Int J Gynecol Obstet,
1992, 39: 197-204
197
International Federation of Gynecology and Obstetrics
Maternal mortality in Assiut S.A. Abdullah”, E.M. Aboloyouna, Departments
of aObstetrics
and Gynecology
H. Abdel-Aleem”, F.M. Moftahb and S. Ismail”
and bCommunity
Medicine,
Faculty of Medicine,
Assiut
University,
Assiut (Egypt)
(Received January 22nd, 1992) (Revised and accepted May 21st, 1992)
Abstract Twenty-nine maternal deaths were identified among 8656 pregnant women residing in Assiut city and three surrounding villages (Upper Egypt). This gives a maternal mortality ratio of 368 per 100 000 live births. Of these maternal deaths 83% were due to direct obstetric causes (hemorrhage, eclampsia, ruptured uterus and sepsis). Logistic regression analysis showed that residence (in villages versus Assiut city), parity (nullipara and grandmultipara) and illiteracy were significantly associated with increased risk of maternal death.
Keywords: Maternal mortality; Maternal health in developing countries; Communitybased study of maternal mortality.
During the reproductive period, maternal mortality is one of the leading causes of death among women in Egypt, but the magnitude of the problem is not always evident from official reports. Unfortunately, hospital based studies are not representative of the community as a whole because hospital cases are selfselected and often have a higher proportion of complicated deliveries. The present study is community-based, with the following aims. OO20-7292/92605.00 0 1992 International
Federation of Gynecology and Obstetrics Printed and Published in Ireland
1. To measure the maternal mortality ratio and risks of dying from pregnancy in women of reproductive age and to compare these with official statistics. 2. To analyze the causes of maternal death. 3. To determine the effect of sociodemographic factors, previous obstetric history, place of delivery and outcome on the risk of mortality using multivariate techniques to adjust for confounding. Methods The study was conducted in Assiut city and in three surrounding villages: Mankabad, Mosha and El-Zawia. The study located as many pregnancies and deliveries as possible occurring in the four localities between 1 January and 31 December 1987, using multiple sources and an elaborate design. A total of 8656 pregnant women identified in 1987 were followed up until delivery or the end of pregnancy, even if this occurred in 1988. Non-residents delivering in the four localities were excluded. Women receiving antenatal care in the hospitals, MCH centres, combined units, or in the rural health unit were identified through frequent visits by a physician and data on previous obstetric complications, preexisting medical conditions and problems in early pregnancy were collected from antenatal care records, health care providers and the women Article
198
Abdullah et al.
themselves. This group comprised 2769 women. Women not receiving antenatal care and those receiving antenatal care in private clinics were primarily identified when they delivered or their pregnancy ended. Identification sites included the hospitals, MCH centres, combined units and the rural health unit. In all four sites, midwives attached to MCH centres and combined units also provided information on home deliveries which they conducted and in the three villages, regular contact with private doctors and traditional birth attendants (dayas) further identitied deliveries and some abortions. The final information source was the civil register of live and stillbirths. The register was used to confirm hospital, clinic and home deliveries identified through other means and to identify any cases not previously located.
These were primarily women in Assiut city receiving care from private doctors or those not receiving care and delivering alone or with a daya. Abortions, ectopic and mollar pregnancies are not registered and registration of stillbirths is largely incomplete. In contrast, registration of live births is ,quite high in Egypt overall and is estimated to cover 90% of all live births [ 121. As a result, the pregnant women most likely to be missed by the study design are those with early terminations of pregnancy who failed to seek health facility care either before or at the time the pregnancy ended and those using private care who have still births or whose neonates die before registration. A schematic representation of the points of identification and the data sources is shown in Table 1. Data were entered using Epi Info [3] and
Table 1. Data sources and quality for women using various sources of maternity care by type of pregnancy outcome. Outcome
Used private ANC or had no ANC
Used facility for ANC
Private doctor
Daya
Alone
Facility
Clinic midwife
Private doctor
Daya
Alone
Facility
Clinic midwife
1
1 1
1 I
1 1
1 1
2 2
2 2
5 3
5 3
5 4
Stillbirths or early neonatal deaths a. Assiut 1 b. Three villages 1
1 1
1
1 1
1 1
2 2
2 2
8 6
8 6
8 I
Abortions, moles or ectopics a. Assiut 1 b. Three villages 1
1 1
1 1
1
1 1
2 2
2 2
11 9
II 9
11 10
Live births a. Assiut b. Three villages
1
1
1
Sources of data from most complete to least complete. In all cases, data could be supplement by information from the woman herself. 1. Clinic records on ANC plus facility delivery records, complete reporting by health centre personnel, plus registration. 2. Facilty delivery records, complete reporting by health centre personnel, plus registration. 3. Some reporting by private health personnel, some reporting by informants, plus complete registration. 4. Local informants plus complete registration. 5. Complete registration. 6. Some reporting by private health personnel, some reporting by informants, plus incomplete registration. 7. Some reporting by informants, plus incomplete registration. 8. Incomplete registration. 9. Some reporting by private health personnel, some reporting by informants. 10. Some reporting by informants. 11. No data. Int J Gynecol Obstet 39
Maternal mortality
Table 2. Frequency and risk of maternal death among pregnant women with various characteristics. Assiut and three villages, 1987-1988. Characteristics
Residence Assiut Mosha Mankabad El Zawia
No. of maternal deaths
10 8 4 7
Total no. of women
0.004 ***
1009 1652 1016
Parity Primipara (0) 11
1967
559
< 0.000 ***
l-4 deliveries 5 5+ deliveries 13
4849 1840
103 707
Age (years) <20 20-39 40+ Education Illiterate Primary school Secondary school+ Employment Unemployed Employed Socioeconomic Low status Moderate or high Used ANC No Yes Birth interval Primipara < 1 year 1-2 years >2 years
Characteristics
No. of maternal deaths
Home 15 Hospital/clinic 9 Birth assistant No delivery GP Obstetrician Nurse/Midwife Daya
Total no. of women
386 7830 440
518 281 1136
27
6281
430
1
844
118
1
1531
65
0.009 ***
0.045 **
297 328
6
870
690
0 9 2 12
54 2697 1340 3695
0 334 149 325
7335 1271
379 79
28
7691
364
1
965
104
22
5887
374
7
2769
253
10
1921
521
2 14 3
1111 4243 1381
180 330 217
Place of delivery Died antepartum 5
859
582
0.302 N.S.
315 2062
0.003 ***
97
Previous perinatal death None 26 One 1 2+ 2
8248 280 128
315 357 1563
0.053*
Previous abortions None 24
7246
331
0.973 N.S.
901 509
333 393
Current complication in pregnancy No 25 7657
326
Yes
One 2+ 28 1
Risk of P death/ 100 000 women
5056 2741
Previous complications in pregnancy 27 No 8559 2 22 5
199
Table 2 (continued)
Risk of P death/ 100000 women
201 793 242 689
in Assuit
2
3 2
0.087 *
0.187 N.S.
0.364 N.S.
0.340 N.S.
0.407 N.S.
Obstetric Medical
400 589
250 509
Current intrapartum fetal death No 20 8502
235
Yes
1 3
9
154
5844
Current antepartum fetal death No 24 8566
280
Yes
0.741 N.S.
5 0.000 ***
< 0.000 ***
5
90
5556
Current abortion No 29 Yes 0
7851 805
369 0
0.080 *
Wastage No
13
7378
176
Yes
16
1278
1252 Article
200
Abdullah et al.
Table 2 (continued) Characteristics
No. of maternal deaths
Current neonatal death No 27
Total no. of women
Risk of P death/ 100 000 women
8425
320
2
231
866
Gestation No delivery
5
865
578
Preterm Term Postterm
3 21 0
415 7348 28
723 286 0
Onset of labor No delivery Spontaneous Induced
5 23
853 7750 53
586 297 1887
0.056 *
200
< 0.000 ***
Yes
1
Intrapartum maternal complications No 15 7515 Yes
14
1141
1227
Mode of delivery No delivery 8 Spontaneous 18 Assisted 3
877 7017 762
912 257 394
0.180 N.S.
0.259 N.S.
0.006 ***
?? P < 0.10. **p < 0.05. ***p < 0.01.
the univariate and bivariate relationships were explored with frequencies and crosstabulations using Epi Info and SAS [9, IO]. Analyses using multiple logistic regression were performed with EGRET [4]. Results
Twenty-nine maternal deaths were identified during the study period giving a risk of death of 335 per 100 000 pregnant women (29/8656 women). Calculated in the conventional way, this gives a maternal mortality ratio of 368 per 100 000 live births (29/7865 live births). Twenty four of maternal deaths were due to direct obstetric causes (83%); in order of imInt J Gynecol Obsrei 39
portance these include hemorrhage (postpartum, 24.1% and antepartum, 13.8%), eclampsia (24.10/o),ruptured uterus (13.8%) and sepsis (6.9%). No deaths associated with complications of abortion were identified. The remaining five deaths were due to indirect obstetric causes (17’), with cardiac diseases as the most common. Table 2 presents the distribution of characteristics among maternal deaths and the total population and shows the risk of maternal death associated with each variable. Each bivariate association was tested using a chi-square statistic or Fisher’s exact test. Residence plays an important role in the risk of maternal mortality, with maternal deaths more likely to occur among women living in Mosha and El-Zawia than in Assiut or Mankabad (P c 0.004). The risk of maternal death was high among women with no previous birth (primipara = 559/100 000) and women with live or more previous births (grandmultipara = 7071100 000), compared with women with l-4 previous births (multipara = 103/100 000) and the differences are statistically significant (P < 0.000). The expected relationship of high risks at younger and older age was also seen (P < 0.009). The risk of death is high in illiterate women compared with those with primary schooling and those with secondary or university education (430/l 00 000, 119/100 000 and 651100 000, respectively; P < 0.045). Use of antenatal care, the length of the previous birth interval and the type of birth attendant did not have a statistically significant impact on the risk of maternal death. A bad previous history, including previous complications during pregnancy (P = 0.003), having two or more previous perinatal deaths (P = 0.053) and having two or more previous abortions (P = 0.973) did appear to increase the risk of maternal death but only the first two were statistically significant. As expected, negative outcomes associated with the current pregnancy increased the risk of maternal death markedly. Many of these variables are not risk factors in the conven-
Maternal mortality in Assuit Table 3. Risk of maternal death associated with various factors predicted using a logistic regression. Assiut and three villages, 1987-1988. Variables
Model 1 Parity 0 Parity l-4 Parity 5+ Model 2 Parity 0 Parity l-4 Parity 5+ Assiut/ Mankabad Mosha/ El Zawia Model 3 Parity 0 Parity l-4 Parity 5+ AssiutI Mankabad Mosba/ El Zawia Complication in previous pregnancy No complication Model 4 Parity 0 Parity l-4 Parity 5+ AssiutI Mankabad Mosha/ El Zawia Complication in previous pregnancy No complication Illiterate Schooled
Odds Ratio
Difference in deviance
D.F.
P
18.95
2
< 0.0001
10.06
1
0.0015
5.73
1
0.017
3.82
1
0.051
5.45 1.oo 6.89
5.80 1.00 6.20 1.00 3.39
6.57 1.00 6.33 1.00 3.46 11.27
1.00
6.84 1.00 5.68 1.00 2.87 11.85
1.00 3.53 1.00
All models were fitted using an intercept term. First order interactions were also fitted but were not statistically significant.
tional sense as they are either signals of the impending death or may even follow the maternal death. No deaths were associated with current abortions or ectopic or molar
201
pregnancies. Intrapartum maternal complications were significantly also associated with a high risk of death compared to women without intrapartum maternal complications (1227/100 000 versus 200/100 000). In order to adjust for potential confounding, a multiple logistic regression analysis was used to control for the effects of several variables simultaneously. Models were tested for all combination of variables and built up on the basis of the best fit of the model by looking at differences in the deviance. The variables tested included residence, parity, age, education, employment, socio-economic status, previous complications in pregnancy and previous perinatal deaths. The steps building up to the final significant model are shown in Table 3. The model shows that the risk of death is best predicted considering parity, residence and education. Once these are adjusted for, adding other variables, such as age, employment, socioeconomic status, use of antenatal care, previous complications in pregnancy and previous perinatal deaths, does not contribute significantly to the model, although several variables such as age show elevated but non-significant odds ratios in the expected direction. First order interactions were fitted but were not statistically significant. Discussion Compared to many other countries in the Middle East and Africa, Egypt has a wealth of published studies on maternal mortality. Only four however, the RAMOS in Menoulia, Lower Egypt [7], the Alexandria Widower study [5], the Giza Mortality study [6] and a previous study on Assiut, Upper Egypt [l], have been community based (Table 4). These studies have mainly measured the prevalence and causes of maternal deaths. The present study reveals a high maternal mortality ratio: 335/100 000 live births. This figure is comparable to the earlier populationbased study in Upper Egypt [I] and shows that the maternal mortality ratio in Upper Article
202
Abdullah et al.
Table 4.
Selected population-based
maternal mortality ratios in Egypt.
Site
Year
Number of maternal deaths
Vita1 statistics Egypt, national Assiut, district
1919 1988
1273 8
78 37
1985-86 1981-83 1963-82
156 385 183
150 190 163
El Kady et al., l989a Fortney et al., 1986 El Ghamry et al., 1984
1984-85 1984-85 1984-85 1987-88
16 23 34 29
178 471 323 368
Abdullah et al., 1985 Abdullah et al., 1985 Abdullah et al., 1985
Population-based studies Lower Egypt Giza study Menoufia (RAMOS) Alexandria Upper Egypt Kusiah (Assiut) Suhag Kena Current study (Assiut + 3 villages)
Egypt is high compared with the rest of the country. This difference could be spurious, if, for example, live births are underestimated in Upper Egypt compared with the rest of the country. Possible explanations for a real difference involve the likelihood that risk factors for maternal mortality, including the level of education, the parity distribution and the socioeconomic status, are unfavorably distributed in Upper Egypt. The Egyptian Demographic and Health Survey suggested for instance that the total fertility rate is higher in Upper Egypt than in Lower Egypt (5.4 versus 4.5) and that women are less likely to have ever used contraception (39% versus 61%) [ 1I]. Underlying poor health of women or lower quality health services are other potential contributors to the higher risk of maternal mortality seen in this area but there are no data to support this conclusion. As has been the case in previous studies from Egypt [ 1,l I], off’cial statistics severely underestimate the maternal mortality ratio at 37 per 100 000 live births. The reason for this is that death certificates fail to mention that the deceased had been pregnant and because Int J Gynecol Obstet 39
Maternal mortality ratio per 100 000 live births
Reference
the death is registered under other causes, such as circulatory diseases [7]. Egypt has recently changed its vital registration forms to specifically ascertain pregnancy status of woman who die. This should improve recognition of maternal deaths. As regards the causes of maternal death, the present study revealed 83% of the maternal deaths to be direct obstetric causes. This proportion of direct deaths is slightly higher than previous studies in Giza and Menoutia which found 63% and 62%, respectively. Nevertheless, the causes seen are those present in developing countries in general [8]. However, among the 29 cases, no cases due to illegal abortion have been detected. This is partly because women always deny abortion and partly because they may present with infection and their deaths will be reported as being due to sepsis. Deaths from ruptured ectopic pregnancy have not been detected in this study, possibly because the three villages included in the study are not far from Assiut city (distance ranged from 6 to 23 km), so cases could be transported to Assiut city where equipped hospitals are available.
Maternal mortality in Assuit
The most frequent variables cited in the literature as risk factors for maternal mortality are age and parity, with young and older women and primiparous and grandmultiparous women at the highest risk of maternal death. Less frequently, studies have examined the risks of maternal mortality associated with socioeconomic and cultural factors including education, employment and marital status and with health service factors such as use of antenatal and/or obstetric services. The study demonstrates that residence plays an important role in the risk of maternal death, as does parity and education. Age, current and previous complications and interventions at the time of delivery are also associated with an increased risk of death. Previous complications in pregnancy significantly increased the risk of maternal death (P = 0.053). Although this is a plausible risk factor, the low overall previous complications reported (1.1%) suggests that information on previous complications was not well recorded and that there may have been some bias favoring reporting of this item for women who died. The use of antenatal care did not significantly influence the risk of death. This may arise either because antenatal care fails to identify and treat or refer high-risk pregnancies to maternity units or hospitals, or alternatively, because women who seek antenatal care may do so because they have an underlying medical problem, which in turn puts them at a higher risk of maternal death. The Egyptian DHS showed for example that in Upper Egypt, 81% of women seeking prenatal care did so because of a medical problem [ 1I]. As regards the place of delivery, the present study demonstrates that hospital deliveries have a higher risk than home deliveries although the difference is not statistically significant. This is likely to be because most hospital deliveries are complicated [2]. The majority of deliveries were at home attended by a daya. 10 out of 29 deaths (35%) occurred at home, 8 of these occurred in villages.
203
Although most deliveries were initially attended by dayas, many were eventually referred to hospital obstetricians for management of complications. Facilities in maternity units, particularly in the rural areas, are not satisfactory. The present study revealed that deliveries attended by GPs or nurse-midwives had low risk compared to those attended by dayas or obstetricians. This is because GPs or nursemidwives referred complicated cases early to hospitals. Unfortunately, many of the risk factors considered, such as age and parity, or education, employment and socio-economic status, are so interrelated that it becomes difficult to estimate the impact of any one variable by itself. This is exacerbated by the relative rarity of maternal death. Our logistic regressions results showed that residence is the most important risk factor, with risks higher in two villages (Mosha and El Zawia) than in Assiut. In Mankabad the risk of death is comparable to Assiut, possibly because it is near to Assiut (6 kilometers) and has many private clinics and more access to health facilities, especially Assiut University Hospital. The finding that parity was associated with the risk of death was not surprising and it is likely that the high degree of correlation between age and parity partially explains why a model including both variables was not statistically significant. Education was the final variable to emerge from the logistic regression. This showed that illiterate women had 3.4 times the risk of death compared with women with some education even after controlling for residence and parity. Implications 1. Complete reporting of maternal deaths is an important policy recommendation, since many obstetric deaths may be preventable. 2. The use of family planning services could prevent deaths among women of unfavorable parity and age. Article
204
AbduNah et al.
status of 3. Raising the socioeconomic women and providing them with a better education would improve their underlying health and enable them to be more receptive to and better use preventive, curative and emergency health services. 4. Motivation for women to use available healthcare facilities and improvement of the quality of health providers is needed. seemed to be Dayas, in particular, associated with negative outcomes and attention should be paid to them. Delivery by an obstetrician was also associated with a high risk of death, but this is probably due to referral of high-risk cases to these specialists. 5. The importance of health services and epidemiological research to find the causes behind the higher maternal mortality rates in Upper Egypt compared with the rest of the country. Acknowledgments The authors acknowledge the Ford Foundation, Cairo, for their financial support and the Maternal and Child Epidemiology Unit, London School of Hygiene and Tropical Medicine for assistance in data analysis. References Abdullah SA, Fathalla MF, Abdel-Aleem AM, Salem, HT and Aly MY: maternal mortality in Upper Egypt. Paper presented to the WHO Inter-regional Meeting on Prevention of Maternal Mortality, November, 1 l-1 5, 1985. Campbell OMR, Rooney CF, Filippi VCA and Graham
Int J Gynecol Obstet 39
3
4
5
6
I
8
WJ: Technology-free obstetrics (letter). The Lancet 337: 1095, 1991. Dean AD, Dean JA, Burton JH and Dicker RC: Epilnfo version 5.0la: a word processing, database and statistics program for epidemiology on microcomputers. Centres for Disease Control, Atlanta, Georgia, 1991. Egret: State of the Art Epidemiology Computing. Seattle, Statistics and Epidemiology Research Corporation, Washington, 1991. El-Ghamry A, El-Sherbini AF, Hussein M, El-Tantawi AS and Hamoud El: The feasibility of getting information about maternal mortality from the husband. Bull High Inst Public Health II: 195, 1984. El-Kady A, Saleh S, Gadalla S, Fortney JA and Bayoumi H: Obstetric deaths in Menoutia Governorate, Egypt. Br J Obstet Gynaecol 96: 9, 1989. Fortney JA, Susanti I, Gadalla S, Saleh S, Rogers SM and Potts M: Reproductive mortality in two developing countries. Am J Public Health 76: 134, 1986. Royston E and Armstrong S (eds): Preventing Maternal Deaths, pp l-1233, World Health Organization, Geneva, 1989. SASlSTAT User’s Guide, Release 6.03, (ed NC Cary), SAS Institute Inc, 1988. SAS Language Guide for Personal Computers, Release 6.03, (ed NC Cary) SAS Institute Inc, 1988 Sayed HAA, Osman MI, El-Zanaty F and Way AA: Egypt Demographic and Health Survey, 1988. Egypt National Population Council, Cairo and Institute for Resource Development/Macro Systems, Inc, Columbia, 1989. US National Academy of Sciences: The estimation of recent trends in fertility and mortality in Egypt. Committee on Population and Demography. National Academy Press, Washington DC, 1982.
Address for reprints:
S.A. Abdullah Department of Obstetrics and Gynecology Faculty of Medicine Assuit University Assuit, Egypt