and staying at a maternity waiting home, were significantly associated with bypassing.

Introduction

Every year approximately 529 000 women die in pregnancy or around the time of childbirth, with the majority of deaths taking place in sub-Saharan Africa (Blum et al. 2006). Most of these deaths can be attributed to low rates of skilled birth attendance and inadequate use of emergency obstetric care (Ronsmans and Graham 2006). There is consensus in both the peer-reviewed literature and in the global public health practice community that most women should deliver in well-equipped and staffed clinics with capacity for basic emergency obstetric care (e.g. active management of third stage, oxytocic drugs to prevent haemorrhage, vacuum extraction) (WHO 2005; Koblinsky et al. 2006).

Many developing countries have designated village-level primary care facilities—variously called dispensaries, maternities or health centres—as the main point of care for uncomplicated delivery (Barnum and Kutzin 1993; Sanders et al. 1998; Campbell et al. 2006). These are often run by non-physician clinicians, such as clinical officers or nurse midwives who are trained to attend deliveries and to refer women with complications to hospitals (Mullan and Frehywot 2007). This pyramidal structure of health care delivery, with many primary-care facilities close to communities and district hospitals designated as referral centres, is seen as an efficient way to expand service coverage in resource-constrained countries with few hospitals and doctor shortages (Koblinsky et al. 2006).

However, research on health care utilization for common illness and preventive care in developing countries suggests that patients frequently bypass first-level facilities in favour of higher-level health centres and hospitals—this despite substantial additional time and financial costs. For example, three population-based studies found that half or more of survey respondents bypassed the nearest (usually lowest level) facility: for antenatal care, immunization and child illness in Kenya (Audo et al. 2005), and for outpatient care of episodic common illnesses in Sri Lanka (Akin and Hutchinson 1999) and Namibia (Low et al. 2001). Bypassing is seen as a powerful expression of people's preference for health care, and high rates of bypassing have important implications for health system efficiency and actual (versus planned) coverage of health services (Leonard et al. 2002). The extent of bypassing for facility delivery in the developing world—which is fundamentally different from preventive and curative care—is not known.

Tanzania is a low-income country in sub-Saharan Africa with a population of 34 million. The most recent estimate of the maternal mortality ratio (MMR) is 950 per 100 000 live births—in line with other countries in sub-Saharan Africa and over 100 times higher than in developed countries (WHO et al. 2007). One of the poorest countries in the world, Tanzania spends only US$7.27 per capita on health (National Bureau of Statistics, Tanzania and Macro International Inc. 2007). In rural Tanzania, village-level dispensaries are designated as the main point of preventive and curative care, including uncomplicated delivery. Dispensaries are typically staffed by one primary care provider—a nurse, clinical officer and/or maternal and child health (MCH) aide—although posts are frequently unfilled and absenteeism is common due to a severe human resource shortage (Kurowski et al. 2007). Health centres, the next level of care, are equipped to provide more complex treatment, including inpatient care. Surgery and referral-level care are provided by district and regional hospitals. The private sector—mainly faith-based organizations or missions—provides approximately one-third of all health services in Tanzania (National Bureau of Statistics, Tanzania and Macro International Inc. 2007).

Despite the low levels of health spending, the government of Tanzania, together with local churches, has established an extensive network of health facilities, with a strong focus on dispensaries. In 2006, there were 4679 dispensaries in the country (government and private) versus 481 health centres and 193 district-level hospitals (National Bureau of Statistics, Tanzania and Macro International Inc. 2007). As a result, an estimated 90% of the population lives within 10km of a health facility—in rural areas, usually a dispensary (National Bureau of Statistics, Tanzania and Macro International 1997). However, there are concerns about the quality of care available in dispensaries and health centres. A recent study in Tanzania found that only 13% of dispensaries provided services 24-hours a day with at least one provider, and that important obstetric equipment was frequently missing (National Bureau of Statistics, Tanzania and Macro International Inc. 2007). Providers in rural areas have also been found to be less skilled than those in urban areas (Leonard and Masatu 2007). It is unclear how women's decisions about facility delivery are influenced by the trade-off between relatively good geographic access and potentially poorer quality of primary care facilities, and how this in turn affects utilization of nearby dispensaries and health centres for childbirth.

The aim of this population-based study was to explore the extent to which women in rural western Tanzania report bypassing the nearest government dispensaries and health centres to deliver at the district hospital or mission facilities. We also sought to identify respondent characteristics associated with the decision to bypass.

Methods

Study area and sampling

Tanzania's Kasulu District is situated within Kigoma Region along the country's western border and has a population of 630 000. The district is primarily rural with one main town, Kasulu (population 33 000), which houses the government district hospital (National Bureau of Statistics 2008). Most of the population belongs to the Muha tribe. Most people in the district are subsistence farmers. Villages within the district are connected by unpaved roads, ranging in quality from dirt paths (tertiary roads) to a relatively smooth, wide road from the regional capital, Kigoma, to Kasulu town (primary road). Many of the roads are impassable during the rainy season between March and May.

As in other rural Tanzanian districts, the majority of health facilities in Kasulu are dispensaries. There are 48 functioning government dispensaries equipped to attend obstetric deliveries. The district has six government health centres, and the Kasulu district hospital, which offers obstetric surgery, is the government referral centre for delivery complications. There are also nine functioning mission dispensaries, two mission health centres and two mission hospitals. In addition, district residents occasionally use two health centres located in refugee camps along the Tanzania-Rwanda border. The survey team obtained the name, location and ownership (government, mission or refugee) of all functioning health facilities from the Kasulu District Health Office.

We selected a three-stage population-representative cluster sample of households from Kasulu District, omitting Kasulu town. Fifty villages were chosen in the first stage, with probability proportional to size, based on the 2002 Tanzania census. One subvillage (approximately 100 households) within each village was then randomly selected, and the leader of that subvillage provided a list of households from which 35 households were selected using random systematic sampling. Inclusion criteria limited participants to women over the age of 18 with a delivery within the previous 5 years. The National Institute for Medical Research in Tanzania and the Institutional Review Board at the University of Michigan provided ethics clearance. Informed consent via signature or thumbprint was obtained from all respondents.

Questionnaire content and fielding

Questionnaires and consent documents were developed in English then translated into Swahili and back translated. Questionnaire content was based on the available literature related to access to and barriers to maternal health care utilization in Africa, and focus groups with Tanzanian health providers. The questionnaires were pretested with rural women living in an adjacent district. Questionnaires included information related to household composition, characteristics and assets (indicators of socio-economic status/wealth), childbirth history, knowledge and perception of the local health care system, and barriers to health care utilization. Information was collected on the women's perception of the quality of care at their nearest dispensary, health centre and hospital using a Likert scale ranging from ‘excellent’ to ‘poor’. A similar Likert scale was used to assess women's trust in health workers at their delivery facility. Lastly, a detailed, to-scale district map was used to determine road distances between sample villages and all health facilities.

The questionnaire was administered in June and July 2007. Face-to-face interviews were conducted by two teams of trained interviewers fluent in Kiswahili and English. Each team also had at least one interviewer fluent in Kiha. Interviews lasted approximately 30 minutes. The quality of the interviewers’ work was monitored by a supervisor who observed two or more interviews per day.

Statistical analysis

Bypassers were defined as women who delivered their most recent child at a health facility other than the nearest facility to their village of residence. The nearest health facility was identified based on the shortest distance from the respondent's village along recognized roads. Non-bypassers were defined as women who delivered their most recent child at the health facility nearest to their village of residence.

A relative index of wealth was constructed based on the method developed by Filmer and Pritchett using reported ownership of household assets (Filmer and Pritchett 2001). We used principal components analysis (PCA) to define weights for ownership of specific assets. Women were then categorized into five wealth quintiles from poorest (quintile 1) to least poor (quintile 5). Four dichotomous ‘reasons for choosing delivery facility’ variables were created from responses to a question in which women were asked to supply all reasons for choosing their delivery facility from a list. This same ‘reasons for choosing delivery facility’ variable was used to identify women who were referred to the facility they ultimately delivered in. In addition, each woman rated her own health on the day of the survey. The cost of each woman's most recent delivery was calculated by asking about specific costs for doctor's fees, drugs, medical tests, transport, maternity waiting home services, and other (specified) costs, and summing these for a total.

We calculated univariate statistics for demographic variables, health and health system perception variables. We performed bivariate logistic regression between bypasser status and a large range of potential determinants of bypassing: age, education, wealth, distance to Kasulu town, previous facility delivery, stay at maternity waiting home, number of antenatal care visits, perceived quality of care at nearest facility, trust in health workers at delivery facility, and perceived importance of facility delivery. We then performed multivariable logistic regression to estimate adjusted associations between potential determinants and bypasser status. The variables selected for the multivariable model were either significant in the bivariate analysis or shown to be significant in previously published studies. We did not include referral by provider in the multivariable model because of collinearity with risk factors, such as advanced age. In addition, we explored differences between bypassers and non-bypassers in various characteristics of their deliveries, such as distance to delivery facility and cost of delivery.

Results

Of the 1322 eligible respondents recruited for the study, 1205 (91.1%) women completed questionnaires. Of the 117 non-\responses, 112 were due to failure to find the respondent at home despite repeated attempts and 5 were refusals to participate. One woman did not provide information on her location of delivery and as a result was excluded from the analysis. Table 1 provides summary statistics for the women surveyed. Seventy-one per cent of the women were between the ages of 18 and 35. Although 60.3% had completed primary school (the equivalent of 7 years of education), only 0.2% had any secondary education. The vast majority of respondents were farmers or fisherwomen (98.5%), Christian (91.1%) and of Muha ethnicity (98.3%). Most (74.5%) lived on a secondary or tertiary road, and 76.1% lived in a village with a functioning health facility. While 99.3% of the women made at least one antenatal care visit for their most recent pregnancy, only 36.4% delivered their most recent child in a health facility. It is worth noting that 59.8% of women who lived in a village with a functioning health facility delivered their most recent child at home. Furthermore, 449 (61.4%) of the 731 women delivering in the home had a government dispensary or health centre in their village (data available from authors).

Table 1

Socio-demographic and health care utilization characteristics of a population-based sample of women from Kasulu District, western Tanzania, 2007 (N = 1204)a

Characteristics n (%) 
Demographics   
Age   
    <25 288 (23.9) 
    25–34 567 (47.1) 
    ≥35 343 (28.5) 
Education   
    No schooling 331 (27.5) 
    Some primary 146 (12.1) 
    Completed primary 724 (60.1) 
    Some secondary or more (0.2) 
Occupation farmer or fisher 1186 (98.5) 
Currently married 1153 (95.8) 
Ethnicity Muha 1184 (98.3) 
Religion   
    Christian 1097 (91.1) 
    Muslim 89 (7.4) 
Number of living children   
    0–1 155 (12.9) 
    2–4 539 (44.8) 
    ≥5 501 (41.6) 
Quality of road in village   
    Primary 307 (25.5) 
    Secondary or tertiary 897 (74.5) 
Household assets   
Electricity (0.6) 
More than 2 meals per day 134 (11.1) 
At least 1 mosquito net 824 (68.4) 
Health care   
Distance to nearest health facility   
    In village 916 (76.1) 
    0.1–4.9 km 78 (6.5) 
    ≥5 km 210 (17.4) 
Antenatal care visits   
    0 (0.7) 
    1–2 197 (16.4) 
    ≥3 998 (82.9) 
Location of delivery   
    Homeb 731 (60.7) 
    Government dispensary 100 (8.3) 
    Government health centre 61 (5.1) 
    Government hospital 72 (6.0) 
    Mission health facilityc 205 (17.0) 
    On the way to a health facility 32 (2.7) 
Total births in a facility   
    0 439 (36.5) 
    ≥1 758 (63.0) 
Characteristics n (%) 
Demographics   
Age   
    <25 288 (23.9) 
    25–34 567 (47.1) 
    ≥35 343 (28.5) 
Education   
    No schooling 331 (27.5) 
    Some primary 146 (12.1) 
    Completed primary 724 (60.1) 
    Some secondary or more (0.2) 
Occupation farmer or fisher 1186 (98.5) 
Currently married 1153 (95.8) 
Ethnicity Muha 1184 (98.3) 
Religion   
    Christian 1097 (91.1) 
    Muslim 89 (7.4) 
Number of living children   
    0–1 155 (12.9) 
    2–4 539 (44.8) 
    ≥5 501 (41.6) 
Quality of road in village   
    Primary 307 (25.5) 
    Secondary or tertiary 897 (74.5) 
Household assets   
Electricity (0.6) 
More than 2 meals per day 134 (11.1) 
At least 1 mosquito net 824 (68.4) 
Health care   
Distance to nearest health facility   
    In village 916 (76.1) 
    0.1–4.9 km 78 (6.5) 
    ≥5 km 210 (17.4) 
Antenatal care visits   
    0 (0.7) 
    1–2 197 (16.4) 
    ≥3 998 (82.9) 
Location of delivery   
    Homeb 731 (60.7) 
    Government dispensary 100 (8.3) 
    Government health centre 61 (5.1) 
    Government hospital 72 (6.0) 
    Mission health facilityc 205 (17.0) 
    On the way to a health facility 32 (2.7) 
Total births in a facility   
    0 439 (36.5) 
    ≥1 758 (63.0) 

aTotals may not add up to 1204 due to missing values.

bOf the 731 women, 717 gave birth in their own home, 11 in another's home, and 3 in a field.

cOf the 205 women, 86 gave birth in a mission dispensary, 71 in a mission health centre, and 48 in a mission hospital.

Figure 1 is a flowchart illustrating how women were distributed according to delivery location. Of the 441 women who delivered their most recent child in a health facility, 237 (53.7%) delivered at their nearest health facility (non-bypassers) while 186 (42.2%) bypassed their nearest health facility. A further 18 women (4.1%) could not be classified: 3 did not give specific information about the facility of delivery and 15 were transferred from one facility to another during labour due to complications, with no information about the first facility. The final sample for analysis was 423 women. Only nine bypassers (4.8%) delivered in a government dispensary or health centre, with the remainder delivering at the government hospital or in mission facilities.

Figure 1

Place of delivery and extent of bypassing of government dispensaries in a population-based sample of women from Kasulu District, Western Tanzania, 2007

Figure 1

Place of delivery and extent of bypassing of government dispensaries in a population-based sample of women from Kasulu District, Western Tanzania, 2007

Table 2 shows bivariate associations between potential determinants of bypassing and bypasser status. Women who bypassed their nearest health facility for delivery were not significantly different from women who delivered at their nearest facility in terms of age, socio-economic status, education, number of antenatal care visits, and distance from Kasulu town. We found significant associations between bypassing and number of living children (0–1: OR 2.1, P = 0.04, compared with 2–4 children), moderate, bad or very bad self-reported health (OR 1.9, P = 0.03), stay at a maternity waiting home (OR 3.4, P ≤ 0.01), reason for choosing delivery facility (closest to home: OR 0.03, P ≤ 0.01; presence of the best providers in the area: OR 2.5, P ≤ 0.01), perceived quality of care at their nearest facility (OR 1.6, P = 0.05), and trust in the workers at the facility they ended up delivering at (OR = 1.6, P = 0.04). Table 2 also permits the calculation of the rate of bypassing of government facilities: among the 303 women for whom the nearest facility was a government primary care facility (dispensary or health centre), 151 (49.8%) bypassed it.

Table 2

Bivariate associations between respondent and nearest health facility characteristics and bypasser status for a population-based sample of women from Kasulu District, Western Tanzania, 2007a

 Non-bypassers
 
Bypassers
 
  
 (n = 237) (%) (n = 186) (%) OR P-value 
Risk factors       
Age       
    <35 174 (73.4) 137 (73.7) ref  
    ≥35 63 (26.6) 49 (26.3) 1.0 0.96 
Number of living children       
    0–1 36 (15.2) 51 (27.4) 2.1 0.04 
    2–4 106 (44.7) 73 (39.2) ref  
    ≥5 94 (39.7) 59 (31.7) 0.9 0.79 
Self-reported health       
    Very good or good 205 (86.5) 146 (78.5) ref  
    Moderate, bad or very bad 30 (12.7) 40 (21.5) 1.9 0.03 
Number of antenatal care visits       
    <4 106 (44.7) 93 (50) ref  
    ≥4 131 (55.3) 93 (50) 0.81 0.38 
Stayed at a maternity waiting home       
    No 123 (51.9) 45 (24.2) ref  
    Yes 113 (47.7) 141 (75.8) 3.4 ≤0.01 
Total births in a facility       
    1–2 130 (54.9) 122 (65.6) ref  
    ≥3 106 (44.7) 61 (32.8) 0.6 0.06 
Demographic factors       
Wealthb       
    1st quintile 44 (18.6) 35 (18.8) ref  
    5th quintile 47 (19.8) 35 (18.8) 0.9 0.84 
Education       
    No schooling 61 (25.7) 45 (24.2) ref  
    Some schooling 176 (74.3) 141 (75.8) 1.1 0.71 
Distance from Kasulu townc, mean (SD) 3.8 (1.6) 3.6 (1.9) 0.9 0.60 
Nearest facilityd       
    Government dispensary 94 (39.7) 97 (52.2) ref  
    Government health centre 58 (24.5) 54 (29.0) 0.9 0.83 
    Mission 85 (35.9) 35 (18.8) 0.4 0.16 
Perception factors       
Reasons for choosing delivery facilitye       
    Closest to home 211 (89.0) 32 (17.2) 0.03 ≤0.01 
    Best doctors, nurses, other staff in the area 49 (20.7) 74 (39.8) 2.5 ≤0.01 
    Has drugs 20 (8.4) 28 (15.1) 1.9 0.07 
    Recommended by relative/friend or good previous experience 50 (21.1) 54 (29.0) 1.5 0.13 
Perceived quality of care at nearest facility       
    Excellent or very good 138 (58.2) 88 (47.3) ref  
    Good, fair or poor 91 (38.4) 91 (48.9) 1.6 0.05 
Trust in health workers at delivery facility       
    Low 95 (40.1) 54 (29.0) ref  
    High 140 (59.1) 129 (69.4) 1.6 0.04 
Stated importance of delivering in a facility       
    Less than very important 48 (20.3) 26 (14.0) ref  
    Very important 189 (79.7) 160 (86.0) 1.6 0.12 
 Non-bypassers
 
Bypassers
 
  
 (n = 237) (%) (n = 186) (%) OR P-value 
Risk factors       
Age       
    <35 174 (73.4) 137 (73.7) ref  
    ≥35 63 (26.6) 49 (26.3) 1.0 0.96 
Number of living children       
    0–1 36 (15.2) 51 (27.4) 2.1 0.04 
    2–4 106 (44.7) 73 (39.2) ref  
    ≥5 94 (39.7) 59 (31.7) 0.9 0.79 
Self-reported health       
    Very good or good 205 (86.5) 146 (78.5) ref  
    Moderate, bad or very bad 30 (12.7) 40 (21.5) 1.9 0.03 
Number of antenatal care visits       
    <4 106 (44.7) 93 (50) ref  
    ≥4 131 (55.3) 93 (50) 0.81 0.38 
Stayed at a maternity waiting home       
    No 123 (51.9) 45 (24.2) ref  
    Yes 113 (47.7) 141 (75.8) 3.4 ≤0.01 
Total births in a facility       
    1–2 130 (54.9) 122 (65.6) ref  
    ≥3 106 (44.7) 61 (32.8) 0.6 0.06 
Demographic factors       
Wealthb       
    1st quintile 44 (18.6) 35 (18.8) ref  
    5th quintile 47 (19.8) 35 (18.8) 0.9 0.84 
Education       
    No schooling 61 (25.7) 45 (24.2) ref  
    Some schooling 176 (74.3) 141 (75.8) 1.1 0.71 
Distance from Kasulu townc, mean (SD) 3.8 (1.6) 3.6 (1.9) 0.9 0.60 
Nearest facilityd       
    Government dispensary 94 (39.7) 97 (52.2) ref  
    Government health centre 58 (24.5) 54 (29.0) 0.9 0.83 
    Mission 85 (35.9) 35 (18.8) 0.4 0.16 
Perception factors       
Reasons for choosing delivery facilitye       
    Closest to home 211 (89.0) 32 (17.2) 0.03 ≤0.01 
    Best doctors, nurses, other staff in the area 49 (20.7) 74 (39.8) 2.5 ≤0.01 
    Has drugs 20 (8.4) 28 (15.1) 1.9 0.07 
    Recommended by relative/friend or good previous experience 50 (21.1) 54 (29.0) 1.5 0.13 
Perceived quality of care at nearest facility       
    Excellent or very good 138 (58.2) 88 (47.3) ref  
    Good, fair or poor 91 (38.4) 91 (48.9) 1.6 0.05 
Trust in health workers at delivery facility       
    Low 95 (40.1) 54 (29.0) ref  
    High 140 (59.1) 129 (69.4) 1.6 0.04 
Stated importance of delivering in a facility       
    Less than very important 48 (20.3) 26 (14.0) ref  
    Very important 189 (79.7) 160 (86.0) 1.6 0.12 

aData are n (%) unless otherwise specified.

b1st quintile corresponds to ‘poorest’ and 5th quintile corresponds to ‘least poor’.

cMeasured in 10 km.

dThere were no individuals for whom the nearest facility was a government hospital.

eResults based on a question allowing multiple responses. Each reason given analysed independently as a dichotomous variable.

Table 3 summarizes differences in characteristics of the delivery between bypassers and non-bypassers. There were significant associations with bypasser status and the following: delivery facility (government hospital: OR 1788.2, P ≤ 0.01; mission dispensary: OR = 21.8, P ≤ 0.01; mission health centre: OR 15.2, P ≤ 0.01; mission hospital: OR 35.8, P ≤ 0.01), transportation to the delivery facility (bicycle: OR 3.0, P ≤ 0.01; car: OR 36.6, P ≤ 0.01; public transport: OR 155.1, P ≤ 0.01, compared with walking), time of travel (OR 6.7, P ≤ 0.01), distance travelled (OR 2.3, P ≤ 0.01), having been referred by a provider (OR 52.8, P ≤ 0.01), and preference for future delivery in a hospital (OR 6.0, P ≤ 0.01). While referral from a health provider was strongly associated with bypassing, only 18.3% of bypassers reported being referred.

Table 3

Bivariate associations between characteristics of the delivery and bypasser status for a population-based sample of women from Kasulu District, Western Tanzania, 2007a

 Non-bypassers
 
Bypassers
 
  
 (n = 237) (%) (n = 186) (%) OR P-value 
Delivery facility       
Government dispensary 94 (39.7) (3.2) ref  
Government health centre 58 (24.5) (1.6) 0.8 0.84 
Government hospital 0b,c (0.0) 61 (32.8) 1788.2 ≤0.01 
Mission dispensary 36 (15.2) 50 (26.9) 21.8 ≤0.01 
Mission health centre 35 (14.8) 34 (18.3) 15.2 ≤0.01 
Mission hospital 14 (5.9) 32 (17.2) 35.8 ≤0.01 
Travel to delivery facility       
Mode of transportation       
    Walking 143 (60.3) 43 (23.1) ref  
    Bicycle 88 (37.1) 80 (43.0) 3.0 ≤0.01 
    Car (personal or borrowed) (1.3) 33 (17.7) 36.6 ≤0.01 
    Public transportation 0b,c (0.0) 23 (12.4) 155.1 ≤0.01 
Time of travel,dmean (SD) 0.7 (1.1) 2.1 (2.3) 6.7 ≤0.01 
Distance and cost       
Distance to delivery facility,emean (SD) 0.5 (4.8) 20.9 (28.3) 2.3 ≤0.01 
Cost of delivery,fmean (SD) 3.2 (12.9) 8.5 (14.4) 1.1 0.12 
Referred to facility by provider 
No 236 (99.6) 152 (81.7) ref  
Yes (0.4) 34 (18.3) 52.8 ≤0.01 
Satisfaction after delivery       
Less than very satisfied 87 (36.7) 56 (30.1) ref  
Very satisfied 149 (62.9) 130 (69.9) 1.4 0.19 
Preference for location of future delivery       
Dispensary 92 (38.8) 43 (23.1) ref  
Health centre 113 (47.7) 57 (30.6) 1.1 0.86 
Hospital 23 (9.7) 64 (34.4) 6.0 ≤0.01 
 Non-bypassers
 
Bypassers
 
  
 (n = 237) (%) (n = 186) (%) OR P-value 
Delivery facility       
Government dispensary 94 (39.7) (3.2) ref  
Government health centre 58 (24.5) (1.6) 0.8 0.84 
Government hospital 0b,c (0.0) 61 (32.8) 1788.2 ≤0.01 
Mission dispensary 36 (15.2) 50 (26.9) 21.8 ≤0.01 
Mission health centre 35 (14.8) 34 (18.3) 15.2 ≤0.01 
Mission hospital 14 (5.9) 32 (17.2) 35.8 ≤0.01 
Travel to delivery facility       
Mode of transportation       
    Walking 143 (60.3) 43 (23.1) ref  
    Bicycle 88 (37.1) 80 (43.0) 3.0 ≤0.01 
    Car (personal or borrowed) (1.3) 33 (17.7) 36.6 ≤0.01 
    Public transportation 0b,c (0.0) 23 (12.4) 155.1 ≤0.01 
Time of travel,dmean (SD) 0.7 (1.1) 2.1 (2.3) 6.7 ≤0.01 
Distance and cost       
Distance to delivery facility,emean (SD) 0.5 (4.8) 20.9 (28.3) 2.3 ≤0.01 
Cost of delivery,fmean (SD) 3.2 (12.9) 8.5 (14.4) 1.1 0.12 
Referred to facility by provider 
No 236 (99.6) 152 (81.7) ref  
Yes (0.4) 34 (18.3) 52.8 ≤0.01 
Satisfaction after delivery       
Less than very satisfied 87 (36.7) 56 (30.1) ref  
Very satisfied 149 (62.9) 130 (69.9) 1.4 0.19 
Preference for location of future delivery       
Dispensary 92 (38.8) 43 (23.1) ref  
Health centre 113 (47.7) 57 (30.6) 1.1 0.86 
Hospital 23 (9.7) 64 (34.4) 6.0 ≤0.01 

aData are n (%) unless otherwise specified.

b0.5 added to cells in calculation of β.

cYates continuity correction in calculation of P value.

dMeasured in hours.

eMeasured in km.

fMeasured in 1000 TZS.

Table 4 shows the results of the multivariable analysis. In terms of pregnancy risk factors, bypassers were more likely to be over the age of 35 (OR 2.5, P ≤ 0.01), to have one or fewer children (OR 2.2, P = 0.03), and to have stayed in a maternity waiting home (OR 4.3, P ≤ 0.01), while they were less likely to have five or more living children (OR 0.5, P = 0.04). Other significant predictors of bypasser status include: a reason for choosing delivery facility related to best providers, drug availability, recommendation by a relative/friend, or a good previous experience (OR 2.1, P ≤ 0.01), perceived less than very good quality of care at the nearest facility (OR 3.1, P ≤ 0.01), and high trust in health workers at the delivery facility (OR 2.7, P ≤ 0.01). Women's perception of the importance of facility delivery was not significant in predicting bypassing.

Table 4

Multivariable associations between participant and nearest health facility characteristics and bypasser status for a population-based sample of women from Kasulu District, Western Tanzania, 2007 (n = 387)

 OR P-value 
Risk factors   
Age   
    <35 ref  
    ≥35 2.5 ≤0.01 
Number of living children   
    0–1 2.2 0.03 
    2–4 ref  
    ≥5 0.5 0.04 
Self-reported health   
    Very good or good ref  
    Moderate, bad or very bad 1.6 0.16 
Number of antenatal care visits   
    <4 ref  
    ≥4 0.8 0.39 
Stayed at a maternity waiting home   
    No ref  
    Yes 4.3 ≤0.01 
Demographic factors   
Wealth   
    1st quintile ref  
    5th quintile 1.0 0.93 
Education   
    No schooling ref  
    Some schooling 1.0 0.91 
Distance from Kasulu town 0.7 0.05 
Nearest facilitya   
    Government dispensary ref  
    Government health centre 0.6 0.37 
    Mission facility 0.2 0.02 
Perception factors   
Reason for choosing facility: best provider, drugs available, recommended by relative/ friend, or good previous experience   
    No ref  
    Yes 2.1 ≤0.01 
Perceived quality of care at nearest facility   
    Excellent or very good ref  
    Good, fair or poor 3.1 ≤0.01 
Trust in health workers at delivery facility   
    Low ref  
    High 2.7 ≤0.01 
Stated importance of delivering in a facility   
    Less than very important ref  
    Very important 1.4 0.24 
 OR P-value 
Risk factors   
Age   
    <35 ref  
    ≥35 2.5 ≤0.01 
Number of living children   
    0–1 2.2 0.03 
    2–4 ref  
    ≥5 0.5 0.04 
Self-reported health   
    Very good or good ref  
    Moderate, bad or very bad 1.6 0.16 
Number of antenatal care visits   
    <4 ref  
    ≥4 0.8 0.39 
Stayed at a maternity waiting home   
    No ref  
    Yes 4.3 ≤0.01 
Demographic factors   
Wealth   
    1st quintile ref  
    5th quintile 1.0 0.93 
Education   
    No schooling ref  
    Some schooling 1.0 0.91 
Distance from Kasulu town 0.7 0.05 
Nearest facilitya   
    Government dispensary ref  
    Government health centre 0.6 0.37 
    Mission facility 0.2 0.02 
Perception factors   
Reason for choosing facility: best provider, drugs available, recommended by relative/ friend, or good previous experience   
    No ref  
    Yes 2.1 ≤0.01 
Perceived quality of care at nearest facility   
    Excellent or very good ref  
    Good, fair or poor 3.1 ≤0.01 
Trust in health workers at delivery facility   
    Low ref  
    High 2.7 ≤0.01 
Stated importance of delivering in a facility   
    Less than very important ref  
    Very important 1.4 0.24 

aThere were no individuals for whom the nearest facility was a government hospital.

Discussion

We found that more than 4 in 10 women in a rural district of Tanzania who delivered in a health facility bypassed their nearest facility to deliver their baby elsewhere. The frequency of bypassing rose to nearly 50% when the nearest facility was a government primary care facility—dispensary or health centre. Sixty-two per cent of bypassers selected mission facilities and 33% selected the government district hospital, with only 5% choosing to deliver at a government dispensary or health centre, despite the fact that government dispensaries and health centres comprise 77% of all the health facilities in the district.

Women who chose to bypass faced substantial obstacles. The women in the sample as a whole were comparatively poor; virtually none had electricity and only 11% reported having more than two meals per day. Women who bypassed travelled an average of 20 km farther (40 km roundtrip) and were 36 times more likely to have to use a car to get to a health facility than women who did not bypass. The high use of mission facilities also meant substantial additional costs given that mission facilities charge for delivery whereas delivery care in government facilities is exempt from official user fees, although under-the-table payments are sometimes charged (Mamdani and Bangser 2004). The total costs of delivery (including medical and transport costs) were substantially higher among bypassers than non-bypassers (mean of 8500 TZS versus 3200 TZS). Lastly, bypassing for delivery is logistically complex as women must either travel while in labour or plan in advance to relocate to a distant village or town to await labour (e.g. in the maternity waiting home or at the home of relatives). Bypassers also incur a substantial opportunity cost due to longer time away from their other children and farming work. In light of these difficulties, the high frequency of bypassing documented here is remarkable.

These findings are consistent with some of the available research on bypassing in the developing world. For example, Audo et al. (2005) reported that between 46.3% and 59.5% of mothers interviewed in a rural district in Kenya bypassed the lowest level municipal (government) facility in favour of district or provincial hospitals when seeking antenatal care, child immunization or other child health services. Akin and colleagues also found very high levels of bypassing in a mixed urban and rural district of Sri Lanka in 1992, where 66.5% of survey respondents with a minor or major illness in the past month reported bypassing their nearest health facility. In particular, primary care facilities that were the planned entry point into the health care system were bypassed more often than chosen (Akin and Hutchinson 1999). On the other hand, in a 1994 study in a rural district of Tanzania, Leonard et al. (2002) found that approximately 12.6% of facility visits for common medical conditions (adult and child) represented bypassing of a closer facility. This figure is lower than ours, possibly due to higher density of health facilities in the district Leonard and colleagues studied, and thus greater availability of preferred facilities nearby—as suggested by the authors’ finding that the average patient travelled only 2.8 km farther (one-way) to bypass (Leonard et al. 2002) This study was conducted 13 years ago and it is possible that the higher prevalence of bypassing observed in our study reflects an increase in bypassing behaviour in Tanzania. This can only be confirmed by longitudinal research.

Factors associated with higher obstetric risk, such as age over 35, having no previous living children, and having stayed at a maternity waiting home, were associated with bypassing. However, risk factors do not tell the whole story. First, only 18% of all bypassers reported being referred to the higher-level facility by a provider. Second, even controlling for obstetric risk, several other factors emerged as important predictors of bypassing. These were largely related to quality of care. For example, in multivariable analysis we found that stronger preference for quality (best provider, drugs, etc.), lower perceived quality of care at the nearest facility, and greater trust in health workers at the facility selected for delivery were all associated with higher odds of bypassing. Living near a mission facility, which have been shown by other researchers to provide better equipment and have more trained providers, reduced the odds of bypassing (Leonard and Masatu 2007). These findings suggest that perceived quality of both technical (drugs, equipment) and non-technical (trust in health workers) aspects of health care have a major influence on women's choice of delivery facility.

Although we did not collect data on the quality of care at the health facilities in the study district, other researchers have documented quality concerns at government dispensaries and health centres in Tanzania, particularly in rural areas. For example, a 2006 national facility survey found that only 7% of dispensaries had all basic delivery room infrastructure (bed, examination light, visual and auditory privacy). Private (including mission) facilities (of all levels) scored much better than government facilities on this indicator, with 34% having all of the inputs versus 6% of government facilities. Only 35% of dispensaries (government and private) had any emergency transportation—a barrier to referral for emergencies that may motivate some women to bypass (National Bureau of Statistics, Tanzania and Macro International Inc. 2007).

In addition, as noted earlier, few dispensaries operate 24 hours per day (despite the expectation that the provider can be called at any hour in case of emergency). The lack of 24-hour services at government dispensaries may have also contributed to bypassing in our study. However, the impact of this on the decision to bypass was likely limited as women labouring at night would have a very difficult time finding transport to allow them to travel to a more remote facility, and thus would be more likely to deliver at home than bypass.

In a recent study of the quality of care at government and non-governmental (primarily mission) facilities in Tanzania, Leonard and Masatu (2007) found worse practice quality (e.g. accuracy of diagnosis and appropriateness of management) among clinicians in rural government facilities than in urban or peri-urban government facilities, whereas rural clinicians at non-governmental facilities performed as well as their urban counterparts. Qualitative work has also found that poor quality of care—both technical (e.g. equipment, drugs) and non-technical (e.g. provider attitude)—at primary care facilities is commonly reported by Tanzanian women (Gilson et al. 1994; Mamdani and Bangser 2004).

Other researchers have also found that bypassing is tied to perceived and objectively observed quality of care at the bypassed and chosen facility. For example, Akin and Hutchinson (1999) reported that in Sri Lanka, facilities with fewer doctors and drugs and in poor structural condition were more likely to be bypassed, controlling for individual characteristics. Similar results were reported in Namibia, where it was found that patients who bypassed were motivated by issues of quality (better facilities and staff), as well as access (facility proximity and hours of operation) (Low et al. 2001). Leonard found that patients tended to bypass facilities that overused injections and overprescribed drugs, and those that had poor consultation practices, as measured by trained observers. This suggests that bypassers have an accurate perception of several important dimensions of care quality. The most frequent reasons for bypassing municipal facilities given by women in Kenya were: poor care (37% of respondents), lack of drugs (30.4%), and lack of laboratory services (21.2%) (Audo et al. 2005).

We did not find any association between wealth and education and bypassing. Although this may seem surprising given the cost and effort involved in travelling to a farther facility, Akin and Hutchinson (1999) also reported no differences in bypassing the nearest government primary care facilities between poorest and least poor groups in Sri Lanka. Audo et al. (2005) found that more educated women tended to bypass municipal facilities in greater numbers than less educated women, but this may have been confounded by distance as more educated women lived closer to town and thus closer to the preferred urban hospitals. Our study population was also relatively homogeneous in terms of asset ownership and demographic characteristics (ethnicity, occupation, etc.), which may in part explain the lack of association between wealth and bypassing.

Women living further from Kasulu Town—the site of the government District Hospital—were somewhat less likely to bypass. Women with five or more living children were less likely to bypass, perhaps indicating a lower perception of risk (despite higher actual risk) and greater comfort with delivering in a primary care facility. There were no differences in the perception of health benefits of facility delivery between bypassers and non-bypassers.

Our study has several limitations. We do not have data on observed or objective quality of care in the government and mission facilities in the study district. Future research combining a population-based with a facility-level survey would be valuable to compare perception of quality with actual quality. However, studies confirm poor quality is problematic in rural health facilities in Tanzania and that patients are generally well aware of quality deficiencies (Leonard et al. 2002; Leonard and Masatu 2007). In addition, other researchers point out that women's perceptions of quality—the focus of this study—are an important determinant of their behaviour, separate and apart from observed quality (Mrisho et al. 2007).

As with other work on bypassing, our analysis was focused on women who used the health system rather than the larger group who did not. However, data from women who delivered in the home support the notion that primary care facilities are not serving this population effectively: 61% of women who delivered at home had a dispensary or health centre in their village. It is likely that at least some of them weighed the quality of care at the dispensary and chose instead to deliver with a traditional birth attendant or relative.

Our findings have several important policy implications. First, the high rates of bypassing in a resource-poor country such as Tanzania are concerning as bypassing shifts health care expenditure away from direct health care costs and into indirect costs such as transport. Bypassing among the poor imposes a particularly large financial burden on the most vulnerable families putting them at risk of deepening poverty. Second, the high rate of bypassing of government dispensaries and health centres combined with high rates of home deliveries in villages with these facilities suggests that the current primary care facilities are not meeting the needs of rural women.

In essence, we document here that in rural western Tanzania broad availability of primary health care facilities does not translate into utilization of these facilities for childbirth. This highlights the challenge facing Tanzania and other countries with high maternal mortality in scaling up facility deliveries to meet the maternal health Millennium Development Goal. This work, together with a growing body of research on patient preferences for health care, suggests that quality is a crucial determinant of women's decision on place of delivery. Investing in improved quality of care in primary care facilities—from provider skills and attitudes to better drug supply—could reduce the financial and logistical burden of bypassing on families and improve overall health system efficiency, while expanding access to life-saving maternal health services.

Acknowledgements

This work was supported by the William Davidson Institute of the Ross School of Business at the University of Michigan and by the Averting Maternal Death and Disability Program (AMDD) of the Mailman School of Public Health at Columbia University. AMDD is funded by the Bill and Melinda Gates Foundation. The funders had no role in the study design, data collection, analysis, interpretation, writing of the paper or the decision to submit the article for publication.

MEK, GM, CWM and SG jointly designed of the study and coordinated the fieldwork. MEK and SG supervised data entry and analysis. PCR led the data analysis. MM assisted in study design and the interpretation of results. All co-authors made critical revisions of the manuscript. The authors declare they have no conflict of interest.

References

Akin
JS
Hutchinson
P
Health-care facility choice and the phenomenon of bypassing
Health Policy and Planning
 , 
1999
, vol. 
14
 (pg. 
135
-
51
)
Audo
MO
Ferguson
A
Njoroge
PK
Quality of health care and its effects in the utilisation of maternal and child health services in Kenya
East African Medical Journal
 , 
2005
, vol. 
82
 (pg. 
547
-
53
)
Barnum
H
Kutzin
J
Public hospitals in developing countries: resource use, cost, financing
 , 
1993
Washington DC
The World Bank
Blum
LS
Sharmin
T
Ronsmans
C
Attending home vs. clinic-based deliveries: perspectives of skilled birth attendants in Matlab, Bangladesh
Reproductive Health Matters
 , 
2006
, vol. 
14
 (pg. 
51
-
60
)
Campbell
O
Graham
WJ
The Lancet Maternal Survival Series steering group
Strategies for reducing maternal mortality: getting on with what works
The Lancet
 , 
2006
, vol. 
368
 (pg. 
1284
-
99
)
Filmer
D
Pritchett
LH
Estimating wealth effects without expenditure data–or tears: an application to educational enrollments in states of India
Demography
 , 
2001
, vol. 
38
 (pg. 
115
-
32
)
Gilson
L
Alilio
M
Heggenhougen
K
Community satisfaction with primary health care services: an evaluation undertaken in the Morogoro region of Tanzania
Social Science and Medicine
 , 
1994
, vol. 
39
 (pg. 
767
-
80
)
Koblinsky
M
Matthews
Z
Hussein
J
, et al.  . 
The Lancet Maternal Survival Series steering group
Going to scale with professional skilled care
The Lancet
 , 
2006
, vol. 
368
 (pg. 
1377
-
86
)
Kurowski
C
Wyss
K
Abdulla
S
Mills
A
Scaling up priority health interventions in Tanzania: the human resources challenge
Health Policy and Planning
 , 
2007
, vol. 
22
 (pg. 
113
-
27
)
Leonard
KL
Masatu
MC
Variations in the quality of care accessible to rural communities in Tanzania
Health Affairs
 , 
2007
, vol. 
26
 (pg. 
w380
-
92
)
Leonard
KL
Mliga
GR
Mariam
DH
Bypassing health centres in Tanzania: revealed preferences for quality
Journal of African Economics
 , 
2002
, vol. 
11
 (pg. 
441
-
71
)
Low
A
de Coeyere
D
Shivute
N
Brandt
LJ
Patient referral patterns in Namibia: identification of potential to improve the efficiency of the health care system
International Journal of Health Planning and Management
 , 
2001
, vol. 
16
 (pg. 
243
-
57
)
Mamdani
M
Bangser
M
Poor people's experiences of health services in Tanzania: a literature review
Reproductive Health Matters
 , 
2004
, vol. 
12
 (pg. 
138
-
53
)
Mrisho
M
Schellenberg
JA
Mushi
AK
, et al.  . 
Factors affecting home delivery in rural Tanzania
Tropical Medicine and International Health
 , 
2007
, vol. 
12
 (pg. 
862
-
72
)
Mullan
F
Frehywot
S
Non-physician clinicians in 47 sub-Saharan African countries
The Lancet
 , 
2007
, vol. 
370
 (pg. 
2158
-
63
)
National Bureau of Statistics. 2008
Tanzania Census.
 , 
2002
Dar es Salaam
National Bureau of Statistics, Tanzania, and Macro International
Tanzania demographic and health survey 1996.
1997
Dar es Salaam: National Bureau of Statistics, and Calverton, MD: Macro International
National Bureau of Statistics, Tanzania, and Macro International
Tanzania Service Provision Assessment Survey 2006.
2007
Dar es Salaam: National Bureau of Statistics
Ronsmans
C
Graham
WJ
Maternal mortality: who, when, where, and why
The Lancet
 , 
2006
, vol. 
368
 (pg. 
1189
-
200
)
Sanders
D
Kravitz
J
Lewin
S
McKee
M
Zimbabwe's hospital referral system: does it work?
Health Policy and Planning
 , 
1998
, vol. 
13
 (pg. 
359
-
70
)
WHO
World Health Report 2005: Make Every Mother and Child Count
 , 
2005
Geneva
World Health Organisation
WHO, UNICEF, UNFPA, World Bank
Maternal mortality in 2005: estimates developed by WHO, UNICEF, UNFPA and The World Bank
 , 
2007
Geneva
World Health Organization