- Email: [email protected]
A community-based nested case-control study of maternal mortality
International Journal of Gynecology & Obstetrics 47 (1994) 247-255
Article
A community-based
nested case-control study of maternal mortality
F.F. Fikree*a, R.H. Gray b,H.W. Berendesc, M.S. Karima ‘Department bDepartment
of Community
of Population
Health
Dynamics,
Sciences. Baltimore.
‘Division
of Epidemiology,
Statistics
and Prevention
National
The Aga Khan
School of Hygiene
Institutes
MD
Research, of Health,
University,
and Public 21202,
Health,
Karachi
74800. Pakistan
The Johns Hopkins
University.
USA
National Bethesda,
Institute MD
of Child Health
20892,
and Human
Development.
USA
Received 25 February 1994; revision received 20 July 1994: accepted 4 August 1994 Abstract Objectives: Population-based information from Pakistan on maternal mortality is inadequate to define the magnitude of the problem or to contribute information on the distribution of clinical causes and risk factors. A population-based survey was conducted in eight urban squatter settlements of Karachi, Pakistan, to address these key issues. Met/to&: Pre-coded questionnaires were administered to the 10 135 households to ascertain deaths of household members in the 5 years preceding the survey (1984- 1989). A verbal autopsy questionnaire was administered to the 121 households where a female death between 15 and 49 years of age was reported. Results: The maternal mortality ratio was estimated as 281 per 100 000 livebirths with hemorrhage, eclampsia and puerperal sepsis as the major causes of maternal deaths. Important risk factors identified were maternal employment, paternal unemployment, and poor prior Inpregnancy history. Women did seek health care but the assistance sought was often inappropriate. Conclusion: tervention programs for improving timely referral and upgrading of hospitals are suggested. Keywords:
Clinical causes; Maternal mortality; Non-clinical causes; Risk factors
1. Introduction There is limited information on levels of maternal mortality and causes of maternal deaths in much of the developing world due to the lack of adequate vital registration systems and poor certification of causes of deaths. Most deaths occur at home, making it more difficult to obtain satisfactory information. Hospital studies on medical l
Corresponding author, Fax: +92 21 4934294
causes of deaths are available, though due to selective admissions they may provide a biased source of data. Therefore, hospital-based maternal mortality data may not be adequate to define the magnitude of the problem or to contribute information on the distribution of clinical causes and risk factors. The World Health Organization (WHO) estimates that half a million women die annually from pregnancy-related causes and that 99% of these deaths occur in developing countries [l]. Women
0020-7292/94/$07.00 0 1994 International Federation of Gynecology and Obstetrics SSDI 0020-7292(94)02216-L
F.F. Fikree et al. /ht.
248
J. Gynecol. Obsret. 47 (1994) 247-255
in developing countries experience 100-200 times higher maternal mortality than their counterparts in Europe and North America, and the lifetime risk of maternal mortality to women from developing countries may be as high as one in 14, while in developed countries it is only one in several thousand [2]. There is no other public health statistic for which the disparity between developed and developing countries is so wide. Recognizing the excessive loss of women in the prime of life, and the catastrophic effects on families when the mother dies in childbirth, a major global effort was proposed in April 1987 to address maternal morbidity and mortality in developing countries [2]. The main causes of maternal deaths in developing countries today, hemorrhage, toxemia and sepsis, are the same as in the developed countries some 50 years ago. Population-based studies to estimate the magnitude and pattern of maternal deaths have been carried out in several developing countries: Bangladesh, Egypt, Indonesia, Ethiopia, India, Jamaica and Kenya [3]. The maternal mortality ratios varied from 108 per 100 000 live births in Jamaica to 7 18 per 100 000 live births in Indonesia [4,5]. The high costs associated with large-scale studies coupled with innumeracy and inadequate recall by respondents, raises questions about the feasibility of such studies in other developing countries. Other methods of estimating maternal mortality using proxy information such as the sisterhood method ’ and ‘networking’ provide simpler and cheaper methods [6,7] although they may not provide information on clinical causes of and risk factors for deaths, including health service utilization patterns. Limited population-based data on maternal mortality exist for Pakistan, a country with poor vital registration data. The sisterhood method was used to estimate maternal mortality ratios in the context of a demographic survey in 1988 [8]. Pakistani health policy makers need such information in order to plan effective use of scarce resources and to evaluate the efficacy of health ’ The
sisterhood method
deriving population-based using the proportion
is a new, indirect estimates
of
technique
maternal
programs aimed at a reduction of maternal deaths. This paper reports on the level and causes of maternal mortality from a sample of low socioeconomic squatter settlements in Karachi, Pakistan, and using a nested case-control design2, this paper also examines various socioeconomic and biologic risk factors for maternal death. 2. Materials and methods A population-based survey was conducted in eight low socioeconomic squatter settlements in Karachi, Pakistan, from August to September 1989, to collect information on the level and causes of maternal and infant mortality, and associated risk factors. The field staff consisted of 30 interviewers, six field supervisors and one field manager. The survey included 10 345 households, with a response rate of 98%. Pre-coded questionnaires were administered to 10 135 households and obtained information on deaths of household members in the 5 years preceding the survey (1984- 1989) household composition, in and out migration, and selected socioeconomic factors. For each death reported, the age and sex of the deceased were obtained. A second questionnaire was administered to 10 647 ever-married women aged 15-54, who were current residents of the sample households. This questionnaire obtained demographic information and a detailed pregnancy history for the 5 years preceding the survey. In addition, information on current contraceptive use as well as intrapartum care was collected. Finally, a verbal autopsy questionnaire was administered in those households reporting female deaths between 15 and 49 years of age. There were 121 such deaths reported. The verbal autopsy interview was conducted by female doctors and, where possible, was administered to female relatives of the deceased. The interview contained detailed questions on the symptoms and signs of the major causes of deaths associated with pregnancy. The definition of maternal death proposed by
for
mortality
of adult sisters dying during pregnancy,
childbirth or the puerperium reported by adults during a census or survey.
’
A nested case-control
study is an investigation
cases and controls are drawn from the population study.
in which
in a cohort
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
the WHO is ‘the death of a woman while pregnant or within 42 days of termination of pregnancy, from any cause related to or aggravated by the pregnancy or its management, but not from accidental causes’ [9]. In our study, the postpartum period was 40 days as this conforms to the recognized Islamic period referred to as chillu. Three obstetricians reviewed the verbal autopsy questionnaires and’independently categorized the 121 deaths as maternal or non-maternal. There were 34 maternal and 87 non-maternal deaths. There was no disagreement in the categorization of deaths, and disagreements in classifying the most probable cause of death were resolved by consensus. Five deaths among pregnant women (two head injuries and three burn cases) were classifted as accidental deaths and were included among the nonmaternal deaths according to the WHO criteria. However, confidential information from additional sources suggested that the trauma was most probably inflicted intentionally in one burn case and in both head injury cases. Also, circumstantial evidence suggested that the pregnancy may have played a role in provoking acts of violence resulting in these three ‘accidental’ deaths. These special cases of possible homicide or manslaughter associated with pregnancy therefore may be considered as non-obstetric (i.e. indirect) maternal deaths. In this paper, we have largely maintained the conservative WHO convention and excluded all accidental deaths from the case definition of maternal mortality, but we have also considered the effects of these three additional deaths on maternal mortality ratios. We estimated the maternal mortality ratio (maternal deaths per 100 000 live births), and the maternal mortality rate (maternal deaths per 100 000 women aged 15-49). Since information was obtained on deaths over a 5-year period, we used the women-years of observation over 5 years as the denominator for the maternal mortality rate. A nested case-control analysis was used to assess risk factors for maternal mortality. The 34 maternal deaths were the cases, and the controls were the surviving women reporting a pregnancy in the preceding 5 years. Since cases included three abortion-related deaths, our controls were not
249
restricted to women who reported only live births or stillbirths. We randomly selected five controls per case from the 6857 surviving women who reported a pregnancy in the 5 years preceding the survey. We limited the number of controls to five per case (i.e. a total of 170 controls) as no further gain in power would be achieved by including more surviving women. The risk factors for the case-control analysis were broadly classified into two categories: socioeconomic and biologic. The socioeconomic variables included maternal and paternal education and occupation, and socioeconomic status indicators such as number of household assets, household facilities and water supply. The household assets (ownership of modem objects) included eight items: radio, television, sewing machine, video-cassette recorder, refrigerator, washing machine, motorbike and car. Household facilities included the provision of modern sewage, electricity and adequate housing material. The biologic factors included maternal age, gravidity, previous pregnancy outcome, health provider at delivery and place of delivery. For the pregnancy-related factors, the index pregnancy for the cases was the fatal pregnancy, while for controls it was the most recent pregnancy. In the current analysis, unadjusted odds ratios (O.R.) and 95% confidence intervals (CL) are presented [9]. The small number of cases limited the utility of multivariate analyses. To assess the potential public health impact of significant risk factors, the population-attributable risk (PAR) was estimated as PAR = P(0.R. - l)/P(O.R. - 1) + 1, where O.R. is the unadjusted odds ratio and P is the prevalence of the risk factor among the control population. Under the assumption of independent effects of each risk factor, the PAR estimates the proportion of cases that might be prevented [9]. 3. Results Table 1 shows the study sample and estimated mortality. The number of currently living women aged 15-49 at the time of the survey (AugustSeptember 1989) was 14 193 and they reported 12 112 live births during the preceding 5 years
250
F.F. Fikrec et al. /hr.
J. Gynecol. Obstet. 47 (1994) 247-255
Table I Study population and level of mortality, Karachi, Pakistan, 1984-1989
Table 2 Distribution of causes of maternal deaths Cause
Status
Women aged 15-49 at time of survey Number of live births in the preceding 5 years Maternal deaths Non-maternal deaths Total deaths Reproductive age mortality rate (per 100 000 women aged 15-49) Maternal mortality rate (per 100 000 women aged 15-49) Non-maternal mortality rate (per 100 000 women aged 15-49) Maternal mortality ratio (per 100 000 live births)
No.
Rates per 100000
I4 193 I2 II2
34 87 I21 I71
Direct maternal deaths Hemorrhage Eclampsia Puerperal sepsis Septic abortion Ruptured uterus Indirect maternal deaths Cardiac disease Tuberculosis Unknown Total
n
u/
I6 7 4 2 I
47.1 20.6 Il.8 5.9 2.9
I I 2
2.9 2.9 5.9
34
100.0
48
123
281
(1984-1989). The 34 maternal deaths were 28.1% of all deaths to women in the reproductive ages. However, inclusion of the three suspect ‘accidental’ deaths among the pregnant women, using our broader definition of maternal deaths, raises this proportion to 30.6%. The estimated maternal mortality ratio was 281 per 100 000 live births (by the WHO definition). Using the broader definition of maternal deaths, which includes pregnancy-related ‘accidental’ causes, the ratio was 306 per ,100 000 live births. The estimated maternal mortahty rate per annum was 48 per 100 000 women aged 15-49 years (or 52 maternal deaths per 100 000 women with accidental deaths included). The main medical causes of direct maternal deaths were hemorrhage (47.1%), eclampsia (20.6%) and puerperal sepsis (11.8%) (Table 2). Of the 16 deaths attributed to hemorrhage, four were antepartum and the remainder (75%) were postpartum, including one as a consequence of abortion. There were two septic abortion deaths identified. Thus, there were three deaths due to abortion complications, or 8.8% of all maternal
deaths. Among the four indirect maternal deaths, one was due to cardiac disease (most probably a consequence of valvular heart disease), one to tuberculosis and no cause could be established for the remaining two. The time lapse between the onset of symptoms or signs of complications and death was more than a day for 42.9% of deaths attributed to hemorrhage, and 14.3% of deaths due to eclampsia. However, nearly 70% of women who later died of hemorrhage eventually reached a hospital. Those who were hemorrhaging in the postpartum period were more likely to reach a hospital (75%) than women with antepartum hemorrhage (50%). All women who died of eclampsia received medical care prior to death. Also, all deaths due to puerperal sepsis, including the two septic abortion deaths, were seen by a physician. Generally, for all pregnancy complications, referral was to a government hospital. We attempted to reconstruct health care seeking behavior and interventions provided by health personnel. Some form of help was sought for 33 out of 34 maternal deaths. Nearly 69% of women who died of hemorrhage sought help and were seen by local health care practitioners (i.e. nonregistered doctors, small maternity homes and dais [traditional birth attendants, TBAs]) where they were given some treatment (generally injections and oral medicine). We were not able to determine the type of injections or oral medicines provided. No women were immediately referred to a hospi-
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994)
Table 3 Selected characteristics, crude O.R. and 95% C.I. of cases (maternal deaths) and controls (current survivors) Characteristics
Cases (n = 34) n
u/o
Controls (n = 170) n
O.R.
95% C.1.
%,
70.6 29.4
0.8 I.0
0.3-1.8
5.3
3.8*
1.1-13.0
94.7
1.0
49.4 50.6
I.9 I.0
0.8-4.3
3.5
5.9;.
1.5-22.5
96.5
I.0
75.9 24.1
I.5 1.0
0.6-4.7
5.3 6.5 7.6 80.6
2.4 1.0 1.3 1.0
0.6-9.5
51.8 48.2
1.3 I.0
0.6-3.0
251
Table 3 (continued) Characteristics
Cases (n = 34) ~ ‘%I
Saciacomakc-es
Maternal education Illiterate 22 64.7 120 Literate (ref.) I2 35.3 50 Maternal occupation Gainfully 6 17.6 9 employed Not gainfully 28 82.4 I61 employed (ref.) Paternal education Illiterate 22 64.7 84 Literate (ref.) I2 35.3 86 Paternal occupation Unemployed 6 17.6 6 Employed (ref.) 28 82.4 164 Ownership of household assets o-3 28 82.4 129 4+ (ref.) 6 17.6 41 Household facilities None 4 II.8 9 One 2 5.9 II Two 3 8.8 13 Three (ref.) 25 73.5 137 Water supply Tap (outside) 20 58.8 88 Tap (inside) I4 41.2 82 (ref.) Maternal bIoIogII charrcteristIcs Maternal age 15-24 I3 38.2 43 25-34 (ref.) I4 41.2 93 35+ 7 20.6 34 Gravidity Primigravid 6 17.6 I9 2-3 (ref.) 6 17.6 47 4-6 8 23.5 56 7+ I4 41.2 48 Outcome of previous pregnancy” Stillbirth/ 4 13.3 9 abortion Live birth 7 23.3 89 (ref.) Primigravida 5 16.7 I9 Not reported 14 46.7 53 Health provider at deliveryb RelativelTBA 16 57.1 105 (ref.)
247-255
Nurse I Doctor IO Others/don’t I know Place of delive$ Home (ref.) I7 Hospital IO Others/don’t I know
Controls (n = 170) -
O.R.
95°K Cl.
‘%I n
n
3.6 35.7 3.6
21 33 II
12.4 19.4 6.5
0.3 0.0-2.3 2.0 0.8-5.2 N.A.
60.7 35.7 3.6
II5 44 II
67.6 25.9 6.5
1.0 I.5 0.6-3.9 N.A.
ref., Reference category. “Index pregnancy (30 cases, 170 controls). b*cIndexpregnancy (28 cases, 170 controls). *P < 0.01; **P < 0.001.
0.2-5. I
25.3 54.7 20.0
2.0 1.0 1.4
0.8-5.0
II.2 27.6 32.9 28.2
2.5 1.0 I.1 2.3
0.6-10.2 0.3-4.2 0.7-7.8
5.3
5.7’
1.0-27.3
52.4
I.0
II.2 31.2
N.A. N.A.
61.8
1.0
0.5-4.0
tal. After receiving local treatment, 70% were later transferred to hospital. In the hospital, only two women who died of postpartum hemorrhage received blood transfusions, while the other women received a combination of injections, medicine or intravenous fluids. Four out of seven women who died of eclampsia (57%) were initially seen by local health care practitioners, but only two of them were immediately referred to a hospital. The remaining two eclamptic cases were treated locally (injections and/or medicines), and only later were transferred to a hospital. 3.1. Risk factors for maternal mortality Table 3 shows the distribution, unadjusted O.R. and 95% C.I. for selected characteristics among 34 maternal deaths (cases) and 170 currently surviving women (controls). Among the socioeconomic factors there were minima1 differences between the two groups with respect to maternal education, household assets, household facilities or water supply. However, cases were more likely to be gainfully employed prior to death (17.6%) compared with 5.3% among controls, whereas their spouses were more often unemployed (17.6% of cases compared with 3.5% of controls), and these differences were statistically significant. The husbands of deceased women were also more frequently illiterate (cases 64.7%, controls 49.4%). Cases were generally younger than controls and
252
F.F. Fikree et al. /hr.
J. Gynecol. Obsret. 47 (1994) 247-255
Table 4 PAR (%) for major predictors of maternal mortality Risk factors
PAR (%)
Gainful maternal employment Paternal unemployment Fewer household assets Pregnancy order Primigravid Grand multigravid Poor previous pregnancy outcome
21.3 14.6 30.6 18.8 12.1 20.7
more likely to be primigravid, but these differences were not statistically significant. The risk of maternal mortality by age and gravidity followed the expected U-shaped pattern with higher risks at the extremes. A poor obstetric history (prior pregnancy resulting in spontaneous abortion or stillbirth) was associated with a markedly increased risk of maternal death (O.R. 5.7, 95% C.I. 1.O-27.3). PARS for selected risk factors such as maternal employment (PAR 21.3%), paternal unemployment (PAR 14.6%), primigravidity (PAR 8.8%) and prior fetal losses (PAR 20.7%) were moderate to substantial (Table 4). The perinatal mortality rate for the study population was 54.1 per 1000 births. Twenty-nine of the 34 maternal deaths were from pregnancies which terminated in the third trimester. The perinatal mortality rate for these 29 pregnancies was 448.3 per 1000 births (10 stillbirths and three early neonatal deaths). In addition to the three early neonatal deaths, there were another four late deaths for an infant mortality rate of 368.4 per 1000 live births. 4. Discussion In this population, most deliveries are conducted by traditional birth attendants (59.4%), the remainder being conducted by doctors (23.1%), nurses (15.8%) or midwives (1.6%). The place of delivery was mainly in the home (68.0%) with
private (14.0%) and public (18.0%) hospitals being the other facilities used. The contraceptive prevalence rate among the study population was only 14%. Among contraceptives, the most common methods used were tubal ligation (47.6%), condoms (18.7%) and the pill (13.5%). We collected no information about prenatal care in this population. The maternal mortality ratio in this communitybased survey of low socioeconomic settlements in Karachi was 281 per 100 000 live births, which is lower than that for Pakistan as a whole. National estimates largely based on hospital reports range from 600 to 800 per 100 000 live births [l]. The sisterhood method reported a national maternal mortality ratio of 270 [8], which, we feel, is underestimated. Recent estimates from provincial areas of Pakistan range from 234 to 745 per 100 000 live births [Dr Farid Midhet, personal communication]. Other rural studies from the subcontinent report rates ranging from 623 per 100 000 live births in rural Bangladesh, to 874 per 100 000 live births in rural India [lO,l 11. The major reason for lower maternal mortality ratios in this urban sample is most likely the geographical proximity of medical services, whereas the lack of services is a serious problem in rural areas. There may also be some underestimation of maternal deaths in our sample due to underreporting of events by female relatives of the deceased women, or due to omission of information resulting from a break-up of the household unit. The clinical causes of deaths from our study in urban Karachi are consistent with those reported from studies in rural and urban areas of other developing countries, with hemorrhage being the leading cause, followed by eclampsia and puerperal sepsis [5,7,12]. This differs from the leading cause of death reported in developed countries some 50 years ago, which was puerperal sepsis [13]. Additionally, our data show only 6% of deaths due to septic abortions, which is similar to reports by Kane et al. [12] in Egypt. Induced abortions are illegal in Pakistan and underreporting of illicit abortions undoubtedly exists, and is probably also present in the study by Kane et al. [12]. Conceivably, therefore, deaths due to septic abor-
F. F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
tions may be higher than those reported here. In contrast, septic abortions account for 20% and 30% of maternal deaths in Bangladesh and Ethiopia, respectively [ 10,141. Many of the acute obstetric emergencies reached the hospital late, despite access by paved roads from any of the eight study sites and distances of 3-12 miles. For example, although local assistance was generally sought shortly after onset of pregnancy complications, the time lapse between onset of hemorrhage and death was more than a day in nearly 43% of maternal deaths. The delay in referral and inappropriate local care clearly suggest a need for improved triage. Furthermore, even after reaching a hospital only two women were given blood transfusions, which further suggests a lack of appropriate emergency care. A hospital-based study in Karachi examined causes of death among 95 women who were dead on arrival at the hospital [ 151.The findings were similar to those of the present study: hemorrhage was reported as the probable cause of death in 43% of cases, with a predominance of postpartum hemorrhage (63%). Late referrals were attributed to various reasons such as lack of transport, hesitancy of the family or absence of the husband. Thus, our study and those of Jafarey [ 151indicate that appropriate and timely triage is inadequate at community level, and an intervention which addresses these current inadequacies is needed to prevent maternal deaths in this urban setting. There is increasing concern regarding violence towards women, especially pregnant women. However, only a few studies have suggested possible homicide or manslaughter as indirect causes of maternal deaths. In a study from Bangladesh, 26% of all deaths among pregnant women were ‘accidental’, mostly due to drowning and homicide (poisoning and violence) with burns accounting for 2.7% of maternal deaths [16]. In our study, 7.7% of such deaths were associated with violence. Measures to prevent acts of violence against women, especially pregnant women, are urgently needed. However, because of the low social status of women, long-term improvements in female education and status may do more to decrease acts of violence than law enforcement.
253
The existing literature on risk factors for maternal mortality is limited. However, the association of extremes of maternal age and parity with higher maternal mortality risk has been clearly demonstrated in several studies including our own [ 12,171. We observed significantly increased risks of maternal death associated with indices of extreme poverty: families in which the women were employed or the husbands were unemployed. Pakistani women are expected to stay at home and take care of the house and children and women from low socioeconomic strata who work, generally do so out of need rather than desire, and are mostly the poorest of the poor. Other socioeconomic indicators such as paternal education, fewer household assets, poorer household facilities and availability of tap water in the house suggested a possible increased risk but the associations were not statistically significant. However, the power to detect statistically significant socioeconomic associations was limited due to small numbers and collinearity between these various indices. We observed no effect of maternal education on maternal mortality, which is consistent with findings reported by Kwast and Liff [ 171.The data indicate that the hypothesized effects of maternal education on household dynamics, which may be important for infant mortality, do not appear to play a similar role in maternal mortality. Prior fetal losses were significantly associated with maternal mortality, which suggests the possible importance of recurrent obstetric complications. The death of a mother profoundly affects family stability, but when it is a maternal death, the health and wellbeing of the infant from that pregnancy is also severely affected. In our study, mortality among infants born alive to mothers who died from maternal causes was nearly 37%, which is similar to estimates from Egypt (22%))but less than those for Bangladesh (95%) [ 12,181. A maternal death also adversely affected perinatal mortality due to an excess of stillbirths. The small sample size and limited power of this study reflect the difficulty of studying relatively uncommon events, such as maternal death, in a general population survey. However, the findings are sufficient to propose public health interven-
254
F.F. Fikree et al. /ht.
J. Gynecol. Obster. 47 (1994) 247-255
tions to reduce maternal mortality. Short-term strategies could focus on steps dealing with inadequate and delayed health care and the urgent need for appropriate triage of obstetric emergencies at the community level in urban settings. Health education has to be targeted especially to the dais and local area health care practitioners, to improve recognition of complications of pregnancy and prompt referral to. appropriate facilities. Health education also needs to target women of childbearing age and their spouses to upgrade their knowledge of pregnancy and pregnancyrelated emergencies, and the need for appropriate care. In addition, public hospitals should become referral centers and be better equipped to handle such emergency cases. At a WHO interregional meeting in 1985, similar program initiatives were suggested [ 191.Rosenfield 1201also stressed health education for recognizing and seeking timely assistance for serious complications in combination with upgrading of referral hospitals (provision on a continuing basis of blood transfusions, parentera1 antibiotics and emergency cesarian sections) as the most important programmatic interventions for preventing maternal mortality. Such pilot projects have been conducted in Brazil and Zaire and underscore the need for community education and improvement in the capability of referral health facilities [21]. Long-term strategies are indicated by the high PARS for maternal employment (21.3%) or paternal unemployment (14.6%), primigravidity (18.8%) or prior fetal losses (20.7%) and suggest targeted programs focusing on antenatal care and screening of such high-risk women.
Acknowledgments We acknowledge members of the MIMS team, for their valuable support and contribution to the study. We are indebted to Drs Jafarey, Rizvi and Setna for reviewing the verbal autopsy questionnaires. This investigation received financial support from the UNDP/World Bank/WHO Special Program for Research and Training in Tropical
Diseases. We gratefully acknowledge help from the Hewlett Foundation and the NICHD/NIH. Support for this study was provided by research grant SFCP 08-024-N. References World Health Organization. Maternal mortality rates: a tabulation of available data. 2nd ed. Geneva: World Health Organization, 1986. Document No.: FHE/86.3. H. The safe motherhood initiative: a call to action. Lancet 1987; 1(329): 668. [31 National Academy of Sciences, Committee on Population and Demography. Contraception and reproduction: health consequences for women and children in the developing world. Washington DC: National Academy Press, 1989.
I21 Mahler
141 Walker GJA, McCaw AM, Ashley DEC. Bernard GW. Maternal mortality in Jamaica. Lancet 1986; l(327): 486.
151 Fortney JA, Susanti I, Gadalla S, Saleh S, Feldblum PJ. Potts M. Maternal mortality in Indonesia and Egypt. Int J Gynecol Obstet 1988; 26: 21.
I61 Graham W, Brass W. Snow RW. Estimating maternal mortality: the sisterhood method. Stud Fam Plann 1989; 20(3): 125. 171 Boerma JT, Mati JKG. Identifying maternal mortality through networking: results from coastal Kenya. Stud Fam Plann 1989; 20(5): 245.
181 Graham W. Results of the application of the sisterhood method for estimating maternal mortality. London: London School of Hygiene and Tropical Medicine, 1990.
[91 SchlesselmanJJ. Case-control studies. Design, conduct, analysis. Oxford: Oxford University Press, 1982.
UOl Khan AR, Jahan FA, Begum SF, Jalil K. Maternal mortality in rural Bangladesh. World Health Forum 1985: 6: 325.
u 11 Bhatia JC. Maternal mortality in Anantapur district. India: preliminary findings of a study. Geneva: World Health Organization, 1985. Document No.: FHE/PMM/ 85.9.16. [I21 Kane IT, El-Kady AA, Saleh S. Hage M, Stanback J. Potter L. Maternal mortality in Giza, Egypt: magnitude, causesand prevention. Stud Fam Plann 1992; 23( I ): 45. 1131 Loudon 1. The transformation of maternal mortality. Br Med J 1992; 305: 1557.
1141 Kwast BE, Rochat RW, Kidane-Mariam W. Maternal mortality in Addis Ababa. Ethiopia. Stud Fam Plann 1986; l7(6): 288. 1151 Jafarey SN. Determinants of maternal mortality. In: Qureshi KY. Qureshi AF. editors. Women in health and development. Report of the International Seminar at The Aga Khan University: 1990 Nov 14-15. Karachi: UNICEF. 1990.
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
(161 Fauveau V, Blanchet T. Deaths from injuries and in-
duced abortion among rural Bangladeshi women. Sot Sci Med 1989; 29(9): 1121. [I71 Kwast BE, Liff JM. Factors associated with maternal mortality in Addis Ababa, Ethiopia. Int J Epidemiol 1988; 17(l): 115. 1181 Chen LC, Gesche MC, Ahmed S, Chowdhury Al, Mosley WH. Maternal mortality in rural Bangladesh. Stud Fam Plann 1974; 5( I I): 334. [I91 World Health Organization. Prevention of maternal mortality. Report of a WHO interregional meeting: 1985
255
Nov 1I-15. Geneva: World Health Organization, 1986. Document No.: FHE/86. I. [20] Rosentield A. Maternal mortality in developing countries. An ongoing but neglected ‘epidemic’. J Am Med Assoc 1989; 262: 376. [2l] Rooks J, Winikoff B. A reassessment of the concept of reproductive risk in maternity care and family planning services. Proceedings of a seminar presented under the Population Council’s Robert H. Ebert Program on Critical Issues in Reproductive Health and Population: 1990 Feb 12-13. New York: The Population Council, 1990.
Article
A community-based
nested case-control study of maternal mortality
F.F. Fikree*a, R.H. Gray b,H.W. Berendesc, M.S. Karima ‘Department bDepartment
of Community
of Population
Health
Dynamics,
Sciences. Baltimore.
‘Division
of Epidemiology,
Statistics
and Prevention
National
The Aga Khan
School of Hygiene
Institutes
MD
Research, of Health,
University,
and Public 21202,
Health,
Karachi
74800. Pakistan
The Johns Hopkins
University.
USA
National Bethesda,
Institute MD
of Child Health
20892,
and Human
Development.
USA
Received 25 February 1994; revision received 20 July 1994: accepted 4 August 1994 Abstract Objectives: Population-based information from Pakistan on maternal mortality is inadequate to define the magnitude of the problem or to contribute information on the distribution of clinical causes and risk factors. A population-based survey was conducted in eight urban squatter settlements of Karachi, Pakistan, to address these key issues. Met/to&: Pre-coded questionnaires were administered to the 10 135 households to ascertain deaths of household members in the 5 years preceding the survey (1984- 1989). A verbal autopsy questionnaire was administered to the 121 households where a female death between 15 and 49 years of age was reported. Results: The maternal mortality ratio was estimated as 281 per 100 000 livebirths with hemorrhage, eclampsia and puerperal sepsis as the major causes of maternal deaths. Important risk factors identified were maternal employment, paternal unemployment, and poor prior Inpregnancy history. Women did seek health care but the assistance sought was often inappropriate. Conclusion: tervention programs for improving timely referral and upgrading of hospitals are suggested. Keywords:
Clinical causes; Maternal mortality; Non-clinical causes; Risk factors
1. Introduction There is limited information on levels of maternal mortality and causes of maternal deaths in much of the developing world due to the lack of adequate vital registration systems and poor certification of causes of deaths. Most deaths occur at home, making it more difficult to obtain satisfactory information. Hospital studies on medical l
Corresponding author, Fax: +92 21 4934294
causes of deaths are available, though due to selective admissions they may provide a biased source of data. Therefore, hospital-based maternal mortality data may not be adequate to define the magnitude of the problem or to contribute information on the distribution of clinical causes and risk factors. The World Health Organization (WHO) estimates that half a million women die annually from pregnancy-related causes and that 99% of these deaths occur in developing countries [l]. Women
0020-7292/94/$07.00 0 1994 International Federation of Gynecology and Obstetrics SSDI 0020-7292(94)02216-L
F.F. Fikree et al. /ht.
248
J. Gynecol. Obsret. 47 (1994) 247-255
in developing countries experience 100-200 times higher maternal mortality than their counterparts in Europe and North America, and the lifetime risk of maternal mortality to women from developing countries may be as high as one in 14, while in developed countries it is only one in several thousand [2]. There is no other public health statistic for which the disparity between developed and developing countries is so wide. Recognizing the excessive loss of women in the prime of life, and the catastrophic effects on families when the mother dies in childbirth, a major global effort was proposed in April 1987 to address maternal morbidity and mortality in developing countries [2]. The main causes of maternal deaths in developing countries today, hemorrhage, toxemia and sepsis, are the same as in the developed countries some 50 years ago. Population-based studies to estimate the magnitude and pattern of maternal deaths have been carried out in several developing countries: Bangladesh, Egypt, Indonesia, Ethiopia, India, Jamaica and Kenya [3]. The maternal mortality ratios varied from 108 per 100 000 live births in Jamaica to 7 18 per 100 000 live births in Indonesia [4,5]. The high costs associated with large-scale studies coupled with innumeracy and inadequate recall by respondents, raises questions about the feasibility of such studies in other developing countries. Other methods of estimating maternal mortality using proxy information such as the sisterhood method ’ and ‘networking’ provide simpler and cheaper methods [6,7] although they may not provide information on clinical causes of and risk factors for deaths, including health service utilization patterns. Limited population-based data on maternal mortality exist for Pakistan, a country with poor vital registration data. The sisterhood method was used to estimate maternal mortality ratios in the context of a demographic survey in 1988 [8]. Pakistani health policy makers need such information in order to plan effective use of scarce resources and to evaluate the efficacy of health ’ The
sisterhood method
deriving population-based using the proportion
is a new, indirect estimates
of
technique
maternal
programs aimed at a reduction of maternal deaths. This paper reports on the level and causes of maternal mortality from a sample of low socioeconomic squatter settlements in Karachi, Pakistan, and using a nested case-control design2, this paper also examines various socioeconomic and biologic risk factors for maternal death. 2. Materials and methods A population-based survey was conducted in eight low socioeconomic squatter settlements in Karachi, Pakistan, from August to September 1989, to collect information on the level and causes of maternal and infant mortality, and associated risk factors. The field staff consisted of 30 interviewers, six field supervisors and one field manager. The survey included 10 345 households, with a response rate of 98%. Pre-coded questionnaires were administered to 10 135 households and obtained information on deaths of household members in the 5 years preceding the survey (1984- 1989) household composition, in and out migration, and selected socioeconomic factors. For each death reported, the age and sex of the deceased were obtained. A second questionnaire was administered to 10 647 ever-married women aged 15-54, who were current residents of the sample households. This questionnaire obtained demographic information and a detailed pregnancy history for the 5 years preceding the survey. In addition, information on current contraceptive use as well as intrapartum care was collected. Finally, a verbal autopsy questionnaire was administered in those households reporting female deaths between 15 and 49 years of age. There were 121 such deaths reported. The verbal autopsy interview was conducted by female doctors and, where possible, was administered to female relatives of the deceased. The interview contained detailed questions on the symptoms and signs of the major causes of deaths associated with pregnancy. The definition of maternal death proposed by
for
mortality
of adult sisters dying during pregnancy,
childbirth or the puerperium reported by adults during a census or survey.
’
A nested case-control
study is an investigation
cases and controls are drawn from the population study.
in which
in a cohort
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
the WHO is ‘the death of a woman while pregnant or within 42 days of termination of pregnancy, from any cause related to or aggravated by the pregnancy or its management, but not from accidental causes’ [9]. In our study, the postpartum period was 40 days as this conforms to the recognized Islamic period referred to as chillu. Three obstetricians reviewed the verbal autopsy questionnaires and’independently categorized the 121 deaths as maternal or non-maternal. There were 34 maternal and 87 non-maternal deaths. There was no disagreement in the categorization of deaths, and disagreements in classifying the most probable cause of death were resolved by consensus. Five deaths among pregnant women (two head injuries and three burn cases) were classifted as accidental deaths and were included among the nonmaternal deaths according to the WHO criteria. However, confidential information from additional sources suggested that the trauma was most probably inflicted intentionally in one burn case and in both head injury cases. Also, circumstantial evidence suggested that the pregnancy may have played a role in provoking acts of violence resulting in these three ‘accidental’ deaths. These special cases of possible homicide or manslaughter associated with pregnancy therefore may be considered as non-obstetric (i.e. indirect) maternal deaths. In this paper, we have largely maintained the conservative WHO convention and excluded all accidental deaths from the case definition of maternal mortality, but we have also considered the effects of these three additional deaths on maternal mortality ratios. We estimated the maternal mortality ratio (maternal deaths per 100 000 live births), and the maternal mortality rate (maternal deaths per 100 000 women aged 15-49). Since information was obtained on deaths over a 5-year period, we used the women-years of observation over 5 years as the denominator for the maternal mortality rate. A nested case-control analysis was used to assess risk factors for maternal mortality. The 34 maternal deaths were the cases, and the controls were the surviving women reporting a pregnancy in the preceding 5 years. Since cases included three abortion-related deaths, our controls were not
249
restricted to women who reported only live births or stillbirths. We randomly selected five controls per case from the 6857 surviving women who reported a pregnancy in the 5 years preceding the survey. We limited the number of controls to five per case (i.e. a total of 170 controls) as no further gain in power would be achieved by including more surviving women. The risk factors for the case-control analysis were broadly classified into two categories: socioeconomic and biologic. The socioeconomic variables included maternal and paternal education and occupation, and socioeconomic status indicators such as number of household assets, household facilities and water supply. The household assets (ownership of modem objects) included eight items: radio, television, sewing machine, video-cassette recorder, refrigerator, washing machine, motorbike and car. Household facilities included the provision of modern sewage, electricity and adequate housing material. The biologic factors included maternal age, gravidity, previous pregnancy outcome, health provider at delivery and place of delivery. For the pregnancy-related factors, the index pregnancy for the cases was the fatal pregnancy, while for controls it was the most recent pregnancy. In the current analysis, unadjusted odds ratios (O.R.) and 95% confidence intervals (CL) are presented [9]. The small number of cases limited the utility of multivariate analyses. To assess the potential public health impact of significant risk factors, the population-attributable risk (PAR) was estimated as PAR = P(0.R. - l)/P(O.R. - 1) + 1, where O.R. is the unadjusted odds ratio and P is the prevalence of the risk factor among the control population. Under the assumption of independent effects of each risk factor, the PAR estimates the proportion of cases that might be prevented [9]. 3. Results Table 1 shows the study sample and estimated mortality. The number of currently living women aged 15-49 at the time of the survey (AugustSeptember 1989) was 14 193 and they reported 12 112 live births during the preceding 5 years
250
F.F. Fikrec et al. /hr.
J. Gynecol. Obstet. 47 (1994) 247-255
Table I Study population and level of mortality, Karachi, Pakistan, 1984-1989
Table 2 Distribution of causes of maternal deaths Cause
Status
Women aged 15-49 at time of survey Number of live births in the preceding 5 years Maternal deaths Non-maternal deaths Total deaths Reproductive age mortality rate (per 100 000 women aged 15-49) Maternal mortality rate (per 100 000 women aged 15-49) Non-maternal mortality rate (per 100 000 women aged 15-49) Maternal mortality ratio (per 100 000 live births)
No.
Rates per 100000
I4 193 I2 II2
34 87 I21 I71
Direct maternal deaths Hemorrhage Eclampsia Puerperal sepsis Septic abortion Ruptured uterus Indirect maternal deaths Cardiac disease Tuberculosis Unknown Total
n
u/
I6 7 4 2 I
47.1 20.6 Il.8 5.9 2.9
I I 2
2.9 2.9 5.9
34
100.0
48
123
281
(1984-1989). The 34 maternal deaths were 28.1% of all deaths to women in the reproductive ages. However, inclusion of the three suspect ‘accidental’ deaths among the pregnant women, using our broader definition of maternal deaths, raises this proportion to 30.6%. The estimated maternal mortality ratio was 281 per 100 000 live births (by the WHO definition). Using the broader definition of maternal deaths, which includes pregnancy-related ‘accidental’ causes, the ratio was 306 per ,100 000 live births. The estimated maternal mortahty rate per annum was 48 per 100 000 women aged 15-49 years (or 52 maternal deaths per 100 000 women with accidental deaths included). The main medical causes of direct maternal deaths were hemorrhage (47.1%), eclampsia (20.6%) and puerperal sepsis (11.8%) (Table 2). Of the 16 deaths attributed to hemorrhage, four were antepartum and the remainder (75%) were postpartum, including one as a consequence of abortion. There were two septic abortion deaths identified. Thus, there were three deaths due to abortion complications, or 8.8% of all maternal
deaths. Among the four indirect maternal deaths, one was due to cardiac disease (most probably a consequence of valvular heart disease), one to tuberculosis and no cause could be established for the remaining two. The time lapse between the onset of symptoms or signs of complications and death was more than a day for 42.9% of deaths attributed to hemorrhage, and 14.3% of deaths due to eclampsia. However, nearly 70% of women who later died of hemorrhage eventually reached a hospital. Those who were hemorrhaging in the postpartum period were more likely to reach a hospital (75%) than women with antepartum hemorrhage (50%). All women who died of eclampsia received medical care prior to death. Also, all deaths due to puerperal sepsis, including the two septic abortion deaths, were seen by a physician. Generally, for all pregnancy complications, referral was to a government hospital. We attempted to reconstruct health care seeking behavior and interventions provided by health personnel. Some form of help was sought for 33 out of 34 maternal deaths. Nearly 69% of women who died of hemorrhage sought help and were seen by local health care practitioners (i.e. nonregistered doctors, small maternity homes and dais [traditional birth attendants, TBAs]) where they were given some treatment (generally injections and oral medicine). We were not able to determine the type of injections or oral medicines provided. No women were immediately referred to a hospi-
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994)
Table 3 Selected characteristics, crude O.R. and 95% C.I. of cases (maternal deaths) and controls (current survivors) Characteristics
Cases (n = 34) n
u/o
Controls (n = 170) n
O.R.
95% C.1.
%,
70.6 29.4
0.8 I.0
0.3-1.8
5.3
3.8*
1.1-13.0
94.7
1.0
49.4 50.6
I.9 I.0
0.8-4.3
3.5
5.9;.
1.5-22.5
96.5
I.0
75.9 24.1
I.5 1.0
0.6-4.7
5.3 6.5 7.6 80.6
2.4 1.0 1.3 1.0
0.6-9.5
51.8 48.2
1.3 I.0
0.6-3.0
251
Table 3 (continued) Characteristics
Cases (n = 34) ~ ‘%I
Saciacomakc-es
Maternal education Illiterate 22 64.7 120 Literate (ref.) I2 35.3 50 Maternal occupation Gainfully 6 17.6 9 employed Not gainfully 28 82.4 I61 employed (ref.) Paternal education Illiterate 22 64.7 84 Literate (ref.) I2 35.3 86 Paternal occupation Unemployed 6 17.6 6 Employed (ref.) 28 82.4 164 Ownership of household assets o-3 28 82.4 129 4+ (ref.) 6 17.6 41 Household facilities None 4 II.8 9 One 2 5.9 II Two 3 8.8 13 Three (ref.) 25 73.5 137 Water supply Tap (outside) 20 58.8 88 Tap (inside) I4 41.2 82 (ref.) Maternal bIoIogII charrcteristIcs Maternal age 15-24 I3 38.2 43 25-34 (ref.) I4 41.2 93 35+ 7 20.6 34 Gravidity Primigravid 6 17.6 I9 2-3 (ref.) 6 17.6 47 4-6 8 23.5 56 7+ I4 41.2 48 Outcome of previous pregnancy” Stillbirth/ 4 13.3 9 abortion Live birth 7 23.3 89 (ref.) Primigravida 5 16.7 I9 Not reported 14 46.7 53 Health provider at deliveryb RelativelTBA 16 57.1 105 (ref.)
247-255
Nurse I Doctor IO Others/don’t I know Place of delive$ Home (ref.) I7 Hospital IO Others/don’t I know
Controls (n = 170) -
O.R.
95°K Cl.
‘%I n
n
3.6 35.7 3.6
21 33 II
12.4 19.4 6.5
0.3 0.0-2.3 2.0 0.8-5.2 N.A.
60.7 35.7 3.6
II5 44 II
67.6 25.9 6.5
1.0 I.5 0.6-3.9 N.A.
ref., Reference category. “Index pregnancy (30 cases, 170 controls). b*cIndexpregnancy (28 cases, 170 controls). *P < 0.01; **P < 0.001.
0.2-5. I
25.3 54.7 20.0
2.0 1.0 1.4
0.8-5.0
II.2 27.6 32.9 28.2
2.5 1.0 I.1 2.3
0.6-10.2 0.3-4.2 0.7-7.8
5.3
5.7’
1.0-27.3
52.4
I.0
II.2 31.2
N.A. N.A.
61.8
1.0
0.5-4.0
tal. After receiving local treatment, 70% were later transferred to hospital. In the hospital, only two women who died of postpartum hemorrhage received blood transfusions, while the other women received a combination of injections, medicine or intravenous fluids. Four out of seven women who died of eclampsia (57%) were initially seen by local health care practitioners, but only two of them were immediately referred to a hospital. The remaining two eclamptic cases were treated locally (injections and/or medicines), and only later were transferred to a hospital. 3.1. Risk factors for maternal mortality Table 3 shows the distribution, unadjusted O.R. and 95% C.I. for selected characteristics among 34 maternal deaths (cases) and 170 currently surviving women (controls). Among the socioeconomic factors there were minima1 differences between the two groups with respect to maternal education, household assets, household facilities or water supply. However, cases were more likely to be gainfully employed prior to death (17.6%) compared with 5.3% among controls, whereas their spouses were more often unemployed (17.6% of cases compared with 3.5% of controls), and these differences were statistically significant. The husbands of deceased women were also more frequently illiterate (cases 64.7%, controls 49.4%). Cases were generally younger than controls and
252
F.F. Fikree et al. /hr.
J. Gynecol. Obsret. 47 (1994) 247-255
Table 4 PAR (%) for major predictors of maternal mortality Risk factors
PAR (%)
Gainful maternal employment Paternal unemployment Fewer household assets Pregnancy order Primigravid Grand multigravid Poor previous pregnancy outcome
21.3 14.6 30.6 18.8 12.1 20.7
more likely to be primigravid, but these differences were not statistically significant. The risk of maternal mortality by age and gravidity followed the expected U-shaped pattern with higher risks at the extremes. A poor obstetric history (prior pregnancy resulting in spontaneous abortion or stillbirth) was associated with a markedly increased risk of maternal death (O.R. 5.7, 95% C.I. 1.O-27.3). PARS for selected risk factors such as maternal employment (PAR 21.3%), paternal unemployment (PAR 14.6%), primigravidity (PAR 8.8%) and prior fetal losses (PAR 20.7%) were moderate to substantial (Table 4). The perinatal mortality rate for the study population was 54.1 per 1000 births. Twenty-nine of the 34 maternal deaths were from pregnancies which terminated in the third trimester. The perinatal mortality rate for these 29 pregnancies was 448.3 per 1000 births (10 stillbirths and three early neonatal deaths). In addition to the three early neonatal deaths, there were another four late deaths for an infant mortality rate of 368.4 per 1000 live births. 4. Discussion In this population, most deliveries are conducted by traditional birth attendants (59.4%), the remainder being conducted by doctors (23.1%), nurses (15.8%) or midwives (1.6%). The place of delivery was mainly in the home (68.0%) with
private (14.0%) and public (18.0%) hospitals being the other facilities used. The contraceptive prevalence rate among the study population was only 14%. Among contraceptives, the most common methods used were tubal ligation (47.6%), condoms (18.7%) and the pill (13.5%). We collected no information about prenatal care in this population. The maternal mortality ratio in this communitybased survey of low socioeconomic settlements in Karachi was 281 per 100 000 live births, which is lower than that for Pakistan as a whole. National estimates largely based on hospital reports range from 600 to 800 per 100 000 live births [l]. The sisterhood method reported a national maternal mortality ratio of 270 [8], which, we feel, is underestimated. Recent estimates from provincial areas of Pakistan range from 234 to 745 per 100 000 live births [Dr Farid Midhet, personal communication]. Other rural studies from the subcontinent report rates ranging from 623 per 100 000 live births in rural Bangladesh, to 874 per 100 000 live births in rural India [lO,l 11. The major reason for lower maternal mortality ratios in this urban sample is most likely the geographical proximity of medical services, whereas the lack of services is a serious problem in rural areas. There may also be some underestimation of maternal deaths in our sample due to underreporting of events by female relatives of the deceased women, or due to omission of information resulting from a break-up of the household unit. The clinical causes of deaths from our study in urban Karachi are consistent with those reported from studies in rural and urban areas of other developing countries, with hemorrhage being the leading cause, followed by eclampsia and puerperal sepsis [5,7,12]. This differs from the leading cause of death reported in developed countries some 50 years ago, which was puerperal sepsis [13]. Additionally, our data show only 6% of deaths due to septic abortions, which is similar to reports by Kane et al. [12] in Egypt. Induced abortions are illegal in Pakistan and underreporting of illicit abortions undoubtedly exists, and is probably also present in the study by Kane et al. [12]. Conceivably, therefore, deaths due to septic abor-
F. F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
tions may be higher than those reported here. In contrast, septic abortions account for 20% and 30% of maternal deaths in Bangladesh and Ethiopia, respectively [ 10,141. Many of the acute obstetric emergencies reached the hospital late, despite access by paved roads from any of the eight study sites and distances of 3-12 miles. For example, although local assistance was generally sought shortly after onset of pregnancy complications, the time lapse between onset of hemorrhage and death was more than a day in nearly 43% of maternal deaths. The delay in referral and inappropriate local care clearly suggest a need for improved triage. Furthermore, even after reaching a hospital only two women were given blood transfusions, which further suggests a lack of appropriate emergency care. A hospital-based study in Karachi examined causes of death among 95 women who were dead on arrival at the hospital [ 151.The findings were similar to those of the present study: hemorrhage was reported as the probable cause of death in 43% of cases, with a predominance of postpartum hemorrhage (63%). Late referrals were attributed to various reasons such as lack of transport, hesitancy of the family or absence of the husband. Thus, our study and those of Jafarey [ 151indicate that appropriate and timely triage is inadequate at community level, and an intervention which addresses these current inadequacies is needed to prevent maternal deaths in this urban setting. There is increasing concern regarding violence towards women, especially pregnant women. However, only a few studies have suggested possible homicide or manslaughter as indirect causes of maternal deaths. In a study from Bangladesh, 26% of all deaths among pregnant women were ‘accidental’, mostly due to drowning and homicide (poisoning and violence) with burns accounting for 2.7% of maternal deaths [16]. In our study, 7.7% of such deaths were associated with violence. Measures to prevent acts of violence against women, especially pregnant women, are urgently needed. However, because of the low social status of women, long-term improvements in female education and status may do more to decrease acts of violence than law enforcement.
253
The existing literature on risk factors for maternal mortality is limited. However, the association of extremes of maternal age and parity with higher maternal mortality risk has been clearly demonstrated in several studies including our own [ 12,171. We observed significantly increased risks of maternal death associated with indices of extreme poverty: families in which the women were employed or the husbands were unemployed. Pakistani women are expected to stay at home and take care of the house and children and women from low socioeconomic strata who work, generally do so out of need rather than desire, and are mostly the poorest of the poor. Other socioeconomic indicators such as paternal education, fewer household assets, poorer household facilities and availability of tap water in the house suggested a possible increased risk but the associations were not statistically significant. However, the power to detect statistically significant socioeconomic associations was limited due to small numbers and collinearity between these various indices. We observed no effect of maternal education on maternal mortality, which is consistent with findings reported by Kwast and Liff [ 171.The data indicate that the hypothesized effects of maternal education on household dynamics, which may be important for infant mortality, do not appear to play a similar role in maternal mortality. Prior fetal losses were significantly associated with maternal mortality, which suggests the possible importance of recurrent obstetric complications. The death of a mother profoundly affects family stability, but when it is a maternal death, the health and wellbeing of the infant from that pregnancy is also severely affected. In our study, mortality among infants born alive to mothers who died from maternal causes was nearly 37%, which is similar to estimates from Egypt (22%))but less than those for Bangladesh (95%) [ 12,181. A maternal death also adversely affected perinatal mortality due to an excess of stillbirths. The small sample size and limited power of this study reflect the difficulty of studying relatively uncommon events, such as maternal death, in a general population survey. However, the findings are sufficient to propose public health interven-
254
F.F. Fikree et al. /ht.
J. Gynecol. Obster. 47 (1994) 247-255
tions to reduce maternal mortality. Short-term strategies could focus on steps dealing with inadequate and delayed health care and the urgent need for appropriate triage of obstetric emergencies at the community level in urban settings. Health education has to be targeted especially to the dais and local area health care practitioners, to improve recognition of complications of pregnancy and prompt referral to. appropriate facilities. Health education also needs to target women of childbearing age and their spouses to upgrade their knowledge of pregnancy and pregnancyrelated emergencies, and the need for appropriate care. In addition, public hospitals should become referral centers and be better equipped to handle such emergency cases. At a WHO interregional meeting in 1985, similar program initiatives were suggested [ 191.Rosenfield 1201also stressed health education for recognizing and seeking timely assistance for serious complications in combination with upgrading of referral hospitals (provision on a continuing basis of blood transfusions, parentera1 antibiotics and emergency cesarian sections) as the most important programmatic interventions for preventing maternal mortality. Such pilot projects have been conducted in Brazil and Zaire and underscore the need for community education and improvement in the capability of referral health facilities [21]. Long-term strategies are indicated by the high PARS for maternal employment (21.3%) or paternal unemployment (14.6%), primigravidity (18.8%) or prior fetal losses (20.7%) and suggest targeted programs focusing on antenatal care and screening of such high-risk women.
Acknowledgments We acknowledge members of the MIMS team, for their valuable support and contribution to the study. We are indebted to Drs Jafarey, Rizvi and Setna for reviewing the verbal autopsy questionnaires. This investigation received financial support from the UNDP/World Bank/WHO Special Program for Research and Training in Tropical
Diseases. We gratefully acknowledge help from the Hewlett Foundation and the NICHD/NIH. Support for this study was provided by research grant SFCP 08-024-N. References World Health Organization. Maternal mortality rates: a tabulation of available data. 2nd ed. Geneva: World Health Organization, 1986. Document No.: FHE/86.3. H. The safe motherhood initiative: a call to action. Lancet 1987; 1(329): 668. [31 National Academy of Sciences, Committee on Population and Demography. Contraception and reproduction: health consequences for women and children in the developing world. Washington DC: National Academy Press, 1989.
I21 Mahler
141 Walker GJA, McCaw AM, Ashley DEC. Bernard GW. Maternal mortality in Jamaica. Lancet 1986; l(327): 486.
151 Fortney JA, Susanti I, Gadalla S, Saleh S, Feldblum PJ. Potts M. Maternal mortality in Indonesia and Egypt. Int J Gynecol Obstet 1988; 26: 21.
I61 Graham W, Brass W. Snow RW. Estimating maternal mortality: the sisterhood method. Stud Fam Plann 1989; 20(3): 125. 171 Boerma JT, Mati JKG. Identifying maternal mortality through networking: results from coastal Kenya. Stud Fam Plann 1989; 20(5): 245.
181 Graham W. Results of the application of the sisterhood method for estimating maternal mortality. London: London School of Hygiene and Tropical Medicine, 1990.
[91 SchlesselmanJJ. Case-control studies. Design, conduct, analysis. Oxford: Oxford University Press, 1982.
UOl Khan AR, Jahan FA, Begum SF, Jalil K. Maternal mortality in rural Bangladesh. World Health Forum 1985: 6: 325.
u 11 Bhatia JC. Maternal mortality in Anantapur district. India: preliminary findings of a study. Geneva: World Health Organization, 1985. Document No.: FHE/PMM/ 85.9.16. [I21 Kane IT, El-Kady AA, Saleh S. Hage M, Stanback J. Potter L. Maternal mortality in Giza, Egypt: magnitude, causesand prevention. Stud Fam Plann 1992; 23( I ): 45. 1131 Loudon 1. The transformation of maternal mortality. Br Med J 1992; 305: 1557.
1141 Kwast BE, Rochat RW, Kidane-Mariam W. Maternal mortality in Addis Ababa. Ethiopia. Stud Fam Plann 1986; l7(6): 288. 1151 Jafarey SN. Determinants of maternal mortality. In: Qureshi KY. Qureshi AF. editors. Women in health and development. Report of the International Seminar at The Aga Khan University: 1990 Nov 14-15. Karachi: UNICEF. 1990.
F.F. Fikree et al. /ht.
J. Gynecol. Obstet. 47 (1994) 247-255
(161 Fauveau V, Blanchet T. Deaths from injuries and in-
duced abortion among rural Bangladeshi women. Sot Sci Med 1989; 29(9): 1121. [I71 Kwast BE, Liff JM. Factors associated with maternal mortality in Addis Ababa, Ethiopia. Int J Epidemiol 1988; 17(l): 115. 1181 Chen LC, Gesche MC, Ahmed S, Chowdhury Al, Mosley WH. Maternal mortality in rural Bangladesh. Stud Fam Plann 1974; 5( I I): 334. [I91 World Health Organization. Prevention of maternal mortality. Report of a WHO interregional meeting: 1985
255
Nov 1I-15. Geneva: World Health Organization, 1986. Document No.: FHE/86. I. [20] Rosentield A. Maternal mortality in developing countries. An ongoing but neglected ‘epidemic’. J Am Med Assoc 1989; 262: 376. [2l] Rooks J, Winikoff B. A reassessment of the concept of reproductive risk in maternity care and family planning services. Proceedings of a seminar presented under the Population Council’s Robert H. Ebert Program on Critical Issues in Reproductive Health and Population: 1990 Feb 12-13. New York: The Population Council, 1990.