Article Text
PDF
Abstract
Objectives This study aimed to explore the effects of short birth spacing (SBS), which is defined as a period of less than 33 months between two successive births, on multiple concurrent forms of child malnutrition (MCFCM) and at least one form of child malnutrition (ALOFCM) using propensity score matching (PSM).
Methods This study used data extracted from the 2017-18 Bangladesh Demographic and Health Survey. PSM with four different distance functions, including logistic regression, classification and regression tree, single hidden layer neural network and random forest, were performed to evaluate the effects of SBS on MCFCM and ALOFCM. We also explored how the effects were modified in different subsamples, including women’s empowerment, education and economic status (women’s 3E index)–constructed based on women’s decision-making autonomy, education level, and wealth index, and age at marriage, and place of residence.
Results The prevalence of SBS was 22.16% among the 4652 complete cases. The matched samples of size 2062 generated by PSM showed higher odds of MCFCM (adjusted OR (AOR)=1.25, 95% CI=1.02 to 1.56, p=0.038) and ALOFCM (AOR=1.20, 95% CI=1.01 to 1.42, p=0.045) for the SBS children compared with their counterparts. In the subsample of women with 3E index≥50% coverage, the SBS children showed higher odds of MCFCM (AOR: 1.43, 95% CI=1.03 to 2.00, p=0.041] and ALOFCM (AOR: 1.33, 95% CI=1.02 to 1.74, p=0.036). Higher odds of MCFCM (AOR=1.27, 95% CI=1.02 to 1.58, p=0.036) and ALOFCM (AOR=1.23, 95% CI=1.02 to 1.51, p=0.032) for SBS children than normal children were also evident for the subsample of mothers married at age≤18 years.
Conclusion SBS was significantly associated with child malnutrition, and the effect was modified by factors such as women’s autonomy and age at marriage.
- Child Abuse
- Epidemiology
Data availability statement
Data are available in a public, open access repository. The Bangladesh Demographic and Health Survey 2017-18 provided the secondary datasets used in this investigation. After requesting the website www.dhsprogram.com, we were able to access data there.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Despite the significant improvement in several public health indices, child malnutrition is still considered a worldwide health concern because of its high prevalence.
WHAT THIS STUDY ADDS
The study revealed a significant association between child malnutrition and factors like women’s autonomy and age at marriage. These findings could aid in reaching United Nations Sustainable Development Goal 3 in Bangladesh.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
The findings suggest the attention of the government and other stakeholders to take necessary interventions or measures to tackle this public health concern.
Introduction
The prevalence of malnutrition among children under 5 years of age is one of the critical public health issues worldwide. Stunting (low height-for-age), underweight (low weight-for-age) and wasting (low weight-for-height) are the recommended three factors to measure malnutrition.1 2 Recent studies indicate that about 45% of children under 5 years of age die due to malnutrition. Moreover, 144 million children under 5 years are affected by stunting, and around 47 million children in the same age group reported wasted.3 Bangladesh, among the Asian countries, has had one of the highest prevalence of malnutrition over the past couple of decades. One-fifth of children under 5 years of age are suffering from multiple concurrent forms of child malnutrition (MCFCM),4 while around one-third of children under 5 years are stunted and/or underweight and one-tenth are wasted in Bangladesh.5–7
According to the World Health Organization (WHO), the recommended interval between a live birth to the next conception or a live birth to the next live birth is at least 24 months or 33 months, respectively.8 The intervals shorter than these are considered short birth spacing (SBS), also known as short birth interval.9 About 25% of live births globally are attributed to SBS.10 In reference to this global percentage, more than a million live births are born in a short interval, which is 26% of the total live births per year in Bangladesh.9
Evidence from reviews and empirical studies indicates that SBS is strongly correlated with child malnutrition.4 The relationship between SBS and child malnutrition is complex. In addition to the complex set of factors that may cause child malnutrition, the relationship depends on the context of the study.10 Several studies revealed a wide range of factors that may influence the abovementioned relationship, such as maternal age, maternal employment, place of residence, child’s sex, division, father’s education, maternal age at marriage, exposure to media, intimate partner violence, terminated pregnancy, birth order, body mass index (BMI) and women’s empowerment, education and economic status (women’s 3E index), which was determined by taking into account the women’s decision-making autonomy, education level, and wealth index.2–4 11
Researchers traditionally implement multivariable regression models to study the effect of birth spacing (or interpregnancy interval) on birth outcomes.12 Chowdhury et al4 used logistic regression to differentiate the risk factors associated with single and multiple concurrent forms of undernutrition among children under 5 in a nationally representative cross-sectional study in Bangladesh.4 Chungkham et al11 explored the relationship between birth interval and undernutrition of children under 5 years of age in India using bivariate and logistic regression model. However, these study designs could be subject to bias from a variety of socioeconomic factors. To reduce the odds of such bias, propensity score matching (PSM) may be an alternative method in studying the effect of birth spacing on child malnutrition.12 In a retrospective cohort study, Congdon et al13 uses one-to-one exact PSM to demonstrate the association between short interpregnancy intervals and adverse birth outcomes. Additionally, they used a traditional logistic regression to compare their results arising from the PSM. Goyal et al14 and Howard et al15 also used PSM in their studies of birth spacing.
PSM is a quasi-experimental method that constructs an artificial controlled group and generates exposed and unexposed groups using statistical matching techniques. Using these experimental subgroups, PSM attempts to estimate the effect of a treatment related to the outcome of interest.16 The PSM explores the actual effect of treatment variable on outcome after balancing the covariates for treated (exposed) and untreated (unexposed) groups. This study used the PSM method to evaluate the effect of SBS on child malnutrition in Bangladesh.
Methods
Subjects
The study subjects were extracted from the nationally representative Bangladesh Demographic Health Survey (BDHS) 2017-18 data, which is publicly available.17 It was a cross-sectional survey and followed a two-stage stratified cluster random sampling to select households to represent the non-institutionalised Bangladesh population. In the first stage, 672 enumeration areas (EAs), each with an average of about 120 households, were chosen from 293 579 EAs using a probability proportional to EAs size. In the second stage, 30 households from each EA that constituted 20 160 households were selected with an equal probability systematic sampling. Finally, a total of 19 457 households were interviewed. The survey collected data on demographics, marriage and sexual activity, reproduction, contraception, pregnancy and postnatal care, maternal and child health, including breastfeeding practices, nutritional status, immunisation, women’s empowerment and non-communicable diseases. Another subsample consisting of one-fourth of households from each EA was chosen for biomarker samples. BDHS was approved by the Institutional Review Board of ICF International, Rockville, Maryland, USA, and Bangladesh Medical Research Council (BMRC), Dhaka, Bangladesh. For additional detail, including the ethical approval number, we refer to the BDHS report.17
From the interviewed households, 8759 children aged less than 5 years were identified. As birth spacing is not applicable to the first child of each mother, we excluded them, which left 5348 children. After excluding twin children, missing cases, out of plausible height limits and flagged cases (information for these individuals was inconsistent and distinct from the bulk of survey responses, which were considered clean and reliable), the final sample had a total of 4652 children. Details of children’s exclusion are presented in the flowchart in online supplemental figure S1. It is noted that the work has been documented according to the Strengthening the Reporting of Observational Studies in Epidemiology checklist, which can be found as online supplemental file.
Supplemental material
Supplemental material
Outcome variable
There were two outcome variables in this study: MCFCM and at least one form of child malnutrition (ALOFCM). The BDHS collected information on height-for-age, weight-for-height and weight-for-age. A child is stunted (yes and no) if height-for-age is 2 SD below the median (−2 SD) value of the WHO reference population.18 The other two indicators, wasted (yes and no) and underweight (yes and no), used the same definition based on weight-for-height and weight-for-age measurements, respectively.18 Finally, the outcome variable MCFCM was defined as yes (1) if a child had at least two positive indicators and no (0) otherwise.4 On the other hand, the outcome variable ALOFCM is defined as yes (1) if a child had at least one of the abovementioned malnutrition issues and no (0) otherwise.4
Treated variable
Birth spacing refers to the time between two successive live births or birth to the next successive pregnancy.8 Following the guidelines of the WHO, this study defined SBS as a dichotomous variable with ‘yes’ if the interval between two successive live births is less than 33 months and ‘no’ if the interval is at or more than 33 months and used it as the treated variable.8
Confounders
Several variables were used to adjust potential confounding effects.4 19–23 These include maternal age (<20 years, 20–34 years and>34 years), maternal employment (no and yes), place of residence (urban and rural), child’s sex (male, female), administrative division (Barisal, Chittagong, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur and Sylhet), father’s education (no education, primary, secondary and higher), maternal age at marriage (<15 years, 15–18 years, >18 years), exposed to media (no and yes), intimate partner violence (no and yes), terminated pregnancy (no and yes), birth order (2, 3 and ≥4), women’s 3E index (coverage<50%, 50%–75% and >75%) and BMI (underweight, normal, overweight and obese).
Women’s 3E index combined women’s three statuses: empowerment, education and economic status. The women empowerment variable was generated based on five questions asked to every woman regarding their participation in the household’s decision-making. These included decisions on (1) spending own earnings, (2) spending partner’s earnings, (3) own healthcare, (4) large household purchases and (5) visits to family or relatives. A mean empowerment score for every woman was then calculated by converting each indicator variable as dichotomous, with “1” if the decision had been taken by women alone or jointly with husbands or partners and “0” for otherwise. Women’s education was dichotomised by combining no and primary education as uneducated with label “0” and secondary and higher as educated with label “1”. For economic status, this study used the BDHS wealth score generated by principal component analysis.17 The women were divided into two groups, namely, poor and rich, based on the median wealth score. Finally, this study used the same definition by Tuz-Zahura et al19 that followed the WHO approach24 to get women’s 3E index, an unweighted mean of three statuses. The mathematical formula of the index was as follows:
We categorised the index into three categories: below 50%, 50%–75% and above 75% coverage. Exposure to media was calculated based on whether mothers watch TV, listen to the radio and read newspapers/magazines. A mother was defined as exposed to media if she had access to at least one of the abovementioned media. A woman was exposed to intimate partner violence if she had justified wife beating for one of the following reasons: (1) wife goes out without telling her husband, (2) wife neglects the children, (3) wife argues with husband, (4) wife refuses to have sex with husband and (5) wife burns the food. Finally, BMI was classified into four categories suggested for Asian people: underweight (<18.5 kg/m2), normal (18.5≤BMI<23 kg/m2), overweight (23≤BMI<27.5 kg/m2) and obese (≥27.5 kg/m2).25 All the variables were measured at the time of interview. Birth order refers to the order of a child among his/her siblings. Terminated pregnancy variable was measured as ‘yes’ if a mother experienced terminated pregnancy within 5 years before the date of interview. However, these variables were not recorded with exact time record as it was cross-sectional study, which was a potential limitation of this study.
Variables to test heterogeneous effects
Although the primary objective of this study was to find the effect of SBS on the outcomes, it is also necessary to investigate the heterogeneous effects that may result from variation of treatment effect over different subsamples.26 We considered three variables, including women’s 3E index, mother’s age at marriage and place of residence, to test whether the effect of the treatment on the outcome varies in different subsamples. Women’s 3E index was further converted to the dichotomous variable with label “0” if it was <50% coverage and “1” if ≥50% coverage. Mother’s age at marriage (≤18 years and ˃18 years) and place of residence (urban and rural) were also defined as dichotomous variables for testing heterogeneous effects.
Statistical analysis
The sample characteristics were summarised by the frequency and per cent distributions. χ2 test was performed to test the association between outcomes and treated variable (SBS) in the unmatched data. The one-to-one PSM approach was used to find the matching treatment and control groups to estimate the effect of exposure on outcome variables. Logistic regression is the most commonly used distance function to estimate propensity score (PS), which uses a binary variable as the dependent variable and a set of covariates as independent variables. Some other distance functions are readily available from the R-package ‘MatchIt’.27 The latest version of ‘MatchIt’ incorporated some machine learning algorithms to estimate PS.27 Among them, we considered four familiar functions: logistic regression without calliper, classification and regression tree, single hidden layer neural network (nnet) and random forest for covariate balance. The nearest neighbourhood matching (NNM) is the most commonly used among several matching methods. Another popular method, optimal matching (OM), to optimise the distance between treated and controlled groups was suggested by Zhao et al.28 However, in the case of balancing covariates, NNM outperformed OM.28 Therefore, we considered NNM for our analysis with four distance functions. However, before conducting the matching, we had to test if the SBS was significantly associated with other covariates. To check this, we conducted a χ2 test of association. Another way to check if the matching is required to find the treatment effect on the outcome of interest is to check the standardised mean difference (SMD) between treatment and control groups.28 29 The SMD greater than 0.1 suggests the implication of matching to balance the covariates over the treated and untreated groups.28 29 A good matching is expected to yield these SMDs less than 0.1. Another way to check the quality of regression-based matching is by comparing the pseudo R2 values of the pre-matching and post-matching models and testing the likelihood ratio test of the fitted models. The lower pseudo R2 value and insignificant goodness of fit test in the matching model compared with the pre-matching model provides evidence of good matching.26 29 Finally, another way to check the matching quality is to check the distribution of PSs between treated and control groups. The overlapping distributions of PSs indicate that mothers who did and did not have SBS had overlapping characteristics. After matching data, we calculated the average treatment effect on the outcome variable. The average treatment effect on treated is the difference in the mean prevalence of the child malnutrition between children with SBS and their counterparts. A propensity weight-based logistic regression, linear probability and weighted least square (WLS) estimate models were used to find the treatment effects.28 30 31 We have calculated both crude (without covariates) and adjusted (with covariates) estimates. We calculated heterogeneous effects considering women’s 3E index (<50% vs ≥50%), maternal age (<20 years vs ≥20 years) and place of residence (urban vs rural) using only binary logistic regression.
Results
Background characteristics of the respondents
A total of 4652 children were extracted from the interviewed 8759 households (online supplemental figure S1). Online supplemental table S1 presents the percentage distribution of these respondents. About one-third of these respondents live in urban areas. About 21% of the children suffered from MCFCM, whereas about 41% reported ALOFCM. Approximately 33% were stunted, 8% were wasted and 23% were underweight. The distributions of a few other factors were also constructed, including exposure to media (60%), intimate partner violence (19%), terminated pregnancy (21%) and parents’ employment status (46%). About 86% of women were married under 18 years of age. Finally, around 22% of the live births had SBS, close to the global rate (25%).
PS estimation
Before conducting PSM, we conducted χ2 test to check the association between outcome variables and treated (short birth spaced) and untreated (not short birth spaced) groups. We found significant differences for both outcomes (p<0.001) (table 1). From the complete data (pre-match), we found maternal age, maternal employment, fathers’ education, administrative division, maternal age at marriage, women’s 3E index, mother’s BMI and exposure to the media were significantly associated with short birth interval (online supplemental table S2). The SMDs for several categories were also greater than 0.1, which implies SBS differed over these confounders. After matching, the study ended with a matched sample of 2062.
χ2 test of association between outcomes and treated variable in the unmatched data
Covariate balance
Among all the distance functions, the logistic classifier outperformed others as all the SMDs within 0.1. It also indicated that the matching balanced the covariates (figure 1). The pseudo R2 values and likelihood ratio statistics with p values from two logistic regression models, one for unmatched data and another for matched data, were reported in online supplemental table S3. We observed that the R2 value decreased from 0.07 (for unmatched) to 0.003 for matched data. The likelihood ratio test suggested that the model for unmatched data had significant goodness of fit (p<0.0001), whereas the model for matched data did not have significant goodness of fit (p=0.6344). Both pseudo R2 values and likelihood ratio test results showed that the matching quality was good, and it reduced the observed selection bias significantly.
Standardised mean differences for unmatched and matched sample, and the distribution of propensity score for treatment and control groups obtained by using logistic regression, classification and regression tree (CART), neural network and random forest.
Finally, we observed the distribution of PS for treated and control groups. The overlapping distributions of PSs indicate that mothers who did and did not have SBS had overlapping characteristics. We observed from figure 1 that density plots for the treatment and control groups were almost fully overlapped, which also provided evidence for overlapping characteristics.
Average treatment effects on treated
The results of average treatment effects based on three models with and without considering covariates are reported in table 2. All three models suggested that SBS significantly associated with child malnutrition. Linear probability model with covariates suggested children with SBS had a 3.8 and 4.2 percentage points more chance of having MCFCM and ALOFCM, respectively, than their counterfactual part. The logistic regression model suggested that odds of MCFCM was 25% (adjusted OR (AOR)=1.25, 95% CI=1.02 to 1.56, p=0.038) higher, and the odds of ALOFCM was 20% (AOR=1.20, 95% CI=1.01 to 1.42, p=0.045) higher for the SBP children compared with their counterparts after adjusting other covariates. The WLS models with covariates also suggested that SBS was significantly positively (for MCFCM: beta=0.029, p=0.048, and for ALOFCM: beta=0.029, p=0.016) associated with child malnutrition.
Average treatment effect on treated using linear probability model (LPM), logit model and weighted least square (WLS) model
Results obtained for heterogeneous effects summarised in table 3. For the women with ≥50% 3E index, SBS had significantly higher odds of MCFCM (AOR: 1.43, 95% CI: 1.03 to 2.00) and ALOFCM (AOR: 1.33, 95% CI: 1.02 to 1.74). In the subgroup of mothers who got married before ≤18 years, the odds of having MCFCM was 27% (p=0.036) higher and the odds of having ALOFCM was 23% (p=0.032) higher for the SBS compared with its counterpart. To test if the effect sizes between subgroups differ significantly, a logistic regression analysis with interaction between SBS and women’s 3E index and between SBS and age at marriage was conducted. None of the interactions were found significant (online supplemental table S4). Moreover, we did not find any significant difference for the place of residence, although a slightly higher treatment effect was observed for urban mothers than the rural mothers.
Heterogeneous effects by women 3E index to maternal age and place of residence
Discussion
This study analysed the impact of SBS on multiple concurrent and ALOFCM after adjusting for a complex set of confounders (socioeconomic, demographic and health factors) using PSM. Effects of treatment on the outcome variable are susceptible to confounders. Thus, in reducing or eliminating these confounding effects, the PSM method was used in our study to identify the effects of SBS on child malnutrition.
A relatively new notion of malnutrition known as MCFCM was first introduced in the 2014 Global Nutrition Report.32 Though MCFCM has not yet been adequately investigated worldwide, its incidence, trends and causes have been studied in a few geographic areas in Asia, Africa and South America. MCFCM refers to a child with more than one nutritional issue present simultaneously, such as stunting, wasting, underweight or all. Based on the result of this study, the multiple concurrent forms and at least one form of malnutrition among Bangladeshi children under 5 years were roughly 20.85% and 40.69%, respectively. That means at least one child in five suffers from multiple concurrent or at least one form of malnutrition. The prevalence of MCFCM among children in emerging nations in Africa and South America, like Ethiopia (26%), Malawi (12%) and Argentina (2%), is comparatively lower than in Bangladesh.33–35 A high prevalence of MCFCM is a problem in Asian nations like Yemen and India (MCFCM: 48% and 39%, respectively, whereas a single form of malnutrition: 21% and 24%, respectively).36 37 The high prevalence of child malnutrition demands that the concerned authorities and other stakeholders of Bangladesh should work jointly to minimise it without further delay.38
This study showed that SBS is significantly associated with an increased risk of MCFCM and ALOFCM. This association may be driven by the mother’s nutrient reserves depletion caused by the SBS that eventually leads to an increased risk of intrauterine growth retardation, adversely affecting infant nutrient stores at birth and nutrient delivery via breast.39–43 Due to SBS, caring for a new child also reduces the devotion time of a mother to the other children, as the subsequent pregnancy may alter care practices that affect the new child’s health.44 Another study supports that longer birth intervals cause a lower risk of child malnutrition. It is also said that when the birth interval is more than 36 months, the reduction in child malnutrition ranges from approximately 10% to 50%.10 Afeworki et al39 demonstrated that the longer length of the preceding birth interval has a modest positive impact on haemoglobin levels in African preschool girls. In an interaction model, a 1-month increase in the birth interval is linked to an average haemoglobin level rise of 0.025 g/L in girls (p=0.001). However, in the case of boys, the effect was not found to be statistically significant.39 A deficiency of haemoglobin in the bloodstream, referred to as anaemia, has become a significant public health issue in both developed and developing countries. A recent report indicates that over 41% of children aged under 5 and one-thirds of women between 15 and 49 years across the globe suffer from anaemia.45 It is worth emphasising that the prevalence of anaemia is more common in stunted children compared with normal children. Rahman et al46 reported that the association between anaemia and stunting in children remains significant even after adjusting the influence of confounding factors. To illustrate, even after adjusting for the impact of gender, children with stunted growth exhibited a 39% higher likelihood of developing anaemia compared with their normally growing counterparts. Therefore, based on the current research and the information available in the literature, it is evident that SBS is associated with malnutritional consequences in children below the age of 5 years.
Moreover, this study examined the effects of SBS on the MCFCM and ALOFCM over the levels of women’s 3E index. For mothers with more than 50% coverage of 3E index, SBS were more likely to have MCFCM and ALOFCM. Most educated and empowered women are engaged in economic activities, which has the apparent consequence that they may have less time to care for their children. In addition, employed mothers are less likely to breastfeed their children than unemployed mothers.47 However, the results of this study about women’s 3E index contradicted findings from some previous studies in Pakistan. One study showed employment status might increase the possibility of getting adequate and better-quality food as it would increase the family income.48 Another study reported that child malnutrition is negatively associated with mothers' educational levels since they are more knowledgeable about nourishing foods and how to raise their children.49 However, this variation might happen as we controlled the confounding effects more sophisticatedly using PSM.
The effect of mother’s age at marriage on the relationship between SBS and child malnutrition was also evaluated in this study. This study discovered a substantial impact of SBS on child malnutrition when mothers married before 18 years of age. Some other studies also support this finding.50 Age at marriage matters since young pregnant mothers need higher nutrition for growing fetuses than mature pregnant women.51 52 Moreover, behavioural immaturity makes young mothers less sensitive to the needs of their children. Mothers marriage at an early age also tends to have less education, have a poorer socioeconomic level, have psychological stress due to lack of resources and may breastfeed for a shorter period than adult mothers.53 54 These factors stunt their kids’ growth and development, leading to multiple concurrent or ALOFCM.
This study showed that the place of residence (urban and rural) did not significantly affect malnutrition due to SBS. The findings from Das et al55 and Saha et al56 are aligned with this finding. In contrast, some studies showed the effects of the place of residence in association between SBS and malnutrition. They found that children in rural areas experience severe malnutrition compared with urban areas.57–59 This difference with our results might result from better dealing with bias and confounders by the PSM. Furthermore, the availability of healthcare and other facilities in rural areas were immensely improved over time, like urban areas. Thus, no difference was observed in the association between SBS and child malnutrition over places of residence.60
Besides several strengths, this study had two limitations. Selection bias may occur in the sample. The effect of treatment on the outcome could be varied based on some covariates. For instance, the weight of a child at birth could be a potential predictor of child malnutrition, but in BDHS 2017-18, most of the children’s weight at birth was not reported. Therefore, we could not consider this variable in our study as it greatly reduced the matching sample. Besides, this study is subject to recall bias as it is a retrospective study. Furthermore, some mothers had repeated measures (two or three birth intervals), therefore, it is expected that children under the same mother will be correlated. But we did not consider this correlation in our analysis, which is also a potential limitation of this study.
Conclusion
This study revealed that SBS was significantly associated with child malnutrition, and the effect was modified by factors such as women’s autonomy and age at marriage. These findings could help to achieve the United Nations Sustainable Development Goal 3 (SDG3) in Bangladesh. Notably, one aim of SDG3 is to curtail neonatal mortality below 12 per 1000 live births and under-5 mortality below 25 per 1000 live births. The current findings suggest the attention of responsible authorities and other stakeholders to take necessary interventions and policy to tackle this public health concern.
Data availability statement
Data are available in a public, open access repository. The Bangladesh Demographic and Health Survey 2017-18 provided the secondary datasets used in this investigation. After requesting the website www.dhsprogram.com, we were able to access data there.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants. The purpose of the study was explained to every participant, and data were only collected with their consent. The National Institute of Population Research and Training of the Ministry of Health and Family Welfare, Bangladesh, provided ethical approval for conducting the survey. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The authors sincerely thank the Demographic Health Survey Program for allowing them access to the dataset for conducting the study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors FA conceptualised and designed the study. FA, MNH, MFH and MTFK cured and analysed the raw data. FA, MTFK and MJH interpreted the analysed data. FA, MMR and MJH critically reviewed and edited the final version of the manuscript. All authors contributed to the manuscript drafting and read and approved the final version of the manuscript for submission of the manuscript. The corresponding authors (FA and MJH) are guarantors of the contents of the published article.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.