A Double-Edged Sword? The Benefits and Hidden Costs of next-gen LLIN-Based Malaria Prevention - RCT Evidence from Benin

The study follows up a cluster-randomized trial in which villages were assigned to receive different types of long-lasting insecticidal nets (LLINs) for malaria prevention in Benin. The intervention consisted of the distribution of LLINs at standard coverage levels by trained health workers under Ministry of Health protocols. This follow-up study does not introduce any new intervention; it collects post-intervention data on households originally assigned to different LLIN types.

2027-07-01

2028-09-30

Health outcomes (malaria RDT results, hemoglobin levels, anthropometric measures), cognition and education outcomes (Raven matrices, digit span tests, school attendance, educational attainment), and net use behavior (declaration, observation and monitoring when possible)

Outcomes are measured using standardized survey instruments, biomarker collection, and cognitive assessments. Health outcomes are constructed from malaria rapid diagnostic tests (RDTs), hemoglobin measurements, and anthropometric indicators. Cognitive and educational outcomes are based on standardized tests of non-verbal reasoning, working memory, attention, and literacy/numeracy, and are aggregated into domain-specific indices after normalization where appropriate. Net use behavior is measured using a combination of self-reported use and direct observation, and is summarized in behavioral indicators and indices. Where relevant, outcomes within each domain are standardized prior to aggregation to facilitate comparison across measures.

Beliefs and perceptions regarding malaria prevention and long-lasting insecticidal nets (LLINs), including perceived effectiveness, safety, comfort, and attitudes toward net use.

Secondary outcomes consist of survey-based measures of beliefs and perceptions regarding malaria prevention and long-lasting insecticidal nets (LLINs), including perceived effectiveness, safety, and comfort. These outcomes are elicited using standardized Likert-scale questions and, where relevant, combined into summary indices after normalization.

This study uses long-term follow-up data from a cluster-randomized trial of long-lasting insecticidal nets (LLINs) conducted in Benin. Villages were originally randomized to receive different LLIN formulations. The present study collects new survey, biomarker, and cognitive data approximately seven to eight years after the original intervention. No new intervention is implemented; the study estimates long-run effects of original treatment assignment on child health, cognitive development, educational outcomes, and behavioral responses.

Not available

The original cohort was selected from approximately 55,000 households using simple random sampling within villages. Within each of 60 villages, 30 children aged 6 months to 9 years were selected using a random number generator, yielding an initial sample of approximately 1,800 children.

Randomization was conducted at the village level, with 60 villages assigned to one of three LLIN arms. Within villages, households and children were sampled for follow-up data collection.

For the Randomized Survey Experiment, second randomization at the household level, within those treatment arms with 3 framings again, with around 170 per framing group per initial treatment arm.

Yes

60 villages.

The original baseline sample included approximately 1,800 children. We expect to re-contact a subset of this cohort in the follow-up survey due to attrition (e.g., migration, mortality, or non-response), yielding an analytical sample of approximately 1,500 children and 1,500 adult caregivers.

20 villages with pyrethroid-only LLINs, 20 villages with pyriproxyfen-pyrethroid LLINs, 20 villages with pyrethroid-chlorfenapyr LLINs.

Accounting for clustering at the village level (60 clusters, ~1,500 children total), the study can detect a minimum standardized effect size of approximately 0.14–0.19 standard deviations for pairwise comparisons of main outcomes at the child level, corresponding to about 14–19% of a standard deviation.