Bypassing primary care clinics for childbirth: a cross-sectional study in the Pwani region, United Republic of Tanzania

Abstract

Objective

To measure the extent, determinants and results of bypassing local primary care clinics for childbirth among women in rural parts of the United Republic of Tanzania.

Methods

Women were selected in 2012 to complete a structured interview from a full census of all 30 076 households in clinic catchment areas in Pwani region. Eligibility was limited to those who had delivered between 6 weeks and 1 year before the interview, were at least 15 years old and lived within the catchment areas. Demographic and delivery care information and opinions on the quality of obstetric care were collected through interviews. Clinic characteristics were collected from staff via questionnaires. Determinants of bypassing (i.e. delivery of the youngest child at a health centre or hospital without provider referral) were analysed using multivariate logistic regression. Bypasser and non-bypasser birth experiences were compared in bivariate analyses.

Findings

Of 3019 eligible women interviewed (93% response rate), 71.0% (2144) delivered in a health facility; 41.8% (794) were bypassers. Bypassing likelihood increased with primiparity (odds ratio, OR: 2.5; 95% confidence interval, CI: 1.9–3.3) and perceived poor quality at clinics (OR: 1.3; 95% CI: 1.0–1.7) and decreased if clinics recently underwent renovations (OR: 0.39; 95% CI: 0.18–0.84) and/or performed ≥ 4 obstetric signal functions (OR: 0.19; 95% CI: 0.08–0.41). Bypassers reported better quality of care on six of seven quality of care measures.

Conclusion

Many pregnant women, especially first-time mothers, choose to bypass local primary care clinics for childbirth. Perceived poor quality of care at clinics was an important reason for bypassing. Primary care is failing to meet the obstetric needs of many women in this rural, low-income setting.

Introduction

Although maternal mortality is declining globally, it remains persistently high in sub-Saharan Africa, with 56% of the 287 000 maternal deaths worldwide in 2010 occurring in this area.¹ Reducing maternal mortality requires that women deliver with a skilled health-care professional who can detect and manage or refer obstetric complications that can arise without warning.^2,3 However, only half of all deliveries in sub-Saharan Africa are attended by health professionals. This is a central obstacle to reaching Millennium Development Goal 5, which calls for reducing maternal mortality by 75% between 1990 and 2015.^4,5 To increase coverage, countries in Africa and other low-income areas have focused on expanding the primary care system, typically through outpatient clinics staffed with nurses and midwives who can provide basic obstetric care.^6,7 These clinics represent the base of a service pyramid in which most women deliver at first-level clinics and women with high-risk pregnancies or those who develop complications are referred to hospitals. This service delivery model is supported by global policy guidance.^8–10

The United Republic of Tanzania is the largest country in eastern Africa. Its population is 34.4 million and life expectancy is 51 years.¹¹ The country has a maternal mortality ratio of 460 deaths per 100 000 live births, similar to the sub-Saharan African average, and a fertility rate of 5.4 children per woman.¹¹ Ninety-six per cent of women attend at least one antenatal care visit but just over half deliver in a health facility, with substantial regional variation.¹¹ Since independence, the United Republic of Tanzania has built up an extensive primary care system, with most people living 5 to 10 km from a clinic.¹² Primary care clinics (referred to locally as “dispensaries”) are outpatient facilities consisting of several rooms and are staffed by nurses and clinical officers who provide basic preventive and curative services, including delivery care. National policy states that primary care clinics should provide basic emergency care for delivery complications.¹³ However, primary care clinics tend to have poor infrastructure, to lack equipment and to be understaffed, particularly in rural areas.¹⁴

In this study, we explored women’s revealed preference for the primary-care-focused model of obstetric care. Specifically, we assessed the extent to which women in rural areas with good access to primary care clinics that offered obstetric services choose to deliver in health centres or hospitals and examined the individual and health system determinants of bypassing. We further explored the costs and quality of care received by women who bypassed. We then discuss the implications of our findings for health system organization.

Methods

Study setting and participants

This survey was based on baseline data collected for a cluster-randomized study that tested models for improving the quality of maternal health care in four districts of Pwani region in the United Republic of Tanzania: Bagamoyo, Kisarawe, Kibaha Rural and Mkuranga (ISRCTN 17107760). Pwani region is in the eastern part of the United Republic of Tanzania, north and east of the largest city, Dar es Salaam. The region is primarily rural and most of the population is employed in agriculture or unskilled manual labour.

The Ministry of Health assigns each village to a local primary care clinic (hereafter referred to as “clinics”). Clinics are open from 7:30 to 15:30, with providers available for emergency after-hours care, including delivery. Staff are trained and expected to perform basic emergency obstetric care and to refer patients with emergency conditions to hospital.¹⁵ In addition, each administrative division has health centres with inpatient services that serve from 6000 to 10 000 people and are staffed by nurses and assistant medical officers. Each district has at least one district hospital staffed by a few physicians and with an operating theatre.

Study facilities were 24 government-managed clinics with the highest volumes of deliveries between January and June 2011 in their districts. The clinics had a mean of four health-care workers, at least one of whom was a medically trained staff member (e.g. clinical officer or nurse midwife).

Women were selected to complete a structured interview from a full census of all households in the study clinic catchment areas (total, 30 076 households). Eligibility was limited to those who had delivered between 6 weeks and 1 year before the interview, were at least 15 years of age and lived within the catchment area of a study facility. Eligible women were informed of the purpose of the study and their right to refuse participation. Interviews were performed after receipt of written consent from the participant or, in the case of minors, upon receipt of assent from the participant and consent from their guardian.

The study was approved by the ethics review boards at Columbia University, the Ifakara Health Institute in the United Republic of Tanzania and the Tanzanian National Institute for Medical Research.

Procedures

The survey composing the structured interview and the consent forms were developed in English, translated to Swahili, back-translated to English and pre-tested to ensure accuracy. Questions covered demographic and household characteristics, child and maternal health and health-care characteristics, health system use and satisfaction and health-care preferences. The survey lasted 45 to 60 minutes and was conducted in Swahili by six teams of local interviewers using hand-held tablet computers with Pendragon Forms VI software (Pendragon Software Corporation, Chicago, United States of America). All interviewers underwent 11 days of training in research ethics, data collection methods and the survey instrument. Supervisors observed one interview per day and conducted partial re-interviews for 3% of interviews. The population-based survey of women, conducted as part of the baseline assessment for the cluster-randomized study, was conducted between 13 February and 28 April 2012.

Assessment of the 24 clinics was conducted from 5 December 2011 to 15 May 2012. This was done using a structured questionnaire adapted from the needs assessment created by the Averting Maternal Death and Disability Program and the United Nations system that has been previously used in more than 30 countries, including the United Republic of Tanzania.⁸ The survey included questions regarding human resources, infrastructure and the services available.

Statistical analysis

We conducted two separate analyses. In the first we explored the determinants of bypassing clinics and in the second we used bivariate analyses to compare the transaction costs and quality of care of bypassers and non-bypassers. Bypassers were defined as women who delivered their youngest child at a health centre or hospital (secondary or tertiary level facility) and were not referred there by a health-care provider; non-bypassers were defined as women who delivered their most recent child at a clinic.

On the basis of a literature review and the utilization model of Andersen, we identified the following determinants of bypassing: demographic and household characteristics, women’s experience with and perceptions of the health system and health system characteristics.^16–20 Variables were categorized to test hypotheses suggested by the extant literature and/or to reflect the distribution of our variables with the intent to create a parsimonious model. Demographic and household variables included age (teenage mothers and older mothers, who are potentially at higher risk, were compared to women aged 20–35 years); marital status; parity, because first-time mothers have been shown to seek higher-level facilities and may be counselled by providers to use those facilities; literacy; and television exposure, because media exposure has been shown to promote urban behavioural norms. A relative index of wealth was constructed using principal component analysis of a set of 18 questions on ownership of household assets.²¹ Women in the 80th wealth percentile or above were compared with those in lower wealth strata to assess whether wealth is an independent enabler of bypassing.

To assess the influence of clinics on bypassing, we also included women’s experience with their local clinics and subjective perceptions of quality of care. We compared respondents who reported a quality of care rating of fair or poor with those who reported alternate ratings. Frequent users of health care were defined as women who used their clinic for any reason more than five times over the previous year (frequency of use, 80th percentile or above). Objective measures indicative of quality of care at clinics were based on the framework first articulated by Donabedian.²² We selected the following structure and process measures: clinic infrastructure, performance of basic emergency obstetric care and participation in community outreach.

To build a multivariate model predicting bypassing, we conducted bivariate logistic regression analyses between bypassing and variables derived from the literature and categorized as described above. For the adjusted model, we retained variables that were significant at an α level of 0.10. We then performed multivariable logistic regression to estimate adjusted associations between potential determinants and bypasser status. We included a fixed effect for the woman’s district to account for between-district differences in population density, availability of a road network and availability and quality of higher-level facilities, all of which are managed by the district. Standard errors were adjusted using robust variance estimation to account for dependence of women living in the same clinic catchment area.

To compare transaction costs and delivery experience of women who bypassed versus those who did not, we asked women about the services they received, their perception of quality, their satisfaction with the services received, payment and the distance they travelled. Questions regarding each subject’s perception of the quality of care were asked using a five-point Likert scale ranging from “excellent” to “poor”. The subjects’ satisfaction with the care received was determined using a four-point Likert scale ranging from “very satisfied” to “very dissatisfied”. Their responses were then dichotomized into “excellent or very good” versus “good, fair or poor” for questions regarding quality and into “very satisfied” versus “somewhat satisfied, somewhat dissatisfied or very dissatisfied” for questions regarding satisfaction.

To determine the cost of each subject’s most recent delivery, we asked about specific costs for relevant goods and services. These costs were converted to United States dollars using the average exchange rate for the year before data collection.²³ We calculated a measure of distance to the delivery facility by using average speeds of travel (5 km/h by foot, 10 km/h by bicycle, 25 km/h by motorcycle and 50 km/h by car/bus) for the subject’s stated transportation method. We then conducted bivariate logistic regression to assess the association between bypasser status and variables related to women’s delivery experience.

Data were imported into Stata, version 12 (StataCorp LP, College Station, USA), for analysis and variables were examined for missing values and outliers. Univariate statistics were calculated for individual and clinic-level characteristics.

Results

We interviewed 3019 of 3238 eligible women (93%). Non-response was predominantly attributable to non-availability of the respondent after three attempts to locate her (204/3238 [6%]) and rarely because of refusal (15 [< 1%]). Of the interviewed women, 71.0% (2144) delivered their most recent child in a health facility; this proportion is consistent with population-representative data from Pwani region (73.1%).¹¹

A total of 246 women who delivered in a facility (11.5%) were referred to the health centre where they delivered. This left 1898 women for inclusion in the analysis. Of these, 794 (41.8%) bypassed their clinic and 1104 (58.2%) delivered at their clinic. Among the 794 interviewees who bypassed, 73.4% delivered in government hospitals. Additional characteristics of the 1898 participants, as well as information about their local health systems, are summarized in Table 1.

Table 1. Characteristics of the study population and local health system, United Republic of Tanzania, 2012.

Characteristic	Women (n = 1898)
Women
Demographic
Age (years)
< 20	248 (13.1)
20–35	1412 (74.4)
> 35	237 (12.5)
Muslim	1528 (80.5)
Currently married or living with partner	1578 (83.1)
Primary education or higher	1231 (64.9)
Literatea	1422 (75.1)
Primary occupation as farmer or homemakerb	1481 (79.0)
Self-rated health as goodc	1332 (70.2)
No. of children, mean (SD)	2.9 (1.8)
First delivery	484 (25.7)
Household assets
Electricity	143 (7.5)
Consumes > 2 meals per day	1890 (99.6)
Listens to the radio	1657 (87.3)
Watches television	456 (24.0)
Has a mobile phone	1466 (77.3)
Health care utilization and quality rating
At least 1 antenatal care visit	1816 (96.4)
≥ 4 antenatal care visits	1251 (66.4)
Has health insurance	116 (6.2)
> 5 visits to local clinic in past year	555 (30.7)
Perceived overall quality of care at local clinic as fair or poor	589 (33.3)
Local clinics
Electricity available	316 (16.6)
Performed ≥ 3 obstetric signal functions in past 3 monthsd	781 (41.1)
Upgrade or renovation within past year	384 (20.2)
Conducts community outreach	1611 (84.9)

SD, standard deviation.

^a Has completed primary school or was able to read all or part of a sentence in Swahili.

^b Includes homemakers, farmers and house cleaners.

^c Based on a “no problem” rating on a scale of all 5 items (mobility, self-care, usual activities, pain or discomfort, and anxiety or depression) in the EQ-5D instrument (EuroQol Group, Rotterdam, Netherlands).

^d Based on a maximum of 7, specified as follows: parenteral antibiotics, parenteral oxytocics, parenteral anticonvulsants, manual removal of placenta, removal of retained products, assisted vaginal delivery and newborn resuscitation with bag and mask.

Note: Data are no. (%) of subjects, unless otherwise indicated. For some rows, denominators differ from 1898 owing to missing data.

Among study participants, 1667 (87.8%) had complete information for all the variables of interest for the final regression model. Because the findings resulting from imputed analysis did not significantly differ from those resulting from non-imputed analysis, results of non-imputed analysis are presented. In the final adjusted model, bypassers were more likely than non-bypassers to be delivering their first child (odds ratio, OR: 2.53; 95% confidence interval, CI: 1.93–3.30; P < 0.001), to have used their clinic for any reason more than five times within the past year (OR: 1.23; 95% CI: 1.00–1.50; P = 0.045) and to have rated the quality of care at their clinic as poor (OR: 1.29; 95% CI: 1.00–1.66; P = 0.049) (Table 2). Bypassers were less likely than non-bypassers to be living in the catchment area of a clinic that had received an upgrade or renovation of some type within the past year (OR: 0.39; 95% CI: 0.18–0.84; P = 0.016) or had performed four or more signal functions (the most common of which was administration of parenteral oxytocics) for obstetric and neonatal care (OR: 0.19; 95% CI: 0.08–0.41; P < 0.001).

Table 2. Multivariable associations between bypasser status and characteristics of subjects and local primary care clinics, United Republic of Tanzania, 2012.

Characteristic	aOR (95% CI)	P
Women
Age (years)
20–35	1.0
< 20	0.83 (0.58–1.20)	0.330
> 35	1.17 (0.89–1.54)	0.252
First delivery	2.53 (1.93–3.30)	< 0.001
Literatea	1.14 (0.85–1.54)	0.387
80th percentile of wealth or above	1.44 (1.02–2.03)	0.036
Watches television	1.19 (0.92–1.54)	0.177
> 5 visits to local health clinic in past year	1.23 (1.00–1.50)	0.045
Perceived overall quality of care at local clinic as fair or poor	1.29 (1.00–1.66)	0.049
Local clinicsb
Electricity available	0.60 (0.27–1.35)	0.217
Upgraded or renovated in past year	0.39 (0.18–0.84)	0.016
Conducts community outreach	0.75 (0.36–1.56)	0.439
Obstetric signal functions performed, no.c
0–1	1.0
2	0.49 (0.23–1.04)	0.065
3	0.37 (0.14–0.99)	0.047
≥ 4	0.19 (0.08–0.41)	< 0.001

aOR, adjusted odds ratio; CI, confidence interval.

^a Has completed primary school or was able to read part of a sentence in Swahili.

^b The primary health facility for which the subject's house falls in the official catchment area.

^c Based on a maximum of 7, specified as follows: parenteral antibiotics, parenteral oxytocics, parenteral anticonvulsants, manual removal of placenta, removal of retained products, assisted vaginal delivery and newborn resuscitation with bag and mask.

Note: A total of 1667 (87.8%) of 1898 women who delivered at a facility had complete information for all variables of interest. The model also includes a fixed effect for district, not shown here, to account for differences in health system and infrastructural characteristics at this administrative level. See Methods for definitions of “bypasser” and “non-bypasser”.

More bypassers than non-bypassers reported receiving a blood transfusion (P = 0.049) and an examination before their own discharge (P < 0.001) and their neonates’ (P = 0.007; Table 3). Although they travelled significantly longer and paid significantly more for delivery, bypassers were more likely than non-bypassers to report being very satisfied with the overall delivery experience (P < 0.001). Bypassers’ ratings of quality of care were significantly higher than those reported by non-bypassers in six of seven quality measures.

Table 3. Bivariate analysis comparing travel, payment, and quality of care among bypassers and non-bypassers, United Republic of Tanzania, 2012.

Characteristica	Bypassersb			Non-bypassersc		cOR (95% CI)	P
Travel to facility
Walked to facility	47	(5.9)		330	(30.1)	1.00
Travelled by bicycle to facility	4	(0.5)		70	(6.4)	0.40 (0.14–1.15)	0.089
Travelled by motorcycle to facility	228	(28.8)		579	(52.7)	2.76 (1.96–3.89)	< 0.001
Travelled by car or bus to facility	515	(64.9)		119	(10.8)	30.38 (21.10–43.77)	< 0.001
Distance travelled (km), mean (SD)	36.8	(53.3)		10.6	(16.9)	1.04 (1.03–1.05)	< 0.001
Payment for delivery
Provider fees (US$), mean (SD)	2.2	(10.7)		2.4	(14.2)	1.00 (0.99–1.01)	0.731
Drugs, supplies, tests (US$), mean (SD)	5.0	(6.6)		3.6	(4.0)	1.06 (1.04–1.08)	< 0.001
Transportation (US$), mean (SD)	6.0	(8.1)		2.0	(4.5)	1.20 (1.16–1.23)	< 0.001
Total costs (US$), mean (SD)d	17.5	(28.4)		8.8	(15.9)	1.07 (1.00–1.08)	< 0.001
Borrowed or sold asset to pay for health care	122	(15.5)		129	(11.9)	1.36 (1.04–1.78)	0.023
Services received during and after delivery
Uterotonic	603	(76.1)		856	(78.3)	0.88 (0.71–1.10)	0.264
IV antibiotic or other drug	177	(22.3)		236	(21.6)	1.05 (0.84–1.30)	0.695
Blood transfusion	21	(2.7)		15	(1.4)	1.96 (1.00–3.82)	0.049
Baby examined before discharge	584	(74.8)		748	(60.0)	1.33 (1.08–1.64)	0.007
Mother examined before discharge	345	(43.5)		357	(32.7)	1.59 (1.31–1.92)	< 0.001
Very good/excellent rating of elements of delivery care
Respectful communication from health worker	381	(48.0)		443	(40.4)	1.36 (1.13–1.64)	0.001
Cleanliness of the delivery room	285	(36.0)		311	(28.5)	1.41 (1.16–1.72)	0.001
Privacy of the delivery room	317	(40.2)		426	(38.9)	1.05 (0.87–1.27)	0.577
Clarity of communication from health worker	310	(39.1)		370	(33.8)	1.26 (1.04–1.52)	0.017
Knowledge of the health worker	352	(44.6)		404	(36.9)	1.37 (1.14–1.65)	0.001
Availability of drugs and modern equipment	276	(35.3)		248	(22.9)	1.84 (1.50–2.25)	< 0.001
Overall quality of care	468	(58.9)		511	(46.7)	1.64 (1.36–1.97)	< 0.001
Very satisfied with the delivery experience	513	(64.6)		543	(49.5)	1.86 (1.54–2.24)	< 0.001

CI, confidence interval; cOR, crude odds ratio; IV, intravenous; SD, standard deviation; US$, United States dollar.

^a For all characteristics, data denote the ratio of the odds of the characteristic being reported by a bypasser (n = 794) to the odds of the characteristic being reported by a non-bypasser (n = 1104).

^b Defined as women who delivered their youngest child at a health centre or hospital (secondary or tertiary level facility) and were not referred there by a health-care provider.

^c Defined as women who delivered their most recent child at a primary care clinic.

^d Includes costs of transportation, provider fees, drugs, supplies, tests and tips. For context, the Tanzanian per capita gross domestic product was US$ 474 in 2012.

Note: Data are for 1898 women. For some rows, denominators differ from 1898, owing to missing data.

Discussion

We found that 41.8% of women who delivered children in a health-care facility in rural parts of the United Republic of Tanzania chose to deliver in a hospital or health centre rather than a local primary care clinic. This is striking because all women lived near a functioning clinic with delivery services and the sample excluded women who were referred to hospital by health providers.

Bypassing involved substantial logistical challenges and higher costs, largely driven by expenses related to transportation. Nearly 16% of bypassers reported having to borrow money or sell household assets to finance delivery, a measure of financial hardship that can lead to impoverishment.²⁴ These data underline that access to higher-level facilities is inequitable, with the wealthiest women more likely to obtain secondary services, and that despite the United Republic of Tanzania’s policy of free delivery, out-of-pocket costs for transport lead to financial hardship for families seeking hospital care.

Among women delivering their first child, the odds of bypassing were more than twice the odds of not bypassing. This may be because women and/or their antenatal care providers consider first deliveries risky and seek out advanced obstetric care for these pregnancies. Other evidence suggests that in the United Republic of Tanzania, first deliveries are associated with higher use of hospitals and, in China, with higher rates of maternal request for Caesarean delivery.^17,25

A total of 66% of subjects made four or more antenatal care visits and 31% had visited their clinic more than five times in the previous year for reasons of any type, but this did not encourage subjects to return to their clinic for delivery. This decision was not mediated by poor self-reported health and thus may have been prompted by having more information about the deficiencies of the clinic. Women’s reported rating of poor quality at their clinic further increased bypassing. Other studies have found that quality of care in hospitals is better than in lower-level facilities^14,26 and that the infrastructure and equipment at primary care clinics in the United Republic of Tanzania and other countries in the region is inadequate.^27,28 Dissatisfaction with quality of care at local facilities has been shown to motivate bypassing in other studies but the sole study that examined bypassing for delivery did not include objective measures of clinic quality.^17–19

In terms of the quality of their health care experience, bypassers were more likely than non-bypassers to be examined before discharge and they rated the quality of care significantly higher than non-bypassers on six of seven measures. Overall, women who delivered in hospitals or health centres were nearly twice as likely to be very satisfied with their birth experience as women who delivered in primary care. Other studies have similarly documented better quality of care in hospitals versus lower-level facilities.^15,29

A strength of this study was the availability of objective data on the quality of the structure and process of care in the primary care facilities. We found that recent facility upgrades in a woman’s catchment clinic reduced her probability of bypassing and that the number of basic emergency obstetric services performed in the clinic in the previous three months was a strong and consistent predictor of women’s behaviour. These results support the notion that Tanzanian women have good information about the obstetric care available in clinics and act on this information.¹⁸

The study has several potential limitations. Our analysis is based on cross-sectional data, which makes it impossible to draw causal inferences about the determinants of bypassing. However, since most predictors preceded the delivery, reverse causation is unlikely. ORs may overestimate likelihood when the outcome is relatively common, as in this analysis. The health system characteristics investigated are indicators of the capacity of the health system available to women and are not linked to the time at which the women delivered. Because our study participants were selected on the basis of good access to primary health care, our results are not generalizable to the rest of the United Republic of Tanzania, where bypassing is even higher. Because data were obtained via survey, recall bias is possible, although the potential for such bias is low because women delivered within one year of the survey. Finally, our data did not permit us to assess the availability of hospitals/health centres and the availability of transportation in bypassing decisions.

Our findings have important implications for health system organization. Although primary care serves many important functions, it may not be the optimal platform for providing high-quality delivery care. Policy-makers should consider gradually shifting the provision of obstetric care to health centres and hospitals, with primary care clinics playing a back-up role. This would have several advantages over the current model. First, it could improve health outcomes for mothers and neonates by concentrating obstetric care in high-volume settings with more competent providers and better equipment. Second, it would better match the expectations of pregnant women and thus strengthen the responsiveness of the health system to users. Finally, it could improve efficiency in the health system and free the funds invested, in maintaining the infrastructure and providers needed to attend (infrequent) deliveries, to be spent on other priorities. Such reorganization would require careful consideration of health system funding, transportation and referral. Health centres and hospitals would probably need to expand their maternity wards and vehicle fleets and improve transport protocols and communication. To promote equitable access to higher-level care, poor women will need to be given transport subsidies, vouchers and other support. Useful models for this exist in other countries.^29–32

Obstetric care needs to be of high quality and should be used universally to reduce maternal mortality. The preferences and experiences of women should thus inform health system design. Looking into the future, as Tanzanian families become wealthier, smaller and more experienced with the health-care system, bypassing is likely to rise. Shifting the focus of delivery care to health centres and hospitals could both save more lives and inspire greater confidence in the health system while making best use of scarce resources.