Exploring Prevalence and Determinants of Early Marriage Through a Cross‐Sectional Study: Insights From Bangladesh Demographic and Health Survey 2022

ABSTRACT

Background and Aims

Getting married young is a common system in developing nations, and it has detrimental effects on the health of the women involved as well as the unborn children. This study aims to identify the key factors affecting early marriage among Bangladeshi women using BDHS‐2022 data, focusing on region, education, residence type, employment status, and religion by considering the best‐fit model among Cox proportional hazard (Cox‐PH) and various accelerated failure time (AFT) survival models.

Methods

This study used survival analysis techniques, such as Cox proportional hazard model and accelerated failure time (AFT) models, using data from the Bangladesh Demographic and Health Survey 2022 (BDHS‐2022). The Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) were used to choose the models, and based on these criteria, the log logistic AFT model was identified as the best‐fitting model.

Results

The women who had an early marriage have a much higher frequency (65.9%) than those who did not have early marriage (34.1%). The log logistic AFT model identified several significant factors influencing early marriage. The hazard ratio of getting early marriage of respondents with primary (HR: 1.017, p < 0.05) and secondary education (HR: 1.024, p < 0.01) had a higher risk compared to those with no education, while the same results are shown in the respondent's husband's education. Respondent's currently working had a lower risk of getting married young (HR: 0.988, p < 0.01). Additionally, rural residence was associated with a higher risk (HR: 1.019, p < 0.001) than urban residence. These results emphasize that the educational level of both male and female, living area, and employment status are influencing factors that affect early marriage.

Conclusion

Early marriage in Bangladesh is influenced by socioeconomic and demographic factors, emphasizing the need for targeted interventions focusing on education, employment, and regional disparities to delay marriage and improve women's social outcomes.

Abbreviations

AFT

accelerated failure time

AIC

Akaike information criterion

BIC

Bayesian information criterion

HR

hazard ratio

PH

parametric hazard

1. Introduction

It is well known that marrying or living with children under the age of 18 is a harmful, discriminatory practice that violates human rights and discriminates against women internationally [1, 2]. Over the past few decades, women's sovereignty and well‐being have received more attention [3, 4]. Child marriage is far more common among females, especially in South Asia and sub‐Saharan Africa, even if it does occur among boys [5, 6]. In many South Asian countries, where a young girl may be perceived as an economic burden, early marriage is a tactic used by low‐income families to cope with financial instability, particularly when school fees or transportation costs are high [6, 7]. Bangladesh has the largest number of marriages involving girls under the age of 15, even though the legal marriage age for men and women is 21 and 18, respectively. Many female marriages occur before the legal age [8]. Early marriage has detrimental repercussions on girls that go beyond their academic achievement [9, 10]. There is a higher likelihood of psychological disadvantage among young married girls, such as depression and low self‐esteem, as well as sexual assault, which increases their risk of HIV and other STDs [11, 12, 13]. Young married ladies also begin having children shortly after marriage, which increases their risk of low birth weight, high infant mortality, pregnancy problems, and delivery deaths [14, 15]. Long‐term violence affects young married girls [16].

Girls' human rights are violated by early marriage because it denies them freedom, opportunities for personal growth, and other rights. Due to international human rights agreements and state constitutions that establish the legal marriage age at 18, many nations, including Ghana, forbid child marriage [17]. Compared to boys and young men, girls and young women, particularly those from the poorest homes, have had less access to education. Gender differences in access and primary and lower secondary completion are particularly noticeable in Sub‐Saharan Africa (SSA) and South West Asia (SWA) [18, 19].

Other negative effects of early marriage for young women include fetal mortality, low birth weight kids, preterm delivery, and physical, sexual, and emotional abuse during the marriage [20, 21]. According to reports, pregnancy‐related deaths are twice as common among girls aged 15–19 years [22, 23, 24]. Consequently, child marriage has detrimental effects on economic survival, human rights, psychological health, education, and achievement of girls, as well as the rate at which the respondent and her child die [25].

In Bangladesh, early marriage tends to overstate the health of teenage girls [26, 27]. Adolescent family transitions are significantly influenced by individual factors, including women's educational attainment, household economic position, living in rural areas, and being in a household headed by a woman [28, 29]. It increases women's chances of becoming pregnant too young, increases divorce rates, and increases the risk of maternal and newborn mortality [30, 31]. Early marriage has been linked to a number of negative social and physical outcomes for young women and their children. These include an increased likelihood of early childbearing, low economic status, stopping school, an increased danger of sexually transmitted infections, an increased rate of divorce, pregnancy complications, and a greater chance of death for the mother and her offspring [32].

Prior research on early marriage in Bangladesh has mostly employed regression models and descriptive analysis [33, 34, 35, 36, 37]. Furthermore, to the best of my knowledge, no survival analysis has yet been conducted using the BDHS 2022 data set. By combining Cox‐PH and AFT survival models with BDHS 2022 data, this study fills these gaps and offers a more thorough knowledge of the factors that influence early marriage.

This study aims to identify the key factors affecting early marriage among Bangladeshi women, focusing on region, education, residence type, employment status, and religion by considering the best‐fit model among Cox PH and various parametric PH and AFT survival models.

2. Methods and Materials

2.1. Study Design

This study examines when to get married young among Bangladeshi women, utilizing data from the Bangladesh Demographic and Health Survey 2022 (BDHS‐2022) using a retrospective cohort technique. Periodically, the BDHS is a nationally representative survey that gathers socioeconomic, health, and demographic information. To ensure representativeness across divisions, urban and rural areas, and various population groups, the survey employs a two‐stage stratified sample design.

2.2. Sample Design

In June 2022–December 2022, the private research group Mitra and Associates conducted the Bangladesh Demographic and Health Survey 2022 (BDHS‐2022). Mitra and Associates stratified the BDHS 2022 sample selection in two stages. Initial stratum selection was independent. Second, 45 households were carefully selected from updated cluster household lists. They surveyed all ever‐married women aged 15–49 who lived in or stayed overnight in the selected houses. They selected 438 rural and 237 urban enumeration units in the first stage of a two‐stage stratified sampling process, using a probability proportional to size! We first listed all households in each enumeration unit. In stage two, systematic sampling picked 30 households from each enumeration subdivision. Researchers interviewed about 30,078 households in 675 groupings (237 urban and 438 rural) from eight Bangladeshi divisions. BDHS data was carried out by the National Institute of Population Research and Training (NIPORT) and supported by the United States Agency for International Development (USAID) [38, 39]. This survey aims to assess healthcare indicators and provide a comprehensive overview of the population, women, and mental health.

The data set used in this study includes 9451 women of reproductive age (15–49 years) who made up the sample. This subset was extracted from a survey for women, focusing on individuals with complete information on the relevant variables. For the survival analysis, the event of interest was the occurrence of early marriage, defined as marriage before turning 18. The time variable was the age of the first marriage, with censoring applied for individuals who had not experienced the event by the time of the survey. Key covariates included division, residence, economic rate, education of the respondent and partner, mass media access, current employment status, and religion.

Statistical analyses followed standard practices in survival modeling, and reporting was guided by recommendations from Assel et al. to ensure transparency and clinical relevance [40].

2.3. Survival Modeling

In the statistical data analysis technique known as survival analysis, the time till an event occurs is the outcome variable of interest. In this study, ‘time’ refers to the age at first marriage, as reported retrospectively by participants in the BDHS‐2022, rather than follow‐up time from a longitudinal study. Death, the beginning of a sickness, recurrence following a time of remission, recovery, or any other particularly specified experience of interest are examples of events.

2.4. Cox Proportional Hazard Model

The Cox Regression model is a widely used regression model for analyzing survival data. The semi‐parametric Cox proportional hazard (Cox‐PH) model is composed of a parametrically estimated collection of covariates and a non‐parametrically determined baseline hazard. The hazard of an individual with zero variables is known as the baseline hazard. The capacity of the Cox‐ PH model to assess the connection between the explanatory factors and the hazard rate without requiring any assumptions regarding he baseline hazard function's form is its unique selling point. To put it another way, this model's strength is that it leaves the baseline hazard function vague. The following is how the Cox‐PH regression model links covariates to the hazard function [41]:

$$h\left(t|X\right)=h_0\left(t\right)\text{exp}\{\beta \prime X\}$$ (1)

Where, The hazard function for an individual whom all covariates included in the model are set to zero, denoted by $h_0(t)$, is known as the baseline hazard function, $\beta =({\beta }_1,{\beta }_2,\text{…},{\beta }_p)\prime$ is a vector of unknown regression coefficients, and a vector of explanatory variables is denoted by $X=(X_1,X_2,\text{…},X_p)\prime$. The maximum likelihood (ML) method is used to estimate the covariates' regression coefficients. A partial likelihood function is maximized to produce ML estimates.

2.5. Accelerated Failure Time (AFT) Models

The survival time's natural logarithm is expressed as a linear function of the covariates to produce the linear model in the AFT model:

$$\log \left(t_j\right)={\beta }^{*\prime }{X}_j+{\epsilon }_j$$ (2)

Where, ${\beta }^{*}=-\sigma \beta$ with $\sigma =1/\alpha ,$ and ${\epsilon }_j$ is an error term with density f(.). The regression model is determined by the error term's distributional form. For instance, setting f(.) equal to the extreme‐value density yields the exponential and Weibull regression models. Unlike proportional hazards models, which assume a constant hazard ratio over time, AFT models assume a constant ratio of survival times over time.

2.6. Model Assumption Verification and Fit Evaluation

To assess the proportional hazards (PH) assumption of the Cox model, we applied the scaled Schoenfeld residuals test, which evaluates whether covariate effects vary over time. A nonsignificant test suggests the PH assumption holds, while significant results indicate time‐varying effects. Additionally, the Cox–Snell residuals test was used to examine the overall model fit by plotting the cumulative hazard function against the residuals. A straight 45° line suggests a good model fit. Deviations from this line suggest model inadequacy. These diagnostics guided the selection of the most appropriate survival model for the data.

2.7. Model Diagnostic

One of the most common problems in statistical practice is choosing and comparing models, and there are a number of ways to do this. This study used the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) to compare the Cox‐PH model and AFT models.

2.8. Data Analysis

Data preparation and initial data management, including coding, labeling, and cleaning of variables, were conducted using IBM SPSS Statistics version 25. All statistical analyses, including descriptive statistics, survival modeling (Cox proportional hazard and accelerated failure time models), model comparison using AIC and BIC, and diagnostics (Schoenfeld residuals, Cox–Snell residuals), were performed in R version 4.4.1. We used relevant R packages such as survival and survminer to conduct the survival analysis. A p value of less than 0.05 was deemed statistically significant, and all statistical tests were two‐sided. To guarantee clear and consistent reporting of findings, the study adhered to the SAMPL (Statistical Analyses and Methods in the Published Literature) principles.

2.9. Ethical Considerations

This study is based exclusively on publicly available, deidentified secondary data obtained from the Bangladesh Demographic and Health Survey 2022 (BDHS‐2022). Before conducting interviews, explicit consent was obtained from all participants. The data collection procedures followed standard Demographic and Health Information Systems (DHSs) protocols, which were thoroughly reviewed and approved by the International Institutional Review Board. As the study involved analyzing pre‐existing secondary data from the public domain, (http://dhsProgram.com/data/available-datasets.cfm), it did not require ethics committee approval. In accordance with the ethical guidelines of my institution, research utilizing such data does not require formal review, approval, or waiver from an Institutional Review Board (IRB). Therefore, no IRB approval or waiver documentation was obtained for this study.

3. Results

3.1. Descriptive Analysis

Table 1 presents an analysis of the association between various sociodemographic factors and the age of marriage, stratified by early marriage (< 18 years) and not early marriage (≥ 18 years). The p values provide insights into the statistical significance of the differences between groups, highlighting key patterns in region, residence, wealth, education, media access, employment, religion, and BMI. Notably, early marriage is most prevalent in Dhaka (14.8%) and Rajshahi (14.6%), whereas delayed marriage is highest in Chattogram (18.4%) and Sylhet (17.4%). This may suggest regional influences on marriage timing, potentially driven by social or economic factors unique to each area. In rural areas, early marriage is more prevalent (69%) than in urban (31%), while delayed marriage is higher in urban settings (42%) than in rural (58%). Early marriages are more common in the lower class (41.4%), while delayed marriage is predominant in the upper class (53.3%). Early marriage is highest among respondents who are illiterate (15%) and those with only a primary level (29.9%). Conversely, those with higher education account for a substantial portion of delayed marriages (34.1%). These findings support education as a key factor in delaying marriage, potentially due to increased career prospects and aspirations for educated individuals. A similar trend is observed with husbands' education levels (p < 0.001), where husbands who are not educated (25.3%) or primary education (30.9%) are more likely to have married early, while those with higher education are associated with delayed marriages (32.9%). The majority of delayed marriages occur among those with media access (61.8%), compared to 55.9% in the early marriage group. This could indicate that media exposure fosters awareness of educational and career opportunities, influencing marriage decisions. A greater percentage of those in the early marriage group are unemployed (66.9%), while employment is lower (25.9%) among those married at 18 or older. Non‐muslims are more represented in the delayed marriage group (15.1%) compared to early marriages (7.9%), suggesting potential cultural differences in marriage timing within religious groups. The distribution of BMI categories is consistent across both groups, with normal BMI being most common (53.6% in early marriage and 52.2% in delayed marriage), followed by overweight and underweight categories.

Table 1.

Distribution of early marriage compared to non‐early marriage using the log‐rank test and survival determinants.

Covariates	Category	Survival status			p value (2‐tailed)
		Early marriage n (%) < 18 years	Not early marriage n (%) > = 18 years	Total n (%)
Region	Barishal	700 (11.2%)	317 (9.8%)	1017 (10.8%)	< 0.001
	Chattogram	797 (12.8%)	594 (18.4%)	1391 (14.7%)
	Dhaka	924 (14.8%)	484 (15%)	1408 (14.9%)
	Khulna	906 (14.6%)	308 (9.5%)	1214 (12.8%)
	Mymensingh	692 (11.1%)	342 (10.6%)	1034 (10.9%)
	Rajshahi	907 (14.6%)	315 (9.8%)	1222 (12.9%)
	Rangpur	858 (13.8%)	305 (9.5%)	1163 (12.3%)
	Sylhet	441 (7.1%)	561 (17.4%)	1002 (10.6%)
Residence	Urban	1929 (31%)	1355 (42%)	3284 (34.7%)	< 0.001
	Rural	4296 (69%)	1871 (58%)	6167 (65.3%)
Wealth index	Lower‐class	2580 (41.4%)	898 (27.8%)	3478 (36.8%)	< 0.001
	Middle‐class	1321 (21.2%)	607 (18.8%)	1928 (20.4%)
	Upper‐class	2324 (37.3%)	1721 (53.3%)	4045 (42.8%)
Respondent's education level	No education	935 (15%)	266 (8.2%)	1201 (12.7%)	< 0.001
	Primary	1863 (29.9%)	548 (17%)	2411 (25.5%)
	Secondary	3051 (49%)	1312 (40.7%)	4363 (46.2%)
	Higher	376 (6%)	1100 (34.1%)	1476 (15.6%)
Husband's education level	No education	1577 (25.3%)	447 (13.9%)	2024 (21.4%)	< 0.001
	Primary	1924 (30.9%)	704 (21.8%)	2628 (27.8%)
	Secondary	2021 (32.5%)	1015 (31.5%)	3036 (32.1%)
	Higher	703 (11.3%)	1060 (32.9%)	1763 (18.7%)
Access to mass media	No	2748 (44.1%)	1233 (38.2%)	3981 (42.1%)	< 0.001
	Yes	3477 (55.9%)	1993 (61.8%)	5470 (57.9%)
Respondent currently working	No	4166 (66.9%)	2390 (74.1%)	6556 (69.4%)	< 0.001
	Yes	2059 (33.1%)	836 (25.9%)	2895 (30.6%)
Religion	Non‐muslim	494 (7.9%)	487 (15.1%)	981 (10.4%)	< 0.001
	Muslim	5731 (92.1%)	2739 (84.9%)	8470 (89.6%)

Note: 5% level of significance.

The results of the log‐rank test in Table 1 indicate that, at the 5% level, the following factors significantly affect the survival time of early marriage: region, residence, economic index, respondent's and husband's educational attainment, access to mass media, respondent's current employment, and religion.

The study's dependent variable, early marriage, is displayed in Figure 1. Two groups are displayed, such as early marriage (considered as complete observations in survival modeling), which is indicated by the pink bar. This group has a much higher frequency (65.9%), meaning that a greater percentage of people were married young. The purple bar, which represents another category of not early marriage (considered as censored observations), has a lower frequency (34.1%), meaning that fewer people did not experience early marriage (censored cases). Given the greater number of complete observations (early marriage) than the censored cases (not early marriage), this figure implies that early marriage is prevalent in the population that is being studied. This finding suggests that further investigation into the reasons for and consequences of early marriage is required.

Figure 1.

Frequency of early marriage.

We start by fitting the data set to the Cox proportional hazard model. The findings showed that the following factors were statistically significant: residence, wealth index, respondent and spouse education levels, access to mass media, and religion. The scaled Schoenfeld residuals test was then used to evaluate the proportional hazards (PH) assumption. The null hypothesis states that the Schoenfeld residuals are independent of time and that there is no relationship between them and ranking event time. The results are given in Table 2.

Table 2.

Application of scaled Schoenfeld residuals to test the PH assumption.

Covariates	χ 2 statistic	p value (2‐tailed)
Region	2.0100	> 0.05
Residence	28.330	< 0.001
Wealth index	62.540	< 0.001
Respondent's education level	117.66	< 0.001
Husband's education level	78.880	< 0.001
Access to mass media	12.350	< 0.001
Respondent currently working	1.2100	> 0.05
Religion	12.710	< 0.001

The plot of Cox–Snell residuals (Figure 2) also suggests that the Cox‐PH model does not fit the data adequately. Thus, based on test results and graphical methods we can conclude that AFT models are more appropriate for the data at hand.

Figure 2.

Plot of Cox–Snell residuals for Cox PH model.

3.2. Model Comparison

The Akaike information criterion (AIC) and Bayesian information criterion (BIC) are commonly used in survival analysis to compare a variety of potential models. Table 3 presents the AIC and BIC values for the semi‐parametric survival (Cox‐PH) model as well as for some AFT models. We can see from the results that the Log logistic parametric AFT model has the smallest information criteria, and thus, is considered the best‐fit model for subsequent analysis.

Table 3.

AIC and BIC values for various survival models.

Model	AIC	BIC
Cox PH	45 579.24	45 633.95
Exponential model (AFT)	52 159.82	52 231.36
Weibull model (AFT)	42 846.68	42 925.37
Log normal model (AFT)	39 552.46	39 631.15
Log logistic model (AFT)	39 415.12	39 493.81
Gaussian model (AFT)	42 225.37	42 304.07
Log Gaussian model (AFT)	39 552.46	39 631.15
Student‐t model (AFT)	41 323.01	41 401.71
Rayleigh model (AFT)	45 461.18	45 532.72
Logistic model (AFT)	41 624.67	41 703.37

Note: Bold values indicate lowest AIC and BIC.

The result of the fitted Log logistic AFT model is present in Table 4, the findings indicate that residence type, respondent's education level, religion and employment status are significant factors influencing the risk of the early marriage.

Table 4.

Results of the fitted log logistic AFT model.

Covariates	Category	Hazard ratio (HR)	95% CI	p value (2‐tailed)
Region	Barishal (ref.)
	Chattogram	1.011	(0.996–1.025)	> 0.05
	Dhaka	0.991	(0.976–1.006)	> 0.05
	Khulna	0.990	(0.974–1.007)	> 0.05
	Mymensingh	0.992	(0.976–1.008)	> 0.05
	Rajshahi	0.998	(0.982–1.016)	> 0.05
	Rangpur	1.005	(0.989–1.022)	> 0.05
	Sylhet	0.975	(0.960–0.990)	< 0.001
Residence	Urban (ref.)
	Rural	1.019	(1.010–1.027)	< 0.001
Wealth index	Lower‐class (ref.)
	Middle‐class	1.005	(0.995–1.017)	> 0.05
	Upper‐class	0.995	(0.984–1.005)	> 0.05
Respondent's education level	No education (ref.)
	Primary	1.017	(1.001–1.034)	< 0.05
	Secondary	1.024	(1.008–1.040)	< 0.01
	Higher	0.975	(0.957–0.993)	< 0.01
Husband's education level	No education (ref.)
	Primary	1.009	(0.996–1.022)	> 0.05
	Secondary	0.998	(0.984–1.011)	> 0.05
	Higher	0.979	(0.963–0.994)	< 0.01
Access to mass media	No (ref.)
	Yes	0.999	(0.992–1.008)	> 0.05
Respondent currently working	No (ref.)
	Yes	0.988	(0.979–0.997)	< 0.01
Religion	Muslim (ref.)
	Non‐muslim	0.982	(0.971–0.993)	< 0.001

Note: 5% level of significance; 95% CI (confidence interval).

Compared to women in Barishal (the reference location), those in Sylhet are less likely to marry young (HR = 0.975, p < 0.001), giving the region a significantly lower risk of early marriage. The p values for the other regions (Chattogram, Dhaka, Khulna, Mymensingh, Rajshahi, and Rangpur) are not significant. Women in rural settings are more likely than those in urban areas to get married early (HR = 1.019, p < 0.001). This suggests that early marriage is somewhat more likely to occur for women living in rural areas. Because the p values for lower, middle, and upper‐class women are all greater than 0.05, there is no significant difference in the likelihood of an early marriage between these groups. The risk of early marriage is slightly higher for those with primary (HR = 1.017, p < 0.05) and secondary (HR = 1.024, p < 0.01) education than for those without, but it is still very low. Higher education is linked to a considerably lower probability of early marriage (HR = 0.975, p < 0.01), indicating that women with higher levels of education are less likely to get married young. Women are less likely to marry young than those whose husbands are uneducated if their husbands have higher levels of education (HR = 0.979, p < 0.01). The husband's degree of primary and secondary education has little effect on the likelihood of an early marriage. There is no identifiable correlation between the chance of an early marriage and mass media access (HR = 0.999, p > 0.05). Working women are less likely to be married young (HR = 0.988, p < 0.01), suggesting that employment may postpone or lower the chance of getting married young. Compared to muslim women, non‐muslim women are marginally less likely to be married young (HR = 0.982, p < 0.001).

4. Discussion

In previous generations, the majority of family members in Bangladesh, particularly the men, made all significant decisions concerning the women, such as whom to marry and when. These changes in the modern era, and women are now choosing for themselves what, when and with whom to get married. This study has illustrated the various triggering important factors of early marriage in Bangladesh. Using data from the 2022 Bangladesh Demographic and Health Survey, this study examined women's time to first marriage using survival analysis techniques.

Early marriage timing determinants were estimated and investigated in this study using univariate analysis, the log‐rank test, and log logistic AFT regression analysis. In a related study, parametric survival analysis, the Kaplan–Meier test, and the Log‐rank test were used. The best‐fitting model for early marriage for both males and females was the log logistic parametric hazard model [42].

This study investigated the socioeconomic and demographic determinants of early marriage among Bangladeshi women, using survival analysis models to identify significant factors. The findings highlight critical insights into the prevalence and predictors of early marriage in the country. According to the findings, Bangladeshi women's early marriage was significantly influenced by their division, place of residence, wealth index, respondent's education level, husband's education level, access to mass media, respondent currently working, and religion.

Early marriage was found to be significantly influenced by education. Compared to women with greater levels of education, those with only a primary or secondary education were more likely to marry young. This could be because, by raising awareness, encouraging career goals, and enhancing decision‐making abilities, higher education frequently postpones marriage. Studies conducted in South Asia and Sub‐Saharan Africa have found a continuous correlation between education and postponed marriage [43]. The intricate relationship between education, social norms, and economic concerns is shown by the fact that, in certain conservative societies, even females with a secondary education may experience familial or cultural pressure to marry young [34].

Employment also significantly influenced early marriage risk, with currently working women at a marginally lower risk. Regional disparities were evident, with Sylhet showing a 2.5% lower risk of early marriage compared to Barishal. Cultural practices, economic conditions, and access to education in Sylhet might contribute to these differences, highlighting the need for region‐specific interventions [44]. Throughout the world, it frequently appears that education and the age at first marriage are related. In this current study, we found that secondary‐educated women were more aware of the negative consequences of an early marriage than were illiterate women. In the meantime, there was no clear correlation between the moment of marriage and the educational attainment of women. Education and fertility are significantly negatively correlated, according to several studies [45].

Religious affiliation also played a role, with non‐muslim women exhibiting a lower risk of early marriage. This aligns with previous studies that identified religious and cultural norms as significant factors shaping marriage patterns [46]. This finding suggests that physical health indicators, often associated with social status, are not primary determinants in marriage timing in Bangladesh, contrary to studies in other contexts [37].

These findings underscore the multifaceted nature of early marriage in Bangladesh. Policy interventions should focus on promoting education, especially secondary and higher education, and addressing cultural norms and regional disparities. Moreover, empowering women economically through skill development and formal employment opportunities could help reduce early marriage rates.

5. Conclusion

This study offers practical insights for focused interventions by identifying important socioeconomic and demographic factors that contribute to early marriage among Bangladeshi women. The results show that while living in a rural area and having less education raises the risk of marriage, higher education levels (secondary and above) significantly postpone marriage. Working women are less likely to get married young, suggesting that employment status is a protective factor. Early marriage was shown to be significantly correlated with religion, with non‐muslim women having a lower likelihood of being married young than their Muslim counterparts. Disparities by region, like the reduced risk in Sylhet as opposed to Barishal, demonstrate the impact of regional cultural and economic settings.

To address early marriage, the following specific measures are recommended:

• 1.Education policies: expand access to secondary and higher education for girls, coupled with awareness campaigns on the health and socioeconomic benefits of delayed marriage.
• 2.Economic empowerment: strengthen vocational training and formal employment opportunities for women to reduce financial pressures that drive early marriage.
• 3.Community engagement: tailor programs to rural areas, addressing norms through community dialogues and involving religious leaders to shift perceptions.
• 4.Legal enforcement: ensure stricter implementation of marriage age laws, with monitoring mechanisms to curb underage unions.

By prioritizing these evidence‐based strategies, Bangladesh can mitigate early marriage's adverse effects, advancing gender equality and health outcomes. Future research should explore the longitudinal impacts of interventions and the role of male education in delaying marriage.

Author Contributions

Md Hazrat Ali: conceptualization and supervision. Farjana Afrin Koasha: writing – original draft, and formal analysis. Md Rakibul Islam: writing – original draft. Hafiza Akter Mim: writing – original draft, writing – review and editing. Md Yusuf Ali: writing – original draft.

Conflicts of Interest

The authors declare no conflicts of interest.

Transparency Statement

The corresponding author, Farjana Afrin Koasha, affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Data Availability Statement

The corresponding author will supply the data that backs up the study's findings if asked reasonably. The website https://www.dhsprogram.com also has data.