Leveraging machine learning to predict mosquito bed net utilization among women of reproductive age in sub-Saharan Africa

Abstract

Background

Malaria remains a major public health challenge, particularly in sub-Saharan Africa, where women of reproductive age are especially vulnerable during pregnancy and childbirth. To identify key predictors and improve predictive accuracy, machine learning algorithms such as Random Forest were applied, along with SHAP analysis, to a large multi-country DHS dataset, with class imbalance addressed using Tomek Links and Random Over-Sampling.

Methods

This study employed a weighted dataset of 153,015 participants from the Demographic and Health Survey (DHS) conducted across ten sub-Saharan African countries. Data preprocessing and analysis were carried out using STATA version 17 and Python 3.10. Feature scaling was applied to standardize numerical variables, ensuring uniform weighting across predictors and improving model stability. An 80:20 data split ratio was applied, and class imbalance was addressed using Tomek Links combined with Random Over-Sampling. Eight models were selected and trained using both balanced and unbalanced datasets. The model performance was evaluated using metrics such as ROC-AUC, accuracy, recall, F1 score, and precision.

Results

The Random Forest algorithm performed best in this study, with an accuracy of 83%, an F1 score of 82%, recall of 80%, precision of 84%, and an AUC of 88%. Fifty-five percent of participants used mosquito nets. The SHAP analysis showed that Age above 34, being employed, frequent social media use, higher education, institutional deliveries, and female-headed households increased bed net use, while fewer ANC visits and being divorced decreased its use.

Conclusion

Age above 34, being employed, frequent social media use, higher education, institutional deliveries, and female-headed households increased bed net use, while fewer ANC visits and being divorced decreased its use. Strengthening social media use for health information, promoting women's education, encouraging institutional delivery, motivate for antenatal care services, and providing support to socially and economically vulnerable women are essential strategies to enhance mosquito net utilization.

Background

Mosquito bed nets serve as a protective physical barrier that prevents mosquito bites during the night [1]. Their high effectiveness, affordability, and ease of implementation make them a cornerstone of malaria prevention strategies [2]. Despite these advantages, malaria remains a major global public health concern, particularly in sub-Saharan Africa, where women of reproductive age (WRA) are disproportionately affected due to increased vulnerability during pregnancy and childbirth [3].

Malaria remains a major public health concern, with sub-Saharan Africa bearing the greatest burden. In 2023, approximately 94% of global malaria cases and 95% of malaria-related deaths occurred in the region [4]. Overall, ninety-four percent of all malaria cases and deaths were reported in Africa. According to the World Health Organization (WHO), one person dies from malaria every minute [5], and the disease causes the deaths of an estimated 200,000 infants and 10,000 women annually in Africa [6].

Malaria remains a serious public health concern, placing nearly half of the global population at risk of infection [7]. In recent years, both malaria cases and fatalities have risen beyond expectations, with evidence indicating that a significant portion of the population continues to live in malaria-endemic areas. In 2023, there were 23 deaths per 10,000 malaria cases [8]. Although mosquito bed nets are proven to be an effective method of malaria prevention, their consistent use remains below the desired level in many endemic regions, particularly in sub-Saharan Africa [9].

In some African countries, such as Uganda, mosquito bed nets are distributed every 3 years as part of ongoing malaria prevention campaigns [10]. However, the effectiveness of these campaigns is often limited by inconsistent usage and behaviour that affect regular bed net utilization [11]. According to previous evidence, lower education levels, limited health awareness, age, and the use of nets for other purposes are factors contributing to poor utilization of bed nets [2, 12]. The use of mosquito bed nets is one of the most cost-effective strategies for malaria control, capable of reducing the risk of malaria infection by up to 50%, improving quality of life, and increasing productivity [13]. Despite their proven effectiveness, the inconsistent use of bed nets among women in sub-Saharan Africa remains a major barrier to achieving the 2030 malaria prevention and control targets, which aim to: (i) reduce malaria mortality rates by at least 90%, (ii) eliminate malaria in at least 35 countries, and (iii) prevent the resurgence of malaria in all malaria-free countries [14].

This issue is particularly critical among women, as evidence indicates that 36% of them are exposed to malaria infection during pregnancy [15, 16], and given that women in Africa are more often responsible for managing their children at home compared to men [17]. Improving mosquito net utilization among women of reproductive age is crucial not only for their own health but also for the protection of their families [18]. Previous studies have given limited attention to women of reproductive age as a distinct group when examining factors influencing mosquito bed net utilization. Additionally, most existing research relies on traditional statistical methods, such as logistic regression [19, 20], which may not fully capture the complex and nonlinear relationships among influencing factors. Advanced machine learning techniques, which are capable of uncovering such hidden patterns [21]. To address this gap, the aim of this study is to identify key predictors and develop accurate predictive models for mosquito net utilization among women of reproductive age in selected sub-Saharan African countries using machine learning approaches.

Methods

Study design and setting

Machine learning was chosen due to its ability to analyze large datasets with multiple predictor variables, improving the accuracy of bed net use prediction and facilitating data-driven decision-making in public health This study was conducted across ten sub-Saharan African countries—Burkina Faso, Burundi, Cameroon, Madagascar, Mozambique, Rwanda, Sierra Leone, Senegal, Uganda, and Zambia using Demographic and Health Survey (DHS). Sub-Saharan Africa bears a significant share of the global malaria burden, with women of reproductive age particularly vulnerable due to biological susceptibility and their role in childbearing and caregiving. The region is characterized by a high prevalence of Plasmodium falciparum, the most deadly malaria parasite, transmitted by Anopheles mosquitoes, which are widely distributed due to favourable climatic and environmental conditions.

Entomological landscapes vary across the selected countries but are generally marked by year-round or seasonal mosquito breeding due to rainfall, stagnant water sources, and inadequate housing. Despite ongoing malaria control efforts, such as mass distribution of insecticide-treated bed nets (ITNs), usage among women of reproductive age remains inconsistent across the region. Some countries have shown steady improvements in bed net utilization, while others still report suboptimal coverage and use, influenced by socioeconomic, cultural, and health system factors. This study aims to analyse these patterns using machine learning to identify predictors of bed net utilization in this high-risk population.

Source and study population

The study population includes a total weighted sample of 153,015 individuals of reproductive age who have available DHS data on mosquito bed net utilization and reside in sub-Saharan Africa.

Study variables

Dependent variable: mosquito bed net utilization: measured among reproductive-age women (15–49 years) who reported sleeping under a mosquito bed net the night before the survey. Responses were recoded as'1'for'yes'and'0'for'no'.

Independent variables: These top independent variables, identified after feature selection and classified and measured based on DHS guidelines and previous studies, include maternal occupation, wealth index, marital status, distance to health facility, maternal age, number of under-5 children, maternal education level, ANC contact, place of residence, husband's occupation, sex of household head, frequency of social media use, and place of delivery [22] (Table 1).

Table 1.

List of variables and their measurements for mosquito net utilization among women of reproductive age in Sub-Saharan Africa

Variable	Category
Age	< 20, 25–34, and > 34
Place of residence	Urban, Rural
Current marital status	Single, Married, Widowed, Divorced
Highest education	No education, Primary, Secondary, Higher
Wealth index	Poor, Middle, Rich
Maternal occupation	Unemployed, Employed
Social media use frequency	Not at all, < once a week, > once a week
Number of under-5 children	One, Two, > Three
Spouse’s occupation	Unemployed, Employed
ANC visits	No contact, < 4 contacts, > 4 contacts
Distance to health facility	Big problem, Not a big problem
Place of delivery	Health institution, Home
Sex of household head	Male, Female

Data processing and management

After a detailed review of the literature and DHS guidelines, data were extracted from the DHS individual record dataset. The data from ten sub-Saharan African countries were then combined using Stata version 17. For data preprocessing, the dataset was exported to Python Colab version 3.10.2. Missing data were addressed using mode imputation for categorical variables to preserve data distribution and KNN imputation for continuous variables to ensure statistical consistency and maintain dataset integrity. Mode imputation is a simple and effective method that preserves the most frequent and representative category in the dataset, avoiding the introduction of unrealistic or rare categories. KNN imputation, on the other hand, preserves the relationships among features and maintains the underlying data structure better than mean or median imputation. Except for tree-based models, most machine learning models are sensitive to feature scaling. Therefore, in this study, after applying label encoding, MinMax scaling was used. For example, this scaling technique was applied to the'educational status'feature.

Feature selection

Recent studies show that irrelevant variables can reduce a model's ability to generalize, increase its complexity, and potentially decrease the overall accuracy of a classifier in machine learning studies [21]. Feature Elimination (RFE) was applied to systematically eliminate irrelevant features, reducing dimensionality while preserving the most predictive variables, thus improving model efficiency and generalizability. After fitting RFE, the feature rankings were obtained using the feature ranking command. RFE removes the least important features in each iteration. See the results in Fig. 1.

Fig. 1.

Feature importance for predicting mosquito bed net usage among women of reproductive age in Sub-Saharan Africa in 2025

Splitting the data

In alignment with the Pareto Principle, an 80:20 train-test split was employed to ensure an optimal balance between model training and evaluation. Allocating 80% of the data for training allows the model to learn underlying patterns effectively, while reserving 20% for testing provides a sufficient and unbiased estimate of the model’s generalization performance [23].

Handling imbalanced data

Imbalanced data refers to datasets with an uneven distribution of observations across target classes, which can result in models that are biased toward the majority class. Effectively addressing this imbalance is essential for developing accurate and reliable machine learning models [24].

Random Over-Sampling combined with Tomek Links is employed for handling imbalanced datasets because it removes only ambiguous or borderline data rather than randomly selected majority samples thereby reducing the risk of generating unrealistic synthetic samples, minimizing overfitting caused by duplicate data, and preserving important information from the dataset.

Model selection

In this study, a diverse set of machine learning models was selected, including Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LGBM), and AdaBoost (ADA). These models were chosen to represent a wide range of learning algorithms with varying strengths. DT and RF offer interpretability and handle both categorical and numerical data effectively. KNN is a simple yet powerful instance-based method useful for capturing local data patterns.

ANN provides the capability to model complex, non-linear relationships. Ensemble methods such as RF, XGBoost, LGBM, and AdaBoost are known for their high predictive performance and ability to reduce overfitting by combining multiple learners. This diverse selection allows for a comprehensive comparison and enhances the robustness of the analysis in predicting bed net utilization.

Hyperparameter tuning

To optimize the performance of the Random Forest (RF) model, hyperparameter tuning was conducted using grid search combined with cross-validation. See Table 2.

Table 2.

Hyperparameter tuning settings for the random forest model used to predict mosquito net utilization

Hyperparameter	Values tested	Recommended value and used	Justification
n_estimators	100, 200, 300, 500	300	Balanced model performance and computational cost; gains plateaued after 300
max_depth	10, 20, 30, None	20	Prevented overfitting while capturing sufficient data complexity
min_samples_split	2, 5, 10	5	Avoided overly specific splits and improved generalization
min_samples_leaf	1, 2, 4	2	Controlled leaf node size to reduce overfitting
max_features	'sqrt','log2', None	'sqrt'	Common default for classification; improves diversity among trees
bootstrap	True, False	True	Maintains ensemble diversity and robustness via sampling with replacement

Model training

After selecting the model, both balanced and unbalanced datasets were used to train the chosen classifiers, with tenfold cross-validation employed to evaluate their performance. The top-performing model was identified through comparison and then retrained using the balanced training data to make final predictions on unseen test data.

Model evaluation

Recent research emphasizes that model evaluation is a crucial step in machine learning, as it allows us to determine how well a trained model performs on new, unseen data [25, 26]. Machine learning (ML) holds significant potential in the healthcare sector, where large and complex datasets can offer valuable insights into areas ranging from patient management to policy development. However, this potential also brings the need to carefully monitor and mitigate biases and overfitting to improve healthcare service quality and produce reliable evidence [27].

F1-Score, Precision & Recall, ROC-AUC Score, and Accuracy were employed to manage false positives and false negatives, reduce prediction, assess the overall strength of the classifier, and to understand the trade-offs inherent to the classification nature of the target variable [28]. These metrics were chosen based on the objective of accurately evaluating model performance in relation to the data characteristics, particularly to minimize false positives [29]. The machine learning model achieved an overall accuracy of 83% in predicting mosquito bed net usage, successfully classifying both users and non-users. It attained a recall of 80%, demonstrating strong sensitivity in identifying individuals who actually use bed nets. Additionally, the model recorded an AUC (Area Under the ROC Curve) of 0.88, indicating robust discriminative capability between users and non-users. See Table 3 and Fig. 2.

Table 3.

Comparative performance of machine learning models for predicting mosquito bed net usage among Sub-Saharan women of reproductive Age, 2025

Model	AUC	Accuracy	F1 score	Recall	Precision
DT	87	0.81	0.80	0.78	0.82
RF	88	0.83	0.82	0.80	0.84
KNN	83	0.78	0.77	0.75	0.79
ANN	83	0.78	0.77	0.76	0.78
XGBoost	86	0.80	0.79	0.77	0.81
LGBM	83	0.78	0.77	0.75	0.79
ADA	76	0.71	0.70	0.68	0.72
GB	78	0.73	0.72	0.70	0.74

The Random Forest (RF) model, highlighted in bold, outperformed the other models

Fig. 2.

ROC curve for predicting mosquito bed net usage among sub-saharan women of reproductive age, 2025

Results

Distribution of mosquito bed net usage among sub-Saharan women of reproductive age, by country distribution

According to the evidence on mosquito bed net usage, Madagascar had the highest percentage of mosquito bed net usage at 12.33%, followed by Uganda at 12.09%. Burkina Faso 11.54%, while Burundi had 11.29%. Other countries include Senegal with 10.80%, Sierra Leone at 10.18%, and Rwanda at 9.56%. Zambia had 8.94%, Cameroon reported 8.90%, and Mozambique had the lowest at 4.36%). See Table 4.

Table 4.

Distribution of mosquito bed net usage among Sub-Saharan women of reproductive age, 2025

Country	Survey years	Frequency	Percentage
Burkina Faso	2021	17,659	11.54
Burundi	2016–2017	17,268.999	11.29
Cameroon	2018	13,615.672	8.90
Madagascar	2021	18,869	12.33
Mozambique	2022–2023	6,678.3045	4.36
Rwanda	2019–2020	14,633.233	9.56
Sierra Leone	2019	15,574.001	10.18
Senegal	2022–2023	16,527.837	10.80
Uganda	2018	18,506	12.09
Zambia	2018	13,683	8.94

Individual characteristics of study participants about mosquito bed net usage among sub-Saharan women of reproductive age, 2025

According to this evidence, majority of participants (48.57%) were between 25 and 34 years old. Most women (66.19%) resided in rural areas, and nearly half (46.95%) were married. In terms of education, 35.67% had attained primary education. Wealth distribution showed that 44.88% of participants were in the rich category, while employment status indicated that 66.57% of women were employed. Social media usage was reported by 89.56% of participants, with 53.70% using it less than once a week. Most women (63.74%) had one child under 5 years of age, and concerning their husband’s occupation, 61.16% were employed.

Access to antenatal care (ANC) was limited, with 57.15% reporting no ANC contact. Regarding access to health facilities, 65.17% stated that distance was not a significant barrier. Most women (64.35%) delivered at health institutions. Finally, 73.11% of households were headed by males. See Table 5.

Table 5.

Individual characteristics of mosquito bed net usage among sub-Saharan women of reproductive age, 2025

Variable	Category	Frequency	Percentage
Age	< 20	34,641.7	22.64
	25–34	74,322.8	48.57
	> 34	44,050.4	28.79
Place of residence	Urban	51,739.5	33.81
	Rural	101,275.5	66.19
Current marital status	Single	46,144.9	30.16
	Married	71,843.2	46.95
	Widowed	21,188.1	13.85
	Divorced	13,838.7	9.04
Highest education	No educated	42,546.5	27.81
	Primary education	54,577.8	35.67
	Secondary education	48,877.2	31.94
	Higher education	7,013.4	4.58
Wealth index	Poor	54,944.5	35.91
	Middle	29,401.3	19.21
	Rich	68,669.0	44.88
Maternal occupation	Unemployed	51,153.3	33.43
	Employed	101,861.7	66.57
Frequency of social media use	Not at all	15,980.4	10.44
	< once a week	82,166.0	53.70
	> once a week	54,868.6	35.86
Number of under-5 children	One	97,533.3	63.74
	Two	35,336.3	23.09
	> three	20,145.3	13.17
Spouse’s occupation	Unemployed	59,424.7	38.84
	Employed	93,590.2	61.16
ANC	< 4 ANC contact	22,406.4	14.64
	> 4 ANC contact	43,167.3	28.21
	No ANC contact	87,441.2	57.15
Distance to health facility	Big problem	53,292.3	34.83
	Not a big problem	99,722.6	65.17
Place of delivery	Health institution	98,461.1	64.35
	Home	54,553.9	35.65
Sex of household head	Male	111,875.9	73.11
	Female	41,139.0	26.89

The Random Forest (RF) model, highlighted in bold, outperformed the other models

Predictors of mosquito bed net usage among sub-Saharan women of reproductive age, 2025

SHAP (SHapley Additive exPlanations) values show the contribution of each feature to mosquito bed net usage. SHAP values quantify each predictor’s contribution to the model’s output, where positive values signify an increased likelihood of Mosquito Bed Net Usage and negative values indicate a protective effect by capturing feature interactions without depending on averaging effects, SHAP significantly enhanced the model’s interpretability. This improvement made it easier to communicate the results to both technical and non-technical audiences, while also fostering transparency, trust, and a deeper understanding of the model's decision-making process.

The random forest machine learning analysis revealed several key predictors influencing mosquito bed net usage among women. Age above 34, being employed, frequent social media use, higher education, institutional deliveries, and female-headed households increased bed net use, while fewer ANC visits and being divorced decreased its use. The result is shown in Fig. 3.

Fig. 3.

SHAP value of bed net usage among reproductive-age women in Sub-Saharan Africa

Discussion

Approximately 94% of malaria cases and deaths worldwide occur in Africa, where women of reproductive age and children under five are disproportionately affected. Bed net use among women is essential to ending the cycle of transmission in households and communities.

This machine learning study offers a powerful tool to enhance understanding of the major predictors that influence bed net utilization. By analysing large study data, it provides evidence that policymakers can use to easily propose and implement effective malaria control strategies. The Random Forest algorithm is the best-performing model, with an accuracy of 83%, F1 score 82%, recall 80%, precision 84% and an AUC of 88%. This is supported by its strong performance in previous machine learning studies using healthcare data [30–33].

A possible explanation for the strong performance of the Random Forest model is its ability to handle complex, non-linear relationships and interactions between variables, which may be common in health-related behavioral data such as bed net utilization. Random Forest is also robust to overfitting, especially when tuned properly, and can manage both categorical and continuous variables effectively. Additionally, its ensemble nature aggregating results from multiple decision trees helps improve predictive accuracy and stability compared to individual models.

According to the findings of this study, 55% [95% CI 54.8–55.2%] of women of reproductive age in sub-Saharan Africa reported using mosquito bed nets. This prevalence is higher than those reported in previous studies from Kenya (51%) [34], Ethiopia (47.05%) [35], Uganda (35%) [36] Zambia (50.8%) [37], but lower than the rates observed in Ghana (61%) [38], Mozambique (68.4%) [39], the Congo (71.4%) [40]. The possible explanation for these differences might include variations in malaria control programmes, access and availability of bed nets, socioeconomic and educational differences, cultural beliefs and practices, media and information exposure, as well as differences in survey timing and methodology.

According to SHAP evidence, female-headed households increase mosquito net usage, which is supported by previous finding from Ghana, Myanmar, and Mozambique [14, 41, 42]. A possible explanation for the higher bed net utilization among female-headed households is that, as primary caregivers and decision-makers, they may be more inclined to weigh the risks and benefits of malaria prevention for both themselves and their children.

Women with primary or higher education levels were more likely to use mosquito nets, consistent with findings from Rwanda, Tanzania, and Cameroon Tanzania, and Cameroon [43–45]. This may be due to better health literacy, stronger decision-making capacity, and increased engagement with health services and malaria prevention campaigns among educated women [46].

As evidenced by this study, women who delivered at health institutions showed increased mosquito net usage. This is supported by earlier studies conducted in Cameroon, Ethiopia, Rwanda, and Myanmar [42, 44, 45, 47]. A possible reason could be that women who deliver at health institutions receive counseling about preventive healthcare services, including malaria prevention, as they have a better understanding of the benefits of bed net utilization [48].

SHAP evidence showed that women aged 34 and above have increased mosquito net usage. This is in line with previous studies conducted in Nigeria, Ethiopia, Rwanda, Cameroon, [44, 45, 47, 49]. The possible justification for this could be that women aged 34 and above have more skills and experience regarding the impact of malaria on their lives and their children. They also tend to have greater decision-making power, gained through their experience in household management, compared to younger women [50].

According to this machine learning analysis, divorced women show decreased use of mosquito bed nets, which is supported by previous evidence from Rwanda, Sierra Leone, Uganda, and Cameroon [44, 45, 51, 52]. Divorced women may live in less stable households with limited net access or motivation for use and may also be less engaged in community health programmes, reducing their access to malaria prevention resources.

This study also showed that those who use social media more than once a week and above have increased mosquito bed net usage this supported by previous evidence conducted in Myanmar, Ghana, and Cameroon [38, 42, 45]. A possible explanation is that social media plays a key role in bed net utilization by increasing exposure to health promotion campaigns. Reproductive-age women who frequently use social media may have better health-seeking behaviour, greater awareness of malaria prevention, and improved knowledge of how and where to use bed nets [53].

This evidence indicates that employed women have higher mosquito net usage than unemployed women, which aligns with previous studies conducted in Ethiopia and Myanmar [42, 47].

Employed women often have greater exposure to social media and increased opportunities to participate in workplace health promotion programmes, both of which can enhance awareness and encourage preventive behaviors. Furthermore, employment is frequently associated with higher levels of education, which improves health literacy and empowers women to make informed decisions about disease prevention. This combination of factors likely contributes to the increased utilization of mosquito nets among employed women.

This study showed that women with fewer than four ANC contacts had decreased utilization of bed nets, a finding supported by studies from Rwanda, Uganda, and India [44, 54, 55]. Frequent ANC visits provide critical opportunities for health promotion and education, during which women are informed about malaria prevention strategies, including the importance of consistent bed net use. In many settings, pregnant women also receive free insecticide-treated nets during ANC visits, further facilitating usage. Therefore, insufficient ANC contact may result in missed chances for both education and access to preventive tools, ultimately reducing bed net utilization.

Strength and limitation of the study

The strength of this study lies in the use of machine learning algorithms, which can analyze large datasets to identify complex and previously unseen patterns that traditional statistical methods might overlook. However, some limitations remain. Despite the power of machine learning, potential biases inherent in the DHS data such as recall bias and underrepresentation of certain sub-populations—may affect the generalizability of the findings. Additionally, models like ANN, XGBoost, and LGBM require considerable computational resources and extensive tuning, which may limit scalability. Furthermore, the study did not include data from all sub-Saharan African countries, limiting the comprehensiveness of the results. Future research should aim to incorporate data from all countries in the region to improve representativeness and robustness.

Conclusion

The Random Forest algorithm performed best in this study, with an accuracy of 83%, an F1 score of 82%, recall of 80%, precision of 84%, and an AUC of 88%. The prevalence of mosquito bed net use among women of reproductive age is 55%, which is lower than previous studies.

The SHAP analysis showed that employed women, frequent social media use, increased age, delivery at health institutions, higher education levels, and female-headed households were positively associated with increased mosquito net usage, whereas fewer ANC contacts and being divorced were associated with decreased usage.

To improve mosquito bed net usage, efforts should focus on promoting female education, distributing free or low-cost bed nets, integrating malaria education into health facility visits, and collaborating with social media corporation to share prevention messages. Special attention should be given to educating young and unmarried women and improving access to health services in rural areas.

Policy implication of the study

The integration of machine learning into predicting mosquito bed net usage presents a significant opportunity to improve healthcare delivery in sub-Saharan Africa. Evidence from this large study is valuable for policymakers in developing policies related to infrastructure investment, educational initiatives, and equitable access to services. For researchers, this study serves as a benchmark for future investigations.

Abbreviations

LGBM

Light gradient boosting machine

DT

Decision tree

XGBoost

Extreme gradient boosting

KNN

K-nearest neighbor

GB

Gradient boosting

LR

Logistic regression

RF

Random forest

AUC

Area under the curve

DHS

Demographic and Health Survey

Author contributions

NDB, TZT, GT, and JMK was responsible for the conceptualization, methodology, data curation, formal analysis, writing of the original draft, and overall supervision of the study. AH, and MGT contributed to the methodology, validation, and review and editing of the manuscript. KAD, and AT was involved in data curation, investigation, and manuscript review and editing. NDB contributed to formal analysis, data visualization, and review and editing of the manuscript. JMK was responsible for providing resources, conducting investigations, and managing project administration. AYA, and MJ contributed to data curation, software development, and manuscript review and editing. JMK participated in validation, investigation, and review and editing of the manuscript. NDB provided supervision, secured funding for the project, and contributed to the review and editing of the manuscript.

Funding

This study did not receive any funding.

Data availability

The datasets analysed in the current study are available in the public domain through the Measure DHS website (www.measuredhs.com).

Declarations

Ethics approval and consent to participate

This study utilized publicly available and fully de-identified Demographic and Health Survey (DHS) data, which can be accessed upon request through the DHS Program website (http://www.dhsprogram.com). As no personal identifiers were included, and ethical approval was obtained by the institutions conducting the original surveys, additional ethical clearance was not required. All analyses adhered to DHS data use policies, and strict measures were taken to maintain data anonymity and confidentiality.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.