Cover Image
close this book06. Can Skilled Attendance at Delivery Reduce Maternal Mortality in Developing Countries?
View the document(introduction...)
View the documentSummary
View the documentIntroduction
View the documentWhat is Skilled Attendance?
View the documentHow Can Skilled Attendance Work at the Individual Level?
View the documentHow Can Skilled Attendance Work at the Population Level?
View the documentReferences

How Can Skilled Attendance Work at the Individual Level?

One of the major rationale often cited for prioritising skilled attendance at delivery is the concentration of maternal deaths around this time, with an estimated two-thirds occurring in late pregnancy through to 48 hours after delivery (AbouZahr 1998a). Assuming therefore that the primary involvement of the skilled attendant is from the onset of labour to the immediate puerperium (up to 48 hours), the four major direct causes of maternal death amenable to intervention are shown in Table 1, along with estimates of their incidence and case-fatality. If this involvement is extended to late pregnancy and the first week after delivery, then the proportion of eclamptic, antepartum haemorrhage and puerperal sepsis cases encountered by the skilled attendant is likely to be increased, but the other major causes - ectopic pregnancy and complicated abortion, would remain excluded.

Table 1. Major causes of maternal death close to the time of delivery, for developing countries ~1990


Incidence (Rate per 100,000 females 15-44)

Case-fatality rate1 (%)

Complication

Highest estimate2

Lowest estimate3

Highest estimate2

Lowest estimate3

Obstructed labour

1422

354

9

0

Eclampsia

1185

442

17

2

Puerperal sepsis

2370

531

11

1

Obstetric haemorrhage

2370

885

17

4

1 Derived from estimated number of deaths divided by estimated number of incident cases.
2 Estimates for sub-Saharan Africa
3 Estimates for China

* SOURCES: Data extracted from AbouZahr 1998b, AbouZahr 1998c, AbouZahr and Guidotti 1998, AbouZahr et al . 1998.

The role of skilled attendance in averting deaths from the causes in Table 1 could be both through primary and secondary prevention. Figure 3 illustrates the main points for intervention along the causal pathway to death. Through appropriate case management (including referral), skilled attendance can prevent complications directly. The scope for primary prevention clearly varies according to the complication, although reliable data on the avoidable fraction is lacking. Table 2 provides first guess-timates of these proportions, emerging from a process comparing published incidence rates between developing and transitional countries, combined with expert clinical opinion, and assuming a basic level of health service infrastructure and limited access to care. Further confirmation of these figures can be regarded as a research need. These preliminary estimates give some broad indication of the potential of skilled attendance to impact not only on maternal deaths through primary prevention, but also on the number of women suffering with these complications - a number which is of course considerably greater than the number dying.

Table 2. Guess-timates of the proportion of complications amenable to primary prevention by skilled attendance1


% complications preventable2

% change between highest developing country estimate and estimate for Former Soviet Union

% change between lowest developing country estimate and estimate for Former Soviet Union

Complication

Optimistic estimate

Pessimistic estimate



Obstructed labour

85

70

-80.1

-20

Eclampsia

40

20

-70.3

-21.4

Puerperal sepsis

70

50

-79.2

-7

Obstetric haemorrhage

50

30

-70.2

-20.2

1 Skilled attendance included skilled attendant (skilled health professional) AND enabling environment of drugs, equipment, supplies, and referral

2 Estimated by reviewing incidence rates, combined with expert opinion, and assuming a basic level of health infrastructure and limited access to care.

SOURCES: Data extracted from AbouZahr 1998b, AbouZahr 1998c, AbouZahr and Guidotti 1998, AbouZahr et al . 1998.


Figure 3. Why do women in poor countries still die or suffer life-long disability owing to pregnancy or childbirth?

A similar approach can be applied to assess the potential impact of skilled attendance through secondary prevention - namely effective, appropriate and timely essential obstetric care. Considering the competencies proposed earlier in Box 2, and assuming that the skilled attendant has access to an enabling environment for BEOC and CEOC, it is possible to model a set of management scenarios or algorithms for averting maternal deaths from the 4 main direct obstetric causes. The approach used here should be regarded as tentative, and research is needed to refine and validate the methodology by using, for example, the Delphi technique. Combining the derived figures with those estimated earlier for primary prevention, a crude indication can be gained of the overall impact on maternal mortality, as shown in Table 3.

Table 3. Estimates from preliminary model of maternal deaths averted by skilled attendance

1

2

3

4

5


CFR (%)

Range of estimates

% complications averted by primary prevention

Model 1 % deaths averted by secondary prevention

Complication




OFA only

BEmOC only

CEmOC only

Obstructed labour (7.5%)

100

Optimistic

85

2

3

95



Pessimistic

70

1.5

2

90

Eclampsia (12.6%)

50

Optimistic

40

2

20

45



Pessimistic

20

0

15

40

Puerperal sepsis (15%)

50

Optimistic

70

20

40

49



Pessimistic

50

10

30

45

Haemorrhage (25.1%)

50

Optimistic

50

10

20

46



Pessimistic

30

5

10

42


6

7

8

9


Model 1
Total % deaths averted with primary and secondary prevention

Model 2
% deaths averted by secondary prevention

Model 2
Total % deaths averted with primary and secondary prevention

% of all maternal deaths averted

Complication

OFA only

BEmOC only

CEmOC only




Obstructed labour (7.5%)

85.3

85.45

99.25

97

99.55

7.47


70.45

70.6

97

94

98.2

5.28

Eclampsia (12.6%)

21.2

32

47

48

48.8

6.15


10

22

42

45

46

1.26

Puerperal sepsis (15%)

41

47

49.7

49

49.7

7.46


30

40

47.5

45

47.5

4.50

Haemorrhage (25.1%)

30

35

48

48

49

12.30


18.5

22

44.4

46

47.2

4.64

Notes:

1 Major complications and % contribution to total maternal deaths

2 Estimated case fatality rates in the absence of any intervention

3 Upper and lower limits of estimates

4 Guess-timates of % complications avoided by primary prevention through skilled attendance

5 Model assuming that women present with the complication to a skilled attendant and receive one of three types of care, but with no referral between them: OFA Obstetric First Aid, BEmOC Basic Emergency Obstetric Care, CEmOC Comprehensive Emergency Obstetric Care

6 Estimates from column 5 combined with the proportion avoided by primary prevention

7 Model 2 assumes women present with complication to a skilled attendant and receive OFA, followed by BEmOC, followed by CEmOC.

8 Estimates from column 7 combined with the proportion avoided by primary prevention

9 Use of highest (optimistic) and lowest (pessimistic) estimates of % averted multiplied by the % of all maternal deaths owing to that complication.

This type of sensitivity analysis is widely used in health economics (Briggs et al . 1994) and, as applied here, suggests that between about 16% and 33% of all maternal deaths could be avoided through skilled attendance, assuming certain competencies as well as the availability of essential, drugs, equipment and referral. The model focuses only on skilled attendance impacting only on the four main causes of maternal death close to delivery, and does not allow either for competing risks nor possible multiplicative effects of skilled attendance. The preliminary nature of these figures and the need for further developmental work must be emphasised. However, the optimistic proportion is comparable to the estimated figure of a third of maternal deaths avoidable by the provision of family planning (Winikoff & Sullivan 1997). To have this level of impact assumes that each woman has access to and utilises skilled attendance, which raises issues related to coverage at the population level, as discussed later.

Given this potential, the key question is what is the evidence that this can be achieved - in other words that the efficacy of skilled attendance can be translated into effectiveness - clinical and cost-effectiveness. The most rigorous approach to answering this question is a randomised-controlled trial (RCT) as it enables systematic bias between the intervention and non-intervention group to be eliminated. Whilst a large number of specific midwifery and obstetric practises have been evaluated using this gold-standard design, few have been conducted in developing countries and none have used maternal death as the primary outcome. In other words, there is no Grade 1 evidence (Gray 1997) to show that women delivering with skilled attendance have a lower risk of dying of maternal causes than women delivering without. However, given the proven effectiveness of specific practices, it would now be regarded as unethical to conduct a study in which the control group of women were specifically denied these procedures. It would on the other hand be possible to use a cluster randomised trial design (CRT) to compare a complete package of skilled attendance (attendant and enabling environment) provided in intervention districts versus the existing maternity care in control districts. A natural opportunity to use this experimental design may arise where countries are committed to skilled attendance as defined above, but are unable to implement this across all districts at the same time. If districts can be randomised in terms of the order in which implementation occurs, then a CRT may be feasible (Graham & Bell 2000a).

In the absence of trial data, there are two main other types of “evidence” which can be used to explore the link between skilled attendance and the risk of maternal death at the individual level: historical and epidemiological. Neither of these can provide proof that there is a lower probability of dying of maternal causes with than without skilled attendance, owing to the difficulty of controlling for confounding factors, such as differences in women’s risk at outset or place of delivery. The majority of historical evidence is in fact relevant to the link between skilled attendance and maternal mortality at the population level, which will be discussed in more detail later. However, the classic work by (Loudon, Death in Childbirth 1992a), provides examples of data spanning an enormous period (1864 -1939) and across several continents, which show the risk of death by type of birth attendant. A selection of these figures is given in Table 4. Not surprisingly, there is no clear pattern and firm conclusions cannot be drawn, particularly as the crude maternal mortality ratios cannot be disaggregated according to, for example, intended place of delivery or booking status. But what perhaps these data do confirm is that the professional label alone is not a good proxy for skills or competencies and that we should not only consider reduced risk but also the elevated risk of maternal death in the presence of unskilled “professional” attendants. This concept of balancing benefit with harm is crucial in all areas of health care (Gray 1997), and in safe motherhood there are both historical and contemporary reviews of maternal deaths which attribute the fatal outcome to the professional attending the delivery (such as Bobadilla et al . 1996, Porges 1985, Egypt Ministry of Health 1994, United Kingdom HMSO 1998). It is of course hard to establish culpability in many situations, and the attendant finally involved in a case may be unfairly blamed when their involvement may have been too late and beyond their control. This sort of detail is lacking from most routine sources of information on coverage and helps to explain some of the patterns observed from correlational analysis which will be discussed later.

Table 4. Selected historical estimates of maternal mortality by place and attendant at delivery*

Year

Country

Location

Attendant

Maternal mortality1

Source table

1864-73

England

Liverpool Lying-in Hospital

Midwives and doctors

1591

12.2

1876-80

England

Birmingham Lying-in Charity (Domiciliary care)

Midwives and General Practitioners

145

12.3

1873

Scotland

Glasgow Maternity (Lying-in) hospital

Midwives and doctors

2500

12.2

1879-81

Scotland

Glasgow Maternity Hospital (Domiciliary care)

Midwives and General Practitioners

875

12.3

1880-84

England

Queen Charlotte’s Hospital

Midwives and doctors

1050

12.5

1885-89

England

Queen Charlotte’s Hospital

Midwives and doctors

420

12.5

1909-14

England

Provincial Lying-in hospital

Midwives and doctors

5680

13.5

1909-14

England

Provincial Hospital (Domiciliary care)

Midwives and General Practitioners

2960

13.5

1929

Canada

Institutional deliveries

Midwives and doctors

1310

6.2

1929

Canada

Non-institutional deliveries

Midwives and General Practitioners

370

6.2

1931

United States (Ohio)

Maternity hospitals

Midwives and doctors

590

174

1931

United States (Ohio)

Domiciliary

Midwives

5142

17.4

1931

United States (Ohio)

Domiciliary

General practitioners

1090

17.4

1931

United States (Ohio)

Maternity hospitals

Obstetricians

734

17.4

1935-39

United States (Michigan)

Home deliveries

Midwives and General Practitioners

130

21.3

1935-39

United States (Michigan)

Hospital deliveries

Obstetricians

1340

21.3

1 No information is available to disaggregate these maternal mortality ratios according to intended place of delivery or booking status.

* SOURCE: Loudon 1992

In terms of epidemiological evidence, this broadly falls into two categories - quasi-experimental and descriptive. In the former case, control groups are used to assess the effects of specific interventions. Perhaps the most famous - and contested - relevant example of this design is the Maternity Care Program in Matlab, Bangladesh, involving both historical (before-and-after comparison) and contemporary (non-intervention) controls. The intervention comprised increasing the number of skilled attendants (government-trained community midwives), together with the creation of the enabling environment of a referral chain and access to a health centre able to provide BEOC (Maine et al . 1996). Although this data set has the potential and the power to examine the complex link at the individual level between maternal death and the delivery attendant, and indeed to unravel the sequence of care-givers prior to death, these results do not appear to be available. Instead, aggregate analysis is used to show that a statistically significant decline in the level of maternal mortality occurred in the intervention area from 1984-86 to 1987-89 but not in the control area (Fauveau et al . 1991). However, subsequent analysis of maternal mortality in a third area - the Comparison Area - found evidence of a fall comparable to that in the intervention area even though it did not receive the Maternity Care Program; this third area did however have access to a district hospital providing CEOC. This suggests that the control area was perhaps not comparable from the outset to the intervention area, and re-emphasises the difficulty of interpreting findings from non-randomly selected groups (Ronsmans et al. 1997).

Descriptive study designs consider patterns of maternal mortality relative to other variables, such as place of delivery and type of attendant, but again cannot prove causal connections owing to their inability to control for confounding factors. One of the most well cited examples relevant to skilled attendance is that reported by (Kaunitz and colleagues 1984). Here maternal death among the Faith Assembly religious group of Indiana was compared with that for the remainder of the state population. A tenet of the sect is that members should not receive medical care, and thus all women give birth without professional obstetric assistance. The pregnant women in this religious group are widely regarded as “low-risk” with regard to demographic and health characteristics. For the period 1975-1982, the maternal mortality ratio among the Faith Assembly women was 872 per 100,000 live births, compared with 9 among the remainder of the Indiana population (Loudon 1992a). In other words, the risk in the former group delivering without skilled attendance was 92 times (95% CI 19-280) greater than that for the latter who had access to (though not necessarily used) modern maternity care services. Although it is hard to identify potential confounding factors that might explain this huge difference, neither can this study be regarded as providing rigorous evidence of the effectiveness of skilled attendance, particularly in view of the small number of deaths considered.

A more recent and developing country example of a descriptive study is provided by (de Bernis and colleagues 2000) as part of the MOMA survey (Bouvier-Colle et al. 1997). Within this cohort study of pregnant women, the component conducted in two different areas of Senegal found that among those delivering in health facilities, there were higher risks of maternal death for those conducted by non-professionals compared to women delivering with health professionals. However, again the number of deaths is extremely small and thus the confidence intervals very wide. Moreover, as for all descriptive studies there is uncertainty about the comparability of the populations in the two areas, since maternal characteristics appeared to differ significantly in a number of respects. Interestingly, for maternal morbidity the reverse pattern was observed - with higher rates recorded in the area where more women delivering in health facilities had professional attendance. Such a finding may be attributed to improved diagnostic techniques among the professionals or to greater iatrogenicity. Further analysis to examine the proportion of women admitted in normal labour and developing complications versus those admitted with complications could help to throw some light on this.

The lack of rigorous data on the effectiveness of skilled attendance in terms of reducing maternal mortality obviously also explains the lack of knowledge on cost-effectiveness. The most frequently used information is that derived from a costing exercise undertaken by WHO in support of the Mother-Baby Package (WHO 1994). This estimated that the maternal component of the package would cost $2 per capita to deliver in low-income countries, with a cost of $230 per mother or infant life saved. These figures are suggested to be “lower than or similar to the cost of many other programmes, such as measles immunisation” (Starrs 1997), but do not enable a judgement to be made about the cost-effectiveness of the skilled attendance component versus other elements of the package.

Normal delivery and essential obstetric care are, however, estimated to account for about 44% of the total costs of the Mother-Baby package (Jowett 2000).

There are sound clinical reasons for believing that the risk of maternal death can be reduced by skilled attendance, particularly as the causal pathways can be elucidated. However, the theory of how skilled attendance can work has not been rigorously tested, and the information available is flawed either owing to weak study designs or inadequate power. The former reflects in part the complexities of mounting controlled trials and thus the resort to quasi-experimental and descriptive approaches which cannot allow adequately for the differences between women receiving or not receiving skilled attendance. Case-control studies, for example, have faced huge difficulties in identifying appropriate controls (Abdulghani 1993). Moreover, the data are often limited to studying the effect of just one type of attendant rather than including all those involved and the timing of their involvement; clearly the risk of maternal mortality can appear higher among doctors simply because they attend the most serious complications. Further explorations of existing data sets may help to establish patterns of attendance involving more than one professional. In summary, we know that skilled attendance could work to reduce maternal mortality at the individual level; we do not know reliably if it can or has.