Perinatal audit using the 3-delays model in western Tanzania

1. Introduction 

Worldwide, approximately 7 million perinatal deaths occur each year, of which 98% occur in low- and middle-income countries [1]. In 2000, 38% of deaths of infants younger than 5 years old occurred in the neonatal period, and 4 million infants each year were estimated to die in the first 4 weeks of life [2]. Three-quarters of these died in the first week of life.

The measurement of perinatal mortality is error prone. Some early neonatal deaths occur after discharge from hospital and are not reported because registration of deaths at home is lacking in most low-income countries [3]. Because of logistic challenges and the almost universal practice of early discharge, very few studies have been conducted involving both hospital and community settings. Where such studies have been conducted, interesting results have been obtained. In northwestern Tanzania, the perinatal mortality rate in the community was found to be 68 per 1000 live births, compared with 96 per 1000 in the nearby hospital [4]. The highest perinatal mortality rate, however, was for high-risk births at home (167 per 1000 live births). Low-risk births in hospital had the lowest mortality rate (14 per 1000 live births). The rate for low-risk births at home (73 per 1000 live births) was similar to the rate for high-risk births in hospital (82 per 1000 live births) [5].

Many studies have explored the causes of perinatal deaths and have shed light on their etiologies in different settings [6], [7], [8], [9], [10], [11]. These studies utilized the Baird, Wigglesworth, or Nordic–Baltic classifications of perinatal deaths, all of which seem to be appropriate, even at district level in low-income countries [12], [13], [14], [15].

Many perinatal deaths could be avoided with appropriate maternal and neonatal health care, and it has been estimated that up to half of these deaths are the result of poorly managed deliveries [16]. Hospital-based studies in low-income countries have identified suboptimal care in 10%–77% of perinatal deaths [17]. Many conditions can be appropriately treated without significant investment in health care and, thus, can be managed in settings with limited resources [18], [19].

In the UK, inquiries into maternal deaths have had a crucial role in reducing the rate of such fatalities [20]. A similar audit approach has been applied to perinatal deaths—a process that has been multinational [21], [22]. Perinatal audit addresses the question of why a perinatal death has occurred. This can be used to detect substandard care and, thus, may promote improvement of the quality of care. The aim of our perinatal audit was to use the “3-delays” model, developed by Thaddeus and Maine to audit maternal deaths, for the analysis of perinatal mortality in a regional hospital in western Tanzania [23].

2. Methods 

The study population included all births in the Department of Obstetrics of the Maweni Regional Hospital in Kigoma, Tanzania, from July 2002 to July 2004. Maweni is the referral hospital for the Kigoma region, which has 1.6 million inhabitants [24].

A detailed questionnaire was used for all perinatal deaths in the hospital. The questionnaire included details of the mothers' backgrounds (age, marital status, and place of residence) and complete medical, obstetric, family, and social histories. To ascertain the duration of labor, attention was paid to the details of parturition that took place prior to admission to the regional hospital. Attention was also paid to fetal viability on admission and the care given from the onset of labor. Information included the time taken to get from home to the institution after the onset of labor, the type and cost of transport, whether the mother was referred from a lower-level health facility, and the diagnosis made at the lower-level health facility (if applicable). The care provided during labor was noted, including institutional delays, type of attendant, decisions on management, mode of delivery, complications, and resuscitation. Partogram use was also assessed. All neonates were examined immediately after birth in the hospital and details were recorded, including weight, sex, presence of congenital anomalies, and most importantly whether fetal death had occurred prior to or during labor. All neonatal deaths (up to 7 days) that occurred in the hospital were recorded; no autopsies were performed. Sick neonates received standard care according to the hospital protocols.

The questionnaire was filled in by one of the authors (CK) after interviewing the mother and checking the records. Counterchecking was done by another author (GM). In addition to the completed questionnaires, prenatal cards and clinical records were studied in detail when available. The data were analyzed by a local team consisting of a consultant obstetrician, a consultant pediatrician, 2 medical officers, and 2 nurse–midwives.

All intrapartum fetal and early neonatal deaths in which the birth weight was ≥2000 g were audited using the 3-delays methodology. This methodology classifies delays as follows: (1) first delay refers to the time spent at home before a decision is made to seek health care; (2) second delay refers to the inability to get to a health facility after having made the decision to go there (due to problems with transport, money, or poor roads); and (3) third delay refers to the time spent waiting for adequate management after arrival at a health facility. In addition to the Thaddeus and Maine model, we considered “mismanagement or incorrect treatment” as a third delay in that it delayed appropriate management.

For each perinatal death, a “process audit” was undertaken, whereby actual practice was compared with standard practice, based on evidence or expert consensus. Fetal monitoring was conducted mainly through intermittent auscultation, and resuscitation of asphyxiated neonates involved clearing the airway, simple drying, ventilation with bag and mask, and immediate skin-to-skin contact with the mother. Neonatal care comprised use of the “Kangaroo method,” breastfeeding and close observation of early signs and symptoms of infection. Sick neonates were seen by a consultant pediatrician.

The research protocol was approved by the Ethics Committee of the Karolinska Institute in Sweden and the National Institute of Medical Research in Tanzania. Written consent was obtained from each participant.

3. Results 

During the study period, 385 perinatal deaths occurred in 10200 hospital births (38 per 1000 live births). Birth weight was less than 2000 g in 185 (48%) of these cases, leaving 132 intrapartum fetal deaths and 68 early neonatal deaths in which the birth weight was ≥2000 g (52%) to be included in the audit. The leading cause of death was birth asphyxia (n=38; 19.0%); followed by prolonged or obstructed labor (n=37; 18.5%); antepartum hemorrhage (n=23; 11.5%); uterine rupture (n=18; 9.0%); cord prolapse (n=15; 7.5%); and breech delivery (n=14; 7.0%) (Table 1). Congenital anomalies accounted for only 7 cases (3.5%). Most infants (n=161; 80.5%) weighed ≥2500 g. There was 1 set of twins, and 3 mothers out of 199 (1.5%) died.

Table 1. Clinicopathologic causes of intrapartum fetal and early neonatal deaths in which birth weight was ≥2000 g (n=200).

	Cause of death	No. (%)
	Birth asphyxia	38 (19.0)
	Prolonged or obstructed labor	37 (18.5)
	Antepartum hemorrhage	23 (11.5)
	Uterine rupture	18 (9.0)
	Cord prolapse	15 (7.5)
	Breech delivery	14 (7.0)
	Malpresentation	10 (5.0)
	(Pre-)Eclampsia	9 (4.5
	Sepsis	9 (4.5)
	Congenital anomalies	7 (3.5)
	Shoulder dystocia	2 (1.0)
	Unknown causes	18 (9.0)

Analysis of the care received in terms of the 3-delays model is provided in the following sections.

3.1. First delays 

First delays occurred in 38 cases (19.0%) (Table 2). Most women were admitted for delivery only after labor had started. Nine women who had been advised to deliver in hospital attended only after complications had occurred at home. In another 38 cases, pregnancy was unwanted and its prevention could have avoided these perinatal deaths.

Table 2. First delaysa among intrapartum fetal and early neonatal deaths in which birth weight was ≥2000 g (n=200).

	First delays	No. (%)
	Wanted to deliver at home, attended after complication occurred	26 (13.0)
	Advised to have hospital delivery, but attended only after complication occurred	9 (4.5)
	Refused intervention	2 (1.0)
	No prenatal care	1 (0.5)
	Total	38 (19.0)

a Thaddeus and Maine [23].

3.2. Second delays 

Second delays occurred in 43 cases (21.5%) (Table 3). The major issues here were unspecified problems with transport after labor had started (27 cases). Six women had long waiting times before embarking on a boat, 5 were transported on a stretcher, and 3 arrived by bus or lorry. In 1 case, a car breakdown caused serious delay and another woman had no money for transport.

Table 3. Second delaysa among intrapartum fetal and early neonatal deaths in which birth weight was ≥2000 g (n=200).

	Second delays	No. (%)
	Unspecified transport problems	27 (13.5)
	Long wait before embarking on a boat	6 (3.0)
	Transported on stretcher	5 (2.5)
	Transported on bus or truck	3 (1.5)
	Car breakdown	1 (0.5)
	No money	1 (0.5)
	Total	43 (21.5)

a Thaddeus and Maine [23].

An example of first and second delays was the case of a grandmultiparous woman with 5 previous stillbirths (1 after cesarean delivery). She had a 13-hour first stage of labor at home, experienced 6 hours of transport delays, and sustained uterine rupture on arrival at the hospital.

3.3. Third delays 

Third delays occurred in 145 cases (72.5%) (Table 4). Cesarean delivery was performed in 67 cases (33.5%). In 19 cases, no fetal heartbeat could be heard at the time of the decision to operate. In 13 women, the indication was for so-called “obstructed” labor, although the membranes were intact at the time of the operation, making such a diagnosis unlikely. Three cases of cesarean delivery were due to huge hydrocephalus, which went undetected during the first stage of labor.

Table 4. Third delaysa among intrapartum fetal and early neonatal deaths in which birth weight was ≥2000 g (n=200).

	Third delays	No. (%) (n=145)
	Delay in dispensary and/or health center	21 (14.5)
	Delay in first stage of labor in hospital	19 (13.1)
	Delay in second stage of labor in hospital	23 (15.9)
	Uterine rupture in a health facility	10 (6.9)
	Clinical mismanagement
	Unskilled breech delivery	14 (9.7)
	Elective cesarean delivery 1 month too early	1 (0.7)
	Cesarean delivery for prolonged or obstructed labor, intact membrane	13 (9.0)
	Cesarean performed when vacuum preferable	23 (15.9)
	Cesarean for hydrocephalus	3 (2.1)
	Cesarean for dead fetus	19 (13.1)
	Undetected twins	5 (3.4)
	Mismanagement of (pre-)eclampsia	4 (2.8)
	Non-standard treatment of neonatal infection	5 (3.4)
	Total (women, not conditions)b	145/200 (72.5)

a Thaddeus and Maine [23]. b Total number of conditions is greater than 145 because different conditions sometimes occurred in the same woman.

Rupture of the uterus occurred in 18 cases (9.0%); 8 occurred in hospital and 2 occurred in dispensaries or health centers. Although health workers often state that women with uterine rupture present late, 10 of 18 cases (55.6%) occurred after admission to a health facility. In 1 case, it remained unclear whether rupture had occurred at home or in a dispensary. Only 3 women had a uterine scar from a previous cesarean delivery.

Partograms were graded into categories of absent (1.5%), non-satisfactory (10.5%), and satisfactory (88.0%). Doctors were called for consultation in 80 cases (40%), and the average time for arrival was 1 hour (range, 0.5–4.0 hours). After daytime working hours, doctors were called from home when abnormalities were noted on the partogram or when specific emergencies were detected (e.g. severe pre-eclampsia, eclampsia, prolonged labor, previous cesarean delivery, or antepartum hemorrhage). In the case of cesarean deliveries, the average time between decision and operation was 1 hour (range, 0.25–2.5 hours).

Occassionally, women with high-risk pregnancies were cared for by low-cadre staff. This resulted in disastrous situations, especially breech deliveries where auxiliary nurses failed to deliver the after-coming head (n=14). Half of the neonatal deaths were due to birth asphyxia, and infections accounted for nearly 20%. Hypothermia was documented in 10% of the neonatal deaths, while cerebral hemorrhage and hypoglycemia were each recorded in 4% of cases. In the remaining 12% of neonatal deaths, the cause of death could not be determined. Most of these early neonatal deaths occurred at night.

4. Discussion 

To our knowledge, the present study is the first in which the 3-delays model has been applied to perinatal mortality. Health workers tend to refer to first and second delays rather than to delays that occur after women enter a health facility (third delay). In this audit, however, the majority of substandard care occurred within the health facilities: third delays occurred in 72.5% of perinatal deaths, compared with second and first delays happening in 21.5% and 19.0% of perinatal deaths, respectively.

By allocating perinatal deaths to etiologic categories with direct clinical implications, one may identify problem areas in care provision [12], [13], [14], [15]. We found that 19% of pregnancies in our study were unplanned, confirming findings from Pakistan [8]. Family planning is an important part of the further reduction of perinatal mortality rates.

First and second delays are important because many births still occur at home and referral practices still cause problems—as observed in this study, where 81 of 226 delays (35.8%) occurred at home owing not only to poor transport and long distance, but also lack of trust in the health system [25], [26]. These delays had important bearings on perinatal outcome. A significant proportion of the pregnant women did attend peripheral health facilities, where some spent a considerable amount of time before health workers decided to refer them to the regional hospital. This may indicate problems ranging from the degree of the peripheral health workers' knowledge to a lack of transport from these areas. Indeed, 61% of uterine ruptures occurred in women who had been admitted to health facilities before rupture occurred. There is a general feeling in the community that pregnant women will go to hospital only when labor is well established or when they perceive complications. Furthermore, it is sometimes the relatives rather than the pregnant woman herself who decide that she should go to hospital [27]. Improving institutional care could have a positive impact on the decision at home to seek care, thereby reducing first delays.

The majority of perinatal deaths that occurred in the regional hospital were associated with third delays. This has been observed in both low-income [17], [28], [29] and high-income countries, where substandard care has been recognized as a major cause of perinatal death [21].

In the present study, 132 deaths (66%) were intrapartum, and the birth weight was ≥2500 g in 161 cases (80.5%), implying that all of these deaths occurred in viable fetuses. This high incidence of normal birth weight intrapartum fetal deaths was similar to findings in Sudan, Kenya, 6 West African countries, and Bangladesh and points to a lack of adequate obstetric care during labor [18], [29], [30], [31], [32]. When we compared perinatal mortality in our study with data from elsewhere, we concluded that there was an excess of intrapartum and neonatal deaths of infants weighing greater than 2000 g [33].

In this hospital-based study, the majority of early neonatal deaths occurred at night. This was at least partly because of a shortage of qualified night staff; on average, only 1 midwife was present and many deliveries were assisted by nurse attendants. In addition, neonatal hypothermia is more likely to occur at night, especially for preterm and low birth weight infants. Emphasis should be on monitoring through regular intermittent auscultation of fetal heart sounds, especially in the late first and second stages of labor. Appropriate action should be taken as soon as abnormalities are detected.

Observation of the partogram and history taking also revealed that prolonged labor both outside and inside the hospital was associated with a substantial number of intrapartum and neonatal deaths. Misinterpretation of clinical signs and mismanagement were also significant contributory factors to intrapartum deaths; this has also been reported in audits performed in other areas of the continent [34], [35].

Audits have been accepted as a systematic critical analysis of medical care. The review of circumstances of adverse health outcomes, especially in cases of maternal and perinatal death, has received increasing attention as a means of improving the quality of care [22]. Benefits have been observed not only in high-income countries, but also increasingly in low-income countries [35], [36], [37]. Audit exercises are not easy to undertake, however, because they require strong commitment and staff time. When audits are performed only by specialists based in urban settings, the contextual factors that contribute to local practices, including underlying reasons for poor performance, may be overlooked [11], [13].

In many instances, cesarean delivery is mentioned as a major intervention to reduce perinatal mortality. In our study, cesarean delivery had been performed in 33% of cases of perinatal death; this seems to be a very poor outcome for an intervention meant to prevent maternal and perinatal mortality. Three maternal deaths occurred: 1 after uterine rupture and 2 after cesarean delivery for obstructed labor with intrauterine fetal death and full dilatation of the cervix. The lives of these latter 2 women might have been saved if birth had been achieved by the vaginal route. When the fetus is already dead in utero, the best mode of delivery is by the vaginal route because the dangers of abdominal delivery are much more profound [38].

The large number of perinatal deaths in our study also indicates the lack of equipment such as Doppler for fetal auscultation, oxygen, and naso-gastric tubes in the hospital studied. Most studies indicate, however, that basic affordable methods do exist that could save substantial numbers of infants who, instead, are stillborn or die during the first days of life. When these technologies are applied with proper evaluation and monitoring, they could save millions of lives in a cost-effective manner [39].

Our inability to analyze perinatal mortality occurring outside health facilities is a limitation of the present study. Home birth is often the preferred way of giving birth, although improved hospital care may influence women's decisions to seek early care in institutions instead of waiting too long and jeopardizing the outcome.

5. Conclusion 

Perinatal mortality can be elucidated in low-resource settings similar to the one investigated through the systematic analysis of perinatal deaths using the simple 3-delays methodology. It is, however, unrealistic to expect reductions in mortality rate at the same pace in all countries because much depends on the economy, individual professional devotion, political will, and deliberate efforts to combat the problems in a defined setting. This methodology appears to be generally useful for analyzing not only maternal but also perinatal deaths.