Association of biological, socioeconomic, and environmental factors with child development at two years of age in Bangladesh, Pakistan, and Tanzania: results from the AMANHI-ACT study

Abstract

Background

The first 1,000 days of life, spanning pregnancy to two years, are critical for development. We investigated the association of maternal, socioeconomic, environmental, nutritional, and childhood characteristics with cognitive, motor, and language development at two years of age.

Methods

The AMANHI-ACT study enrolled pregnant women between 8 and 19 weeks of gestation from Bangladesh, Pakistan, and Tanzania. Data was collected on sociodemographic factors, maternal BMI and co-morbidities during pregnancy, and weight and gestational age at birth. Child characteristics included vaccination history, infections, and physical growth at predefined intervals up to two years. Child development at two years was assessed using the Bayley Scales of Infant and Toddler Development Third Edition (Bayley-III). The home environment was assessed using the Home Observation for Measurement of the Environment Infant-Toddler (HOME-IT) tool. We performed linear regression and reported site-adjusted regression coefficients with 95% confidence intervals.

Results

We assessed 3,062 children at two years of age to identify factors associated with their cognitive, language, and motor development scores. Cognitive scores were significantly lower among children whose mothers had no formal education (Adjusted coefficient = -3.8; 95% CI: -5.8, -1.8) and those who had not received any vaccinations (Adjusted coefficient = -6.0; 95% CI: -9.3, -2.6). Language development was similarly affected: children with mothers who had no education (Adjusted coefficient = -4.7; 95% CI: -6.8, -2.5) or only primary-level education (Adjusted coefficient = -4.5; 95% CI: -6.5, -2.5) showed markedly lower scores. Language scores were also reduced in children whose fathers had no education (Adjusted coefficient = -2.4; 95% CI: -4.2, -0.5) or only primary education (Adjusted coefficient = -2.4; 95% CI: -4.1, -0.6). For motor development, children who were stunted (Adjusted coefficient = -2.2; 95% CI: -2.9, -1.4), wasted (Adjusted coefficient = -2.8; 95% CI: -4.3, -1.4), or unvaccinated (Adjusted coefficient = -4.7; 95% CI: -8.0, -1.4) at two years exhibited significantly lower scores.

Conclusion

Preterm births, vaccination history, parental education, and physical growth significantly affected the first 1000 days of child development, thus identifying areas for targeted intervention to optimize long-term health outcomes in LMICs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12887-025-06060-2.

Introduction

Globally, it is estimated that 52.9 million children under five years live with a developmental disability, and 94.9% of this burden lies in low- and middle-income countries (LMICs) [1]. Despite improvements in healthcare, Sub-Saharan Africa and South Asia report nearly 15 million preterm births annually, leading to long-term consequences such as developmental disabilities [1, 2]. These regions also account for approximately 32.8 million stunted children under five years, according to the Joint Child Malnutrition Estimates (JCME) [3, 4].

Early childhood development (ECD), particularly from conception till two years of age, is a period with rapid progression of perception, motor, cognitive, language, socio-emotional, and self-regulation domains. These developmental milestones are shaped by a complex interplay of socioeconomic, nutritional, and environmental factors, which may have lifelong implications [2–4]. Previous studies that attempted to derive causal relationships between birth characteristics, physical growth, and childhood development often fail to account for factors such as parental education, home environment, illnesses such as diarrhea and fever, vaccination status, and frequency of hospitalizations.

To study the relationship between these factors and developmental domains, we established a sub-cohort within the Alliance for Maternal and Newborn Health Improvement (AMANHI) initiative across three low- and middle-income countries in sub-Saharan Africa and South Asia [5]. The sub-cohort comprised children born within the AMANHI cohort between 2016 and 2018, who were longitudinally followed for two years. We aim to investigate the association between various maternal, socioeconomic, environmental, and nutritional characteristics with cognitive, motor, and language development.

Methods

Study population and design

We carried out a longitudinal follow-up of children initially enrolled in the Alliance for Maternal and Newborn Health Improvement (AMANHI) biorepository cohort between 2016 and 2018 in Pemba (Tanzania), Sylhet (Bangladesh), and Karachi (Pakistan). A detailed description of the cohort has been published previously [5, 6]. The biorepository study aimed to collect and store high-quality maternal and newborn biological samples to identify early-life biomarkers linked to adverse outcomes such as preterm birth, small-for-gestational-age (SGA), and neurodevelopmental delays. To recruit study participants, households in Sylhet, Karachi, and Pemba Island had their geographical coordinates recorded and linked to a database using unique identification numbers. Trained community health workers visited each household in the predefined catchment areas every 2–3 months to inquire about pregnancies among women of reproductive age. If a woman reported or suspected a pregnancy, the CHWs determined the gestational age using the date of her last menstrual period and confirmed the pregnancy with a urine test. Pregnant women who provided consent underwent a screening ultrasound. Women with early pregnancies (between 8 and 19 weeks of gestation), who intended to remain in the study areas for the entire follow-up period and consented to the collection of both epidemiological data and biological samples, were enrolled in the cohort. The CHWs visited the women thrice during pregnancy, at 24–28 weeks, 32–36 weeks, and 38–40 weeks, to collect data on maternal age, body mass index (BMI), parity, and clinical information on the history of hypertension, preeclampsia, eclampsia, antepartum and postpartum infection, antepartum, and postpartum hemorrhage. CHWs visited twice in the postnatal period at 1–6 days and 42–60 days after delivery to capture birth outcomes and collect biospecimens.

The AMANHI biorepository study aimed to enroll approximately 6,000 pregnant women (around 2,000 per site) by April 2016. Given the uncertainty in effect sizes and the prevalence of predictors in discovery-focused biomarker research, we used a sensitivity analysis approach rather than fixed a priori power calculations. Assuming a 10% incidence of preterm birth and small-for-gestational-age (SGA) at each site, and accounting for ~ 15% attrition and 5–10% loss to follow-up or mortality by age three, a cohort of 2,000 mother–infant pairs per site was expected to yield at least 200 cases of each outcome and around 1,500 children with complete follow-up.

As a part of the AMANHI- All Children Thrive (ACT) secondary objective, all children born between 2016 and 2018 in the cohort were eligible for developmental assessments at two years of age, as those born earlier had already exceeded the follow-up window [6]. Our inclusion criteria were children aged 2 years whose mothers were enrolled in the AMANHI cohort during early pregnancy (before 20 weeks), consented to data collection, and were present in the study area for the duration of the follow-up. Trained community health workers performed three-monthly household visits till 24 months of age to collect data on factors that influence child development (Supplementary Table S1) [7].

Measures

We used standardized tools to collect information on evidence-based determinants of child development from birth to 24 months. These included birth characteristics, antenatal history, child’s physical growth, home environments, sociodemographic details, and a history of illnesses.

Child Development: Child development was assessed at 24 months of age using Bayley Scales of Infant and Toddler Development Third Edition (Bayley-III) [8, 9]. Bayley-III raw scores for motor, language, and cognitive domains were calculated by adding the number of successfully performed items for which the child scored “1 or yes” along with the number of unadministered items before the start point based on chronological age. The raw scores were then converted into scaled scores and subsequently composite scores, using the Bayley-III manual’s conversion Table [10]. These composite scores, with ranges of 55–145 for the cognitive domain and 45–155 for the motor and language domains, were retained in continuous form as outcome variables for further analysis.

Birth characteristics: We used ultrasound to determine the gestational age at enrollment. Preterm birth was defined as delivery before 37 weeks. Birth weight, measured within 72 h of delivery, was categorized as low if it was less than 2500 g.

Maternal factors: Maternal Body Mass Index (BMI) was calculated at enrollment (8–19 weeks gestation) and three antenatal visits. It was categorized as underweight (BMI < 18.5 kg/m²), normal weight (BMI 18.5–24.9 kg/m²), overweight (BMI 25.0–29.9 kg/m²), and obese (BMI ≥ 30 kg/m²). A maternal morbidity index was constructed to account for all comorbid complications each woman reported during pregnancy and up to 42 days postpartum.

Child anthropometric and nutritional indicators: Anthropometric measurements were converted into Z-scores for weight‐for‐age (WAZ), height‐for‐age (HAZ), and weight‐for‐height (WHZ) based on WHO cut-offs [11]. A cut-off of − 2 Z‐scores for these indices was used for classifying children for moderate to severe wasting (WHZ < − 2SD), underweight (WAZ < − 2SD), and stunting (HAZ < − 2SD). Exclusive breastfeeding was assessed through maternal recall at 3 months postpartum, with exclusivity defined as only breast milk, and no additional food or drink.

Socio-demographics: We collected data on household demographics, parental occupation, and education. We generated asset indices using principal component analysis, which were ranked and divided into socioeconomic quintiles, assigning each woman a wealth quintile based on her household’s asset profile. The asset inventory included ownership of durable goods (television, bicycle), dwelling characteristics (flooring, wall, and roof materials), utility access (electricity, water source, sanitation), and livestock or land ownership. We used the FANTA Household Hunger Scale to record household poverty at 24 months of age. It consisted of six questions that assessed the household’s food insecurity during the past 30 days. Each question was scored from 0 to 2, with 2 indicating the highest frequency. The scores were summed to calculate the HHS (0–6), with households categorized as ‘little to no hunger’ (0–1) or ‘moderate to severe hunger’ [2–6, 12].

Child stimulation at home: We used the culturally adapted version of the Home Observation for Measurement of the Environment Infant-Toddler (HOME-IT) to assess the child’s home environment and behavior at 24 months. The HOME-IT inventory tool assigned binary scores (yes/no) to 49 items across seven domains, with the cumulative score indicating the level of enrichment in the home environment [13, 14].

Childhood illnesses from birth to 24 months: Diarrhea and fever episodes in children were quantified as the cumulative number of maternal reports of diarrhea or fever within the first 24 months.

Training and standardization

Trainers from all sites underwent standardized training for anthropometric measurements and administration of the Bayley-III. A centralized training session for site trainers was conducted in Pemba, Tanzania, in September 2016. Anthropometry specialists from the WHO Department of Maternal, Newborn, Child, and Adolescent Health and Aging served as master trainers. Site trainers conducted anthropometric measurements on ten children in two rounds, which were first evaluated for intra-rater agreement (precision) and then against the measurements taken by the master trainers for inter-rater agreement (accuracy). During data collection, predefined margins set the acceptable range for differences between measurements by two assessors. Trainers who did not meet the required standards in the initial round were recommended for additional rounds for standardization.

For Bayley-III, specialists provided didactic sessions with item-wise explanations, followed by practice sessions where teams received feedback on administration and scoring. Trainers were certified once they achieved a high Pearson correlation ≥ 0.90) across 7–10 children, and they then served as master trainers.

Quality control

Harmonized standard operating procedures (SOPs) were developed to ensure quality control. These included weekly team meetings to address concerns related to the administration of study tools or challenges encountered during implementation; weekly calls with investigators and consultants to promote a consistent understanding of the SOPs across sites and to discuss emerging issues; data recording on hard copies followed by a double entry with logical checks in the data entry software to ensure accuracy; random spot checks conducted by the study manager or master trainers using a checklist used to provide feedback to assessors; and quarterly refresher trainings for data collectors.

Statistical analysis

We presented sociodemographic and pregnancy characteristics of the three sites as absolute frequencies with percentages for categorical variables, mean with standard deviations for continuous variables with normal distribution, and median with interquartile ranges for non-normally distributed variables. We assessed exposure variables including socioeconomic status, household hunger status, maternal age, BMI, parental education, parity, maternal morbidity, child sex, birthweight, gestational age at birth, nutritional and vaccination history, child anthropometric measurements at 3,6, 12,18, and 24 months, infections in the past two years, and HOME-IT domains taken as a continuous variable in relation to Bayley-III cognitive, language, and motor domains in univariate analyses. The factors with a significant association (p-value of < 0.10) were considered for higher-order analyses. We followed a complete-case analysis approach. Multicollinearity among independent variables was assessed using a Variance Inflation Factor (VIF) at a cut-off point of > 10,and all VIF values were below the threshold. For multivariable model building, variable selection was performed using a forward stepwise approach, retaining variables with a p-value of < 0.05. We included parental education and SES as independent predictors in the regression model as impact developmental domains differently. We assessed the interaction between household hunger and wealth quintile variables to determine if the combined effect of these socioeconomic factors differed from their individual effects. Despite testing these interactions, none were found to be statistically significant. Therefore, we did not include these interaction terms in the final multivariable models. We adjusted for any site-specific factors by including them as a fixed effect in the final linear regression model. Adjusted regression coefficients with 95% confidence intervals (CIs) represent the adjusted effect of each predictor on child development scores (i.e., cognitive, language, and motor) domains. All analyses were performed using Stata software version 16.0.

Ethical considerations

The AMANHI study received ethical approval from the local and institutional ethics committees of all three sites. These included Zanzibar Health Research Ethics Committee (formerly ZAMREC) (ZAMREC/0002/OCTOBER/013) for Tanzania, ICDDR, B (PR12073), and Johns Hopkins University (IRB 00004508) for Bangladesh, and Aga Khan University (2790-paeds-ERC-13) for Pakistan. In addition, the protocols for the biorepository study were approved by the WHO Ethics Committee (RPC 532), and continuing approvals were sought yearly. Informed consent was obtained from pregnant women at the time of enrollment, and later, at childbirth, they consented to their child’s participation in the study. Consent forms were provided in the mothers’ preferred languages to ensure comprehension. In cases where mothers expressed concerns about their child’s development, or where the child showed faltered physical growth or developmental delays, referrals were made to the outpatient department of the study center. Physicians at the OPD then assessed the child and, if necessary, referred them to nearby hospitals for further evaluation and management.

Results

Between January 2017 and December 2019, a total of 4,006 live births were recorded and subsequently followed for two years. Of these, 3,062 children (76.4%), including 878 from Bangladesh, 838 from Pakistan, and 1,349 from Tanzania, were assessed for cognitive, motor, and language development at the age of two. The flow of the participants in the AMANHI ACT cohort is displayed in Fig. 1. Of the 3,065 children eligible for analysis, 571 (18.6%) were excluded from the cognitive score model, 591 (19.3%) from the language score model, and 512 (16.7%) from the motor score model due to missing data. Supplementary tables S2-S4 compare the baseline characteristics of participants excluded versus included in the models.

Fig. 1.

Flow of study participants

The detailed distribution of demographic, maternal, and child characteristics across the three study sites is presented in Table 1. In Bangladesh, 20% of the pregnant women were under 20 years of age, which was higher than in Pakistan (7.4%) and Tanzania (4.3%). The majority of women across all sites were between 20 and 29 years old, although Tanzania had a higher proportion of women aged ≥ 30 years (44.0%). Nearly half of the women in both Bangladesh and Tanzania had completed secondary education, but considerably lower in Pakistan (25.8%), where nearly half the mothers (47.6%) had no formal education. In all three sites, more than half of the women had a BMI of 18.5 or higher. While over 30% of women in Bangladesh were underweight, this was less common in Pakistan (22.6%) and Tanzania (5.2%). Paternal education showed a similar trend, with over half of fathers in Pakistan (52.7%) reporting no education, compared to just 12.8% and 12.5% in Bangladesh and Tanzania, respectively.

Table 1.

Sociodemographic characteristics of the study participants

	Bangladesh	Pakistan	Tanzania	Total
Socioeconomic status at enrollment	N = 878	N = 838	N = 1349	N = 3065
Poorest	171 (19.5%)	168 (20.0%)	286 (21.2%)	625 (20.3%)
Poor	178 (20.3%)	168 (20.1%)	277 (20.5%)	623 (20.3%)
Middle	180 (20.5%)	167 (19.9%)	274 (20.3%)	621 (20.3%)
Rich	173 (19.7%)	168 (20.0%)	249 (18.5%)	590 (20.0%)
Richest	176 (20.0%)	167 (20.0%)	263 (19.5%)	606 (20.0%)
Household hunger status*
No or little hunger	838 (95.4%)	641 (76.5%)	1,241 (92.0%)	2,720 (88.7%)
Moderate to severe hunger	40 (4.6%)	197 (23.5%)	108 (8.0%)	345 (11.3%)
Father’s education
No education	112 (12.8%)	442 (52.7%)	169 (12.5%)	723 (23.6%)
1–6	480 (54.7%)	133 (15.9%)	400 (29.7%)	1,013 (33.1%)
7–10	212 (24.1%)	219 (26.1%)	743 (55.1%)	1,174 (38.3%)
≥ 11	74 (8.4%)	44 (5.3%)	36 (2.7%)	154 (5.0%)
Maternal education
No education	33 (3.8%)	399 (47.6%)	196 (14.5%)	628 (20.5%)
1–6 years	331 (37.7%)	193 (23.0%)	439 (32.6%)	963 (31.4%)
7–10 years	454 (51.7%)	216 (25.8%)	683 (50.7%)	1,353 (44.2%)
≥ 11 years	60 (6.8%)	30 (3.6%)	30 (2.2%)	120 (3.9%)
Maternal age at enrollment
< 20 years	178 (20.3%)	62 (7.4%)	58 (4.3%)	298 (9.7%)
20–29 years	601 (68.5%)	544 (64.9%)	698 (51.7%)	1,843 (60.1%)
≥ 30 years	99 (11.3%)	232 (27.7%)	593 (44.0%)	924 (30.1%)
Maternal Body Mass Index (BMI) at enrollment	N = 878	N = 838	N = 1299	N = 3015
Underweight (BMI < 18.5 kg/m2)	267 (30.4%)	189 (22.6%)	68 (5.2%)	524 (17.4%)
Normal weight (BMI 18.5–24.9 kg/m2)	549 (62.5%)	414 (49.4%)	671 (51.7%)	1,634 (54.2%)
Overweight (BMI 25.0–29.9 kg/m2)	54 (6.2%)	172 (20.5%)	332 (25.6%)	558 (18.5%)
Obesity (BMI ≥ 30 kg/m2)	8 (0.9%)	63 (7.5%)	228 (17.6%)	299 (9.9%)
Parity
≤ 2	737 (83.94%)	583 (69.57%)	503 (37.31%)	1823 (59.5%)
3–5	116 (13.21%)	167 (19.93%)	342 (25.37%)	625 (20.4%)
> 5	25 (2.85%)	88 (10.5%)	503 (37.31%)	616 (20.1%)
Maternal morbidity index†
0	808 (92.0%)	440 (52.5%)	1,076 (79.8%)	2,324 (75.8%)
1	58 (6.6%)	276 (32.9%)	258 (19.1%)	592 (19.3%)
≥ 2	12 (1.4%)	122 (14.6%)	15 (1.1%)	149 (4.9%)
Child sex
Male	426 (48.5%)	409 (48.8%)	685 (50.8%)	1,520 (49.6%)
Female	452 (51.5%)	429 (51.2%)	664 (49.2%)	1,545 (50.4%)
Low birthweight (< 2500 g)	N = 786	N = 798	N = 1349	N = 2933
No	547 (69.6%)	614 (76.9%)	1271 (94.2%)	2432 (82.9%)
Yes	239 (30.4%)	184 (23.1%)	78 (5.8%)	501 (17.1%)
Gestational age at birth
Term (≥ 37 weeks)	781 (88.95%)	724 (86.4%)	1,247 (92.4%)	2,752 (88.8%)
Preterm (< 37 weeks)	97 (11.05%)	114 (13.6%)	102 (7.6%)	313 (10.2%)
Exclusive breastfeeding (till 3 months)	N = 854	N = 786	N = 1341	N = 2981
Yes	607 (71.1%)	324 (41.2%)	985 (73.4%)	1916 (64.3%)
No	247 (28.9%)	462 (58.8%)	356 (26.5%)	1065 (35.7%)
Child vaccination status at 2 years ‡	N = 862	N = 837	N = 1315	N = 3014
Unvaccinated	0 (0%)	20 (2.4%)	25 (1.9%)	45 (1.49%)
Partially vaccinated	44 (5.1%)	261 (31.2%)	269 (20.5%)	574 (19.0%)
Fully vaccinated	818 (94.9%)	556 (66.4%)	1021 (77.6%)	2395 (79.5%)
Moderate to severe wasting (WHZ < − 2SD) in children at age
3 months	95 (12.6%)	73 (9.4%)	87 (7.0%)	255 (9.2%)
6 months	74 (9.8%)	98 (12.3%)	50 (3.9%)	222 (7.9%)
12 months	88 (11.6%)	162 (20.1%)	49 (3.9%)	299 (10.6%)
18 months	81 (10.7%)	134 (16.9%)	52 (4.0%)	267 (9.4%)
24 months	76 (9.8%)	103 (12.7%)	37 (2.8%)	266 (7.5%)
Moderate to severe stunting (HAZ < − 2SD) in children at age
3 months	135 (17.9%)	192 (24.7%)	199 (15.9%)	526 (18.9%)
6 months	147 (19.4%)	171 (21.5%)	261 (20.5%)	579 (20.5%)
12 months	218 (28.8%)	213 (26.5%)	417 (33.0%)	848 (30.0%)
18 months	401 (52.9%)	384 (48.3%)	576 (44.9%)	1,361 (48.0%)
24 months	432 (55.6%)	475 (58.9%)	596 (45.8%)	1503 (52.1%)
Moderate to severely underweight (WAZ < − 2SD) children at age
3 months	197 (25.9%)	211 (26.9%)	141 (11.3%)	549 (19.6%)
6 months	175 (22.9%)	209 (26.2%)	120 (9.4%)	504 (17.8%)
12 months	200 (26.0%)	238 (29.5%)	129 (10.2%)	567 (20.0%)
18 months	261 (33.9%)	314 (38.7%)	157 (12.2%)	732 (25.5%)
24 months	295 (37.9%)	345 (42.1%)	148 (11.3%)	788 (27.1%)
Diarrhea episodes reported by caregiver during study visits
0	42 (4.8%)	169 (20.2%)	1037 (76.9%)	1248 (40.7%)
1	117 (13.3%)	161 (19.2%)	187 (13.8%)	465 (15.2%)
≥ 2	719 (81.9%)	508 (60.6%)	125 (9.3%)	1352 (44.1%)
Fever episodes as per caregiver recall
0	621 (70.7%)	44 (5.2%)	773 (57.3%)	1438 (46.9%)
1	200 (22.8%)	108 (12.9%)	380 (28.2%)	688 (22.4%)
≥ 2	57 (6.5%)	686 (81.9%)	196 (14.5%)	939 (30.6%)
Hospital admissions reported during study §
0	644 (73.3%)	744 (88.8%)	1,303 (96.6%)	2,691 (87.8%)
1–5	204 (23.2%)	78 (9.3%)	24 (1.8%)	306 (10.0%)
≥ 6	30 (3.4%)	16 (1.9%)	22 (1.6%)	68 (2.2%)
Child home environment measured by HOME-Inventory-at 24 months	N = 870	N = 646	N = 1344	N = 2860
Emotional and verbal responsiveness of caregiver (score ranges 0–11)	10.16 ± 1.02	9.99 ± 0.82	10.70 ± 0.89	10.37 ± 0.97
Avoidance of restriction and Punishment (score ranges 0–5)	4.35 ± 0.51	4.20 ± 0.59	4.58 ± 0.86	4.43 ± 0.73
Caregiver promotes child development (score ranges 0–5)	4.99 ± 0.11	3.90 ± 0.56	4.82 ± 0.56	4.66 ± 0.63
Organization of physical and temporal environment (score ranges 0–12)	10.52 ± 1.36	8.52 ± 1.78	10.34 ± 1.67	9.98 ± 1.79
Provision of appropriate play materials (score ranges 0–3)	2.25 ± 0.64	1.52 ± 0.64	1.82 ± 1.31	1.88 ± 1.04
Opportunities for variety in daily stimulation (score ranges 0–9)	7.59 ± 1.20	7.22 ± 1.15	7.90 ± 1.16	7.65 ± 1.20
Cleanliness of the child (score ranges 0–3)	3.00 ± 0.00	2.84 ± 0.66	2.81 ± 0.70	2.87 ± 0.58
Total home score (score ranges 0–48)	42.85 ± 2.80	38.19 ± 3.36	42.98 ± 4.45	41.86 ± 4.26

WAZ: weight-for‐age; HAZ: height‐for‐age and WHZ: weight‐for‐height Data presented as mean ± SD and n (%)

* Household went hungry (moderate to severe hunger on FANTA Household Hunger Scale) at any point during the last two years

† Number of complications: Hypertension, pre-eclampsia, eclampsia, antepartum infection, postpartum infection, postpartum hemorrhage

‡ BCG, OPV1, OPV2, OPV3, PENTA1, PENTA2, PENTA3, and Measles

§ Hospital admission is defined as an overnight stay at the hospital

The gender distribution among the children was nearly equal, with 49.6% males and 50.4% females. Bangladesh and Pakistan had higher proportions of preterm births and low birth weight than Tanzania. Exclusive breastfeeding rates at 3 months were highest in Tanzania (73.4%) and Bangladesh (71.1%), and lowest in Pakistan (41.2%). Household food insecurity was high in Pakistan, where 23.5% of participants reported moderate to severe hunger, in contrast to 4.6% in Bangladesh and 8.0% in Tanzania. By 2 years of age, the proportion of fully vaccinated children was highest in Bangladesh (94.9%), followed by Tanzania (77.6%), and lowest in Pakistan (66.4%). Moderate to severe wasting at 12 months was most common in Pakistan (20.1%), followed by Bangladesh (11.6%) and Tanzania (3.9%). By 24 months, moderate to severe stunting affected more than half the children in Bangladesh (55.6%) and Pakistan (58.9%), while still impacting 45.8% in Tanzania. Similarly, underweight prevalence at 24 months was highest in Pakistan (42.1%) and Bangladesh (37.9%) versus 11.3% in Tanzania.

As per maternal recall, Bangladesh reported the highest prevalence of diarrhea (82%) and fever episodes (30.6%) in children under two. Parity was higher in Tanzania, where over 62% of women had three or more children, compared to just 16% in Bangladesh. Maternal morbidity was highest in Pakistan, with 47.5% of women reporting one or more morbidities, including 14.6% with two or more.

The mean Bayley-III composite score for cognitive development across all three sites was 84.8, while the mean composite scores for the language and motor domains were 87.8 and 90.8, respectively. Pakistan had comparatively lower mean scores for both cognitive (76.6 ± 10.8) and language (79.8 ± 11.39) development, whereas the mean motor development scores were similar for all three sites (Table 2).

Table 2.

Distribution of composite development scores across sites

Domains	Measures	Bangladesh	Pakistan	Tanzania	Total
Cognitive	Overall Mean (SD), n	88.44 (6.91), 878	76.69 (10.80), 838	87.60 (9.72), 1349	84.86 (9.34), 3065
	Median (IQR)	90 (85–95)	80 (70–85)	90 (80–95)	85 (80–90)
Language	Overall Mean (SD), n	91.26 (8.68), 878	79.86 (11.39), 838	88.82 (8.94), 1349	87.07 (9.6), 3065
	Median (IQR)	91 (86–97)	79 (74–86)	89 (83–97)	89 (79–94)
Motor	Overall Mean (SD), n	93.01 (6.72), 878	87.25 (11.51), 838	91.69 (10.06), 1349	90.86 (9.68), 3065
	Median (IQR)	94 (88–97)	88 (82–94)	91 (85–97)	91 (85–97)

Cognitive development at 24 months

There was strong evidence of positive association between cognitive scores and maternal obesity, child’s home environment, particularly the ‘Provision of appropriate play materials’, ‘Organization of physical and temporal environment’, and ‘Child Cleanliness’ domains.

Children born to mothers with no education or only primary and secondary education levels had lower cognitive scores compared to those with 10 or more years of maternal education. Cognitive scores were negatively associated with preterm birth, unvaccinated status, stunting at 18 months, wasting at 6 months, and household hunger. (Table 3)

Table 3.

Factors associated with cognitive development

	Regression coefficients for cognitive scores (95% CI) (N = 2,494)	P-value
Gestational age at birth
Term (≥ 37 weeks)	Ref	Ref
Preterm (< 37 weeks)	−1.44 (−2.62, −0.27)	0.016
Child vaccination status
Unvaccinated	−6.02 (−9.36, −2.67)	< 0.001
Partially vaccinated	−1.17 (−2.12, −0.22)	0.016
Fully vaccinated	Ref	Ref
Child nutritional status
WHZ at 6 months
Normal	Ref	Ref
Moderate to severe wasting (WHZ < − 2SD)	−1.69 (−3.01, −0.36)	0.012
HAZ at 18 months
Normal	Ref	Ref
Moderate to severe stunting (HAZ < − 2SD)	−2.19 (−2.91, −1.48)	< 0.001
Maternal education
No education	−3.87 (−5.84, −1.9)	< 0.001
1–6 years	−3.46 (−5.32, −1.6)	< 0.001
7–10 years	−2.12 (−3.94, −0.3)	0.022
≥ 11 years	Ref	Ref
Maternal BMI at Enrollment
Underweight (BMI < 18.5 kg/m2)	−0.89 (−1.88, 0.11)	0.081
Normal weight (BMI 18.5–24.9 kg/m2)	Ref	Ref
Overweight (BMI 25.0–29.9 kg/m2)	0.75 (−0.2, 1.69)	0.12
Obesity (BMI ≥ 30 kg/m2)	1.75 (0.55, 2.96)	0.004
Household Hunger
No or little hunger	Ref	Ref
Moderate to severe hunger	−1.64 (−2.81, −0.47)	0.006
Child home environment measured by the HOME-Inventory at 24 months
Organization of physical and temporal environment	0.47 (0.23, 0.71)	< 0.001
Provision of appropriate play materials	0.86 (0.49, 1.23)	< 0.001
Cleanliness of the child	1.23 (0.6, 1.86)	< 0.001

CI Confidence Interval, WHZ Weight-for-Height Z-score, HAZ Height-for-Age Z-score, BMI Body Mass Index

Adjusted regression coefficients with 95% confidence intervals (CIs) represent the adjusted effect of each predictor on child development scores (i.e., cognitive, language, and motor) domains. The models were adjusted for categorical variables, socioeconomic status, household hunger, maternal age, BMI, and education, parity, maternal morbidity, child sex, birthweight, gestational age, nutritional and immunization history, child anthropometrics at 3, 6, 12, 18, and 24 months, infections in the past two years, and HOME-IT domains as a continuous variable.

Language development at 24 months

The multivariable model showed strong evidence of a positive association between language scores with maternal BMI during pregnancy and home environment domains of ‘Organization of physical and temporal environment’.

Children whose mothers had no formal education or education only up to primary or secondary levels had lower language scores compared to those whose mothers had more than 11 years of education. Similarly, no paternal education or only till primary level was negatively associated with language scores.

Underweight at 12 months and 24 months, stunting at 6 months, partial vaccination status, maternal reports of two or more fever and diarrhea episodes in children were also negatively associated with language scores. (Table 4)

Table 4.

Factors associated with Language development

	Regression coefficients for language scores (95% CI) (N = 2,474)	P-value
Child sex
Male	Ref	Ref
Female	1.01 (0.29, 1.72)	0.006
Gestational age at birth
Term (≥ 37 weeks)	Ref	Ref
Preterm (< 37 weeks)	−1.77 (−3, −0.53)	0.005
Child vaccination status
Unvaccinated	−2.52 (−5.87, 0.83)	0.14
Partially vaccinated	−1.18 (−2.18, −0.19)	0.02
Fully vaccinated	Ref	Ref
Child nutritional status
WAZ at 12 months
Normal	Ref	Ref
Moderate to severe underweight (WAZ < − 2SD)	−1.48 (−2.64, −0.32)	0.012
WAZ at 24 months
Normal	Ref	Ref
Moderate to severe underweight (WAZ < − 2SD)	−1.65 (−2.68, −0.62)	0.002
HAZ at 6 months
Normal	Ref	Ref
Moderate to severe stunting (HAZ < − 2SD)	−0.93 (−1.9, 0.03)	0.057
History of fever reported by mother during study visits
No	Ref	Ref
Yes	−2.45 (−3.57, −1.34)	< 0.001
History of diarrhea reported by mother during study visits
No	Ref	Ref
Yes	−1.11 (−2.11, −0.12)	0.028
Maternal education
No education	−4.69 (−6.82, −2.57)	< 0.001
1–6 years	−4.5 (−6.48, −2.52)	< 0.001
7–10 years	−3.39 (−5.3, −1.48)	0.001
≥ 11 years	Ref	Ref
Father’s education
No education	−2.34 (−4.2, −0.48)	0.014
1–6 years	−2.32 (−4.08, −0.57)	0.01
7–10 years	−1.32 (−3.06, 0.42)	0.137
≥ 11 years	Ref	Ref
Maternal BMI at enrollment
Underweight (BMI < 18.5 kg/m2)	−1.31 (−2.34, −0.27)	0.013
Normal weight (BMI 18.5–24.9 kg/m2)	Ref	Ref
Overweight (BMI 25.0–29.9 kg/m2)	0.54 (−0.45, 1.52)	0.284
Obesity (BMI ≥ 30 kg/m2)	1.68 (0.42, 2.95)	0.009
Child home environment measured by the HOME-Inventory at 24 months
Organization of physical and temporal environment	0.53 (0.3, 0.76)	< 0.001

CI Confidence Interval, WHZ Weight-for-Height Z-score, HAZ Height-for-Age Z-score, BMI Body Mass Index

Adjusted regression coefficients with 95% confidence intervals (CIs) represent the adjusted effect of each predictor on child development scores (i.e., cognitive, language, and motor) domains. The models were adjusted for categorical variables, socioeconomic status, household hunger, maternal age, BMI, and education, parity, maternal morbidity, child sex, birthweight, gestational age, nutritional and immunization history, child anthropometrics at 3, 6, 12, 18, and 24 months, infections in the past two years, and HOME-IT domains as a continuous variable.

Motor development at 24 months

Females scored 1.64 points lower than males for motor development. Motor development was positively associated with the HOME inventory tool domains-'Provision of appropriate play materials’, ‘Organized physical and temporal environment’, and ‘Child cleanliness’. Lower motor scores were associated with underweight status at 6 months, stunting and wasting at 24 months, unvaccinated status, hospitalization exceeding two days in the past two years, and household hunger. (Table 5)

Table 5.

Factors associated with motor development

	Regression coefficients for motor scores (95% CI) (N = 2,553)	P-value
Child sex
Male	Ref	Ref
Female	−1.64 (−2.34, −0.93)	< 0.001
Child vaccination status
Unvaccinated	−4.72 (−8.01, −1.43)	0.005
Partially vaccinated	−1 (−1.97, −0.03)	0.042
Fully vaccinated	Ref	Ref
Child nutritional status
HAZ at 24 months
Normal	Ref	Ref
Moderate to severe stunting (HAZ < − 2SD)	−2.22 (−2.96, −1.48)	< 0.001
WHZ at 24 months
Normal	Ref	Ref
Moderate to severe wasting (WHZ < − 2SD)	−2.86 (−4.31, −1.42)	< 0.001
WAZ at 6 months
Normal	Ref	Ref
Moderate to severe underweight (WAZ < − 2SD)	−1.92 (−2.94, −0.9)	< 0.001
Hospitalization during the study period
No	Ref	Ref
Yes	−1.3 (−2.46, −0.15)	0.027
Household Hunger
No or little hunger	Ref	Ref
Moderate to severe hunger	−1.74 (−2.94, −0.54)	0.005
Child home environment measured by the HOME-Inventory tool at 24 months
Organization of physical and temporal environment	0.35 (0.1, 0.59)	0.005
Provision of appropriate play materials	0.63 (0.26, 1.01)	0.001
Cleanliness of the child	0.85 (0.2, 1.51)	0.01

CI Confidence Interval WHZ Weight-for-Height Z-score HAZ Height-for-Age Z-score BMI Body Mass Index

Adjusted regression coefficients with 95% confidence intervals (CIs) represent the adjusted effect of each predictor on child development scores (i.e., cognitive, language, and motor) domains. The models were adjusted for categorical variables, socioeconomic status, household hunger, maternal age, BMI, and education, parity, maternal morbidity, child sex, birthweight, gestational age, nutritional and immunization history, child anthropometrics at 3, 6, 12, 18, and 24 months, infections in the past two years, and HOME-IT domains as a continuous variable.

Discussion

This multi-country cohort study identified maternal, child, and household-level factors associated with child development Bangladesh, Pakistan, and Tanzania. We found that preterm birth, undernutrition, household food insecurity, low parental education, incomplete vaccination, and limited home stimulation were associated with poor developmental outcomes at two years of age. These findings are consistent with a growing body of evidence that highlights the complex, multisectoral determinants of Early Childhood Development (ECD) in LMICs. Preterm birth was negatively associated with cognitive and motor domains but not the language domain. This aligns with neurodevelopmental studies suggesting that late-gestation myelination and hippocampal immaturity impair processing speed, coordination, and memory later in life [15]. However, language neural circuits retain prolonged plasticity, which could adapt when supported by enriched verbal environments postnatally [16].

Optimum nutritional intake reflected by physical growth indicators, was positively associated with developmental outcomes across cognitive, motor, and language domains. Maternal BMI during enrolment showed an association with only cognitive and language development. Previous evidence suggests that both maternal undernutrition and obesity can affect fetal neurodevelopment via placental nutrient transfer and inflammatory pathways, respectively [17, 18]. Interestingly, maternal BMI was not associated with motor development, suggesting that postnatal environment and stimulation may play a more critical role in shaping gross motor trajectories [19]. Household hunger was associated with lower cognitive and motor scores but not language development. This may reflect the dual impact of nutritional deficits and caregiver stress, the latter potentially disrupting responsive caregiving and learning opportunities. These findings are in line with previous studies linking food insecurity to impaired developmental outcomes through both biological and psychosocial pathways [20].

Maternal education was a predictor of improved cognitive and language development. Carneiro et al. report that each additional year of maternal education enhances supportive and stimulating environments at home [21]. Consistent with other studies, we saw no association of parental education with motor milestones, suggesting that genetic and physical environmental factors more strongly influence motor milestones [22, 23]. Similarly, we found an organized physical and temporal home environment to be positively associated with child development. While the availability of age-appropriate play materials is known to boost fine-motor skills and cognitive scores, they do not directly affect verbal interactions needed for phonemic development, which may explain our lack of correlation with language scores.

Frequent illness and incomplete vaccination were associated with poorer developmental outcomes, emphasizing the role of preventive healthcare in supporting early neurodevelopment. Hospitalizations were particularly associated with motor delays, likely due to restricted mobility and limited rehabilitative services in LMIC hospital settings [24].

Country-level differences in developmental scores corresponded to each setting’s nutritional, health, and educational indicators. In Pakistan, where nearly half of the mothers had no formal education, only 66% received complete vaccination, one-third of the households experienced food insecurity, and undernutrition remained prevalent, the mean cognitive, language, and motor scores were lowest among the three sites. In contrast, Bangladesh reported the highest vaccine coverage, a higher maternal secondary education attainment, coinciding with the highest mean language scores and better cognitive and motor outcomes. However, persistent undernutrition across all sites remains a significant developmental risk.,Our study identified several factors that influence early childhood development. It is imperative to promote holistic Early Childhood Development (ECD) through policies that integrate nutrition, education, and caregiver support to prevent developmental delays, especially in disadvantaged communities. Previous research has shown that programs that target psychosocial stimulation and health interventions for children under 3 years in low-resource settings are key to improving developmental outcomes [25]. In addition, national action plans must aim at expanding access to the provision of quality early childhood education (ECE) and addressing socio-economic barriers to enhance children’s cognitive and social-emotional growth [26].

Our study’s strength lies in its extensive prospective cohort, which includes women and children from three LMICs. Well-trained study workers determined gestational age using the gold standard ultrasound. However, there were limitations. Data on illness episodes and clinical history were largely based on caregiver recall, introducing potential bias. We performed a complete-case analysis to handle missing data, which may have introduced selection bias by disproportionately excluding children at the highest nutritional and socioeconomic risk, potentially leading to an underestimation of the observed associations. We used Pearson correlation to assess interrater agreement, which does not reflect true agreement. While malaria is nearing eradication on Pemba Island, where our research site is located, we acknowledge that this factor was not assessed in our study. The prevalence of malaria may be low in all three study sites, and considering its potential impact on developmental outcomes could still be relevant for future research.

Conclusion

This study shows a broad range of modifiable factors that significantly influence child development in resource-limited settings. The results reinforce the urgency of integrating nutrition, education, and caregiver support into national ECD strategies. Future interventions must be context-specific and equity-driven to ensure all children have the opportunity to reach their developmental potential.

Author contributions

The study was conceptualized and designed by principal investigators of the three sites (AB (Bangladesh), FJ (Pakistan), and SS (Tanzania)), AM, YS, and RB. All authors from three sites (Bangladesh, Pakistan, and Tanzania) conducted the acquisition of data. MIN, FJ and AN wrote the first draft of the manuscript that was reviewed by all authors. All authors read and approved the final manuscript. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Funding

This work was supported by the Bill & Melinda Gates Foundation through a grant to the World Health Organization, Grant Number (I64438). The funders have played no role in the drafting of the manuscript and the decision to submit it for publication.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval and consent to participate

The AMANHI study received ethical approval from the local and institutional ethics committees of all three sites. These included Zanzibar Health Research Ethics Committee (formerly ZAMREC) (ZAMREC/0002/OCTOBER/013) for Tanzania, ICDDR, B (PR12073), and Johns Hopkins University (IRB 00004508) for Bangladesh, and Aga Khan University (2790-paeds-ERC-13) for Pakistan. In addition, the protocols for the biorepository study were also approved by the WHO Ethics Committee (RPC 532), and continuing approvals were sought yearly. Informed consent was obtained from pregnant women at the time of enrollment, and later, at childbirth, they consented to their child’s participation in the study. Consent forms were provided in the mothers’ preferred languages to ensure comprehension. In cases where mothers expressed concerns about their child’s development, or where the child showed faltered physical growth or developmental delays, referrals were made to the Outpatient Department (OPD) of the study center. Physicians at the OPD then assessed the child and, if necessary, referred them to nearby hospitals for further evaluation and management.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.