The study was carried out in Nyabondo, a plateau area located in Upper Nyakach Division of Kisumu County, about 30km North-East of Lake Victoria. Nyabondo lies between an altitude of 1,520m and 1,658m above sea level, and 0° 23′ 0 S and 34° 58′ 60 E. The area is host to an estimated 34,000 people with a high population density of nearly 368 persons per square kilometer (km). The main livelihood activities in the area include subsistence farming, mainly of maize, cassava and sweet potatoes, small scale livestock rearing, as well as brick-making [31, 15]. Malaria is endemic in the Lake Victoria region, with a reported average prevalence of about 27% in 2015 [1, 4]. Previous entomological surveys in Nyabondo found that larval Anopheles mosquitoes bred in both temporary and permanent habitats with An. arabiensis being the main malaria vector species (99.3%), followed by An. gambiae (0.7%) [32 33]. The overall poverty incidence in Nyabondo is approximately 61%, perpetuated by inadequate agricultural technology, poor roads, water and sanitation systems . Poor infrastructure, extreme weather conditions and poor resource management are among other constraints in this area. The majority of houses in Nyabondo were of local design, constructed with mud walls and iron sheets and a plaster finishing made by mixing ash, mud and cow dung .
The study enrolled a total of 16 villages, with a ratio of 1:1 villages among the control and treatment sites. In each village, there were 10 houses enrolled with 80 houses in each of the two study arms (control and intervention) for both entomological and epidemiological assessment. 80 out of the 160 houses were randomly selected for later screening of the eaves while the other 80 were designated as the control houses. Randomization was carried out by use of computer-generated list, in permuted blocks of ten houses and 16 village blocks in the study site, with treatments in the ratio of 1:1 (control vs treatment). For entomological assessment, all indoor resting mosquitoes were collected monthly for two years using CDC light traps between 19:00hours and 07:00hours. For epidemiological assessment, all persons living in the house at the time of the survey were tested for the presence of malaria parasite using RDT . Malaria prevalence survey was done for four consecutive survey times (every three months) during the two years and included one baseline survey.
Sampling of adult mosquitoes was conducted from September 2017 to November 2018. The sampling was spread out over a period of 16 days each month, with all the ten houses in a particular village being sampled in one night. Baseline data was collected during the first such sampling occasion while the post-intervention sampling was conducted during the subsequent months. Sampling was conducted using CDC light traps with one light trap set up in an occupied bedroom per house and left to run for 12 hours overnight between 19:00hours and 07:00hours. Mosquitoes collected in the morning were killed using chloroform and morphologically identified in the field station as either belonging to anopheline or culicine group. Subsamples of anophelines were further identified morphologically to species level using keys of Gullies and De Mellon  with Gullies and Coetzee et al.  into sexed separated by physiological state, unfed and blood fed. Collections were done simultaneously for two years in both intervention and control houses and scheduled to span in both dry and wet seasons. The numbers of Anopheles gambiae vectors collected per trap night were used as a primary endpoint in assessing the efficacy of eave screening in reducing indoor vector densities.
Malaria prevalence surveys
Household malaria prevalence surveys were conducted after every three months from November 2017 to November 2018. The observed malaria prevalence was used as the epidemiological endpoint and diagnosis of Plasmodium falciparum was done by Rapid Diagnostic Tests (RDTs), using SD-Bioline malaria antigen P.f® test as recommended by the National Malaria Control Programme (NMCP) in Kenya . The tests were performed following RDTs instructions guides [34, 37]. Individual socio-demographic information of the household members were collected in addition to the house characteristics. Consent to participate in this study was requested from the participant during the study. Individuals were asked whether they had taken any anti-malaria medication prior to the survey day. Participants found to be positive for malaria parasites were treated by the Ministry of Health staff according to the National Guidelines for malaria management in Kenya. RDTs were performed by a trained staff of the Ministry of Health following the manufacturer instructions [34, 37]. Individual socio-demographic information of the household members was collected in addition to the house characteristics. Individuals were asked whether they had taken any anti-malaria medication prior to the survey day. Participants found to be positive for malaria parasites were treated according to the WHO Guidelines for malaria treatment and management .
Eave screening procedure
After randomization, and getting consent from household owners, screening houses were selected and fitted with a white polyester netting material designed to tightly and firmly fit onto pre-measured eave openings. An elastic cloth lining was sewn onto the edges of netting material by a tailor hired from the community and fixed onto position in the house eave using one inch nails as a harness. The screening work and roll-out roster was undertaken by project staff and local project youths, trained by an experienced consultant who was familiar with screening houses. Household occupants were encouraged during the entire study period to close windows and doors early in order to reduce mosquito entry into houses. Both Informed, verbal and written consent were sought from the head of each household before screens were installed into houses.
Data management and statistical analysis
Data was entered in MS Excel, cleaned and checked for errors by an independent person. Adult mosquito relative density was defined as the number of female adult anopheles mosquitoes per house per night. The 95% confidence intervals (CIs) for the mean adult mosquito relative density were estimated using negative binomial regression model adjusted for household clusters. Mean mosquito population densities and the relative abundance of different vector species were compared between 2017 and 2018. The effect of house screening intervention on adult mosquito density was estimated using generalized estimating equations (GEE), allowing for within-subject correlation using robust variance estimator to calculate standard errors (SEs). From the GEE model, we reported the incidence rate ratios (IRRs), the control arm (unscreened houses) was used as the reference against the experimental (screened houses).
For malaria epidemiological survey, Plasmodium falciparum infection was defined as a positive rapid diagnostic test (RDT) result. Proportion of individuals infected with malaria was calculated and the 95% confidence intervals (CIs) estimated using binomial logistic regression model that accounted for household clusters. The impact of house screening on infection prevalence was calculated and odds ratios (ORs) estimated using multilevel mixed effects logistic regression model while accounting for household clusters. All statistical analyses were performed using STATA version 14.1 (STATA Corporation, College Station, TX, USA) .