Factors associated with overweight and obesity among school children aged 6-12 years in the South Tongu District of Volta Region, Ghana

¹Department of Educational Leadership, College of Health and Society, Miami University, USA, ²Dodowa Health Research Center, Ghana Health Service, Dodowa, Greater Accra, Ghana

^&Corresponding author
Peter Kofi Dabie, Department of Educational Leadership, College of Health and Society, Miami University, USA

Introduction    

Overweight and obesity in childhood are defined as abnormal or excessive fat accumulation that presents a risk to health, commonly assessed using Body Mass Index (BMI)-for-age Z-scores based on World Health Organization (WHO) reference standards [1]. Globally, the prevalence of overweight, including obesity, among children and adolescents aged 5-19 years rose from 8% in 1990 to 20% in 2022, when more than 390 million children and adolescents in this age group were overweight, and 160 million were living with obesity [2]. Obesity in this age group has quadrupled since 1990, and the steepest increases are now occurring in low- and middle-income countries (LMICs) undergoing rapid nutrition transition [3]. In sub-Saharan Africa (SSA), the number of overweight or obese children nearly doubled from 5.4 million in 1990 to 10.6 million in 2014, signaling a rapid nutritional transition that now threatens LMICs alongside existing undernutrition burdens [4].

The health consequences of childhood obesity have been well documented. Children with excess weight are at an increased risk of developing hypertension, dyslipidemia, insulin resistance, type 2 diabetes, and early markers of cardiovascular disease (CVD) [4,5]. Childhood overweight is also strongly predictive of adult obesity, making early prevention essential for reducing the burden of non-communicable diseases (NCDs) in later life [6]. Beyond physical health, obesity affects children's psychological well-being, educational attainment, and quality of life [5].

In Ghana, the prevalence of childhood overweight and obesity varies considerably by geographic and socioeconomic contexts. A scoping review of 24 studies covering 23,663 in-school children and adolescents aged 5-19 years in Ghana reported overweight/obesity prevalence estimates ranging from 0.5% to 47.1%, with girls consistently more affected than boys [7]. Studies in urban centers such as Accra and Kumasi have reported combined overweight/obesity rates as high as 17% [8], and 16.4% has been reported among children aged 8-11 years in the Adentan Municipality of Greater Accra [9]. In comparison, semi-rural and rural districts tend to report lower estimates ranging from 3% to 10.9% [7,10]. This pediatric burden is emerging in a country where excess weight is already common among adults: the 2022 Ghana Demographic and Health Survey found that more than 40% of women of reproductive age were overweight or obese, so a large proportion of Ghanaian children now grow up in households where excess weight is the norm [11]. However, there is growing evidence that peri-urban and rapidly urbanizing districts are experiencing a rise in childhood obesity driven by dietary shifts, sedentary lifestyles, and changing household socioeconomic dynamics [7,9].

Despite this emerging burden, evidence on the determinants of overweight and obesity among school-age children in the Volta Region remains sparse, and peri-urban and semi-rural districts have received far less attention than the metropolitan centers of Accra and Kumasi [10,11]. The South Tongu District illustrates this gap. Located in the southern part of the Volta Region, with a population of approximately 102,000, the district is undergoing rapid peri-urbanization along the Accra-Aflao Road corridor, combines farming with a substantial salaried civil service, and has growing access to markets that supply processed and energy-dense foods, conditions under which childhood weight gain may escalate undetected. District-level evidence on BMI-for-age and its determinants among primary school children remains limited, so neither the size of the problem nor the exposures that drive it are well characterized for the district health directorate. In particular, it is not known whether the exposures implicated elsewhere in Ghana operate in this setting: maternal and paternal employment and the caregiving and food-provisioning arrangements that follow from it; reliance on food bought from school vendors rather than prepared at home, and the frequency of such purchases; participation in the school feeding program; low fruit and vegetable intake; and low levels of physical activity and active transport. Each of these exposures is potentially modifiable by the school, the household or the district, yet none has been examined against BMI-for-age in this population [12].

Generating such evidence is timely because Ghana already possesses the policy instruments that district-level data could direct. The National Policy for the Prevention and Control of Chronic Non-Communicable Diseases (NCDs) commits the health sector to primary prevention of diet-related NCDs across the life course [13]; the Excise Duty (Amendment) Act, 2023 introduced a 20% excise tax on sweetened beverages as the country's first major fiscal measure against unhealthy diets [14]. The 2023 Ghana STEPS survey has renewed national attention on modifiable NCD risk factors. Within schools, the Ghana School Feeding Program and the School Health Education Program provide delivery platforms that reach public primary schools in districts such as South Tongu, and the Ghana Education Service regulates the vendors permitted to trade on school premises [15]. What these instruments lack at district level is evidence on which exposures to prioritize and which children to target; without it, public health responses risk being misaligned with the dietary, behavioral, and sociodemographic determinants that operate locally [16].

This study therefore aimed to identify the factors associated with overweight and obesity among school children aged 6-12 years in the South Tongu District of the Volta Region. Specifically, it sought: (i) to determine the prevalence of overweight and obesity; (ii) to describe the sociodemographic, dietary and physical activity characteristics of the children; (iii) to examine the association between sociodemographic characteristics and overweight/obesity; (iv) to examine the association between dietary and physical activity factors and overweight/obesity; and (v) to identify the factors independently associated with overweight and obesity after mutual adjustment [17]. The null hypothesis, tested at the 5% level of significance, was that none of the sociodemographic exposures (maternal and paternal occupation, parental education, household size and the number of children in the household), dietary exposures (participation in the school feeding program, buying food at school and its frequency, meal frequency, snack and sweet intake, and fruit and vegetable intake) or physical activity exposures (participation in sport, mode of transport to school and leisure-time activity) is associated with BMI-for-age status. By generating locally grounded evidence, the study is intended to inform school-based nutrition education, the regulation of school food vendors and the inclusion of childhood obesity in district health action plans, while contributing district-level data to the national evidence base on the double burden of malnutrition.

Methods     

Study setting, design and population: an analytic cross-sectional study design was used. The study was conducted in the South Tongu District of the Volta Region of Ghana, which is divided into six sub-districts: Sogakope, Dorkploame, Sotewu, Dabala-Adutor, Dordoekope, and Agorta-Gamenu. The district has 111 public primary schools spread across the sub-districts. Data collection was conducted from July to August 2019. The study was restricted to public primary schools because they enroll the majority of primary school-age children in the district and are the schools covered by the Ghana School Feeding Program and the School Health Education Program, which are the platforms through which any resulting intervention would be delivered; private schools were not sampled, and the implications of this are addressed in the Limitations. The interval between data collection in 2019 and publication is acknowledged, and its bearing on the current applicability of the findings is also addressed in the limitations.

The study population comprised children aged 6-1 years who were enrolled in public primary schools in the South Tongu District. Children were eligible if they were: (i) aged 6-12 years; (ii) attending a public school in the district; (iii) had written assent and parental and headmaster consent to participate; and (iv) were not ill at the time of the study.

Sampling and sample size: the minimum sample size was estimated using the formula:

Based on a prevalence of 29.4% [18], a 95% confidence interval (Z = 1.96), and a margin of error of 5%. In this formula, n is the minimum required sample size; Z is the standard normal deviate corresponding to the chosen confidence level (1.96 for a 95% confidence level); P is the expected prevalence of overweight/obesity among school children aged 6-12 years, expressed as a proportion (0.294); and e is the margin of error, i.e., the desired level of absolute precision (0.05). This yielded a minimum sample of 319, which was adjusted upward by 10% to accommodate nonresponse, giving a final minimum sample size of 351. The 29.4% figure is the national overweight/obesity prevalence reported for Ghana [18]; no district- or region-specific estimate for school-age children was available at the time of the study.

A multistage sampling approach was employed. In the first stage, the district's six sub-districts served as clusters, and sample sizes were allocated proportionately to the sub-district population size (Table 1). In the second stage, two public primary schools were selected from each sub-district by simple random sampling: a sampling frame of the public primary schools in the sub-district was obtained from the South Tongu District Education Directorate, each school on the list was assigned a number, and two numbers were drawn per sub-district by lottery without replacement. Systematic selection was not used at this stage because only two schools were drawn from each list, a situation in which a sampling interval offers no advantage over simple random selection. In the third stage, pupils were selected within each sampled school by systematic sampling from the class registers, which are arranged alphabetically: the sampling interval k was obtained by dividing the number of eligible pupils in the school by the number of participants allocated to that school, a random start was drawn between 1 and k, and every kth pupil was selected until the school's allocation was reached. Systematic sampling was appropriate at this stage because the number of eligible pupils per school was large.

Data collection and measurements: a structured questionnaire was used to collect data on sociodemographic characteristics (age, sex, parental occupation and education, household size, religion), dietary habits (school feeding status, frequency of buying food in school, frequency of meals per day, snack intake, sweet intake, fruit and vegetable consumption), physical activity patterns (sporting activities, means of transport to school, leisure time activities), and a 24-hour dietary recall. The questionnaire was pre-tested on 10 children from a community outside the study area to assess its clarity and comprehension. A trained local interpreter was available for children who did not understand English.

Anthropometric measurements were taken by trained data collectors using a calibrated bathroom scale (for weight in kilograms) and a stadiometer (for height in meters). BMI was calculated as weight (kg) divided by height (m²). BMI-for-age Z-scores were computed using WHO reference standards; children with Z-scores between +1 and +2 SD were classified as overweight, and those with Z-scores >+2 SD were classified as obese [1].

The outcome of interest was BMI-for-age status, dichotomized as overweight/obese versus not overweight. Exposures were considered at two levels. Distal, household-level exposures comprised maternal and paternal occupation, parental education, household size, and the number of children in the household. These were expected to act on a child's weight indirectly, through the money available for food purchases, the time available for home food preparation and for supervision after school, and the food environment of the home. Proximal, child-level exposures comprised the behaviors through which those household conditions are expressed: participation in the school feeding program, buying food from school vendors and the frequency of doing so, meal frequency, snack and sweet intake, fruit and vegetable intake, participation in sport, mode of transport to school and leisure-time activity. Because exposures and outcomes were measured at the same point in time, the design supports statements of association only. Temporality cannot be established and reverse causation - for example, a change in a child's food purchasing following weight gain - cannot be excluded. Hence, the factors reported here are risk factors in the sense of measured associations rather than demonstrated causes.

Selection bias was limited by probability sampling at every stage and by including all six sub-districts with allocation proportional to population. Information and recall bias in the dietary and activity data were limited by pre-testing the questionnaire on children of the same age outside the study area, by asking about a short and recent reference period rather than about usual intake over months, by supplementing the frequency questions with a 24-hour dietary recall, and by providing a trained local-language interpreter so that questions were understood as intended. Interviewer bias and measurement error in the anthropometric data were limited by training all data collectors in a single standardized procedure and by using calibrated instruments. Residual bias remains, however. Dietary and activity information was reported by children aged 6-12 years and was not corroborated by parents or guardians, so under- and over-reporting are both likely; social desirability may have inflated reported fruit, vegetable and sporting activity; and a Hawthorne effect cannot be excluded, since children knew their weight and height were being measured and may have reported more favorable behaviors as a result. Such misreporting is likely to be non-differential with respect to BMI status, which would tend to bias the observed odds ratios towards the null rather than away from it.

Statistical analysis and ethical considerations: data were entered using Epi Data version 7 and exported to Stata version 14.0 for data cleaning and analysis. Descriptive statistics (frequencies and percentages) were used to summarize the sociodemographic and risk factor variables. Chi-square tests were used to assess bivariate associations between independent variables and BMI status (normal weight versus overweight/obese). Crude odds ratios (COR) and adjusted odds ratios (AOR) with 95% confidence intervals (CI) were estimated using binary logistic regression analysis. Variables with p < 0.20 in the bivariate analyses were entered into the multivariable model. Statistical significance was set at P < 0.05. Questionnaires were checked for completeness in the field at the end of each school visit, and inconsistencies were resolved before the class was released. All analyses were carried out on complete cases, and no values were imputed; the number of participants contributing to each variable is shown in the tables, and a denominator smaller than 351 reflects either a question that applied only to a subgroup - for example, the frequency of buying food per week, which applies only to the 288 children who reported buying food at school - or missing responses to that item. Children whose BMI-for-age Z-score was at or below +1 SD are referred to throughout as normal weight; this group includes any children with thinness, who were not analyzed separately.

Potential confounders were specified in advance from the literature. They comprised the child's age and sex, maternal and paternal occupation, parental education, household size, the number of children in the household, and participation in the school feeding program. These were handled analytically: every variable associated with BMI status at p < 0.20 in bivariate analysis was entered simultaneously into the multivariable logistic regression model, so that each adjusted odds ratio reported is mutually adjusted for the other variables in the model. Variables that did not reach this screening threshold, including age, sex and fruit intake, were not retained, and residual confounding by them cannot be excluded.

Two further sources of confounding were only partly addressed. Children who were ill on the day of data collection were excluded at recruitment, but chronic conditions capable of affecting body weight - endocrine disorders, physical disability limiting activity, or long-term use of medicines such as corticosteroids - were not systematically screened for, and no clinical records were consulted. Household food patterns were captured only through the child's own reported intake and a 24-hour recall; the frequency with which energy-dense or processed foods are purchased for and prepared in the home, which is plausibly associated with both parental occupation and the child's weight, was not measured. Both are potential sources of residual confounding and are acknowledged in the limitations.

Ethical approval was obtained from the Research Ethics Committee of the University of Health and Allied Sciences (UHAS-REC A.518-19). Written consent was obtained from the school principals and parents/guardians, and written assent was obtained from all child participants. Participation was voluntary, and children with nutritional abnormalities received counseling. All data were kept confidential and securely stored.

Results     

Prevalence of overweight and obesity: approximately one in four children carried excess weight. Of the 351 pupils assessed, 82 had a BMI-for-age Z-score above +1 SD, a prevalence of overweight and obesity of 23.4% (95% CI 19.0-28.1); 72 children (20.5%) were overweight, and 10 (2.8%) were obese, while the remaining 269 (76.6%) were classified as normal weight.

Characteristics of participants: participants were drawn from all six sub-districts in proportion to their populations, with Sogakope contributing the largest share (31.1%) (Table 1). Girls slightly outnumbered boys (52.7% versus 47.3%), the mean age was 9.66 ± 1.73 years, and more than half of the children (54.4%) were aged 10-12 years. Trading was the predominant maternal occupation (60.4%), whereas paternal occupations were more evenly distributed, the civil service being the most common (30.5%). Nearly all children were Christian (96.0%), and most lived in households of five to seven persons (58.4%) (Table 2).

Dietary and physical activity profile: buying food at school was close to universal: 82.1% of children bought food from school vendors, and 71.5% of these buyers did so daily. Participation in the Ghana School Feeding Program, under which the school provides a cooked meal on each school day, was reported by 62.7% of children; buying food at school refers to purchases from vendors trading on or immediately outside the school premises, made with money given to the child by a parent or guardian. It is distinct from participation in the feeding program. Most children ate three meals a day (79.8%), but fewer than half (49.3%) consumed vegetables daily, and 36.7% ate sweets every day. Physical activity was common but largely informal: 85.2% of children took part in sporting activities, most commonly 'Ampe' (a traditional Ghanaian game, played mainly by girls, in which players jump, clap and thrust one foot forward in rapid repeated rounds, making it a high-intensity activity; 49.8% of participants in sport) and football (46.2%); 77.8% walked to school; and reading was the most frequently reported leisure-time activity (48.1%) (Table 3).

Sociodemographic factors and overweight/obesity: overweight and obesity were more common among children whose mothers were civil servants: 17 of the 46 children of civil-servant mothers (37.0%) were overweight or obese, compared with 43 of the 212 children of traders (20.3%) (χ² = 9.638, p = 0.047). Religion was also associated with BMI status (χ² = 4.445, p = 0.035). Still, this result rests on an empty cell, since none of the 14 non-Christian children was overweight or obese, and it should not be interpreted substantively. Father's occupation showed no association (χ² = 1.592, p = 0.810), with the proportion of children whose father was a civil servant being identical (30.5%) in the two groups (Table 4). Age group (p = 0.662), sex (p = 0.956), mother's education (p = 0.877), father's education (p = 0.352), household size (p = 0.930), the number of children under five (p = 0.107) and the number of children aged 6-12 years in the household (p = 0.220) showed no statistically significant association with BMI status, with p-values ranging from p = 0.107 to p = 0.956. Results for parental education and the number of children in the household are not presented in Table 4 and are available from the corresponding author.

Dietary and physical activity factors and overweight/obesity: children who bought food at school were less likely to be overweight or obese than those who did not (20.8% versus 34.9%; χ² = 5.73, p = 0.017). Among buyers, the frequency of purchase was associated with BMI status (χ² = 10.508, p = 0.015): 12 of the 28 children who bought food only once a week (42.9%) were overweight or obese, compared with 35 of the 206 daily buyers (17.0%). Vegetable consumption frequency was also associated with BMI status (χ² = 10.119, p = 0.038), prevalence being lowest among daily consumers (19.1%) and highest among children who ate vegetables three or more times a week but not daily (36.1%) (Table 5). None of the remaining dietary or physical activity exposures was significantly associated with BMI status, with p-values ranging from p = 0.076 to p = 0.973: school feeding status (p = 0.347), meals per day (p = 0.076), snack intake (p = 0.722), sweet intake (p = 0.729), fruit intake (p = 0.266), involvement in sport (p = 0.683), means of transport to school (p = 0.804), leisure-time activity (p = 0.519) and dietary diversity score (p = 0.973). Results for meal frequency, snack and sweet intake, means of transport, leisure activity and dietary diversity are not presented in Table 5 and are available from the corresponding author.

Applying the p < 0.20 screening criterion across both sets of bivariate analyses, seven variables qualified for multivariable analysis: mother's occupation (p = 0.047), religion (p = 0.035), buying food at school (p = 0.017), frequency of buying food per week (p = 0.015), vegetable consumption frequency (p = 0.038), meals per day (p = 0.076) and the number of children under five in the household (p = 0.107). Religion could not be modeled because the empty cell for non-Christian children precludes estimation of an odds ratio, and buying food at school is represented in the model by the frequency of purchase. Meals per day and the number of children under five, although meeting the criterion, were not retained in the final model presented in Table 6.

Factors independently associated with overweight and obesity: after mutual adjustment, three factors remained independently associated with overweight and obesity. Children of civil-servant mothers had more than twice the odds of being overweight or obese compared with children of traders (aOR 2.43, 95% CI 1.056-5.584, p = 0.037). Infrequent purchasing of school food carried the highest odds: children who bought food at school only once a week had three times the odds of daily buyers (aOR 3.06, 95% CI 1.303-7.173, p = 0.010). No significant association was observed between buying food two or three times per week and childhood overweight/obesity status. Children who consumed vegetables three or more times a week but not daily had more than three times the odds of those who ate vegetables daily (aOR 3.33, 95% CI 1.396-7.928, p = 0.007), whereas maternal farming (aOR 2.20, 95% CI 0.979-4.963) and vegetable consumption once a week (aOR 1.85, 95% CI 0.789-4.350) were associated with elevated but non-significant odds (Table 6).

Discussion     

This study identified three independent risk factors significantly associated with overweight and obesity among school children aged 6-12 years in the South Tongu District: maternal occupation as a civil servant, buying food in school at a frequency of once per week, and consuming vegetables less frequently than daily. These findings are contextually relevant to the ongoing nutritional transition in peri-urban districts of Ghana.

Maternal occupation and childhood obesity: children whose mothers were civil servants were approximately 2.4 times more likely to be overweight or obese than those whose mothers were traders [19]. This finding aligns with evidence from a systematic review of childhood overweight in sub-Saharan Africa, which found that formal maternal employment was associated with increased odds of childhood overweight, particularly in Ghana and Kenya [8,20]. A plausible mechanism involves the reduced availability of home-cooked meals when mothers work formal daytime hours. Civil servant mothers who report to work early and return late may leave their children less supervised and more likely to purchase energy-dense foods from school vendors and street hawkers [13]. This is consistent with evidence that out-of-home food consumption contributes a substantial share of daily energy intake among schoolchildren [13,21]. Furthermore, a Ghana-specific study in the Adentan Municipality found that higher socioeconomic household status, which often correlates with formal parental employment, was independently associated with increased odds of being overweight (OR: 2.58, 95% CI: 1.41-4.70) [8].

Maternal occupation is unlikely to act on a child's weight directly; it is better read as a marker of the household arrangements around the child, and family structure was not measured in this study. Whether a child lives with both parents, with the mother alone, or with a grandparent or other guardian, and who prepares meals and supervises the child after school while the mother is at work, may account for more of the association than employment status itself. Formally employed mothers are more likely to live apart from their children during the working week or to delegate daily care to a guardian or house help, in which case what is being detected here is a caregiving arrangement rather than an occupation. The absence of these variables is a limitation of this study, and future work in this setting should measure household composition, the identity of the child's principal caregiver, and that caregiver's own food practices alongside parental occupation.

Paternal occupation and household provisioning: both parents' occupations were recorded, and the civil service was also the most common paternal occupation (30.5%), yet only maternal occupation was associated with a child's BMI status in this sample. Father's occupation showed no significant association at the bivariate level (χ² = 1.592, p = 0.810) and was therefore not retained in the multivariable model. This asymmetry is plausible in the Ghanaian context. In most households, the father is the principal earner. At the same time, the mother decides what is bought, prepared, and served, and how much money the child is given for food at school, so maternal employment is more closely tied to a child's daily food environment than paternal employment, even where the father contributes the larger share of household income. Paternal occupation is at best a crude proxy for that income, which was not measured directly, and the contrast between the two parents should therefore be interpreted cautiously and tested in studies that measure household income, caregiving roles and food provisioning separately.

Out-of-home food purchasing and obesity risk: children who bought food at school only once per week (compared to daily buyers) had a threefold elevated risk of overweight/obesity (aOR: 3.06, 95% CI: 1.303-7.173). This counterintuitive finding may reflect a pattern where infrequent school food buyers consume larger quantities of often calorie-dense, nutrient-poor foods on the days they do buy, consistent with studies showing that episodic purchasing of fast foods and street-vended snacks tends to involve high-calorie, high-fat food items [22]. This contrasts with children who buy food daily, who may develop more regular and perhaps moderate purchasing habits as they grow older. This finding is consistent with the literature demonstrating that in peri-urban Ghana, children have easy access to energy-dense ready-to-eat meals that are high in fat, sugar, and salt [6]. A scoping review of childhood overweight and obesity among in-school children in Ghana (2024) also confirmed that the consumption of unhealthy out-of-home foods is a leading dietary exposure associated with obesity [9].

The reverse side of this finding also requires explanation, since children who bought food every day carried the greatest cumulative exposure to vendor food yet were not at higher risk. Several accounts are plausible. Daily buyers may spend a small, fixed amount on a staple item such as rice, kenkey, or porridge that substitutes for a meal. In contrast, an occasional purchase may be treated as a treat and spent on fried snacks, pastries and sweetened drinks, so that frequency and energy density move in opposite directions. Daily purchasing may also mark households that send no food from home, a pattern associated with tighter rather than looser household food budgets and, in this setting, with lower rather than higher body weight. Reverse causation cannot be excluded in a cross-sectional design: parents of a child who is visibly gaining weight may reduce the money given for school purchases, in which case infrequent buying is a consequence of overweight rather than a cause of it. Finally, the estimate rests on 28 children who bought food once a week, of whom 12 were overweight or obese, and the confidence interval (1.303-7.173) is correspondingly wide. This result is best treated as hypothesis-generating and tested in studies that record what is bought and its energy content, not only how often it is bought.

Low vegetable consumption and obesity risk: children consuming vegetables three or more times per week, but not daily, were over three times more likely to be overweight or obese compared to daily vegetable consumers (aOR: 3.33, 95% CI: 1.396-7.928). This finding suggests that the absence of daily vegetable consumption creates a dietary pattern susceptible to caloric excess, as children fill the nutritional gap with higher calorie alternatives. Low vegetable consumption is associated with inadequate dietary fiber, micronutrients, and satiety signaling, which may facilitate the overconsumption of energy-dense foods [7,10]. A systematic review of childhood obesity risk factors in SSA found that dietary patterns characterized by low fruit and vegetable consumption were significantly associated with increased overweight and obesity risks [23]. In Ethiopia, Kumma et al. similarly reported that dietary patterns low in vegetables were associated with an adverse cardiometabolic risk profile [24]. The findings of this study are particularly important given that farming is the primary economic activity in the South Tongu District, yet fresh vegetable consumption among school-age children remains insufficient daily, suggesting a need for nutrition education and school garden initiatives.

Vegetable intake was measured as a frequency rather than as a quantity or as a specific nutrient, and it is interpreted here as a marker of overall diet quality rather than of any single constituent. Two mechanisms are usually proposed. The first is dietary fiber and the low energy density of vegetable-rich meals, which increase satiety and displace energy-dense foods; the second is micronutrient adequacy, which bears less directly on energy balance. If the fiber pathway is the operative one, fruit intake becomes a potential confounder, since fruit is a competing source of fiber and of low-energy-density bulk in this setting. Fruit intake was measured in this study but was not associated with BMI status at the bivariate level (χ² = 5.216, p = 0.266) and so did not meet the p < 0.20 threshold for entry into the multivariable model. The vegetable association reported here is therefore not adjusted for fruit intake, and residual confounding by fruit intake cannot be excluded. Studies that quantify intake in grams per day and separate fiber from micronutrient intake would be needed to distinguish the two pathways.

Non-significant risk factors: contrary to the findings of some studies, sex, age, physical activity participation, means of transport to school, TV watching, and snack or sweet intake frequency were not significantly associated with BMI status in this sample [25]. This may reflect the relatively modest sample size, limiting the statistical power to detect moderate effect sizes, or it may reflect genuine contextual differences. For instance, most children (77.8%) walked to school, which reduced the variation in active transport as a discriminating variable in the study. Similarly, although 85.2% of children engaged in sporting activities, the intensity and duration may have been insufficient to produce a detectable protective effect on BMI [26-29]. Previous studies in Ghana have reported mixed evidence: urban studies have found low participation in sport and more than two hours of daily television viewing to be associated with overweight [7, whereas the national scoping review of in-school children reports inconsistent findings for physical activity across settings [30].

Limitations: this study had several limitations. The cross-sectional design precludes causal inference between the identified risk factors and overweight/obesity, and reverse causation - a change in food purchasing or in diet following weight gain - cannot be ruled out. Recall bias may have affected the accuracy of the dietary frequency data, particularly among younger children. Dietary and physical activity information was reported by the children themselves and was not corroborated by parents or guardians, so social desirability and a Hawthorne effect may also have influenced responses; any resulting misclassification is likely to be non-differential and to have biased estimates towards the null. The study was limited to public schools, potentially excluding private school children who may have different dietary and activity profiles from public school children. Family structure, the identity of the child's principal caregiver, and the household food environment - in particular the purchase and preparation of energy-dense foods at home - were not measured, and residual confounding by these factors is likely. Chronic medical conditions and medications capable of affecting body weight were not screened for; only acute illness on the day of data collection was an exclusion criterion.

Some strata were small: 28 children reported buying food once a week, and 36 reported eating vegetables three or more times but not daily, so several adjusted odds ratios rest on few events and carry wide confidence intervals. Children who were not overweight or obese were analyzed as a single normal-weight group, so coexisting thinness was not examined, and the double burden of malnutrition in this population cannot be described from these data. Additionally, variables such as genetic predisposition, sleep duration, and household income, which are documented correlates of childhood obesity, were not assessed. Despite these limitations, the study's population-based sampling approach across all six sub-districts strengthens the representativeness of the findings within the study area. A further consideration is the interval between data collection in July-August 2019 and publication. Three points bear on the current applicability of the findings. First, the structural conditions that generated these exposures - peri-urbanization along the Accra-Aflao corridor, the mixed farming and salaried occupational profile of the district, the school feeding program and the vendor-based school food environment - have not been reversed since 2019. Hence, the exposures examined here remain present. Second, childhood overweight has continued to rise nationally and globally over the same period, so the prevalence reported here is more likely to understate than to overstate the current district burden. Third, and against this, the COVID-19 pandemic and the school closures, income shocks and food price inflation that followed it fell between data collection and publication and are likely to have altered both diet and physical activity in ways these data cannot capture. The findings are therefore best read as a 2019 district baseline against which a repeat survey should be set, and such a survey is warranted.

Conclusion     

This study identified three independent risk factors for overweight and obesity among school children aged 6-12 years in the South Tongu District of Ghana's Volta Region: having a mother employed as a civil servant, infrequent (once per week) out-of-home food purchasing, and consuming vegetables less than daily. Approximately one in four children in the sample was overweight or obese. These findings highlight the critical role of maternal employment type and children's dietary behaviors, particularly patterns of out-of-home food consumption and vegetable intake, in shaping the nutritional status of school-age children in a rapidly urbanizing district. Because the design is cross-sectional and the data were collected in 2019, these associations should be read as priorities for confirmation and action rather than as established causes. Public health interventions should prioritize school-based nutrition education programs targeting children, parents (particularly civil servant mothers), and school food vendors. Promoting daily vegetable consumption and home-prepared meals while discouraging energy-dense out-of-home food purchases are key strategies to curb the growing obesity burden. Further longitudinal research incorporating income data, sleep patterns, and genetic factors is needed to deepen our understanding of the determinants of childhood obesity in peri-urban Ghanaian communities.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Faustina Abor-Adjei conceptualized and designed the study, developed the data collection instruments, conducted the literature review, coordinated and supervised field data collection across all six sub-districts, performed data entry and cleaning, and drafted the initial and revised versions of the manuscript; Peter Kofi Dabie performed statistical analysis, interpreted the findings, and supervised the entire study; provided expert guidance on conceptualization and methodology, and reviewed and approved the final manuscript for submission. All the authors read and approved the final version of this manuscript.

Acknowledgments     

The authors acknowledge the principals, parents, and school children of the South Tongu District who participated in this study, as well as the staff of the South Tongu District Health Directorate for their support.

Tables     

Table 1: distribution of study participants by sub-district, South Tongu District, Ghana, July to August 2019

Table 2: sociodemographic characteristics of participants, South Tongu District, Ghana, from July to August 2019

Table 3: dietary and physical activity risk factors of participants, South Tongu District, Ghana, from July to August 2019

Table 4: demographic characteristics and body mass index (BMI) status among school children in South Tongu District, Ghana, from July to August 2019

Table 5: dietary and physical activity risk factors and body mass index (BMI) status among school children in South Tongu District, Ghana, from July to August 2019

Table 6: binary logistic regression, independent risk factors of overweight/obesity among school children in South Tongu District, Ghana, from July to August 2019

References