1 Introduction

1.1 Research Problem

Malaria remains a leading cause of morbidity and mortality among children under five in sub-Saharan Africa. The use of insecticide-treated nets (ITNs) is a proven, cost-effective intervention for malaria prevention. In Kenya, despite mass distribution campaigns and high reported household ownership, consistent use among the most vulnerable group—children under five—is not universal. Identifying the factors that influence whether a child actually sleeps under an ITN is critical for refining programmatic strategies and moving beyond ownership to effective use.

1.2 Hypothesis

This study was guided by a primary hypothesis that mosquito net use among children under five is determined by a combination of perceived need (malaria transmission risk) and household capacity (socioeconomic status). Secondary hypotheses posited that maternal characteristics and household structure would also play significant roles, independent of cultural and regional demographics.

2 Methods

2.1 Study Design

A cross-sectional secondary data analysis was conducted using nationally representative data from the most recent Kenya Demographic and Health Survey (KDHS) available at the time of analysis (2022). The DHS employs a two-stage stratified sampling design to ensure representativeness at national and sub-national levels.

2.2 Study Population

The analysis focused on children under five years of age (0-59 months) from surveyed households. The analytical sample included all eligible children with complete data for the variables of interest.

2.3 Study Variables

Outcome: Binary variable indicating whether the child slept under an ITN the night before the survey.
Explanatory Variables: Variables were selected based on the Andersen-Newman behavioral model of health services utilization, encompassing predisposing, enabling, and need factors.
- Predisposing: Maternal age, education, ethnicity, religion.
- Enabling: Household wealth quintile (reported as wealth index), place of residence (urban/rural).
- Need: Malaria transmission risk zone (categorized per Kenya’s epidemiological strata), number of living children.
- Household Structure: Number of household members.

2.4 Statistical Analysis

To account for the complex survey design (clustering, stratification, and sampling weights), all analyses were performed using survey methods. A survey-weighted logistic regression model was fitted using the svyglm function in R, specifying a quasibinomial family. Adjusted Odds Ratios (aORs), 95% Confidence Intervals (CIs), and p-values were calculated to assess the strength and significance of associations.

Multicollinearity was assessed using generalized variance inflation factors (GVIFs), with interpretation based on the adjusted GVIF^(1/(2·Df)) measure for multi-level categorical variables. Linearity of the logit for continuous predictors (household members and number of living children) was evaluated using restricted cubic splines and design-based Wald tests. Model fit was assessed using the pseudo R² statistic. The statistical significance level was set at α = 0.05.

3 Results

The final model included data from 12,616 children. The pseudo-R² was 0.107, indicating a modest improvement in model fit over the null model, consistent with expectations for survey-weighted analyses of behavioral outcomes.

Model diagnostics indicated no problematic multicollinearity among predictors. However, tests of the linearity of the logit revealed significant non-linear associations for both household size and number of living children (p < 0.001). As a result, the reported odds ratios for these variables represent average marginal associations across their observed ranges, rather than constant per-unit effects.

	Results
Determinants of Mosquito Net Use Among Children Under Five in Kenya
At the national population level, accounting for DHS sampling design, what variables are associated with net use?
Variable	OR	95% CI	p-value
(Intercept)	0.62	0.33, 1.16	0.14
Respondent’s Age Group
15-19	—	—
20-24	0.97	0.75, 1.24	0.8
25-29	1.01	0.78, 1.32	>0.9
30-34	0.85	0.63, 1.14	0.3
35-39	0.76	0.55, 1.05	0.10
40-44	0.66	0.44, 0.98	0.038
45-49	0.55	0.31, 0.96	0.036
Place of Residence
Urban	—	—
Rural¹	1.00¹	0.80, 1.25	>0.9¹
Malaria Zone
Low Risk / Malaria-Free	—	—
Seasonal / Low Transmission	1.54	1.17, 2.03	0.002
Highland Epidemic-Prone²	3.86²	2.74, 5.43	<0.001²
Coastal Endemic²	3.92²	2.62, 5.84	<0.001²
Lake Endemic²	7.81²	5.68, 10.7	<0.001²
Wealth Index
Poorest	—	—
Poorer³	1.43³	1.21, 1.68	<0.001³
Middle³	1.99³	1.64, 2.42	<0.001³
Richer³	1.98³	1.57, 2.50	<0.001³
Richest³	3.29³	2.33, 4.63	<0.001³
Highest Educational Level
No education	—	—
Primary¹	1.01¹	0.72, 1.41	>0.9¹
Secondary¹	1.39¹	0.99, 1.96	0.057¹
Higher¹	1.41¹	0.96, 2.08	0.079¹
Respondent’s Ethnicity
Other	—	—
Embu	1.61	0.93, 2.77	0.088
Kalenjin	1.22	0.88, 1.69	0.2
Kamba	1.63	1.20, 2.21	0.002
Kikuyu	0.90	0.62, 1.32	0.6
Kisii	1.67	1.12, 2.49	0.012
Luhya	1.03	0.73, 1.45	0.9
Luo	0.83	0.60, 1.16	0.3
Maasai	1.20	0.80, 1.79	0.4
Meru	1.21	0.85, 1.72	0.3
Mijikenda/Swahili	2.24	1.50, 3.36	<0.001
Somali	1.08	0.55, 2.15	0.8
Taita/Taveta	1.88	0.95, 3.72	0.071
Respondent’s Religion
Other	—	—
Catholic	0.89	0.61, 1.29	0.5
Protestant	1.00	0.68, 1.47	>0.9
Evangelical churches	0.93	0.63, 1.37	0.7
African instituted churches	1.03	0.69, 1.55	0.9
Orthodox	0.45	0.14, 1.46	0.2
Islam	0.83	0.50, 1.37	0.5
Hindu	0.06	0.01, 0.45	0.006
Traditionists	0.50	0.13, 1.93	0.3
No religion/atheists	0.87	0.44, 1.71	0.7
Household Members⁴	0.86⁴	0.83, 0.89	<0.001⁴
Living Children⁴	1.12⁴	1.06, 1.18	<0.001⁴
¹ After adjustment, place of residence and education were not significant predictors, suggesting contextual malaria risk outweighs urban–rural or schooling differences.
² Mosquito net use was strongly associated with malaria endemicity, with substantially higher odds in lake, coastal, and highland epidemic-prone zones.
³ Household wealth showed a clear positive gradient, indicating socioeconomic inequalities in net use.
⁴ Household structure mattered: larger households had lower odds of net use, while households with more living children had higher odds.
Abbreviations: CI = Confidence Interval, OR = Odds Ratio
Source: DHS Data Program 2022

Figure 1: Adjusted Associations Between Selected Characteristics and Mosquito Net Use Among Children Under Five in Kenya.

Key Findings

Perceived malaria risk emerged as the dominant determinant of mosquito net use.
After accounting for household, socioeconomic, and cultural factors, malaria transmission intensity showed the strongest and most consistent association with net use. Children residing in the Lake Endemic zone had nearly eight times higher odds of sleeping under a mosquito net compared to those in low-risk or malaria-free areas (aOR = 7.81, p < 0.001).
Household socioeconomic status demonstrated a clear and monotonic relationship with net use.
Net use increased steadily with rising household wealth, independent of malaria risk and other covariates. Children from the richest households had more than three times the odds of net use compared to those from the poorest households (aOR = 3.29, p < 0.001).
Household composition exerted a nuanced and non-linear influence on net use.
Larger household size was associated with reduced odds of net use, consistent with competition for limited sleeping spaces or nets within households. In contrast, a greater number of living children was associated with increased odds of net use, suggesting preferential allocation of nets toward young children. The statistically significant spline terms indicate that these relationships were non-linear, with diminishing marginal effects at higher levels of household size and parity, highlighting the complexity of intra-household decision-making.
Maternal age showed an independent association with child net use.
Children of older mothers (aged 40 years and above) had significantly lower odds of net use compared to those born to adolescent mothers (15–19 years).
Cultural and ethnic differences remained relevant after adjustment.
Significant heterogeneity in net use was observed across ethnic groups, with children from Mijikenda/Swahili, Kamba, and Kisii households exhibiting higher odds of net use compared to the reference category.
Several commonly cited determinants were not independently associated with net use.
Place of residence (urban versus rural) and maternal educational attainment did not show statistically significant associations in the fully adjusted model.

4 Discussion

4.1 Interpretation

The findings strongly support the primary hypothesis. The dramatic gradient by malaria zone aligns with the perceived need construct, where households in high-transmission areas are more motivated to use nets due to immediate, tangible risk . The wealth gradient supports the household capacity construct, where economic advantage may facilitate net acquisition, maintenance, and intra-household allocation, even in a context of mass distribution .

The negative association with household size, combined with evidence of non-linearity, suggests that competition for limited nets within households intensifies rapidly as household size increases, followed by diminishing marginal effects. Similarly, the non-linear positive association with the number of living children indicates prioritization of young children for net use, with the strongest effects observed at lower parities. The lack of a rural-urban disparity is noteworthy and may reflect the success of equitable national distribution campaigns, shifting the primary determinants to household-level factors.

4.2 Limitations

This study has limitations inherent to cross-sectional DHS data, including an inability to establish causality. Net use was self-reported and subject to recall and social desirability bias. Important unmeasured confounders, such as net durability, specific campaign exposure, and intra-household decision-making dynamics, were not available.

Additionally, non-linear associations were identified for key household composition variables; although average effects are reported, these may mask variation across the distribution of household size and parity.

5 Conclusion

5.1 Conclusion

At the national population level in Kenya, the use of mosquito nets by children under five is predominantly determined by geographical malaria risk and household socioeconomic status. Household composition and cultural factors provide additional, significant explanatory power. This suggests that while distribution has been widespread, equitable use is influenced by persistent structural and contextual determinants.

5.2 Recommendations

Programming: Malaria control programs should move beyond blanket distribution toward stratified, risk-based social and behavior change communication aligned with Kenya’s National Malaria Strategy. Messaging in high-risk areas should reinforce the protective benefit, while in lower-risk areas, it should focus on sustaini
Policy: Efforts to reduce wealth-based disparities are essential. This could include complementary poverty-alleviation programs or ensuring that free net distribution mechanisms are truly accessible to the poorest households without hidden costs.
Research: Future studies should employ mixed methods to explore the qualitative “why” behind the quantitative associations found here, particularly regarding intra-household allocation and the role of maternal age and ethnicity. Longitudinal research is needed to understand how net use changes over time and with repeated distributions.

By systematically adhering to this research framework, the findings provide credible, evidence-based insights that can inform more effective and equitable malaria prevention strategies in Kenya and similar contexts.

5.0.0.1 References

1. WHO. (2023). World Malaria Report. Geneva.

2. Andersen, R.M. (1995). Revisiting the behavioral model and access to medical care: does it matter? Journal of Health and Social Behavior, 36(1), 1-10.

3. Larsen, D.A., et al. (2020). Do household wealth and household size moderate the equity impact of insecticide-treated net distribution campaigns in Malawi? Malaria Journal, 19, 219.

4. Koenker, H., et al. (2018). Quantifying seasonal variation in insecticide-treated net use from household survey data in Tanzania. Malaria Journal, 17, 247. 5. Kenya National Bureau of Statistics (KNBS) & ICF. (2023). Kenya Demographic and Health Survey 2022. Nairobi, Kenya, and Rockville, Maryland, USA.