Knowledge of gestational diabetes mellitus and its sociodemographic determinants among pregnant women in Bangladesh: a nationwide facility-based cross-sectional study

Abstract

Objectives

To assess knowledge regarding gestational diabetes mellitus (GDM) among pregnant women attending antenatal care facilities in Bangladesh and identify sociodemographic factors independently associated with good knowledge.

Design

Nationwide facility-based cross-sectional study.

Setting

Seventy diabetes screening centres and affiliated antenatal care facilities of the Diabetic Association of Bangladesh (BADAS) across urban and rural Bangladesh.

Participants

Of 6631 women retained in the parent-study analytical cohort, 6120 women aged 18–45 years with an observed knowledge outcome were included in the present analysis. Participants were recruited consecutively during routine antenatal care visits between November 2019 and October 2020.

Primary outcome measures

Knowledge regarding GDM was assessed using an interviewer-administered questionnaire comprising six pregnancy-specific domains adapted from a previously validated Bangladeshi instrument (Cronbach’s α=0.72). Responses were scored from 0 to 6, and good knowledge (score>5) was analysed using population-averaged generalised estimating equations (GEEs) with centre as the clustering unit.

Results

Among 6120 women with an observed knowledge outcome, 23.2% had poor knowledge, 66.9% had average knowledge and 9.9% demonstrated good knowledge regarding GDM. In centre-clustered multivariable GEE analysis, middle household expenditure (adjusted OR 1.25, 95% CI 1.02 to 1.54), urban residence (adjusted OR 1.77, 95% CI 1.02 to 3.07), family history of diabetes (adjusted OR 1.22, 95% CI 1.02 to 1.45), receipt of preconception care (adjusted OR 1.26, 95% CI 1.02 to 1.55) and absence of previous pregnancy complications (adjusted OR 1.26, 95% CI 1.02 to 1.55) were independently associated with good knowledge. Maternal age, educational attainment and occupation were not independently associated with good knowledge.

Conclusions

Knowledge regarding GDM among pregnant women attending BADAS-affiliated antenatal care facilities in Bangladesh was suboptimal, with significant socioeconomic and urban–rural disparities. Strengthening structured GDM counselling within routine antenatal and preconception care may improve maternal health education and support earlier detection and management of GDM.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study included 6120 pregnant women with observed knowledge outcomes from 70 antenatal care facilities across all administrative divisions of Bangladesh.

• Gestational diabetes mellitus knowledge was assessed using a standardised, previously validated questionnaire with acceptable internal consistency (Cronbach’s α=0.72).

• Generalised estimating equations accounted for within-centre clustering, with both unadjusted and multivariable-adjusted estimates reported.

• The cross-sectional design precluded causal inference, and facility-based recruitment may limit generalisability to women not attending antenatal care facilities.

• Knowledge was assessed using a dichotomous scoring approach, which may have reduced sensitivity to partial understanding.

Introduction

Gestational diabetes mellitus (GDM) is a significant and growing public health concern, particularly in low- and middle-income countries (LMICs), where healthcare systems often face constraints in providing comprehensive antenatal care. GDM, defined as glucose intolerance with onset or first recognition during pregnancy, is associated with adverse maternal outcomes such as pre-eclampsia, operative delivery and future type 2 diabetes (T2DM), as well as neonatal complications including macrosomia, neonatal hypoglycaemia and birth trauma.^1–3 In addition to its immediate obstetric consequences, GDM has important long-term intergenerational implications, increasing the lifetime risk of obesity, glucose intolerance and cardiometabolic disorders in both mothers and their offspring.^{2 3} Consequently, improving awareness of GDM during pregnancy is considered an important component of strategies aimed at early detection, timely management and prevention of future diabetes.

In Bangladesh, reported GDM prevalence ranges from approximately 9% to over 15%, depending on diagnostic criteria and study setting.^4–6 Similar increasing trends have been reported across South and Southeast Asia, where rapid urbanisation, nutritional transition and increasing obesity are contributing to a growing burden of hyperglycaemia during pregnancy.^{4 7} Despite this increasing burden, awareness and understanding of GDM among pregnant women remain limited in many LMICs, potentially contributing to delayed screening, reduced uptake of preventive interventions and poorer maternal and neonatal outcomes.

Awareness and knowledge of GDM are essential for promoting timely healthcare-seeking behaviour, informed decision-making and adherence to recommended screening and management during pregnancy. Women with better knowledge are more likely to recognise risk factors, participate in recommended screening, comply with lifestyle advice and engage in appropriate follow-up after delivery. Nevertheless, studies from Bangladesh and other LMICs have consistently reported considerable variation in GDM knowledge according to age, education, socioeconomic status and urban–rural residence.^8–15 However, existing evidence from Bangladesh is derived primarily from single-centre or geographically restricted facility-based studies, including a previous urban-based study conducted by our group.⁸ Consequently, the extent to which knowledge varies across women attending antenatal care services in different geographic and socioeconomic settings remains insufficiently characterised.

A nationwide assessment using a standardised approach across multiple antenatal care facilities may provide a broader understanding of knowledge gaps and sociodemographic inequalities among women accessing pregnancy care in Bangladesh. Such evidence is important for informing targeted antenatal education programmes, strengthening routine counselling services and supporting implementation of GDM prevention and early detection strategies within maternal healthcare programmes.

Therefore, this nationwide facility-based cross-sectional study aimed to assess knowledge regarding GDM among pregnant women attending the Diabetic Association of Bangladesh (BADAS) diabetes screening centres and affiliated antenatal care facilities across Bangladesh and to identify sociodemographic factors independently associated with good knowledge. By including women from geographically diverse urban and rural facilities, this study sought to generate robust evidence to inform maternal health education and GDM prevention strategies among women accessing antenatal care in Bangladesh.

Methods

Study design and reporting standards

This nationwide facility-based cross-sectional study represents a secondary analysis of data collected through a broader nationwide survey conducted between November 2019 and October 2020. The parent survey primarily investigated the prevalence and determinants of GDM and diabetes in pregnancy (DIP) among pregnant women attending antenatal care facilities across Bangladesh. The present secondary analysis was undertaken to assess knowledge of GDM and its sociodemographic determinants. The design, setting, participant recruitment and data-collection procedures relevant to the present analysis are described below. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cross-sectional studies.¹⁶

A written protocol for the parent study was developed and approved by the BADAS Ethical Review Committee before data collection. The study was not prospectively registered, and no separate formal statistical analysis plan was registered or published for this secondary analysis. Before undertaking the present association analyses, the primary knowledge-outcome definition and covariates for the adjusted model were specified based on clinical relevance and previously published literature; these decisions were not documented in a time-stamped public registration.

Study setting

The study was conducted through the nationwide network of the BADAS, the largest non-governmental organisation providing diabetes prevention, diagnosis and treatment services in Bangladesh. In collaboration with the Directorate General of Health Services, BADAS-affiliated centres provide routine antenatal care, diabetes screening during pregnancy and referral services.

Data collection was coordinated through 70 BADAS-affiliated centres purposively selected to ensure broad geographic coverage across all administrative divisions of Bangladesh, including urban and rural populations. Selection of participating centres was based on predefined operational criteria, including geographic representation, antenatal attendance, capacity to conduct standardised GDM screening and availability of trained healthcare personnel. Collectively, these centres provide antenatal services for women residing in all 64 districts of Bangladesh (online supplemental table 1). The centre was therefore the natural clustering unit; administrative division was used only to describe geographic coverage and was not specified as the analytical clustering unit.

Because participants were recruited from antenatal care facilities rather than through probability sampling of the general population, this study should be considered a nationwide facility-based survey. Accordingly, the findings are generalisable primarily to women attending participating antenatal care facilities and should not be interpreted as nationally representative of all pregnant women in Bangladesh.

Study population and participant recruitment

Trained interviewers approached potentially eligible pregnant women attending routine antenatal care at participating BADAS-affiliated centres. Women were recruited consecutively throughout the study period at the individual level within each centre; this was not probability-based cluster sampling. Women were eligible if they were 18–45 years of age, had a confirmed ongoing pregnancy and attended one of the participating centres. Trained interviewers provided and explained the participant information and consent form, answered participants’ questions and obtained written informed consent before administering the face-to-face questionnaire and before any clinical or laboratory assessment. Because women attending the same centre could share characteristics or care-related exposures, the analysis accounted for potential within-centre correlation. The complete case report form and study questionnaire are provided in online supplemental file 1. Participant data were anonymised before analysis.

Women were excluded from the knowledge-outcome analyses if information required to calculate the knowledge score was missing. Missing occupation data for six women with observed knowledge outcomes were retained as an explicit missing category in the adjusted model.

Of 7300 eligible pregnant women approached, 6631 (90.8%) were retained in the parent-study analytical cohort after exclusion of 669 women (9.2%) with incomplete questionnaires or missing key sociodemographic information. Among the 6631 women, 6120 had an observed knowledge outcome and contributed to the present knowledge analyses; 511 had missing knowledge-category data. Participant flow is presented in figure 1.

Figure 1. Participant flow diagram. The flow diagram illustrates participant selection for the present secondary analysis. Between November 2019 and October 2020, 7300 eligible pregnant women attending 70 BADAS-affiliated antenatal care centres were approached. Of these, 669 women (9.2%) were excluded because of incomplete questionnaire responses (n=412) or missing key sociodemographic variables (n=257), leaving 6631 women (90.8%) in the parent-study analytical cohort. A further 511 women with missing knowledge-outcome data were excluded from the present knowledge analysis. The final analysis included 6120 women, representing 92.3% of the parent-study analytical cohort and 83.8% of all women approached. Percentages at each stage were calculated using the corresponding preceding cohort as the denominator. BADAS, Diabetic Association of Bangladesh.

To assess potential selection bias, available baseline characteristics of included and excluded participants were compared (online supplemental table 4). No clinically meaningful differences were observed for age, area of residence or other available baseline characteristics.

Data collection procedures

Data were collected using a structured interviewer-administered quantitative questionnaire during face-to-face interviews conducted before clinical assessment and laboratory investigations. The questionnaire formed part of the standardised data collection procedures developed for the nationwide GDM survey.

Before study implementation, all interviewers received centralised training covering study objectives, participant recruitment, informed consent, questionnaire administration and research ethics. Standard operating procedures were used across all participating centres. Field supervisors regularly monitored data collection, reviewed completed questionnaires for completeness and consistency, and provided ongoing quality assurance throughout the study period.

The questionnaire was originally developed in English, translated into Bangla using a forward–backward translation procedure and independently reviewed by experts in diabetology, obstetrics and public health to ensure conceptual equivalence. The Bangla version was pilot tested among 50 pregnant women before nationwide implementation to evaluate clarity and cultural appropriateness. Minor linguistic modifications identified during pilot testing were incorporated without changing questionnaire content or scoring.

Variables and measurements

Sociodemographic and clinical variables

Information was collected on maternal age, area of residence (urban/rural), educational attainment (0–5, 6–10 and >10 completed years of formal education), occupation (housewife, student, service holder or other), monthly household expenditure (<10 000, 10 000–20 000 and >20 000 Bangladeshi Taka (BDT; 1 US$ = 85 BDT in 2020)), family history of diabetes (yes/no), receipt of preconception care (yes/no) and previous pregnancy complications (present/absent).

Monthly household expenditure was used as a proxy indicator of socioeconomic status because household income is frequently underreported in Bangladesh and household expenditure provides a more stable measure of long-term economic position.

Assessment of knowledge regarding GDM

Knowledge regarding GDM was assessed using a structured questionnaire adapted from a previously validated instrument developed among pregnant women in Bangladesh.⁸ The questionnaire was selected because it was developed within the same sociocultural setting and demonstrated acceptable psychometric properties.

The final questionnaire comprised six domains: (1) awareness that diabetes can occur during pregnancy; (2) knowledge of GDM risk factors; (3) recommended timing of GDM screening; (4) maternal complications; (5) fetal and neonatal complications and (6) GDM management. The complete questionnaire and scoring system are presented in online supplemental file 1 and online supplemental table 2.

Each correct response received one point, whereas incorrect or ‘do not know’ responses received zero points. Individual scores were summed to generate a total knowledge score ranging from 0 to 6, with higher scores indicating greater knowledge regarding GDM. Internal consistency of the questionnaire in the present study was acceptable (Cronbach’s α=0.72).

Knowledge classification

The mean (SD (SD)) knowledge score was 2.6 (2.1). Consistent with the original validation study,⁸ participants were classified as having poor knowledge (score <1), average knowledge (score 1–5) or good knowledge (score >5).

For regression analyses, good knowledge (score >5) was prespecified as the primary outcome and analysed as a binary variable. This threshold was adopted from the original Bangladeshi validation study from which the questionnaire was adapted,⁸ enabling comparison with previous studies while identifying women demonstrating comprehensive knowledge across all six core domains of GDM. Sensitivity analyses using alternative knowledge classifications and continuous knowledge scores confirmed that the primary findings were robust (online supplemental tables 3 and 5).

Study size

The present study is a secondary analysis of a nationwide survey designed primarily to estimate the prevalence of GDM and DIP in Bangladesh. For the parent study, an expected GDM prevalence of 9.7%,⁵ a 95% confidence level and an absolute precision of 1% yielded an initial sample size of 3365 women. A design effect of 1.2 was applied at the planning stage to allow for potential correlation among women recruited within the same participating centre, increasing the required sample to 4038; allowing 10% for non-response or exclusion produced a minimum target of approximately 4442 women. The design effect reflected the multicentre structure and did not imply probability-based cluster sampling of individual participants.

All eligible women attending the 70 participating centres during the study period were approached consecutively without imposing an upper recruitment limit. Of 7300 women approached, 6631 were retained in the parent-study analytical cohort and 6120 with an observed knowledge outcome contributed to the present analysis, exceeding the minimum target and providing adequate precision for the planned analyses.

Statistical analysis

Continuous variables are presented as mean (SD) or mean (95% CI), whereas categorical variables are presented as frequencies and percentages.

Differences in mean knowledge scores across participant characteristics were assessed descriptively using independent-samples t-tests for binary variables and one-way analysis of variance for variables with more than two categories. Associations between categorical variables were examined using the χ² test.

The primary outcome was good knowledge of GDM (score >5). Because participants were recruited consecutively at the individual level within 70 participating centres and observations from the same centre could be correlated, centre—not administrative division—was specified as the clustering unit. Population-averaged generalised estimating equation (GEE) models with a binomial distribution, logit link and exchangeable working correlation structure were fitted with robust sandwich standard errors. Univariable models estimated unadjusted ORs, and a multivariable model estimated adjusted ORs. Both estimates are reported with centre-robust 95% CIs and two-sided p values, in accordance with STROBE recommendations.

Covariates were selected a priori based on clinical relevance and published literature rather than statistical significance in univariable analyses. The adjusted model included maternal age, monthly household expenditure, educational attainment, occupation, area of residence, family history of diabetes, receipt of preconception care and previous pregnancy complications. The GEE analyses included 6120 women with observed knowledge outcomes from all 70 centres; six women with missing occupation data were retained as an explicit missing category.

Multicollinearity among model covariates was assessed using variance inflation factors, and no problematic collinearity was identified.

Sensitivity analyses evaluated the robustness of the findings using alternative definitions of good knowledge and by modelling the knowledge score as a continuous outcome. These analyses produced no material change in the overall interpretation of the primary findings (online supplemental tables 3 and 5).

All analyses were performed using PASW Statistics V.21 (SPSS Inc., Chicago, Illinois, USA). Statistical tests were two-sided, and p<0.05 was considered statistically significant.

Bias and quality assurance

Several procedures were implemented to minimise potential bias. Consecutive recruitment of eligible participants reduced selection bias, while standardised data collection procedures, centralised interviewer training and regular field supervision promoted consistency across participating centres.

Information bias was minimised through use of a structured interviewer-administered questionnaire, standardised interviewing procedures and a predefined scoring guide. Translation, pilot testing and routine data verification further strengthened data quality.

Patient and public involvement

Patients and members of the public were not involved in developing the research question, study design, outcome selection, analysis or interpretation of the findings. Healthcare professionals and programme managers from BADAS contributed to the development and implementation of the study procedures. Findings will be disseminated through peer-reviewed publications, scientific meetings and engagement with relevant healthcare and public health stakeholders

Results

Participant characteristics

Between November 2019 and October 2020, 7300 eligible pregnant women attending 70 BADAS-affiliated antenatal care centres were approached. After 669 women (9.2%) with incomplete questionnaires or missing key sociodemographic information were excluded, 6631 women (90.8%) were retained in the parent-study analytical cohort. Of these, 6120 women with an observed knowledge outcome from all 70 centres contributed to the present analysis; 511 had missing knowledge-category data (figure 1). Available baseline characteristics of included and excluded participants are compared in online supplemental table 4.

The mean (SD) age of participants was 25.4 (5.3) years. Overall, 57.7% of women resided in urban areas and 42.3% in rural areas, while the largest age group was 25–29 years (30.8%). Most participants were housewives (84.7%), one-third reported a family history of diabetes (33.4%) and approximately one-third had received preconception care. The mean (SD) knowledge score was 2.6 (2.1). Detailed participant characteristics are presented in table 1.

Table 1. Characteristics of the study participants (number = 6631).

Variables	Descriptive measures
Age (years), mean±SD	25.42±5.3
Age categories, % (number)
18–20 years	23.2 (1537)
21–24 years	22.1 (1464)
25–29 years	30.8 (2044)
≥30 years	23.9 (1586)
Residence, % (number)
Urban	57.7 (3824)
Rural	42.3 (2807)
Monthly expenditure, % (number)
Low income (<10 000 BDT)	40.7 (2592)
Middle income (10 000–20 000 BDT)	44.5 (2831)
High income (>20 000 BDT)	14.8 (944)
Education, % (number)
0–5 years	26.2 (1656)
6–10 years	43.3 (2736)
>10 years	30.5 (1926)
Occupation, % (number)
Housewife	84.7 (5328)
Students	2.7 (172)
Service holders	8.9 (559)
Others	3.7 (233)
Family history of diabetes, % (number)	33.4 (2171)
Preconception care, % (number)	34.0 (2114)
History of previous pregnancy complications, % (number)*	24.6 (987)
Knowledge score, mean±SD	2.6±2.1

^*Among women with ≥1 previous pregnancy (n=4,012); 24.6% (987) reported complications.

BDT, Bangladeshi Taka (1 US$=85 BDT in 2020).

Knowledge regarding GDM

Knowledge regarding individual domains of GDM varied considerably (table 2). Awareness that diabetes can develop during pregnancy was relatively high (63.5%), whereas knowledge regarding risk factors (35.6%), recommended timing of GDM screening (37.6%), maternal complications (42.8%), fetal and neonatal complications (40.4%) and appropriate management (40.1%) was substantially lower. Overall, fewer than half of participants correctly answered five of the six knowledge items, indicating important gaps in understanding of GDM beyond basic awareness.

Table 2. Knowledge regarding gestational diabetes (GDM) among pregnant women.

	Correct answer	Incorrect answer
(1) Diabetes can develop during pregnancy	3881 (63.5)	2228 (36.5)
(2) Risk factors of GDM	2172 (35.6)	3928 (64.4)
(3) Time for GDM screening	2287 (37.6)	3791 (62.4)
Problems in GDM mothers	2610 (42.8)	3493 (57.2)
Problems in babies of GDM mothers	2452 (40.4)	3624 (59.6)
Management of GDM	2441 (40.1)	3646 (59.9)

Data are presented as number (%). Denominators vary due to missing item responses.

Distribution of knowledge levels

Among 6120 women with an observed knowledge outcome, 23.2% were classified as having poor knowledge (score <1), 66.9% as having average knowledge (score 1–5) and 9.9% as having good knowledge (score >5). The distribution of knowledge categories, with 95% CIs, is illustrated in figure 2.

Figure 2. Level of knowledge regarding gestational diabetes (GDM) among pregnant women. Bars represent the percentage of pregnant women classified as having poor, average or good knowledge of gestational diabetes mellitus based on a composite knowledge score. Error bars indicate 95% CIs. Percentages are shown above each bar.

For GEE analyses, good knowledge (score >5) was treated as the primary binary outcome, with poor or average knowledge classified as non-good knowledge.

Knowledge score and proportion with good knowledge by participant characteristics

Mean knowledge scores and the proportion of participants with good knowledge (score >5) varied across sociodemographic groups (table 3).

Table 3. Knowledge score and proportion with good knowledge by participants’ characteristics (n=6631).

Variables	Mean knowledge score (95% CI)	P value	% (n/N) with good knowledge (score >5)*	P value
Age categories		<0.001		0.879
18–20 years	2.37 (2.26 to 2.49)		8.7 (134/1537)
21–24 years	2.53 (2.41 to 2.64)		8.8 (129/1464)
25–29 years	2.69 (2.60 to 2.79)		9.5 (194/2044)
≥30 years	2.77 (2.66 to 2.87)		9.4 (149/1586)
Socioeconomic status		<0.001		<0.001
Low income (<10 000 BDT)	2.21 (2.12 to 2.29)		6.8 (176/2592)
Middle income (10 000–20 000 BDT)	2.87 (2.78 to 2.95)		11.8 (333/2831)
High income (>20 000 BDT)	2.83 (2.69 to 2.97)		7.5 (71/944)
Education		<0.001		0.962
0–5 years	2.09 (1.98 to 2.19)		8.9 (148/1656)
6–10 years	2.59 (2.51 to 2.68)		8.6 (235/2736)
>10 years	2.97 (2.88 to 3.07)		8.5 (164/1926)
Occupation		<0.001		0.377
Housewife	2.54 (2.48 to 2.60)		8.7 (465/5328)
Students	3.07 (2.77 to 3.37)		11.0 (19/172)
Service holders	3.07 (2.89 to 3.25)		10.6 (59/559)
Others	2.17 (1.88 to 2.47)		10.3 (24/233)
Area of residence		0.241		<0.001
Urban	2.57 (2.50 to 2.65)		10.3 (395/3824)
Rural	2.64 (2.56 to 2.72)		7.5 (211/2807)
Family history of diabetes		0.007		0.031
No	2.55 (2.49 to 2.62)		9.4 (375/3995)
Yes	2.71 (2.62 to 2.80)		10.2 (222/2171)
Preconception care		0.006		<0.001
No	2.53 (2.47 to 2.60)		9.5 (325/3417)
Yes	2.70 (2.60 to 2.79)		11.3 (238/2114)
Previous history of pregnancy complications		0.002		<0.001
Absent	2.76 (2.67 to 2.84)		10.6 (368/3467)
Present	2.50 (2.38 to 2.63)		5.9 (69/1167)

Data are presented as mean (95% CI) and % (n/N).

^*Good knowledge defined as score >5 on a 0–6 scale. This column now shows within-category prevalence with explicit denominators.

BDT, Bangladeshi Taka (1 US$=85 BDT in 2020).

Mean knowledge scores increased significantly across age categories (p<0.001), rising from 2.37 (95% CI 2.26 to 2.49) among women aged 18–20 years to 2.77 (95% CI 2.66 to 2.87) among those aged ≥30 years. However, the proportion with good knowledge did not differ significantly by age group (p=0.879).

Socioeconomic status showed a strong gradient. Mean knowledge scores were lowest in the low-income group (2.21; 95% CI 2.12 to 2.29) and highest in the middle-income group (2.87; 95% CI 2.78 to 2.95) (p<0.001). Similarly, the proportion with good knowledge was significantly higher among middle-income women (57.4%) compared with low-income (30.3%) and high-income groups (12.2%) (p<0.001).

Education level was significantly associated with mean knowledge scores (p<0.001), with higher scores observed among women with >10 years of schooling (2.97; 95% CI 2.88 to 3.07). However, education was not significantly associated with the proportion classified as having good knowledge (p=0.962).

Students and service holders had higher mean knowledge scores compared with housewives (p<0.001), although occupation was not significantly associated with good knowledge in categorical analyses (p=0.377).

Mean knowledge scores did not differ significantly between urban and rural residents (p=0.241). In contrast, the proportion with good knowledge was significantly higher among urban women (65.2%) than rural women (34.8%) (p<0.001).

Participants with a family history of diabetes had higher mean knowledge scores (2.71 vs 2.55; p=0.007) and a higher proportion of good knowledge (37.2% vs 62.8% without family history; p=0.031). Similarly, receipt of preconception care and absence of previous pregnancy complications were associated with both higher mean scores and greater likelihood of good knowledge (all p<0.01).

Knowledge score according to participant characteristics

Mean knowledge scores differed significantly according to age, monthly household expenditure, educational attainment, occupation, family history of diabetes, receipt of preconception care and previous pregnancy complications (all p<0.01), whereas no significant difference was observed according to area of residence (p=0.241) (table 3).

When knowledge was examined using the binary outcome of good knowledge, significant differences were observed according to monthly household expenditure, area of residence, family history of diabetes, receipt of preconception care and previous pregnancy complications (all p<0.05). Women from middle-income households, those residing in urban areas, women with a family history of diabetes, those who had received preconception care and women without previous pregnancy complications had a significantly higher prevalence of good knowledge. In contrast, age, educational attainment and occupation were not significantly associated with the prevalence of good knowledge.

Factors independently associated with good knowledge

In centre-clustered univariable GEE analyses, middle household expenditure, service-holder occupation, urban residence, family history of diabetes, receipt of preconception care and absence of previous pregnancy complications were associated with increased odds of good knowledge (table 4).

Table 4. Population-averaged GEE associations between participant characteristics and good knowledge of gestational diabetes mellitus.

Variable	Unadjusted OR (95% CI)	P value	Adjusted OR (95% CI)	P value
Age
18–20 years	Reference		Reference
21–24 years	0.97 (0.84 to 1.12)	0.695	0.94 (0.77 to 1.15)	0.568
25–29 years	1.19 (0.99 to 1.44)	0.067	1.14 (0.96 to 1.35)	0.141
≥30 years	1.00 (0.87 to 1.15)	0.990	0.93 (0.75 to 1.16)	0.509
Monthly household expenditure
Low (<10 000 BDT)	Reference		Reference
Middle (10 000–20 000 BDT)	1.18 (1.01 to 1.38)	0.043	1.25 (1.02 to 1.54)	0.034
High (>20 000 BDT)	1.28 (0.90 to 1.81)	0.168	1.09 (0.81 to 1.46)	0.584
Education
0–5 years	Reference		Reference
6–10 years	1.05 (0.91 to 1.21)	0.477	1.02 (0.89 to 1.17)	0.795
>10 years	1.28 (0.99 to 1.65)	0.058	1.12 (0.80 to 1.58)	0.500
Occupation
Housewife	Reference		Reference
Student	1.03 (0.84 to 1.26)	0.805	0.98 (0.50 to 1.93)	0.950
Service holder	1.42 (1.05 to 1.91)	0.023	1.22 (0.88 to 1.68)	0.229
Other	1.24 (0.86 to 1.79)	0.252	1.23 (0.86 to 1.76)	0.264
Area of residence
Rural	Reference		Reference
Urban	2.04 (1.09 to 3.81)	0.025	1.77 (1.02 to 3.07)	0.042
Family history of diabetes
No	Reference		Reference
Yes	1.21 (1.01 to 1.45)	0.035	1.22 (1.02 to 1.45)	0.029
Preconception care
No	Reference		Reference
Yes	1.45 (1.15 to 1.84)	0.002	1.26 (1.02 to 1.55)	0.031
Previous pregnancy complications
Present	Reference		Reference
Absent	1.20 (1.01 to 1.46)	0.047	1.26 (1.02 to 1.55)	0.036

Population-averaged generalised estimating equation (GEE) models specified a binomial distribution, logit link and exchangeable working correlation structure, with robust sandwich standard errors and centre as the clustering unit. The adjusted model included 6120 participants with observed knowledge outcomes from all 70 centres; 606 participants had good knowledge. Six participants with missing occupation data were retained as an explicit missing category. Adjusted estimates included maternal age, monthly household expenditure, educational attainment, occupation, area of residence, family history of diabetes, preconception care and previous pregnancy complications. Good knowledge was defined as a score >5 on a 0–6 scale.

BDT, Bangladeshi Taka.

In the centre-clustered multivariable GEE model, middle household expenditure remained associated with good knowledge (adjusted OR 1.25, 95% CI 1.02 to 1.54; p=0.034). Urban residents had higher odds of good knowledge than rural residents (adjusted OR 1.77, 95% CI 1.02 to 3.07; p=0.042). Family history of diabetes (adjusted OR 1.22, 95% CI 1.02 to 1.45; p=0.029), receipt of preconception care (adjusted OR 1.26, 95% CI 1.02 to 1.55; p=0.031) and absence of previous pregnancy complications (adjusted OR 1.26, 95% CI 1.02 to 1.55; p=0.036) were also independently associated with good knowledge (table 4).

Maternal age, educational attainment, occupation and high household expenditure (>20 000 BDT) were not independently associated with good knowledge after multivariable adjustment.

Sensitivity analyses

Sensitivity analyses using alternative definitions of good knowledge produced findings that were consistent with the primary analysis, with middle household expenditure, urban residence, family history of diabetes, receipt of preconception care and previous pregnancy complications remaining significantly associated with better knowledge (online supplemental table 3).

When knowledge score was analysed as a continuous outcome using multivariable linear regression, the overall pattern of associations was broadly consistent with the primary logistic regression model (online supplemental table 5). Although increasing age was associated with a modest increase in mean knowledge score, age was not independently associated with the binary outcome of good knowledge, indicating that the primary findings were robust to alternative modelling approaches.

Discussion

This nationwide facility-based cross-sectional study provides one of the most comprehensive assessments of GDM knowledge among pregnant women attending antenatal care services in Bangladesh. Although awareness that diabetes can occur during pregnancy was relatively high, comprehensive knowledge regarding GDM remained poor, particularly with respect to risk factors, screening, complications and management. Women from middle-income households, urban areas, those with a family history of diabetes, those who had received preconception care and women without previous pregnancy complications were more likely to demonstrate good knowledge after multivariable adjustment. Maternal age, educational attainment and occupation were not independently associated with good knowledge, although age showed a modest positive association with the continuous knowledge score in sensitivity analyses.

These findings are consistent with previous studies from Bangladesh and other LMICs reporting inadequate knowledge of GDM, particularly regarding risk factors, screening recommendations and disease management.^6–9 Unlike previous single-centre or urban-based studies,¹⁷ this study included women attending 70 BADAS-affiliated antenatal care centres across Bangladesh, providing a broader assessment of knowledge among women receiving routine antenatal care. Although participants were recruited from healthcare facilities rather than through population-based sampling, the wide geographic coverage strengthens the relevance of these findings for antenatal care programmes within Bangladesh.

The observed knowledge gaps are concerning given the increasing burden of hyperglycaemia during pregnancy in Bangladesh and across South Asia.^2–7 Limited knowledge may delay participation in recommended screening, reduce adherence to lifestyle advice and impair postpartum follow-up, thereby limiting opportunities to prevent future T2DM.^10–15 These findings emphasise that improving awareness should remain an integral component of maternal healthcare and diabetes prevention programmes.

Socioeconomic position emerged as one of the strongest determinants of good knowledge. Women from middle-income households were significantly more likely to demonstrate good knowledge than women from lower-income households, whereas the highest expenditure group did not remain significant after adjustment. Similar socioeconomic gradients have been reported in Bangladesh and other LMICs.^{8 9} This finding suggests that financial resources alone may not ensure adequate knowledge and that differences in healthcare utilisation, health literacy and access to structured antenatal counselling may also contribute.

Urban residence was independently associated with better GDM knowledge, consistent with findings from studies conducted in other low- and middle-income settings.^{11 13 15} National survey data from Bangladesh also demonstrate urban–rural inequalities in the use of maternal healthcare services, including antenatal care.¹⁸ These differences in healthcare contact and opportunities to receive health information may partly explain the higher level of GDM knowledge observed among urban participants; however, healthcare access and exposure to counselling were not measured directly in the present analysis. Strengthening GDM education through rural antenatal services and community-based programmes may help address this knowledge gap.

Receipt of preconception care was also independently associated with better knowledge, highlighting the potential value of incorporating GDM education into preconception and early pregnancy services. Likewise, women with a family history of diabetes demonstrated better knowledge, consistent with previous evidence suggesting that personal or familial exposure increases health awareness and information-seeking behaviour.¹⁹ Family-centred educational approaches may therefore represent an effective strategy for improving GDM awareness within the Bangladeshi context.

Interestingly, educational attainment was not independently associated with good knowledge after adjustment. Although women with higher education achieved higher mean knowledge scores in unadjusted analyses, this association disappeared in the primary multivariable model. Similarly, sensitivity analyses indicated that increasing age was associated with slightly higher mean knowledge scores but not with the likelihood of achieving good knowledge. These findings suggest that exposure to structured antenatal counselling and healthcare services may be more influential than formal education alone in determining clinically meaningful knowledge regarding GDM.

The findings have important public health implications. Bangladesh is experiencing a growing burden of diabetes and hyperglycaemia during pregnancy.^2–7 While knowledge alone is unlikely to change health outcomes, it is an essential prerequisite for informed healthcare-seeking behaviour, participation in screening programmes and adherence to recommended management. Integrating structured GDM counselling into routine antenatal and preconception care, particularly in rural and socioeconomically disadvantaged communities, should therefore be prioritised. Community health workers, family-centred education and digital health approaches may further improve awareness and support early detection and management.

The principal strengths of this study include its large sample size, broad geographic coverage through 70 BADAS-affiliated centres, use of a previously validated knowledge instrument, standardised data collection procedures and comprehensive quality assurance. The inclusion of prespecified multivariable analyses, assessment of missing data and sensitivity analyses further strengthens the robustness of the findings.

Several limitations should be acknowledged. First, the cross-sectional design precludes causal inference, and residual confounding cannot be excluded. Second, because participants were recruited from healthcare facilities, the findings are generalisable primarily to women attending participating antenatal care services rather than to all pregnant women in Bangladesh. Third, interviewer-administered questionnaires may have introduced recall and social desirability bias. The categorical classification of knowledge may also have reduced sensitivity to detect subtle differences, although sensitivity analyses using the continuous knowledge score produced similar findings. Finally, the study did not evaluate attitudes or behavioural practices related to GDM, limiting assessment of how knowledge translates into preventive behaviours.

In conclusion, knowledge regarding GDM among pregnant women attending antenatal care services in Bangladesh remains inadequate, particularly regarding risk factors, screening and disease management. Significant socioeconomic and urban–rural disparities persist. Strengthening structured GDM counselling within routine antenatal and preconception care, supported by community-based and family-centred educational strategies, may improve awareness and contribute to earlier detection and better management of GDM.

Data availability statement

Data are available upon reasonable request.