Two Level Logistic Regression Model of Factors Influencing in Early Childbearing and its Consequences on Nutritional Status of Bangladeshi Mothers: Nationally Representative Data

: Background : Early marriage and early pregnancy is a social as well as a medical problem in developing countries, which may have an impact on the health and nutritional status of teenage mothers. Therefore, the objective of this study was to determine the influencing factors of early childbearing (ECB) and its consequences on the nutritional status of Bangladeshi mothers. Methods : Data was extracted from Bangladesh Demographic and Health Survey (BDHS-2014). Women who delivered their first baby before the age of 20 years are considered ECB mothers. Nutritional status was measured by body mass index (BMI). Chi-square test and both univariable and multivariable logistic regressions, and z-proportional test were used in this study. Results : The prevalence of ECB among currently non-pregnant mothers in Bangladesh was 83%. The logistic regression model provided the following six risk factors of ECB: (i) living location (division) (p<0.01), (ii) respondents’ education (p<0.05), (iii) husbands’ education (p<0.05), (iv) household wealth quintiles (p<0.01), (v) respondents’ age at first marriage (p<0.05), and (vi) number of family members (p<0.05). Still, 17.6% of mothers were undernourished in Bangladesh; among them, 18.5% and 13.4% were ECB and non- ECB mothers respectively. ECB mothers had a greater risk to be undernourished than non-ECB mothers [COR=1.26, 95% CI: 1.11-1.43; p<0.01]. Conclusions : In this study, some modifiable factors were found as predictors of ECB in Bangladesh. ECB mothers were more prone to become under-nourished. These findings can be considered to reduce the number of ECB mothers in Bangladesh consequently improve their nutritional status.

Bangladesh [4]. The pregnancy rate among adolescents was higher than adult mothers particularly due to their higher fertility rate, poor reproductive knowledge, inexperience and insufficient use of family planning practice and finally inadequate engagement with care providers [5]. ECB mothers need extra attention to consider their health, babies' growth and development during pregnancy and lactation and long term consequences in developing and underdeveloped countries including Bangladesh [6,7]. Adolescent mothers with poor health status due to nutritional deficiencies are at higher risk for various diseases [8,9]. The pelvic bone of young mothers may be vulnerable which increases the risk of obstructed labour [10]. The risk of pregnancy-related mortality was double before the age of 15 years than the mother's age 15 to 19 years [11,12]. Poverty is the basic cause of nutritional deficiency throughout the world. A study on the African-American community in Chicago reported that adolescent mothers were more likely to be unemployed, lived in poverty and be dependent on social welfare [13]. The cultural, religious, economic and other social factors (i.e. education) also impacts early pregnancy and nutritional deficiency; even some community create pressure to probe fertility on a stillgrowing body [14][15][16][17]. Other studies established that socio-economic factors such as education, wealth index, age at marriage, number of ever-born children, residence, religion, occupation, place and type of delivery are linked to poor health and nutritional status [18][19][20][21].
The Government of Bangladesh (GOB) achieved some health-related millennium development goals (MDGs) such as reducing child mortality and improving maternal health. Now, GOB focuses on Sustainable Development Goals (SDGs) to achieve integrated targets by 2030 [22]. In spite of considerable progress, Bangladesh is still facing trouble with malnutrition that adversely affecting the children and their mother's health.
One study found that the mean age of marriage for Bangladeshi women was 15.69 ± 2.97 years which increased the risk of being early mothers [24]. Studies reported that the prevalence of underweight (BMI<18.5 kg/m 2 ) was higher among poor families in both rural (38.8%) and urban (29.7%) areas [24,25].
Recently, one study was done on early childbearing among Bangladeshi mothers using the BDHS-2011 dataset. Authors considered only early childbearing mothers and did not compare with non-early childbearing mothers to find the effect of socioeconomic and demographic factors on nutritional status [1]. It is needed to identify the risk factors of early childbearing and its effect on their nutritional status and then comparing it with non-early childbearing mothers using nationally representative last dataset. Therefore, the aim of the present study was to determine the factors influencing ECB, in addition, we also looked at the effect of ECB on nutritional status among mothers in Bangladesh.

Research Questions
There are three research questions in this study; these are (i)  Bangladeshi married women ages ranged from 15 to 49 years. This is a nationally representative survey that covered all administrative divisions of Bangladesh [22].

Sampling
BDHS-2014 utilized two stages of stratified cluster sampling for selecting a sample from Bangladeshi married women aged 15-49 years. With this design, the survey selected 18,000 residential households, which were expected to result in completed interviews with about 18,000 ever-married women. Data from a sample of 17,863 married Bangladeshi women were collected. Outliers and missing values and currently pregnant women were excluded from the data. Finally, we analyzed 15,015 data after excluding outliers and missing values and currently pregnant women.

Outcome Variable
Age at childbearing of mothers was considered as the first outcome variable in this study, and it was classified into two classes as (i) early childbearing (ECB) (delivery first baby before age 20; code =1), (ii) non-early childbearing (non-ECB) (delivery first baby after age 20; code=0) mothers. Nutritional status was the second outcome variable of this study which was measured by body mass index (BMI). BMI was categorized into three categories on the basis of cut-off points; (i) undernutrition (BMI<18.5 kg/m 2 ) (code, 1) (ii) healthy (normal weight) (18.5≤BMI<25 kg/m 2 ) (code, 2) and (iii) over nutrition (BMI≥25 kg/m 2 ) (code, 3) [26].

Independent Variables
Various socio-economic and demographic factors were used as independent variables such as divisions (living location), place of residence, parents' education level, number of family members, household wealth index, total ever born children and age at first marriage. All selected independent variables with their categories and codes are described in Table 1.

Statistical Analysis
Frequency distribution was used to determine the prevalence of early childbearing mothers in Bangladesh. Z-proportion test was used to find the difference in each category of nutritional status between ECB and non-ECB mothers. The Chi-square test was used for selecting independent factors for multilevel logistic regression models. In our dataset the observations came from several levels of hierarchy, it was possible to obtain a clustering effect in the outcome variable. If we are having a clustering effect, we first need to remove it otherwise we could not get actual results, and could not apply a single-level statistical model for analyzing this type of data set [26]. Two levels of multiple logistic regression analysis were used to remove the clustering effect for getting the effect of socio-economic and demographic factors on ECB mothers in Bangladesh.
To check the existence of the clustering effect in the outcome variable, we used the median odds ratio (MOR). The following formula is used for calculating MOR value: There is no evidence of clustering effect if MOR=1, but if MOR>1, there is cluster variation of outcome variable [28]. We obtained from our dataset, the MOR=1.372142 > 1, there was a clustering effect of the outcome variable (ECB mothers). We selected a twolevel binary logistic regression model for analyzing our data. Level I and II were considered for individual and clusters (EAs) levels respectively. In this study, the magnitude of the standard error (SE) was used to detect the multicollinearity problem, if the magnitude of the SE is less than 0.5, it was judged that there was no evidence of multicollinearity [28]. The goodness of fit of the model was checked by Hosmer and Lemeshow test. Finally, multinomial logistic regression was used to find the effect of early childbearing on the nutritional status of Bangladeshi mothers. For statistical analysis of data, sampling weight was necessary but unfortunately, proportional allocation technique was not used for selecting samples by BDHS-2014. Sampling weights were calculated for univariate, bivariate and multivariate analysis for this study. Statistical significance was accepted at p<0.05. All statistical analyses were performed using SPSS, IBM (version 23.0) and STATA (version 11).

RESULTS
Out of 15,015 mothers, 83% were early at the childbearing (ECB) stage. The highest number of early  Table 2).
The magnitude value of SE for every independent variable in a multivariable two-level logistic model was less than 0.50, there was no evidence of multicollinearity problems among independent variables. After removing the clustering effect of childbearing age, and controlling the effect of other factors on ECB, the two-level model demonstrated that   Table 3). Table 4. It was noted that 25.5% of mothers in Bangladesh were overnourished, while 17.6% of mothers were undernourished. Z-proportional test demonstrated that significantly a greater number of ECB mothers (18.5%) were undernourished than that of non-ECB mothers (13.4%) (p<0.05), while the number of overnourished mothers among non-ECB (33.9%) was significantly higher than ECB mothers (23.7%) (p<0.01). Also, it was observed that the number of healthy among ECB mothers (57.8 %) was higher than that of non-ECB mothers (52.7%) (p<0.01) (

DISCUSSION
Globally around 17 million adolescent girls give birth each year comprising 11.0% of all births and 90% of them occurs in low and middle-income countries. The most prevalent countries are Bangladesh, Brazil, the Democratic Republic of the Congo, Ethiopia, India, Nigeria and the United States [29]. Early pregnancy and motherhood are still common in Bangladesh. This study investigated the factors which influence ECB among Bangladeshi mothers because ECB creates a poor nutritional status of mothers. Some studies in Bangladesh found the effects of various factors on the nutritional status of ECB mothers in Bangladesh using BDHS 2007 and 2011 data [1,2]. Moreover, they did not compare with non-ECB mothers. In the present study, we investigated the factors influencing ECB among Bangladeshi mothers, using Bangladesh  Demographic Health Survey data 2014 (BDHS 2014) which was the latest dataset. However, our findings showed that the overall prevalence of ECB among Bangladeshi mothers was 83% which were slightly lower than the previous study (86%) from BDHS 2011 [1]. The percentage of ECB among mothers has decreased due to the increase in the female education and household wealth quintile during the last two decades in Bangladesh and these may be the most important reason for the development of ECB [23].

Effect of Socio-Economic and Demographic Factors on ECB
We analyzed the divisional distribution of ECB as regional coverage. The prevalence of ECB mothers was varied among divisions' and the highest number of ECB was observed in Rangpur and lowest in Sylhet division due to differentiation of customs, cultures and economic conditions among the regions [30][31][32]. Therefore, these facts could be potential reasons for ECB. A study in Brazil found a strong association between childbearing at young ages and the mother's subsequent economic and social indicators. Young mothers were less likely to complete high school, less likely to participate in the labour force, more likely to have lower earnings, than women who did not bear children as teenagers [33]. The effect of education may be explained by the likelihood that educated mothers are more able to receive and understand health promotion messages. This study also exhibited that a positive association exist between a mother's education level and ECB. The illiterate are more likely to be ECB than educated mothers. Subsequently, the higher educated mothers were 0.797 times less likely to be ECB than uneducated mothers. This could be depicted as the fact that uneducated mothers were mostly served as housewives with improper knowledge of the effect of early pregnancy. Our results revealed that mothers from rural areas were more likely to commence early pregnancy than urban areas. Similar results were observed among other south Asian mothers in India and Nepal [8,34]. The other necessary socio-economic characteristics were also considered in this study such as the number of the family members, total ever born children, age at first marriage and household wealth quintile [1].

Impact of ECB on Nutritional Status
In this study, we found that Bangladeshi mothers had faced the dual burden of malnutrition; still, more than 17% of mothers were suffering from chronic energy deficiency while 25% of mothers were overnourished. The trends in the number of overnourished mothers in Bangladesh have been increasing but the number of undernourished mothers has been decreasing with increasing the household wealth quintile during the last two decades in Bangladesh [23]. The prevalence of undernutrition among ECB mothers was significantly higher than that of non-ECB mothers. This may indicate the impact of ECB with other mentioned contributing factors. Interestingly, the prevalence of healthy among ECB mothers was also significantly higher than their counterparts may illustrate the optimum utilization of nutrients by young mothers. Alternatively, the prevalence of overnutrition was lower among ECB mothers than non-ECB mothers. This also indicates the imbalance of intake and utilization of nutrients by adult mothers. These results were also supported by the previous study based on Bangladesh Demographic and Health Survey, 2011 [1] and also a cross-sectional study among the nutritional and health risks among adolescent pregnant mothers in Bangladesh [35].

Strength and Limitation of the Study
Perhaps, it is the first time to investigate the associated factors of ECB mothers in Bangladesh and analyze the effect of ECB on their nutritional status compared with non-ECB mothers. We analyzed also the risk factors were identified after removing the clustering effect of ECB using multilevel binary logistic regression analysis. However, there were some limitations of this study. We used secondary crosssectional data derived from BDHS-2014, women's parental related factors were not available such as the household wealth quintile of women's parents before their marriage, the education level of parents which were more important predictors for early marriage as well as ECB. Sampling weight was essential to analyze BDHS-2014 data for ensuring national representation of the survey, but in this survey, we did not consider a proportional allocation technique for selecting samples from different regions such as divisions. We only controlled the effect of some socio-economic, demographic and anthropometric factors for determining the effect of ECB mothers on their nutritional status, but physiological factors which may also impact nutritional status such as level of physical activities, level of energy intake and behavior patterns like dietary habits, smoking habits, weight maintenance, methods of weight-loss etc [36,37] due to unavailability of these factors in BDHS-2014 dataset. However, the mentioned factors would be the potential clues for researchers to conduct more research on ECB mothers in Bangladesh.

CONCLUSIONS
We investigated the factors of ECB among nonpregnant mothers in Bangladesh aged 15-49 years and compared with non-ECB non-pregnant mothers, in addition, the effect of ECB on nutritional status were also observed. Two-level and multinomial logistic regression models were applied to find the effect of socio-economic and demographic factors on ECB and the effect of ECB on nutritional status of mothers respectively. The prevalence of ECB mothers in Bangladesh was 83%. It was found that living location (division), respondents and their husbands' education, household wealth quintile, age at first marriage, number of family members were the predictors of ECB among mothers in Bangladesh. Still, 17.6% of mothers were suffering from chronic energy deficiency; among them, 18.5% and 13.4% were ECB and non-ECB mothers respectively. On the other hand, 25.5% of mothers in Bangladesh were overnourished, among them 23.7% and 33.9% were ECB and non-ECB mothers respectively. ECB mothers had a higher risk to be underweight than non-ECB mothers. The health authorities of the Bangladesh government and nongovernment organizations should special attention to childbearing mothers' health and nutritional status due to their potential contribution to the family and productivity. We hope our findings could be considered to reduce the ECB mothers in Bangladesh as well as to improve their nutritional status.

DECLARATIONS Ethics Approval and Consent to Participate
The 2014 BDHS received ethics approval from the Ministry of Health and Family Welfare, Bangladesh. The 2014 BDHS also received written informed consent for participation in the study was obtained where participants were children (under 16 years old) from their parent or guardian.

Consent for Publication
Not applicable for this study.