Perinatal mortality in southern Brazil: a population-based study of 7392 births

Bulletin of the World Health Organization, 65 (1): 95-104 (1987) © World Health Organization 1987 Perinatal mortality in southern Brazil: a population-based study of 7392 births F. C. BARROS,l C. G. VICTORA,2 J. P. VAUGHAN,3 & H. J. ESTANISLAU4 Perinatal, fetal and early neonatal mortality rates were determined in a population of 7392 babies born in hospitals in Pelotas (total population, 260 000) during 1982. These babies represented over 99% of all births in the city in that year. The perinatal mortality ratefor singletons was 31.9per 1000 total births, thefetal mortality rate being 16.2 and the early neonatal mortality rate 15.9 per 1000 total births. The most important variable influencing perinatal mortality was birth weight; low- birth-weight babies were 17 times more likely to die in the perinatal period than those weighing 2500 g or more. Perinatal mortality was also strongly influenced by socio- economic status, which was measured by family income. Babies belonging to the poorest families were 3 times more likely to die during the perinatal period than those in families with the highest incomes. Other variables significantly associated with perinatal mortality were gestational age, maternal age, maternal weight and height, reproductive history, attendance in antenatal clinics, parity, and birth interval. Information on most of these, risk factors can easily be obtained by health workers at the beginning ofpregnancy and should be taken into consideration when allocating women to the proper level of care. Although most epidemiological studies of the peri- natal period have been performed in industrialized countries (1-4), there is evidence that perinatal events are very important determinants of infant and child health in the less developed countries as well (5). It is known that birth weight strongly influences infant nutrition (6-8), morbidity (9), and mortality (5). As low birth weight is more common in the developing countries (10), its consequences will be more marked in these areas. One reason why many of these countries still lack sound studies in perinatal epi- demiology is that they still face high post-perinatal mortality rates and therefore tend to concentrate their preventive efforts in this period. Also, epi- demiological studies need to be properly designed and funded, and these conditions are not easily available; medical education is mainly concentrated on curative care, and financing agencies are less interested in such projects. ' Associate Professor, Department of Maternal and Child Health, Universidade Cat6lica de Pelotas, Brazil. Requests for re- prints should be sent to Dr F. C. Barros, Av. Dom Joaquim 982, 96.060-Pelotas-RS, Brazil. 2 Associate Professor, Department of Social Medicine, Universi- dade Federal de Pelotas, Brazil. 3Reader in Health Care Epidemiology, Evaluation and Planning Centre for Health Care, London School of Hygiene and Tropical Medicine, London, England. 4Assistant Professor, Department of Maternal and Child Health, Universidade Cat6lica de Pelotas, Brazil. In Brazil, a vast country with over 130 million in- habitants and considerable social disparities, the few studies of perinatal epidemiology available are either based on official statistics or restricted to births from particular hospitals. These studies are flawed, in the first case by the under-registration of deaths and in the second by the lack of representativeness of the sample under study. The present paper is based on a large study of peri- natal health carried out in the city of Pelotas, southern Brazil.a It is, to our knowledge, the first population- based study of perinatal health in this country. Its main objective was to estimate perinatal mortality rates and to identify risk factors with a view to prevention. METHODOLOGY All 7392 hospital births occurring in Pelotas (popu- lation, 260 000) in the year 1982 were studied. Access to hospital confinement is almost universal and a recent household census revealed that less than 1% of the births occurred at home in that year (11). This perinatal research was carried out at three levels: in BARROS, F. C. The epidemiology ofperinatal health in southern Brazil. A study of perinatal mortality, low birth weight and the utilisation ofhealth care (Ph.D. thesis, University of London, 1985). 4756 -95- F. C. BARROS ET AL. maternity hospitals, through home visits, and at local registries for births and deaths. Hospital study All mothers giving birth in the three city hospitals were interviewed with the help of a previously tested and structured questionnaire focused on demo- graphic, environmental and socioeconomic variables and on health care utilization; at the same time a de- tailed history was taken about previous pregnancies. The mothers were measured and weighed upon arrival, and the newborns were weighed on the first day of life. The interviews were conducted by paedia- tricians, obstetricians, and senior medical students previously trained during two pilot studies. All newborns were followed up during their hospi- tal stay, special attention being paid to morbidity and mortality. The criteria for their admission into the study were (a) when the birth weight was equal to 500 g or more and there were signs of life (liveborns), and (b) in the case of stillbirths, when the gestational age was 28 weeks or more (if the period of gestation was unknown, fetuses weighing 1000 g or more were included). These limits were adopted because for still- borns 1000 g is usually considered to be the weight of fetuses at 28 weeks, whereas some live babies weighing 500-1000 g are known to survive the peri- natal period and must therefore be taken into consid- eration when this period is studied. Home visits Based on a random sample of approximately 1501o of the births, 1093 babies were visited at home after the first week of life. These visits had three main objectives. Firstly, as the perinatal period ends on the seventh day and hospital discharges occurred very early (often less than 24 hours after birth), a sample of the babies was followed up to check their health status. Secondly, this was a way of double-checking some of the information collected in hospital. And thirdly, information relating to the homes and other environmental conditions could be gathered. These home visits yielded only one extra death which failed to reach the official registration system. Monitoring of death certificates All death certificates of the babies born in 1982 were checked every month and were later linked to the information provided by the hospital question- naires. Variables studied The outcome variables were the perinatal mortality rate (PMR, or the number of perinatal deaths per 1000 total births), the fetal mortality rate (FMR, or number of fetal deaths per 1000 total births), and the early neonatal mortality rate (ENMR, or number of deaths in the first week per 1000 live births). The following variables were studied: -Birth weight. Babies were weighed during the first 24 hours of life with calibrated scales; low birth weight (LBW) was defined as less than 2500 g. -Gestational age was calculated in complete weeks from the first day of the last menstrual period (LMP) to the date of birth; this information was not avail- able in 21.40o of births. Babies with a gestational age of less than 37 weeks were considered as pre-term. -Family income was expressed as minimum wages (MW) per month (currently about US$ 50). -Smoking during pregnancy. Mothers were classi- fied as ever-smokers or non-smokers. -Maternal age was expressed in years. -Maternal weight at the beginning of pregnancy was expressed in kilograms, based on the figure given in the antenatal card or as remembered by the mother; this information was not obtained for 15.70o of the mothers. -Maternal weight at the end of pregnancy was obtained from either the hospital admission card or the antenatal card, if the mother had been weighed in the previous 10 days; this information was not avail- able for 16.6%7o of the mothers. - Weight gain during pregnancy was the difference between maternal weights at the end and at the begin- ning of pregnancy; in 27.50% of cases this could not be calculated. -Maternal height was expressed in centimetres; only 2.3%o of the mothers were not measured. -Previous reproductive history included previous abortions, stillbirths, neonatal deaths, LBW babies and caesarean sections. -Parity indicated the number of previous births. -Birth interval was given by the number of months since the last birth. -Number of attendances at the antenatal clinic (ANC) was obtained from the antenatal card or, if unavailable, by asking the mother. Data analysis The statistical package for the social sciences (SPSS) (12) was used for descriptive analysis and con- tingency tables. Logistic regression analyses were per- formed with the GLIM (generalized linear interactive modelling) system (13). RESULTS Perinatal mortality rates and causes of death The perinatal outcome of single, multiple, and all 96 PERINATAL MORTALITY IN BRAZIL 97 Table 1. Percentage incidence of low birth weight (LBW) births and the fetal, early neonatal and perinatal mortality rates of single, multiple and total births, Pelotas, 1982 Early No. LBW Fetal neonatal Perinatal of (%) mortality' mortality' mortality' births Single 8.1 16.2 (118)" 15.9 (114) 31.9 (232) 7267 Multiple 58.0 32.0 (4) 107.4 (13) 136.0 117) 125 All births 8.8 16.5 (122) 17.5 (127) 33.7 (249) 7392 ' Rates are given per 1000 births for fetal and perinatal mortality and 1000 live births for early neonatal mortality. " Figures in parentheses are the number of deaths. births and the proportions of low-birth-weight babies are shown in Table 1. The overall perinatal mortality rate was 33.7 per 1000 total births, nearly equally split between fetal and early neonatal deaths, and 31.9 per 1000 singleton births. The causes of death are shown in Table 2. In 407o of these deaths it was impossible to ascertain the cause with confidence; the majority were stillbirths. Preterm-related problems, such as immaturity and hyaline membrane disease, were responsible for nearly 407o of deaths with a known cause. Birth weight and gestational age as risk factors The study of risk factors was restricted to single- tons, as proposed by Macfarlane et al. (14). Table 3 shows the perinatal, fetal and early neonatal mortality rates for various birth weight groups. More than one in five babies weighing less than 2500 g died in the perinatal period. This was 17 times higher than the PMR of babies weighing >2500 g. More than 60%0 of all perinatal deaths, and 71.2% of deaths in the first week occurred in LBW babies, which gives a relative risk of 28 compared with babies weighing ) 2500 g for early neonatal mortality. Regarding gestational age, almost 20% of the pre- term babies died during the perinatal period. Term babies of 37-38 weeks of gestation were twice as likely to die in the perinatal period than those of 39-41 weeks (Table 3). Babies with an unknown length of gestation presented a PMR of 54.0 per 1000 total births. The relationship between perinatal mortality, gestational age and birth weight was also studied; 42'7o of the LBW babies were preterm and the remain- ing 58070 had a gestational age of 37 weeks or more, and were probably growth-retarded. Table 3 also shows that one third of the LBW, preterm babies died during the perinatal period, whereas only one in 10 died among LBW babies born at term. Deaths were evenly distributed between fetal and early neonatal periods for the LBW, preterm babies, but occurred predominantly during the fetal period for the intra- uterine growth-retarded group. Babies weighing ) 2500 g presented a low perinatal mortality rate irrespective of their gestational age, which suggests that low birth weight was a much stronger deter- minant of perinatal mortality than gestational age. However, as large preterm babies comprised a small group with only 3 deaths, the pattern of mortality de- scribed for this group should be judged with caution. Table 2. Number of perinatal deaths and mortality rates by cause (all births included), Pelotas, 1982 Deaths in Mortality Cause of death Fetal deaths first week Total rates' Immaturity 0 38 38 (15.3)" 5.2 Asphyxia 18 10 28 (11.2) 3.8 Malformation 5 19 24 (9.7) 3.2 Hyaline membrane disease 0 19 19 (7.6) 2.6 Infection 2 10 12 (4.8) 1.6 Trauma 0 5 5 (2.0) 0.7 Other causes 11 12 23 (9.2) 3.1 Ill-defined 86 14 100 (40.2) 13.5 Total 122 127 249 (100.0) 33.7 "Rates per 1000 births. "Figures in parentheses are percentages. 98 F. C. BARROS ET AL. Table 3. Perinatal, fetal and early neonatal mortality rates by different birth weight groups (singletons only), gestational age, and combined birth weight and gestational age categories, Pelotas, 1982 Early Perinatal Fetal neonatal No. of mortality" mortality' mortality' births Birth weights (g): < 1000 958.3 41.7 956.5 24 1000-1499 690.1 322.9 541.7 71 1500-1999 219.5 138.2 94.3 123 2000-2499 97.1 47.4 52.2 422 2500-2999 16.6 9.8 6.8 1629 3000-3499 13.1 7.6 5.5 2750 3500-3999 9.0 6.2 2.8 1768 )4000 17.0 14.8 2.1 471 <2500 218.7 95.3 136.4 640)2500 13.1 8.3 4.8 6618 All groups 31.9 16.2 15.9 7267 Gestational age (weeks): >37 188.0 99.7 98.1 351 37-38 22.0 14.7 7.4 1226 39-41 10.4 6.4 3.9 3567 )42 31.7 19.4 12.6 567 All groups 25.9 15.2 10.8 5711 X2 401.08 186.02 241.40 P 0.00001 0.00001 0.00001 LBW, pretermh (< 2500 g, < 37 weeks) 337.0 171.3 200.0 181 LBW, term (IUGR)" (<2500g, >37 weeks) 98.6 71.7 29.0 223 Large, preterm (>2500g, <37 weeks) 18.0 12.0 6.1 167 ABW, term" (>2500g, 37 weeks) 11.3 7.0 4.3 5133 X2 804.70 372.42 544.32 P 0.000001 0.000001 0.000001 "Rates per 1000 births for fetal and perinatal mortality and 1000 live births for early neonatal mortality. "LBW (low birth weight); IUGR (intra-uterine growth retardation); ABW ("appropriate" birth weight). Social factors Family income. Table 4 shows that babies in the poorest families (1 minimum wage (MW) or less per month) were more than 3 times more likely to die in the perinatal period than babies from the richest families (over 10 times the minimum wage per month). This difference was nearly fivefold for the fetal mortality rate, which showed a steady decline in mortality with rising income. Smoking. The perinatal mortality rates of babies of smokers and non-smokers were exactly the same (31.9 per 1000), even though the babies of smokers were twice as likely to be of low birth weight than those of non-smokers (11 .4Gb and 6.5%, respectively). Table 4. Perinatal, fetal and early neonatal mortality rates according to family income, measured in minimum wages (MW) per month,' Pelotas, 1982 Early Family income Perinatal Fetal neonatal No. of (MW/month) mortality" mortality" mortality" births A 1 44.6 24.2 20.8 1772 1.1-3 31.9 15.9 16.2 3382 3.1-6 20.5 12.3 8.3 1218 6.1-10 22.5 6.7 15.8 444 >10 13.2 5.2 7.9 379 All groups 31.5 16.2 15.5 7195 X2 20.07 13.66 4.54 P 0.0005 0.008 0.02 a About USS 50 per month. h See footnote a in Table 1. PERINATAL MORTALITY IN BRAZIL Biological factors Maternal age. Fig. 1 shows a J-shaped curve for the perinatal mortality rates, which were lowest for the babies of mothers aged 25-29 years. The PMR for babies of teenagers was virtually the same as that for the total population (33.2 per 1000), the rates increas- ing markedly after 30 years of age. On the other hand, the lowest fetal mortality rates were seen among babies of adolescents (11.3 per 1000), the rates increasing progressively with age to 48.1 per 1000 in the oldest group. Deaths in the first week (early neonatal) were more common in the extreme age groups. The association between maternal age, family income, antenatal care and perinatal mortality was explored further through logistic regression analysis. Fig. 2 shows the shapes of the curves relating mater- nal age to the relative risks for perinatal mortality, the value of 1 being given to the risk for adolescent mothers. While adolescents are at higher risk than mothers in their twenties in the unadjusted curve, after adjusting for family income and number of antenatal visits the adolescents become the group at a z 4 0 I ir tL0. n 4 trI' 4 I-_ 0 2 N1-1 . / S / // /w I,, "I I I, If ENMR * --*~~~~~~~1 I711 1 .1 lowest risk. The observed differences in the shapes are due to changes in the relative position of mothers aged under 25 relative to older women, for whom the slope of the curve remains virtually unchanged. A test for a linear trend of perinatal mortality with age becomes more significant in the adjusted curves (P<0.001) than in the unadjusted data (P=0.01). After adjustment, women aged 35-39 years are at a significantly higher risk than teenagers (odds ratio, 2.01; 950/ confidence interval, 1.18-3.41), as well as those aged 40 years or more (odds ratio, 2.59; 95%I confidence interval, 1.31-5.11). Maternal weight at the beginning of pregnancy. There were no significant differences in perinatal or early neonatal mortality rates according to maternal weight (Table 5). On the other hand, fetal deaths in- creased with maternal weight (P= 0.05). Maternal weight at the end of pregnancy. Babies whose mothers weighed less than 55 kg were almost 4 times more likely to die in the first week of life than those of mothers weighing 75 kg or more (Table 5). Weight gain in pregnancy. The perinatal and early neonatal mortality rates decreased smoothly with in- creases in maternal weight gain, and were lowest in the group that gained 16 kg or more during preg- nancy (Table 5). :3 14_ 0 2 -J 4 IL cr 0 Cl) at !i 30 2.5 20 1.5 2 1.0 0.9 0.8* 0.7 06 05* UNADJUSTED ADJUSTED FOR INCOME --- ADJUSTED FOR INCOME AND ANTENATAL CARE < 20 20-24 25-29 30-34 35-39 (e4045 (Ita44} 0 95iff) (1,9)} (got) MATERNAL AGE (NUMBER OF BIRTHS) Fig. 1. Perinatal, fetal and early neonatal mortality rates (PMR, FMR and ENMR) according to maternal age, Pelotas, 1982. < 20 20-24 25-29 30-34 35-39 40+ (i145) (2284) (1955) (14198) (552) (191) AGE OF THE MOTHER IN YEARS (NUMBER OF CHILDREN) Fig. 2. Comparison of relative risks for perinatal mor- tality by maternal age, using unadjusted data and after adjustment for family income and the combined effects of antenatal care and family income, Pelotas, 1982. .1 99 PMR .\ I F. C. BARROS ET AL. Table 5. Perinatal, fetal and early neonatal mortality rates according to maternal weights at the beginning and end of pregnancy, the weight gain during pregnancy, and maternal height, Pelotas, 1982 Early Perinatal Fetal neonatal No. of mortality' mortality" mortality" births Initial weight (kg): < 49.0 27.6 8.0 19.7 1378 49.0-53.9 30.1 15.7 14.6 1461 54.0-60.9 28.9 15.0 14.1 1660 >61.0 31.9 20.3 11.9 1628 All groups 29.7 15.0 14.9 6127 P NS" 0.05 NS Final weight (kg): < 55.0 30.4 7.1 23.4 559 55.0-64.9 31.3 15.1 16.3 2111 65.0-74.9 20.7 11.5 9.2 2154 > 75.0 22.8 16.2 6.7 1288 All groups 25.7 13.4 12.4 6112 P NS NS 0.004 Weight gain (kg): < 7.0 45.6 15.9 31.3 877 7.9- 9.9 32.7 14.9 18.1 1008 10.0-12.9 22.4 12.7 9.8 1340 13.0-15.9 18.8 10.9 8.0 1011 ,16.0 16.8 11.5 5.4 1129 All groups 26.3 13.0 13.6 5365 P 0.0003 NS 0.00001 Height (cm): < 1 50 27.0 6.4 20.7 777 150-154 34.0 19.1 15.1 1883 155-159 33.8 16.7 17.4 2217 160-164 32.5 18.2 14.6 1535 > 165 11.6 10.2 1.5 687 All groups 30.7 15.9 15.0 7099 P 0.03 NS 0.02 See footnote a in Table 1. NS, not significant. Maternal height. The perinatal mortality rate was significantly lower among babies of taller mothers (Table 5). However, when family income was con- trolled through logistic regression analysis the as- sociation between maternal height and perinatal mortality ceased to be significant (likelihood ratio test (LRT), 2.968; P>0.05). The fetal mortality rate was lowest for the shortest mothers (6.4 per 1000), but this was not statistically significant and may be the result of small numbers. The early neonatal mortality rate was lowest among the women over 164 cm, with only one death in 687 births. Previous reproductive history. The perinatal and fetal mortality rates were significantly increased for mothers with a previous abortion (Table 6), owing to a higher risk of having a fetal death. Mothers with a history of a previous fetal death were almost 4 times more likely to have another stillbirth compared with those without such a history. Mothers with a previous LBW baby had twice the risk of having a perinatal death than those without a previous LBW baby. The perinatal mortality rates for mothers with or without a previous caesarean section were almost significantly different (44.0/1000 and 31.3/1000, respectively) (P = 0.07). Parity. The fetal mortality rate was lowest among babies of parity-2 mothers (12.9 per 1000) and highest for babies whose mothers had 3 or more previous births (Table 7). The same trend was observed for perinatal deaths, but the differences were not quite significant (P= 0.07). Birth interval. The lowest early neonatal mortality rate was seen for babies born after a birth interval of 36 to 47 months (7.9 per 1000), and the highest rates were for the groups with a birth interval of less than 36 months (Table 7). Health care Antenatal care. There was a highly significant trend towards lower rates of perinatal, fetal and early neonatal mortality with increase in the number of antenatal visits. The perinatal mortality rate among babies of non-attenders was 56.2 per 1000 and for those with 10 or more visits it was 16.2 per 1000 (Table 7). When the effect of family income was con- trolled through logistic regression analysis, the asso- ciation of number of ANC visits and perinatal mortality remained highly significant (LRT, 14.232; P<0.001). Type of delivery. Although the perinatal mortality rate was higher for babies delivered by caesarean section (38.4 per 1000) compared with vaginal delivery (29.4 per 1000), this difference was not quite significant (P=0.06) (Table 7). Deaths in the first week were significantly more frequent in the caesar- ean section group than after a vaginal delivery (21.8/1000 and 13.7/1000, respectively). DISCUSSION The perinatal, fetal and early neonatal mortality rates described in this study are high, considering that our birth weight distribution compares favourably with countries showing lower perinatal mortality rates. For example, in Cuba a LBW incidence of 100 PERINATAL MORTALITY IN BRAZIL Table 6. Perinatal, fetal and early neonatal mortality rates of babies whose mothers had had a previous abortion, still- birth, low-birth-weight baby, neonatal death, or caesarean section, Pelotas, 1982 Early Perinatal Fetal neonatal No. of mortality' mortality mortality' births Previous abortion: Yes 48.5 29.5 19.5 1319 No 27.7 12.4 15.5 3387 P 0.0005 0.0001 NS" Previous stillbirth: Yes 77.8 61.1 17.7 180 No 32.0 15.7 16.6 4526 P 0.001 0.00001 NS Previous LBW baby: Yes 54.0 25.6 29.1 741 No 27.5 12.8 14.9 3595 P 0.0003 0.01 0.01 Previous neonatal death: Yes 38.4 12.8 25.9 234 No 33.5 17.7 16.1 4472 P NS NS NS Previous caesarean section: Yes 44.0 20.9 23.6 909 No 31.3 15.5 14.9 3798 P NS NS NS See footnote a in Table 1. NS, not significant. 10.8% and a PMR of 26.9 per 1000 have been de- scribed (4), and in Cardiff (Wales) the respective values were 8.707o and 22.1 per 1000 in the period 1972-76 (15). Although the birth weight distribution in Pelotas in terms of socioeconomic and other factors can be re- garded as reasonably good, low birth weight was a leading risk factor for perinatal mortality and LBW babies were 17 times more likely to die in the perinatal period than babies weighing 2500 g or more. One way of lowering the perinatal mortality is theoretically, therefore, to improve the birth weight distribution, especially among socially deprived groups, where the incidence of LBW is highest. In fact, the striking variations observed in the PMR according to family income strongly suggest that social inequalities are important determinants of perinatal problems, largely by influencing biological factors (including birth weight) and the utilization of health care. Peri- natal mortality was 3.3 times higher among babies of the poorest families compared with the richest. This difference is much greater than that observed in de- veloped countries, such as the United Kingdom, where there is a twofold increase from social class I to V (16). Another important variable influencing perinatal mortality was gestational age, with the PMR reaching almost 200 per 1000 among preterm babies. In addition, term babies did not make up a homo- geneous group, and babies after 37-38 weeks of gestation had a twofold increase in the PMR com- pared with those after 39-41 weeks. This finding has important applications in Brazil, where elective caesarean sections are frequently performed. In the present population, for example, 28% of all deliv- eries were by caesarean section (17, 18). One interesting finding of this study was that the babies of smoking mothers did not show an increased perinatal mortality despite a twofold increase in low birth weight, and the fact that smoking mothers were significantly poorer and lighter than non-smokers. It appears that this comparatively lower mortality of babies of smokers occurred because smoking has a stronger effect on birth weight than on gestational age, so that when the newborns of the same birth weight are compared, those of smoking mothers tend 101 F. C. BARROS ET AL. Table 7. Perinatal, fetal and early neonatal mortality rates according to parity, birth interval, number of antenatal clinic (ANC) attendances, and type of delivery, Pelotas, 1982 Early Perinatal Fetal neonatal No. of mortality' mortality' mortality births Parity: 0 28.9 14.1 15.0 2380 1 31.1 15.1 16.2 2054 2 26.7 12.9 13.9 1161 3 44.4 23.1 21.7 518 ,>4 45.7 28.0 18.2 678 All groups 31.9 16.2 15.9 7241 P NS" 0.05 NS Birth interval (months): < 24 36.3 12.6 24.0 1330 24-35 35.7 15.8 .20.3 936 36-47 24.9 17.5 7.9 629 >48 33.5 22.6 11.2 1511 All groups 33.6 17.4 16.5 4406 P NS NS 0.01 ANC attendance: None 56.2 23.4 33.5 427 1-4 46.9 24.1 23.3 1449 5-9 27.1 14.0 13.3 4343 >10 16.2 7.1 9.2 983 All groups 31.4 15.7 15.9 7202 P 0.00001 0.003 0.0005 Type of delivery: Vaginal 29.4 16.0 13.7 5260 Caesarean section 38.4 16.9 21.8 2005 Both groups 31.9 16.2 15.9 7265 P NS NS 0.01 " See footnote a in Table 1. hNS, not significant. to show lower levels of prematurity. As it is known that for babies of the same birth weight the mortality is higher for those of lower gestational age, this could be one reason for the findings described above. How- ever, the literature is still controversial on this subject (2, 19-22) and further studies are needed, where the confounding effects of gestational age, family income, and other variables are taken into consider- ation. The association of a higher number of antenatal care attendances with a better perinatal outcome was also an important finding. The fact that this associ- ation remained quite strong even after controlling for the possible confounding effects of family income reinforces the impression that frequent visits to ante- natal care clinics should be encouraged. However, it is still not known whether antenatal care per se increases the likelihood of a better perinatal perfor- mance or if the women who attend more often are a self-selected group. Also, if antenatal care accounts for the difference, what are the specific aspects of such care that are responsible for the better outcome? Although further studies in this area are certainly needed, efforts to persuade pregnant women to attend antenatal clinics should continue, especially those wtho are at increased risk. This study has identified a number of risk factors for perinatal, fetal and early neonatal mortality that could be used by health workers when allocating pregnant women to the appropriate level of care. These factors are summarized in Table 8. Those that can be identified at the outset of pregnancy are low family income, low maternal weight and height, unfavourable maternal ages (adolescents or older 102 PERINATAL MORTALITY IN BRAZIL Table 8. Significance of different social, biological and health care variables on perinatal, fetal and early neo- natal mortality, Pelotas, 1982 Early Perinatal Fetal neonatal Variables mortality" mortality" mortality" Family income + + + Birth weight + + + Gestational age + + + Maternal age + + NS Maternal weight (initial) NS + NS Maternal weight (final) NS NS + Maternal weight gain + NS + Maternal height + NS + Previous abortion + + NS Previous stillbirth + + NS Previous LBW infant + + + Previous caesarean section NS NS NS Parity NS + NS Birth interval NS NS + Antenatal clinic attendance + + + Type of delivery NS NS + " The + sign indicates that P is S0.05; NS means P>O0.05. mothers), a history of perinatal problems in previous pregnancies, multiparity, and a short birth interval. Poor mothers were more likely to be smokers and of younger ages. Poverty thus presented a set of unfavourable characteristics, in terms of the woman's nutritional status, age, smoking, not reporting for antenatal care, and reproductive history which increased the risk of a perinatal death. Thus, it appears that family income might be used as one of the major indicative risk factors for perinatal death, which can be assessed early in pregnancy. Finally, the study showed that in 40% of the peri- natal deaths the cause was ill-defined mainly because post-mortem examinations are not routinely per- formed. The diagnoses are almost always presump- tive, and the cause of death of stillbirths cannot often be determined. Clearly, a better understanding of the causes of perinatal death is necessary, and the utiliz- ation of systems for ascertaining the causes of death that do not depend on high rates of post-mortems, like the one developed by Wigglesworth (23), should help. ACKNOWLEDGEMENTS This study was supported by the International Development Research Centre (IDRC), Ottawa, Canada. RESUMt MORTALITt PERINATALE DANS LE SUD DU BRtSIL: ETUDE SUR 7392 NAISSANCES L'etude de la mortalite perinatale a consiste a suivre tous (260 000 habitants), au sud du Bresil, a porte sur les 7392 enfants nes en milieu hospitalier. Toutes les femmes de cette ville ont acces a l'h6pital pour accoucher et un recensement des menages a revel que, cette annee-1a, moins de I Go des accouchements avaient eu lieu a domicile. L'etude de la mortalite perinatale a consiste a suivre tous les nouveau-nes pendant leur sejour a l'h6pital et, une semaine apres leur naissance, un echantillon de 15% d'entre eux a fait l'objet d'une visite a domicile, visite rendue neces- saire du fait que 23% des femmes quittent l'h6pital dans les 24 heures suivant la naissance et que la periode perinatale s'acheve le septieme jour. Les certificats de deces ont en outre e contr6les chaque mois dans les bureaux du Secre- tariat a la Sante et compares a l'information contenue dans les questionnaires des hopitaux. Le taux de mortalite perinatale etait de 33,7 pour 1000 naissances vivantes et de 31,9 pour 1000 naissances uniques. Le taux de mortalite foetale pour les naissances uniques etait de 16,2 pour 1000, et le taux de mortalite neonatale precoce de 15,9 pour 1000 naissances vivantes. La variable qui influait le plus sur la mortalite perinatale etait le poids a la naissance et le risque de deces pendant la periode perinatale etait 17 fois plus eleve pour les nouveau- nes de poids insuffisant a la naissance que pour ceux qui pesaient 2500 g ou plus. La mortalite perinatale dependait aussi largement du statut socio-economique, mesure d'apres le revenu familial. Le risque de deces pendant la periode perinatale etait trois fois plus eleve chez les enfants des families les plus pauvres que chez ceux des familles les plus aisees. Une autre variable etroitement liee a la mortalite peri- natale etait l'age gestationnel, la proportion d'enfants nes avant terme approchant de 200 pour 1000 naissances. Parmi les enfants nes a terme, le risque de deces pendant la periode perinatale etait deux fois plus eleve chez ceux qui etaient nes apres 37 ou 38 semaines de gestation que chez ceux qui etaient n6s a 39 ou 41 semaines (22,0 pour 1000 et 10,4 pour 103 104 F. C. BARROS ET AL. 1000 respectivement). Cette observation a des incidences importantes au Bresil oui les cesariennes sont souvent prati- quees systematiquement. Dans la population etudiee, par exemple, 27,6% de tous les accouchements ont eu lieu par cesarienne. Au nombre des autres variables associees de pres a la mortalite perinatale figurent l'age de la mere, le poids et la taille de la mere, les grossesses anterieures, la frequen- tation des dispensaires prenatals, la parite et l'intervalle entre les naissances. Les agents de sante peuvent facilement obtenir des renseignements sur ces facteurs de risque au debut de la grossesse et ils doivent en tenir compte lorsqu'ils decident du niveau de soins qui convient a chaque mere. REFERENCES 1. Better perinatal health. A survey from the Lancet. London, Lancet, 1980. 2. BUTLER, N. R. & ALBERMAN, E. Perinatal problems. London, Livingstone, 1969. 3. CHAMBERLAIN, R. ET AL. British births 1970. Vol. 1. The first week of life. London, Heinemann, 1978. 4. Social and biological effects on perinatal mortality: a WHO report. Saudi medical journal, 4 (Suppl. 1): 5-202 (1983). 5. PUFFER, R. R. & SERRANO, C. V. Patterns ofmortality in childhood (PAHO Scientific Publication No. 263). Washington, Pan American Health Organization, 1973, pp. 73-90. 6. BABSON, S. G. Growth of low-birth-weight infants. Journal ofpediatrics, 77: 11-18 (1970). 7. FITZHARDINGE, P. M. & STEVEN, E. M. The small-for- date infant. I. Later growth patterns. Pediatrics, 49: 671-681 (1972). 8. MATA, L. J. The children of Santa Maria Cauque: a prospective field study of health and growth. Massa- chusetts, MIT Press, 1978, pp. 167-201. 9. MCCORMICK, M. C. ET AL. Rehospitalization in the first year of life for high-risk survivors. Pediatrics, 66: 991-999 (1980). 10. The incidence of low birth weight. A critical review of available information. World health statistics quar- terly, 33: 197-224 (1980). 11. VICTORA, C. G. ET AL. [Longitudinal study of children born in 1982 in Pelotas, Rio Grande do Sul: method- ology and preliminary results.] Revista desauidepublica (Sdo Paulo), 19: 58-68 (1985) (in Portuguese). 12. NIE, N. H. ET AL. Statistical package for the social sciences. 2nd ed. New York, McGraw Hill, 1975. 13. BAKER, R. G. & NELDER, J. A. The GLIM system (Release 3). Oxford, Numerical Algorithms Group, 1978. 14. MACFARLANE, A. ET AL. The role of standardization in the interpretation of perinatal mortality rates. Health trends, 12: 45-50 (1980). 15. ALBERMAN, E. Prospects for better perinatal health. Lancet, 1: 189-192 (1980). 16. OFFICE OF HEALTH ECONOMICS. Perinatal mortality in Britain: a question of class (OHE Briefing No. 10). London, 1979, pp. 1-8. 17. BARROS, F. C. ET AL. Perinatal risk in Third World cities. World health forum, 6: 322-324 (1985). 18. BARROS, F. C. ET AL. Why so many caesarean sections? The need for a further policy change in Brazil. Health policy and planning, 1: 19-29 (1986). 19. MEYER, M. B. ET AL. Perinatal events associated with maternal smoking during pregnancy. American journal of epidemiology, 103: 464-476 (1976). 20. RUSH, D. & KASS, E. H. Maternal smoking: a reassess- ment of the association with perinatal mortality. American journal ofepidemiology, 96: 183-196 (1972). 21. YERUSHALMY, J. The relationship of parents' cigarette smoking to outcome of pregnancy -implications as to the problem of inferring causation from observed asso- ciations. American journal of epidemiology, 96: 443-456 (1971). 22. RANTAKALLIO, P. Groups at risk in low-birth-weight infants and perinatal mortality. Acta paediatrica Scan- dinavica, 193 (Suppl.): 1-71 (1969). 23. WIGGLESWORTH, J. S. Monitoring perinatal mortality: a pathophysiological approach. Lancet, 2: 684-686 (1980).