Risk Factors for Low Birthweight in Zimbabwean Women: A Secondary Data Analysis

Background Low birth weight (LBW) remains the main cause of mortality and morbidity in infants, and a problem in the care of pregnant women world-wide particularly in developing countries. The purpose of this study was to describe the socio-demographic, nutritional, reproductive, medical and obstetrical risk factors for delivering a live LBW infant at Harare Maternity Hospital, Zimbabwe. Methods A secondary data analysis from data obtained through a questionnaire and delivery records was conducted. Linear regression models with a complimentary log-log link function were used to estimate the relative risks for all LBW, term LBW and preterm LBW. Results The frequency of LBW was 16.7%. Lack of prenatal care (adjusted relative risk [ARR] 1.69, 95% CI 1.44, 1.98), mother’s mid-arm circumference below 28.5 cm, (ARR 1.35, 95% CI 1.19, 1.54) and rural residence (ARR 1.22, 95% CI 1.04, 1.40) increased the risk of LBW. Eclampsia, anemia, and ante-partum hemorrhage, were associated with LBW (ARR 2.64, 95% CI 1.30, 5.35; ARR = 2.63, 95% CI 1.16, 5.97; and ARR = 2.39, 95% CI 1.55, 3.68), respectively. Malaria increased the risk of LBW (ARR = 1.89, 95% CI 1.21, 2.96). Prenatal care, infant sex, anemia, antepartum hemorrhage, premature rapture of membranes and preterm labor were associated with the three LBW categories. History of abortion or stillbirth, history of LBW, malaria, eclampsia, and placenta Previa, were associated with all LBW and preterm LBW, while pregnancy induced hypertension, and number of children alive were associated with all LBW and term LBW. Conclusions LBW frequency remains high and is associated with nutritive, reproductive, medical and obstetrical factors. Preterm LBW and term LBW have similar and also different risk factors. Understanding the role of different risk factors in these different LBW categories is important if the goal is to reduce LBW frequency, and its complications, in Zimbabwe.

limited data for this important problem of natality in Zimbabwe persists [4,17]. The methods have been described elsewhere [24]. Briefly, all women delivering a singleton infant that survived the first hour of life were eligible to participate. For the 3,722 women who delivered during the study period, 527 (14.2%) were ineligible including 198 (5.3%) multiple pregnancies, 26 (0.7%) very ill babies 248 (16.7%) stillbirths and 52 (1.4%) early neonatal deaths. Of the 3,195 eligible women, 20 (0.6%) refused to participate, 27 (0.8%) could not be interviewed before discharge, 30 (0.9%) had incomplete records, and 8 (0.2%) did not have information on birth weight, leaving 3,110 (97.3%) eligible women for the LBW analysis.
In this present day analysis, the main outcome of interest was LBW, which was later categorized into term LBW, and preterm LBW infants, based on information collected from the medical records. Women were identified as having delivered a singleton LBW live infant if the infant weighed 500 grams, but below 2,500 grams at birth, irrespective of gestational age. Women who delivered a"normal weight, term" infant were all women who delivered a liveborn singleton baby weighing 2,500 grams and more at term (37 weeks of gestation and above). Term LBW infants were defined as term infants who weighed more than 500 grams, but below 2,500 grams. Preterm LBW infants were defined as preterm infants who weighed more than 500 grams, but below 2,500 grams.
Based on original study [24], each day at 8 am and 2 pm, a list of women was made from the delivery logbook. Eligible women who agreed to participate and signed the consent form, had their medical records abstracted, completed a short interview regarding demographic and lifestyle factors, and their baby was examined for maturity. Six research assistant were used for data collection, and to administer the Ballard method of assessing gestational age [25].
For socio-demographic factors, age of mother was calculated as the number of years from her date of birth to her previous birthday. Information of marital status (currently married or living as married, never married, separated, divorced or widowed), education of mother and father (less than primary education, having achieved primary education, secondary education and above), employment status of mother and father (yes/no), residence (urban/rural), if they had electricity (yes/no), had water supply (yes/no), and had a toilet (yes/no), were obtained by interview.
For lifestyle factors, information on alcohol drinking (chibuku, beer, spirits or wine), or drank home brew (mahewu), were obtained through the questionnaire. Mahewu is a local nonalcoholic nutritious beverage made from corn meal, rapoko or sorghum, soya beans and sugar [24]. Chibuku is a locally brewed alcoholic drink, which could also be nutritious, containing soya beans, and sugar apart from the intoxicants. Women were also asked individually (yes/ no), if they drank mahewu, chibuku, beer, spirits or wine during pregnancy. Women were asked about smoking during pregnancy (yes/no).
For anthropometric measurements, weight and height at first prenatal care visit of the mother were collected from medical records. Body mass index (BMI) was calculated as weight of mother at first contact, in kilograms divided by height in meters squared. Mid-arm circumference (MUAC) was obtained by measuring the length between shoulder and the elbow with arm bend, with circumference measured at the midpoint [26]. MUAC was later categorized, using Jellife standards at cut off of more than 28.5 centimeters (cm), less denoting under nutrition [27].
Mothers were grouped into those who attended at least one visit of prenatal care during pregnancy or otherwise. Parity (0, 1 to 2 and more than 2 pregnancies), prior history of abortion (delivery before 20 weeks of gestation, or infant weighing less than 500 grams at birth), stillbirth or LBW birth (yes/no) sex of the infant, and infant birth weight were abstracted from obstetrical records [24]. Information on a diagnosis of chronic medical condition or obstetrical complication including diabetes, hypertension, anemia, pregnancy induced hypertension, eclampsia, cardiovascular disease, ante-partum hemorrhage, premature rapture of membranes (PROM), preterm labor (PTL) with current pregnancy and placenta previa, based on diagnosis by attending medical doctor, was abstracted from obstetrical records [24]. History of infections during pregnancy of malaria, urinary tract infection, syphilis or gonorrhea, were also obtained from obstetrical records.

Ethics Statement
When one is carrying out research on humans in Zimbabwe, it is required to get approval from and register the project with the Medical Research Council of Zimbabwe, which in turn gets the Research Council of Zimbabwe's approval. This process involves seeking permission from the Permanent Secretary of Health and from the departmental heads of any institutions defined in the proposal. The principal investigator was responsible for processing this approval. The University of Michigan Institutional Review Board permission was also obtained prior to this study. The medical record review did not use personal identities and we requested exempt status for that portion of the study. A consent form, translated into Shona, was signed by each individual study participant. The study therefore, was approved by the University of Michigan Institutional Review Board and the Medical Research Council of Zimbabwe and permissions were obtained from the Ministry of Health and Harare Central Hospital. This article is based on secondary analysis of these data, therefore exempt for IRB.

Statistical analysis
Since a complete population of live births over a 3-month period within the hospital was collected, we estimated the relative risks of LBW, and its sub-sets (as common in Reproductive and Perinatal Epidemiology). In univariable analysis, crude relative risks, 95 percent confidence intervals and chi-square tests were calculated from cross-tabulations with each pair of the outcome variables (all LBW, term LBW and preterm LBW) and each risk factor (determinants or exposures) to assess the association between each potential risk factor and either all LBW, term LBW, or preterm LBW using EPINFO version 7. We grouped risk factors into major subsets including, socio-demographic; anthropometric and nutritional factors; reproductive factors; medical and obstetrical complications; and infections, in our crude analysis. We then fitted generalized linear regression models with a complimentary log-log link function, because of the nature of our data, to estimate the adjusted relative risks of all LBW for these risk factors [28][29]. We limited our adjusted analysis to all LBW infants as numbers for term LBW and preterm LBW were too small, and were giving unstable estimates. We built a model of socio-demographic, anthropometric and nutritional factors and reproductive factors, using a cut off of a p-value of 0.10 for variables to be included and also taking into account their contribution to the model, or if they were established potential confounders for this outcome. We also adjusted for whether the mother was referred to Harare Maternity Hospital or not. In the second multivariable model, we examined the risk associated with medical and obstetrical complications and for infections after adjustment for socio-demographic, reproductive factors, anthropometric and nutritional factors. We did not analyze data for VLBW (births below 1,500 grams in weight) and ELBW (births below 1,000 grams in weight) as the numbers were small, and gave unstable estimates. Data were analyzed using SAS version 9.3 (SAS Institute, Cary, NC).
Age of mother ranged from 13 to 49 years (mean = 24.4 years). Very few mothers (7.3%) were above 34 years old, in Table 1. About three quarters of the mothers had attained a secondary level of education and almost all women were married, the majority lived in urban areas. Very few women (13.4%) were employed, and 82.2% of fathers had some form of employment.
The mean maternal weight (and standard deviation) was 64.7(±11.4) kilograms (range = 25-135 kilograms), BMI 26.0 (±4.6) (range = 11.4-36.8) and MUAC was 27.5(±3.2) cm (range = 17.5-43 cm). In crude analysis, women with BMI 25 were 91% less likely to have a LBW, and 55% less likely to deliver a term LBW infant compared with women with BMI 18.49 to less than 25, in Table 2. Women with a BMI less than 18.49 had a 98% increased crude risk of LBW. Women with a MUAC of less than 28.5 cm compared to women with MUAC of 28.5 and more, had greater risk of delivering all three LBW categories of LBW, 75%, 2.36-fold increase, and 62% increase in the risk for all LBW, term LBW and preterm LBW infants, respectively, in Table 2.
During pregnancy, about 27.1% of the women drank alcohol, and most women (90.2%) reported drinking a local non-alcoholic nutritional beverage (mahewu). Alcohol drinking was associated with increased risk of LBW, and preterm LBW in the crude analysis, in Table 2. Drinking mahewu or chibuku was not associated with all the three forms of LBW reported in this study. Only 10 (0.3%) of women smoke, and smoking were not associated with LBW.

Reproductive factors and infections
About 87% of women received prenatal care, in Table 3, and lack of prenatal care was associated with a 2.29, 1.77, 3.30-fold increases in the risk of all LBW, term LBW, and preterm LBW, respectively. Almost 86% of women were referred to Harare Maternity Hospital for delivery with 78% coming from Harare Municipal clinics. Referral was not associated with risk of all the three categories of LBW, in Table 3. Parity ranged from 0 to 9 with almost half of the women having their first child, and parity was not associated with all forms of LBW in this present day study. Having no live children was associated 18% and 43% increased risk of all LBW and term LBW, in Table 4. About 6.1% reported having a previous history of abortion, 0.3% stillbirth, 10.6% previous LBW birth. Women reporting prior history of abortion or stillbirth had a 47% increased risk of LBW and 54% increase preterm LBW in crude analysis, in Table 4. Women with prior history of delivering a LBW infant had a 70% increase in risk of all LBW, and 2.34-fold increase in delivering a preterm LBW infant. Delivering a female infant was associated with a 36%, 40% and 33% decrease in risk of delivering all three forms of LBW: all LBW, term LBW and preterm LBW, respectively, in Table 4. Therefore, male infants were less likely to be LBW, irrespective of gestational age. The frequency of mothers diagnosed with malaria during pregnancy or urinary tract infection was less than 3%, in Table 3. History of malaria was associated with a 1.93-fold increase in all LBW, and a 2.66-fold increase in preterm LBW in crude analysis, in Table 4. Urinary tract infection was not associated with all forms of LBW reported in this study. Nearly 1% (n = 27) of the women were diagnosed with syphilis, but almost half (41.6%) were not tested, in Table 3. About 5.3% women were diagnosed with gonorrhea, and were less likely to deliver a LBW infant. Women who were not tested (93.4%) for gonorrhea, were less likely to deliver a preterm LBW infant, in Table 4.

Medical factors and obstetrical complications
The frequency of mothers diagnosed with medical or obstetrical complications, except for pregnancy induced hypertension and PROM was low, in Table 5. Less than 1% of the women had a diagnosis of anemia, diabetes, cardiovascular disease, eclampsia, or placenta Previa, and less than 10% had a diagnosis of hypertension, ante-partum hemorrhage, or history of PTL. Still, even with the small numbers, anemia was associated with a 3.51-fold increase risk of all LBW, increasing to 4.44-fold increase for term LBW and 5.15-fold increase for preterm LBW infant, in Table 6.
Hypertension had a 47% increase in the risk of all LBW infants only. On the other hand, pregnancy induced hypertension was associated with 31% increase in all LBW, and 53% increase in term LBW infants. Also, eclampsia was associated with 3.25-fold increase in all    Table 7 presents adjusted relative risks for reproductive and nutritive factors, as well as obstetrical complications and history of infections after adjusting for relevant demographic, reproductive and nutritive factors. Maternal age and rural residence were modestly associated with all LBW, in adjusted analysis. Prenatal care, history of abortion or stillbirth, and MUMC, remained significantly associated with all LBW. Obstetric complications, except for placenta Previa, remained significant factors for all LBW in adjusted models. Malaria remained significantly associated with all LBW, while the urinary tract infection remained insignificant in adjusted analysis.

Discussion
This paper, using a secondary data analysis, with a full data complement for over a 3-month period, estimated and examined risk factors for all LBW, term LBW and preterm LBW, among mothers delivering live births at the largest referral center in Harare, Zimbabwe. Our results suggest that the frequency of LBW is high; that nutritional factors, prenatal care, history of abortion, previous LBW, stillbirth, obstetric complications including anemia, hypertension, pregnancy induced hypertension, eclampsia, ante-partum hemorrhage are important predictors of LBW in this population. Having PTL and PROM were associated with LBW. Infection with malaria during pregnancy was also associated with LBW.
The prevalence of LBW among live-births at Harare Maternity Hospital over a three-month period of 167/1000 live births is comparable to the 168/1000 live births observed at the same hospital [10], based on delivery log data. More recently Feresu et al described a frequency rate of 19.9% and 24.3% at Harare Maternity Hospital [4,[9][10][11][12]. The prevalence of LBW for Zimbabwe from the UNICEF and WHO was 11.0% for 1999 [1,6,17]. Sanders et al had reported an incidence of LBW of 10.8% [9], while WHO annual reports incidence of 11.0% as late as 2013 [4,17]. The prevalence and incidence are varying without a clear trend. There is need for more studies to depict the estimates and patterns of LBW, since it is an important contributor to infant mortality in Zimbabwe [4,[9][10][11][12][13][14]. The LBW rate in a population is a good indicator of a public health problem that includes long-term maternal malnutrition, ill health and poor health care. On an individual basis, LBW is an important predictor of newborn health and survival. Limited data from Zimbabwe suggests that LBW is a common obstetrical problem [4,[9][10][11][12][13][14], and an important contributor to infant mortality [9][10][11][13][14]. Levels of infant mortality in Zimbabwe (73 per1000 live births) are high compared to South Africa (55 per 1000 live births), mid-income countries such as Mexico (25 per 1000 live births), and to developed countries including the USA, UK, or Sweden (7, 6 and 3 per 1000 live births respectively) [6][7].
We note that some conditions such as, prenatal care, infant sex, anemia, antepartum hemorrhage, PROM and history of preterm labor were associated with the three LBW categories; all LBW, term LBW and preterm LBW. Conditions like history of abortion or stillbirth, history of LBW, malaria during pregnancy, eclampsia, and placenta Previa, were associated with all LBW and preterm LBW. While, pregnancy induced hypertension, and number of children alive tend to be associated with all LBW and term LBW infants. Infections tend to be associated with preterm LBW. Chronic conditions tend to be associated with term LBW, while acute conditions tend to be associated with preterm LBW. Some risk factors or determinants are common for both term and preterm LBW. These observations may have to do with the differences in etiology for intrauterine growth restriction versus, preterm birth, although both area subset of  LBW. We discuss this phenomenon, in our previous studies [10,13,24]. Further studies to discern these differences, will help handing of LBW in its different forms in pregnancy. Poor nutrition measured through BMI and MUAC was adversely associated with LBW, as in previous studies [30][31], confirming the role nutrition in pregnancy [32]. Anemia was associated LBW in this population [10][11]13], similar to other studies [33][34][35]. More than a 5.9 § based on data obtained from the records † denotes raw percent within group π history of premature rupture of membranes with current pregnancy # history of pre-term labor with current pregnancy * 211 observations had no information on anemia ** 328 observations had no information on diabetes mellitus *** 333 observations had no information on cardiovascular disease **** 338 observations had no information on hypertension quarter of mothers in our study drank alcohol, as was in previous studies [36], a modest number (10.1%) drank chibuku a local brew, while less than 3% each drank wine, or bottled beer during pregnancy. But, about 90.2% of the women drank mahewu, a nutritious drink, which in previous studies seemed to be associated with reduced risk of preterm birth [24]. In our present day study, using the same dataset, mahewu was not associated with LBW, and therefore needs further investigation.
Although Harare is an urban setting, malaria appears to remain an important determinant of LBW in this population as has been shown previously [24], and in African populations [20,  . Malaria was not endemic in this setting, thus women were not likely to be screened for parasites, raising concerns about missed cases, particularly among women arriving from rural endemic areas. In Zimbabwe, screening for syphilis or gonorrhea during pregnancy is poor [24,49], consistent with our study. Infections were not often recorded in the medical records and are not consistently screened for, thus under-diagnosis and inadequate treatment for these conditions is likely. Screening for infections such as syphilis, gonorrhea and malaria as a whole is poor, and thus we could not evaluate these important risk factors. Our present day study did not evaluate the role of HIV infection. Recent studies have tried to evaluate the role of HIV infection for perinatal outcomes including LBW and PTD, in this population, especially with the use of antiretroviral drugs [20][21]23], however more studies are still needed to further characterize these relationships. In this study we described the risk factors for LBW, a contributor of infant mortality, for women delivering at HMH. However, some potential limitations in our study warrant consideration. Our study focused on live-births as WHO uses this rate [1,3,6], but excluding stillbirth and early neonatal deaths underestimates the true burden of LBW and may have biased our results towards the null hypothesis. Focusing solely on births within Harare Maternity Unit raises concerns about potential selection bias, however, the risk of LBW did not differ between those referred and not referred, and referral status was controlled for in the final model on demographic, nutritive and reproductive factors. Estimates of LBW are within range of other studies [1,[9][10][11][12]17]. Exclusion of twin deliveries and stillbirth somewhat underestimates the rate of LBW at this hospital. Prenatal care in this study was defined as attending care irrespective of gestational age. We did not have proper estimates of timing and number of visits. Lack of adequate data on the number of prenatal care visits limits our ability to assess the adequacy of prenatal care. Our study relied on medical records, raising concerns of incomplete data. A culmination of poor reporting, lack of adequate screening for medical conditions, and infections, and ensuing possibility of missed exposed cases, raises concerns about non-random misclassification. Our small sample size did not allow us to further explore term, term LBW, and preterm LBW birth relationships with risk factors in adjusted analysis, nor could we analyze for very low and extremely birthweight infants. Albeit, we were also able to examine crude risks for term and preterm LBW infants, a distinction that would be important for maternity care in this population. Importantly, we have shown that traditionally established risk factors including nutrition, prenatal care, maternal risk factors, medical conditions, obstetric complications and malaria remain important risk factors for LBW in this urban Zimbabwean population.

Conclusion
Our study is among few studies evaluating risk factors for LBW in Zimbabwe. We have demonstrated similar and different risk factors for subsets of LBW. This step is important for targeted interventions. Programs aimed at improving women's health focused on improving nutritional status of women remain of critical importance for this population. Perhaps most importantly, adequate and focused screening and evaluation of the role of infections, such as malaria, syphilis, and HIV comorbidity, shown to be a risk factor for LBW in previous studies [20-21, 23, 50], is warranted. More studies are imperative for Zimbabwe, as more prevention efforts are needed, if the goal is to ultimately reduce infant mortality and morbidity.