Factors associated with low birthweight in North Shewa zone, Central Ethiopia: case-control study

The study was carried out in three secondary-care public hospitals – Fiche, Kuyu and Dera –found in North Shewa zone, Oromia region, central Ethiopia. As of 2016, North Shewa zone had an estimated population size of 1.5 million. The zone is administratively divided in to 13 districts and has the aforementioned 3 functional hospitals, 62 health centers, 268 health posts. In 2016, the total health facility deliveries in the zone were 38,131.

Unmatched case-control study with controls-to-case ratio of 4:1 was conducted from January 01 to March 30, 2017.

Singleton live births in the three hospitals during the study period, irrespective of the duration of pregnancy and mode of delivery, were considered eligible for the study. Birthweight of every child was measured and newborns who weigh less than 2.5 kg were taken as cases; whereas, a similar group of children with a birthweight of 2.5 to 4.0 Kg were categorized as a controls. Multiple births, macrosomic babies (birthweight greater than 4.0 kg), mothers or newborns in critical medical conditions and babies weighed more than an hour after birth were excluded.

Optimal sample size was determined via the online OpenEpi statistical program [24]. The computation was made using double population proportion formula assuming 95% confidence level, 80% power, control-to-case quotient of 4, and odds ratio (OR) of 2 to be detected as significant. The calculation was separately made for four potential predictors (maternal age (>/< 18 years), place of residence, birth interval (>/< 2 years) and maternal MUAC (>/< 22 cm)) of LBW and the maximum was taken as the ultimate sample size of the study. The expected proportions of controls exposed for the aformentioned factors were extracted from a study conducted in Southeastern Ethiopia [25]. Ultimately the sample size 94 cases and 376 controls was determined. Between Jan to Mar 2017, cases and controls were recruited on an ongoing basis until the required sample size was fulfilled for both groups.

The data were collected by interviewing the mothers, reviewing medical records and measuring the anthropometry of the mothers and the newborns.

Six trained midwives working in the delivery wards of the three hospitals collect the data using structured and pretested questionnaire prepared in Afan Oromo language. Eligible mothers were interviewed face to face within 24 h after delivery. Socio-demographic and economic information was assessed using standard questions extracted from the DHS questionnaire [26].

The medical records of the mothers were reviewed and relevant information including last-normal menstrual period and ultrasound dating of pregnancy were extracted to the questionnaire.

The frequency of consumption of eleven major food groups during the pregnancy was measured using a Food Frequency Questionnaire (FFQ) based on the mothers’ recall. On the other hand, the level of household food insecurity was assessed using the Household Food Insecurity Access Scale (HFIAS) of the Food and Nutrition Technical Assistance (FANTA) project. The scale categorized the subjects into four ordinal groups – secure; mild, moderate and severe insecurity [27].

The weight of the newborns was measured within the first hour of birth using a calibrated Seca scale and rounded to the nearest 100 g. MUAC of mother was measured to the nearest 0.1 cm using MUAC tape. Anthropometric measurements were taken in duplicates by an observer and ultimately the average of the duplicates was registered.

The independent variables of the study include socio-demographic factors (maternal age, education, occupation, wealth index, residence, marital status, religion, ethnicity), reproductive factors (gestational age, prior history of LBW, parity, birth-to-birth interval, utilization of antenatal care (ANC), reported illness during pregnancy), nutritional factors (maternal MUAC, household food security status, exposure to nutrition counseling during the pregnancy, frequency of consumption of major food groups, restriction of diet during pregnancy due to food taboo, history prenatal iron supplementation), work load during pregnancy and infant’s sex. The dependent variable was birthweight status dichotomized into LBW or normal birthwieght.

The collected data were checked for completeness, coded and entered into Epi info version-3.5, and then exported to the Statistical Package for Social Sciences (SPSS), version 20 for analysis.

Wealth index was computed as a composite indicator of living standard based on ownership of selected household assets, size of agricultural land, number of livestock owned, materials used for housing construction, and ownership of improved water and sanitation facilities. The analysis was made using the Principal Component Analysis (PCA). The generated principal component was divided into three wealth classes.

The socio-demographic and other background profiles of the cases and controls were compared using chi-square test. Prior to analysis the assumptions of chi-square test were checked. When smaller expected frequencies were encountered, re-categorization of variables or merger of the levels was made.

Factors associated with LBW were identified using bivariable and multivariable logistic regression models. Independent variables that demonstrated near to statistically significant association (p-value less than 0.25) with the outcome variable in the bivariable models, were considered as candidate variables for the multivariable logistic regression models. In order to reduce over adjustment bias, direct and indirect predictors of LBW were fitted separately into two multivariable models [3]. In the ultimate multivariable models the level of multicolinearity was evaluated using variance inflation factor and found within a tolerable range. The goodness-of-fit assessed using Hosmer–Lemeshow test.

The study was approved by the Institutional Review Board (IRB) of College of Medicine and Health Sciences, Hawassa University. Data were collected after taking informed consent from the mothers.

A total of 470 mother-newborn dyads comprising 94 cases and 376 controls were included in the study. The mean (±SD) age of the mothers of the cases was 27.4 (±6.6) years and that of mothers of the controls was 29.0(±6.4) years. About two-fifth (41.5%) of the mothers of the cases and about a quarter (27.7%) of that of the controls were in the age group 15–24 years and the difference was significant (p = 0.027). Higher proportion of the mothers of the cases (53.2%) had no formal education than that of the controls (42.0%) (p = 0.002). The two groups were significantly different in terms of ethnicity, religious affiliation and marital status (p < 0.05). However, there were no significant variations based on  household wealth index, place of residence and maternal employment status (p > 0.05) (Table 1).

Regarding the reproductive profile, more cases (16.0%) were preterm births than the controls (5.9%) (p = 0.001). Among mothers who had at least two births, short birth-to-birth interval (less than 2 years) was observed in 52.8% of the cases and 21.2% of the controls (p < 0.001). Among mothers of the cases, 25.5% had no ANC visits as compared to 13.8% in that of the controls (p < 0.001). Significant differences were also observed between the groups in the number of ANC consultations (p < 0.001). However, the mothers of cases and controls were not significantly different in terms of parity, presence of pregnancy related complications (p = 0.403) and infant’s sex (p = 0.781) (Table 2).

Nearly half (44.9%) of mothers of the cases and 58.5% of that of the controls received no nutrition counseling during the pregnancy (p = 0.019). In contrast to 42.3% of the mothers of the control babies, 28.0% of mothers of the cases had taken prenatal iron supplements (p = 0.009). About half (52.1%) of the households of the controls were food secure but the corresponding figure was significantly lower (26.6%) in the cases group (p < 0.001). The two groups were not significantly different based on reported practice of food taboo during pregnancy and prevalence of thinness (MUAC less than 210 mm) (Table 3).

The overall dietary intake of the mothers was assessed using a FFQ. However, the cases and controls were not significantly different in the frequency of consumption of 11 food groups. The food groups which considered in the study were: cereals, roots and tubers, legumes, milk and milk products, flesh foods, eggs, vitamin A rich fruits and vegetables, other fruits, other vegetables, sweets and condiments.

As it was stated earlier, risk factors of LBW were identified by fitting two different multivariable regression models for the direct and indirect predictors of birthweight. The variables were classified into the two blocks according to the Kramer’s framework [3].

For the indirect model, six socio-demographic variables (maternal age, place of residence, maternal education, occupation of the mother, household wealth index, religion, ethnicity and marital status) were considered. Among these, based on bivariable logistic regression analyses, six variables presented in Table 5 had p-value less than 0.25 hence fitted into the multivarible model.

The ultimate analysis showed the odds of LBW were increased by two fold in the mothers with no formal education as compared to their counterparts. Those who were not married had three times elevated odds of LBW as compared to married mothers. Protestant Christians had three times increased odds of LBW than Orthodox Christians (Table 4).

For the direct model, ten variables (namely gestational age, parity, number of ANC visits, presence of pregnancy related complications, infant’s sex, restriction of diet due to food taboos, dietary counseling in the index pregnancy, use of iron supplements in the pregnancy, MUAC of the mother, and household food security status) were considered. Among them, based on bivariable analysis, five variables presented in Table 5 had p-value less than 0.25 and hence subjected to the multivariable analysis.

The ultimate model showed that mothers who have no history of nutrition counseling during the pregnancy had more than three folds increased odds of delivering LBW babies than those who got counseling. Mothers from food insecure households had about four times higher odds of LBW as compared to food secure mothers. Taking mothers who had the recommended four or more ANC visits as reference, those who were not booked had three times increased odds of giving to LBW baby (Table 5).