Housing modifications for heat adaptation, thermal comfort and malaria vector control in rural African settlements

Climate change and associated extreme weather events are adversely affecting the health of the human population globally¹. Sub-Saharan Africa (SSA) is projected to experience some of the highest increases in temperatures^2,3, which are expected to result in severe consequences due to the region’s limited adaptive capacity and preexisting health burdens⁴.

Although outdoor exposures to extreme heat are widely studied, household indoor environments, where people spend most of their time, are often overlooked⁵. In rural SSA, houses are often constructed with materials and designs that are poorly suited to dissipate heat, such as metal roofing, limited shading and inadequate ventilation⁶. Furthermore, indoor heat gain is largely driven by raised outdoor temperatures and solar radiation⁷ as well as internal heat generated by building occupants, installed equipment and activities such as cooking. Elevated indoor temperatures not only lead to heat stress and reduced productivity but also increase reliance on active cooling systems such as electric fans, thereby increasing energy demands and the use of fossil fuels⁸.

In recent decades, several validated indices have been developed to assess heat-related health risk; one of these indices is the HI, which combines temperature and humidity^9,10. In addition, to mitigate heat stress resulting from environmental heat load, it is important to ensure that the ambient temperature in an indoor living space is maintained within the thermal comfort zone (TCZ), which is the range within which most of the occupants are predicted to feel thermally comfortable. The TCZ is based on temperature, humidity, radiant temperature and airflow, according to different complex models¹¹.

Passive cooling strategies are critical, particularly where active systems are unavailable and/or unaffordable⁸. A common architectural approach to indoor cooling is the inclusion of openings in a building´s exterior surface—for example, doors, windows and vents⁸. These openings facilitate indoor and outdoor heat transfer through ventilation—that is, both cross-ventilation and stack-ventilation. The size, number and orientation of these openings relative to the building layout substantially influence their cooling effectiveness. In practice, however, many households in SSA have openings that are often undersized, poorly oriented or absent due to cost constraints, cultural preferences or structural concerns, limiting their capacity to mitigate indoor heat accumulation.

Although essential for ventilation, these openings also serve as entry points for mosquitoes, particularly Anopheles species that transmit malaria^12,13, further risking the health of occupants. The world observed an increase of 11% in malaria burden in 2023, above the 2022 levels¹⁴. The SSA region bears the highest burden, accounting for 94% of malaria cases and 95% of estimated mortality globally¹⁴. Although the exact causes of this rise are multifactorial, climate change is a key driver, altering temperature, humidity and rainfall patterns in ways that expand mosquito habitats, enhance vector survival and increase malaria transmission risk¹⁵. Additionally, climate-induced temperature increases further reduce long-lasting insecticidal net (LLIN) use during hot nights¹⁶, compounding malaria risk.

Malaria-transmitting mosquitoes are highly adapted to feeding indoors at night^17,18. Field studies from western Kenya have shown that human−vector interactions that lead to malaria transmission still occur mostly indoors, throughout the night and morning, despite widespread use of LLINs^19,20. Therefore, house designs that simultaneously prevent mosquito entry and maintain thermal comfort is a priority in SSA²¹.

House modifications involving the screening of doors, windows and eaves with insect mesh were previously demonstrated to reduce malaria transmission^22,23. Consequently, the World Health Organization has given an interim recommendation for the use of untreated insect screens as a supplementary control measure against malaria²⁴. However, addition of screens to these openings could potentially reduce airflow and increase indoor temperatures, further jeopardizing the comfort of the occupants. An experimental study in The Gambia assessing how house design affects malaria mosquito density, temperature and humidity observed that screening of eaves of both thatched and metal-roofed houses increased indoor temperatures, whereas increasing ventilation in metal-roofed houses made them as cool as thatched houses with open eaves²¹.

Critical considerations on house design and its impact on indoor thermal comfort and mosquito control are necessary in every building design. This pilot, mixed-methods study, conducted in Siaya County, western Kenya, tested the feasibility and acceptability of low-cost, sustainable house modifications aimed at improving passive cooling and reducing mosquito entry. The study also evaluated implementation costs, community engagement strategies and participant recruitment processes. Findings from this pilot study will inform a forthcoming larger cluster randomized trial (Wellcome Trust grant number 226750/Z/22/Z), involving approximately 300 households, which will assess the epidemiological and environmental impacts of scalable housing interventions on malaria transmission and indoor thermal comfort.

Housing modifications

Rural houses were typically rectangular in shape, with walls made of mud and sticks and roofs of corrugated iron sheets. Unmodified (control) houses had open eaves, unscreened or absent windows, unscreened doors, no ceilings and unpainted roofs (Fig. 1a). Mat-ceiling houses were fitted with horizontally installed papyrus reed mats across the roof space, above the eaves, as ceiling for cooling (Fig. 1b). Cool-roof houses received white reflective paint on the roofs to reduce heat gain (Fig. 1c). Cross-ventilation houses received new, screened windows to enhance indoor airflow (Fig. 1g,f). All intervention houses with cool-roof, mat-ceiling and cross-ventilation received screened doors (Fig. 1d) and eave screening (Fig. 1e) for mosquito control, whereas control houses did not receive any vector screening. For cool-roof and mat-ceiling houses, no additional windows were added; existing ones were screened for mosquito control.

Baseline assessment of house characteristics

Table 1 presents a description of the study structures before modification. All structures had corrugated iron roofs, with most (97.5%, 39) having mud walls. Most of the houses (67.5%, 27) did not have windows, 10.0% (4) had one window, 15.0% (6) had two windows and 7.5% (3) had three windows. Most of the houses (97.5%, 39) had open eaves on all sides, and 2.5% (1) had open eaves on two sides of the wall, typical for gable roof design. Most of the houses had double rooms (62.5%, 25), followed by single rooms (27.5%, 11) and three rooms (10.0%, 4). The houses were evenly distributed across the study arms, with no significant differences in wall type, roof type, number of windows, presence of eaves and number of rooms.

Indoor thermal environment

We measured the indoor HI, temperature and relative humidity a few days before the start of the modification (Extended Data Table 1). These data indicate a high degree of homogeneity of the indoor environmental conditions, regardless of whether data from a whole day, during daytime hours or at night were considered.

Figure 2 shows daily mean values (±95% confidence interval), temperature and relative humidity for the different study arms, after the housing modifications. Data shown correspond to the representative post-intervention period—May—1 month after implementation.

HI and temperature were higher during the day (7:00−19:00; Fig. 2a,c) than at night (Fig. 2b,d). By contrast, humidity was observed to be higher at night than during the day (Fig. 2e,f). The HI was highest during the day in houses with cross-ventilation (mCV) and in control houses (mCL), whereas, at night, mat-ceiling houses (mMC) consistently recorded the highest HI values across all groups (Fig. 2a,b). Cool-roof houses (mCR) showed the lowest HI and the lowest indoor temperature but also the highest relative humidity, regardless of time of day (Supplementary Tables 1 and 2). Table 2 reports descriptive statistics for HI across study arms.

The findings suggest that cool-roof houses reduced indoor heat out of all interventions. In addition, houses with cross-ventilation and mat-ceiling had the highest indoor heat during the day and night, respectively.

Thermal comfort

Next, we measured thermal comfort of the housing designs using psychrometric charts. The psychrometric charts were based on the indoor environment data, collected on each day (24 hours) throughout the month of May, across all the study arms (Fig. 3). Extended Data Fig. 1 shows the same data separately for daytime (7:00–19:00) and nighttime (19:00–7:00) measurements. Each point in the charts represents a paired measurement of dry-bulb temperature and humidity ratio, and the red polygons indicate the TCZ. Across the 24-hour measurement periods over multiple days, most houses with cool-roof and mat-ceiling were within the TCZ, whereas control and cross-ventilation houses were mostly above the TCZ. Similarly, during daytime measurements, cool-roof and mat-ceiling houses were largely within the TCZ (Extended Data Fig. 1a). At night, mat-ceiling houses fitted best with TCZ (Extended Data Fig. 1b), maintaining the warmest indoor conditions.

Extended Data Table 2 summarizes the predicted mean vote (PMV) values across study arms for the same time window. PMV is a thermal comfort index that predicts the mean thermal sensation of a group of people on a seven-point scale from −3 (too cold) to +3 (too hot), with 0 being neutral. The daily PMV values were higher than +3 for control, mat-ceiling and cross-ventilation houses and lower for cool-roof houses. Additionally, Extended Data Fig. 2 provides a graphical description of the PMV with respect to daytime and nighttime data. During the day, houses with cool-roofs had PMV values closest to zero (0 represents neutral or comfortable thermal sensation). At night, houses with mat-ceiling had PMV values closest to the comfortable thermal sensation zone. This observation, therefore, coincides with the psychrometric charts.

The results suggest that houses with cool-roofs or mat-ceilings provided the greatest thermal comfort, in contrast to control and cross-ventilation houses, which did not. From a health perspective, these findings suggest that houses with cool-roof or mat-ceiling designs could be better interventions to prevent heat-related illness.

Mosquito data

We assessed the entry of both Anopheles and Culex mosquitoes, before and after housing interventions. Anopheles gambiae s.s, A. funestus and Anopheles arabiensis of the genus Anopheles are the primary species that transmit malaria in the region. Both A. gambiae s.s. and A. arabiensis are part of the A. gambiae s.l. (complex). No further analysis was performed to distinguish between the two species and are reported as A. gambiae s.l. Culex is the genus of mosquitoes of several species that transmit a number of arbovirus infections. A total of 8,297 Anopheles and 2,840 Culex mosquitoes were collected indoors by Centers for Diseases Control and Prevention (CDC) light traps. More A. funestus were collected prior to any house modification (n = 6,958; 84.2%) than after the intervention (n = 1,303; 15.8%). On the contrary, fewer Culex species were collected before modification (n = 974; 34.3%) than afterwards (n = 1,866; 65.7%); however, the numbers were higher in unmodified houses compared to modified ones. Only two A. gambiae s.l. were collected in the pre-modification period, and 34 were collected in the post-modification period. Therefore, the species was not subjected to further analysis because the numbers were too low for any meaningful statistical analysis and interpretation.

The baseline assessment of mosquito numbers showed no difference in indoor vector densities among the different study arms. After any modification, screening significantly reduced the number of female A. funestus mosquitoes and lowered the counts of Culex mosquitoes. In modified houses, the mean nightly density of female A. funestus was 1.3 mosquitoes per house compared to 5.6 in control houses, representing a 77% reduction (P < 0.001). Similarly, the density of Culex spp. decreased from an average of 4.8 (before modification) to 2.0 (after modification) mosquitoes per house, a 58% reduction (P = 0.004), as reported in Fig. 4 and Extended Data Table 3.

Cost of modification

Extended Data Table 4 presents the costs incurred in housing modification for vector proofing (screening) and passive cooling. The cost of modifying a single house for passive cooling and vector proofing was estimated at US $189.12. Modification for passive cooling using cool-roof was the most expensive at approximately US $145.14 per structure, followed by mat-ceiling at US $79.00. The cost of a single window for cross-ventilation was estimated at US $37.36, and the cost per structure varied by the size of the house and the number of windows required. The exchange rate for US Dollars to Kenya Shillings was US $1 = KES 133.66 as of April 2023 when the modifications were conducted.

Community knowledge, attitude and perception on house modification

Knowledge, attitude and perception surveys were conducted on 28 and 26 households before and after modification, respectively. The participants interviewed were mostly women: 75.0% (21/27) before modification and 76.9% (20/26) after modification. The mean age of the participants was 52 years in the pre-modification period and 54 years in the post-modification period, with the median age being 51 years and 58 years, respectively. The highest level of education attained by the study participants was primary education. The main source of income for the study population was small-scale farming (53.6% (15/28)) and small-scale businesses (25.0% (7/28)). Other sources of income included skilled labor at 7.1% (2/28), donations at 7.1% (2/28) and charcoal burning at 7.1% (2/28). The assessment of community knowledge, attitude and perception of housing modification for mosquito control and thermal comfort was targeted at the 26 intervention houses. At the beginning of the survey, 42.9% (12/28) of the households reported having windows. This increased to 61.5% (16/26) in the post-modification period as some houses received windows for cross-ventilation.

All 26 respondents surveyed in the post-modification period reported that house modification reduced indoor temperatures and mosquito numbers. Most of the respondents (76.9%, 20/26) perceived a substantial reduction in indoor temperatures after the modifications. An additional 15.4% (4/26) reported moderate reduction, and 7.7% (2/26) reported only a slight reduction. Most respondents in both the pre-modification (64.3%, 18/28) and post-modification (65.4%, 17/26) surveys mentioned the importance of building designs that reflect their cultural values, especially when building the initial ‘starter’ home. This first structure is typically built within a day, guided by specific cultural practices, and is considered temporary, with the intention of replacing it afterwards with a more permanent dwelling. Before modification, 14.3% (4/28) of the respondents, and 19.2% (5/26) of the respondents after modification, felt that the reflection of cultural values in building design was somewhat important. In addition, 21.4% (6/28) and 15.4% (4/26) of respondents in the pre-modification and post-modification surveys, respectively, reported that they did not prioritize cultural considerations in housing design.

The majority (71.4% (20/28) and 96.2% (25/26)) of the respondents in the pre-modification and post-modification periods, respectively, expressed willingness to adopt new house designs that are different from their cultural preferences if they help to improve thermal comfort and mosquito control. The proportion of respondents willing to use their family resources increased from 78.5% (22/28) in the pre-modification survey to 84.6% (22/26) in the post-modification survey.

The main importance of windows, identified by participants, was to allow fresh air and light into the house. Additional benefits mentioned included access to the house when the door is locked, visibility of the outdoors and the ability to use windows as an outlet for business purposes if part of the house served as a shop. The main reason given for not having windows installed was inadequate finances. Other reasons provided included: ‘The current house is a temporary structure before building the desired house’; ‘I did not think about it then’; ‘it is a cottage, so it has no window’; ‘I just built it without windows’; ‘I left to fix later’; and ‘fear of house collapsing if the windows are installed after construction’. The disadvantages of having windows stated during the pre-modification survey were insecurity, mosquito entry, allowing dust into the house and superstitious beliefs, such as being used by witches to harm people in the house. However, after building new windows in the intervention phase, most participants noted that there were no real disadvantages to having windows. Among the houses that had windows, 66.7% (8/12) of respondents at baseline reported opening them daily compared to 92.9% (13/14) after the modification.

This study demonstrates that targeted housing modifications can effectively address the dual challenges of extreme heat and malaria transmission in rural African settings. The integration of passive, sustainable, cooling housing modifications with vector proofing resulted in significant reductions in indoor HI and mosquito densities and improved thermal comfort levels compared to controls. These findings inform climate change adaptation and malaria control in resource-limited settings. Climate models project that equatorial SSA will experience temperature increases of at least 1−2 °C in the coming decades²⁵. Although this temperature rise may seem small, it has important implications for human health and functioning. This is because the human body must maintain its core temperature within a narrow range of around 37 °C ± 0.5 °C. Even slight deviations of ±1.5 °C from this core temperature can substantially affect physical and mental performance, and persistent larger deviations exceeding 3 °C can be life-threatening^26,27. The body can tolerate additional heat stress for only brief periods. Consequently, even small but sustained increases in environmental temperature, especially when protracted for a long time, directly impact human health, causing a range of conditions from dehydration and cardiovascular stress to physical and mental impairment, exhaustion and heat stroke. In rural African communities, where active cooling options are limited, passive cooling approaches that reduce indoor heat exposure represent a critical adaptation strategy for protecting health.

Among the passive cooling options evaluated, cool-roofs effectively reduce HI during the day, but they are unlikely to improve comfort at night because the thin mud walls have low thermal mass and do not store much heat. Mat-ceilings appear to better insulate houses and, therefore, maintain consistent temperatures during the day and night. However, they are likely to be ineffective in warmer months as the ceiling prevents heat from radiating out at night. Mat-ceilings might, therefore, induce heat stress and perform the worst, especially due to heat trapping and reduced indoor space for air circulation. The use of cross-ventilation for cooling in this setting would benefit from co-creation on window design and community sensitization to promote appropriate use of windows for indoor cooling.

The application of white reflective coating on iron sheet roofs, which were the hottest part of the house during the day, effectively reflected solar radiation, reducing indoor mean temperatures by 2.8 ± 0.2 °C compared to control houses. This finding aligns with studies from other tropical regions: in North Australia, light-colored roofs showed 30% lower total heat gain than dark-colored ones, and, in The Gambia, white-roofed houses were consistently cooler and more comfortable than those with bare metal roofs^28,29. Consequently, passive cool-roofs are observed as a means of reducing energy cooling loads to satisfy human comfort requirements in hot climates³⁰. The measure of how much solar energy a building’s exterior surface reflects away has been applied on walls as well to reduce daytime heat gain³¹. The effectiveness of cool-roofs is particularly relevant in SSA, where corrugated metal roofing is widespread in both rural and urban buildings. These roofs typically allow substantial solar energy flux into living spaces³². With global warming intensifying this heat burden, cool-roofs offer a practical, passive solution for thermal comfort in resource-poor communities. Consistent with other studies, the data presented here demonstrate the benefit of reflective roofs in lowering indoor HI in resource-poor, rural African communities.

Although cool-roofs had the lowest indoor temperatures, they had over 10% higher relative humidity levels compared to control. Additionally, relative humidity—the ratio of the current absolute humidity to the maximum possible absolute humidity at a given air temperature—was observed to increase when air temperature levels decreased. It is generally recommended to keep relative humidity indoors between 40% and 60% for comfort and health³³; studies have shown that higher relative humidity levels of between 50% and 70% are fairly tolerated at lower temperatures of between 26 °C and 30 °C³⁴. But when temperatures are elevated above 37 °C at 70% relative humidity, heat stress sets in with increased heart rate, respiratory rate and mean skin temperature³⁴. Although high relative humidity at lower temperatures is tolerable, it also promotes dampness and the growth of fungi and mites and creates moisture problems in indoor building materials with poor air quality, which altogether increases health risks. It is, therefore, essential to balance both temperature and humidity levels to achieve thermal comfort indoors. A combination of cool-roofs and natural ventilation has been recommended to increase indoor thermal comfort³⁰ by balancing both temperature levels and humidity. In this study, six of the 10 houses that received cool-roofs did not have windows, hence contributing to the high humidity levels observed in such houses. Sufficient ventilation in combination with other cooling options is, therefore, critical in achieving thermal comfort indoors.

Mat-ceilings reduced daytime HI by 1.3 °C but increased nighttime HI by 2.0 °C compared to control houses. Although the mats provided daytime insulation from roof heat, they likely trapped heat released by walls at night. Despite reduced headroom and air circulation, houses with mat-ceilings maintained the most stable daily temperature (23.2−29.9 °C) and humidity (56.5−72.7%) ranges, closest to human comfort levels (22−27 °C and 40−60%). Performance could be improved through roof space venting, better mat alignment and enhanced cross-ventilation.

Cross-ventilation was not achieved due to behavioral and structural barriers. Residents frequently closed windows for security reasons, particularly at night, and 67.5% of houses lacked windows altogether, at baseline. Additionally, the study was conducted during colder months of the year, and the residents are more likely to maximize the use of cross-ventilation for cooling in the hotter months of the year. These factors potentially prevented meaningful cooling, with cross-ventilated houses showing higher nighttime HI than controls, likely due to eave screening that reduced natural airflow²¹. The limited effectiveness highlights challenges in real-world housing interventions where security concerns and cultural practices influence resident behavior. Future designs should incorporate permanently open, secure ventilation systems or combine cross-ventilation with other passive cooling strategies³⁵.

House modifications significantly reduced indoor mosquito densities, with female A. funestus numbers 77% lower in screened versus unscreened houses. This effectiveness aligns with findings from multiple studies across SSA demonstrating that screening eaves, windows and doors successfully reduce mosquito entry^36,37. Although mosquito control in the Global North primarily focuses on outdoors due to widespread indoor screening³⁸, house modifications are particularly crucial in regions like western Kenya where mosquito biting remains high indoors despite bed net use^19,20. The importance of house screening is highlighted by recent findings showing that 87% of A. funestus and 88% of A. gambiae bites still occur indoors even when bed nets are used¹⁹. Other studies have demonstrated that mosquito activity peaks in the early morning as people leave their bed nets and extends into other indoor spaces such as schools^19,39. Because open eaves are the primary route for mosquito entry when doors and windows are closed at night^12,18, comprehensive screening of all entry points is essential for effective vector control.

However, screening can potentially reduce airflow and increase indoor temperatures, as demonstrated in The Gambia where screened houses were warmer than unscreened ones⁴⁰. Our study addresses this challenge by combining vector proofing with sustainable passive cooling strategies; the aim is to create living spaces that are both mosquito free and thermally comfortable. With the limitations of being an in-field pilot study (for example, small sample size, heterogeneity of houses and absence of airflow measures), future studies would need to demonstrate whether such an integrated approach could yield broader co-benefits beyond our measured outcomes, including sleep quality, economic productivity, heat stress, malaria incidence and other health outcomes.

The adoption of such interventions would depend not only on technical effectiveness but also on community uptake. In our study, acceptance of the modifications was high, with 85% of households expressing a willingness to invest their resources on similar improvements. This represents an increase from the 79% observed during pre-intervention, suggesting that experiencing the benefits firsthand strengthened community buy-in. Furthermore, 96% of the respondents expressed willingness to adopt new house designs that differ from traditional preferences if they improved thermal comfort and mosquito control. This openness to innovation, while maintaining respect for cultural values, suggests promising potential for scaling up such interventions. Implementation costs and structural challenges present important considerations for scaling. The total cost of screening per house (US $189.12) protects approximately four individuals per house throughout the day and night and lasts longer, making it relatively more cost-effective compared to repeated distributions of bed nets every 3 years at approximately US $5−7 per net for every two people with only partial protection. However, challenges encountered during installation, particularly with door modifications and existing structural limitations, highlight the need for standardized approaches and skilled implementation. Future interventions might benefit from incorporating these modifications into initial construction rather than retrofitting existing structures.

This study has several limitations. Although airflow is critical for thermal comfort and a major functionality of cross-ventilation, indoor airflow measurements were not performed in this pilot study, due to study constraints. We will consider this aspect in the upcoming larger randomized controlled trial. Direct airflow measurement in this setting presented technical challenges due to behavioral factors—residents frequently closed windows for security and space needs, which would have confounded continuous airflow monitoring. Future studies should incorporate standardized airflow measurement protocols that account for these behavioral patterns. The use of preexisting structures that vary in size and shape may have resulted in indoor climate variability among the houses. This may limit the generalizability of the findings from this evaluation. However, retrofitting existing houses is less costly and most suitable for local context. Furthermore, the approach promotes community participation and ownership and preserves the community’s cultural values and practices in housing construction.

Another limitation is that the study was conducted during a cooler period (April−June) rather than during the hottest months (January−March), which may have underestimated the full thermal stress experienced by residents during peak heat periods. The seasonal timing was chosen for logistical reasons, which might limit the generalizability of findings to year-round thermal comfort assessment. Due to cultural predisposition to leave doors open during the day, some households were observed to prop their door screens open, which potentially reduced the effectiveness of the screened doors in limiting mosquito entry into the house.

Additionally, to provide a more comprehensive understanding of thermal comfort, we provided psychrometric charts and PMV based on ASHRAE Standards 55. However, we could not determine if such standards, developed in high-income countries where residential buildings are different (for example, larger and made of concrete/cement), might not be fully relevant to rural SSA homes (for example, smaller and mostly made of mud), due to relevant structural differences. The short evaluation period also limited our ability to assess long-term durability and performance across different seasons. Although we observed significant reductions in vector populations, the study duration precluded assessment of epidemiological impacts on malaria transmission. Collection and counting of mosquitoes were performed by the same staff; hence, blinding was not possible. However, potential bias in mosquito data was minimized by standardized measurement protocol and instrument, consistent timing of collections and data validation procedures. Due to the small sample size and the short duration of this pilot study, a cost-effectiveness evaluation and detailed analysis of confounding factors on mosquito catch sizes were not included. The limitations observed in this feasibility pilot evaluation will be addressed in a cluster randomized trial, with the most effective modification being applied in approximately 300 houses.

In conclusion, housing modifications that combine passive cooling with vector proofing represent a sustainable approach to improving health outcomes in rural communities. However, successful scaling will require careful consideration of humidity control, structural challenges and implementation costs. For future scale-up and implementation, the housing modifications would need to be optimized for different climatic zones, evaluating long-term durability and cost-effectiveness and the epidemiological impacts considered, in order to facilitate broader adoption of these interventions. The dual benefits of reduced heat stress and mosquito control, coupled with strong community acceptance, suggest the potential for climate change adaptation and malaria control in SSA.

Study site description and population characteristics

This study was conducted in Kadenge Ratuoro village (0.0242° N, 34.1749° E) with an altitude of 1,140 m, in Alego Usonga sub-county, Siaya County. The study area is within the Kenya Medical Research Institute (KEMRI)/CDC Health and Demographic Surveillance System (HDSS)⁴¹. The residents are of the Luo ethnic group, subsist on farming, fishing and trade and live in small houses, clustered into family social units of relatives called compounds. Supplementary Fig. 1 is a pictorial and technical representation of an unmodified house in the study area. Houses are typically rectangular and constructed of stick frames (wattle), compacted soil or cement foundation and dirt or cement floor^37,42. The walls are either mud or cemented with an average thickness of approximately 0.20 m. The roofs are mostly of corrugated galvanized iron sheets. A few houses have either thatched or clay tile roofs. Most of the houses have two rooms, with an average internal floor area of 19.88 m² and a headroom of 2.24 m. Doors are approximately 0.85 × 2.02 m (1.71 m²) and are either unframed or framed with wood to create a jam and sash. Windows vary greatly in size, with an average size of 1.37 m² each; however, most of the houses do not have windows. For context, 67.5% of the houses in this study did not have windows at the start of the study. The eaves are usually open and are approximately 3.33 m² (0.22 m height × 19.40 m perimeter). The open eaves, in addition to doors and windows, provide ventilation into the houses, allowing entry of light and fresh air. Unfortunately, eaves are also the main route for unlimited entry of mosquitoes into the houses.

Malaria transmission in Siaya County is stable throughout the year, with a prevalence of 37% in children between 6 months and 14 years of age¹⁶, with the Alego Usonga sub-county having a prevalence of 50%. A. gambiae, A. funestus and A. arabiensis are the main malaria vector species in the region. Both A. gambiae and A. funestus are considered the primary malaria vectors in SSA because they feed more frequently indoors and on humans^43,44. A. funestus is capable of sustaining malaria transmission even in dry seasons due to its preference for permanent breeding habitats⁴⁵. A. arabiensis, on the other hand, is a more opportunistic^43,44 feeder with lower vectoral capacity compared to the first two. The vector is, however, capable of sustaining residual malaria transmission where both A. gambiae and A. funestus have been controlled⁴⁶. The region has a bimodal rainfall pattern, with long rains between March and May and short rains between October and December. The region has typical SSA tropical climatic conditions. A temperature suitability index for malaria transmission shows that the western Kenya region has ambient temperature and adequate rainfall suitable for endemic malaria transmission^16,47. The average wet-bulb temperature range within the HDSS has been estimated at between 17 °C and 35 °C.

Study design

We conducted a pilot, randomized field study assessing the impact of housing modification on indoor thermal comfort and mosquito numbers. The 40 houses were randomly allocated to mCV, mCR, mMC or mCL, n = 10 houses per arm. Quantitative and qualitative data collections were conducted before and after house modification.

Mobilization and recruitment of study households

Extended Data Fig. 3 illustrates the selection and randomization of the study houses. Home visits were conducted to enumerate and characterize houses within a section of the study village. A total of 47 compounds with 84 houses where people slept were enumerated. For every active structure in each compound, structural features, including wall type, roof type, presence or absence of ceiling, eave type and the number of windows, doors and rooms, were recorded. From the 47 compounds with a pool of 84 houses, to standardize data collection, n = 40 houses, all with identical characteristics—mud walls, open eaves, iron roofs and not more than three rooms—were selected for the study. A single house that met the selection criteria was identified per compound after a discussion with the compound head.

Randomization

Representatives of the 40 selected structures were invited to a meeting where a random, transparent and fair allocation of houses to the different study arms was conducted. Forty raffle tickets, labeled with the different study arms—mCR, mCV, mMC or mCL (10 tickets for each study arm)—were provided in a transparent container. The tickets were folded to conceal labels and were mixed up in the container. The house representatives were invited to draw a single ticket each. The houses were allocated to different treatment arms based on the selection of the house representative.

Structural modification

Structural modification of houses was conducted by a building professional identified based on previous experience with similar modifications. After the randomization of houses into the various study arms, the specific house characteristics, including floor area, numbers and sizes of doors and windows and the presence and sizes of open eaves, were collected to guide modification. The building expert established a workshop within the study area where all materials, including doors, windows and pieces of timber for eaves and ceiling modifications, were fabricated before installation in the various houses. Modifications were conducted based on the randomization for the three passive cooling options as follows (Fig. 1). For all houses that received new windows for cross-ventilation, a standard-size window was made based on the average size of rooms in the study houses. The modified houses were assessed by the structural engineer on the project to ascertain the quality of the changes. At the end of the data collection, all the control houses were modified with screened doors, windows and eaves and a cool-roof for indoor cooling.

Cross-validation

Cross-validation (mCV) was achieved by installing screened windows on the opposite walls of each room. This involved a complete overhaul of the existing windows and/or the creation of new ones if no windows existed in a house. The windows were made of cypress battens to create two batten window leaves, each measuring 360 × 600 mm with an area of 0.46 m². Each window leaf hung on two 50-mm ordinary hinges and was secured at the center using a barrel bolt latch. A fiberglass insect mesh (Streme Limited 12) laid between two sheets of coffee tray mesh (ALS Limited, Trading Division) was attached to the window frame for insect screening. The insect screen was installed outwards, whereas the window panels opened inwards. To install the windows, a section of the wall was cut to create space for the window if none existed before or adjusted if the original window was smaller. After the installation of the modified window, the remaining gaps in the wall were filled with mud to achieve the same finish as the original wall (Supplementary Fig. 2).

Cool-roof system and insect-proof housing

Iron-roofed houses were painted with a reflective white coat to reduce the amount of heat conducted into the house, hence lowering internal temperatures. Two coats of paint were applied to the roofs. Crown Roofmaster (Crown Paint Industries), an extremely durable, weather-resistant, self-priming acrylic resin-based paint with a waterborne topcoat and matte finish, was used.

Mat-ceiling and insect-proof housing

Locally made papyrus mats were installed horizontally, covering the roof space just above the eaves. Locally sourced round poles were used as structural bearers and cross-binders in the ceiling substructure (brandering) to support the mat finish. Timber battens were nailed below the mats to fasten them securely to the round poles forming the brandering (Supplementary Fig. 3).

Mosquito proofing

All houses that received passive cooling options were screened for mosquito control. The doors were modified by introducing wooden frames and panels in addition to the originally existing door. The existing doors opened inwards, whereas the newly introduced screened doors opened outwards. The screened door panels were made of wooden frames and fiberglass insect mesh laid between two sheets of coffee tray mesh (CTM). The doors hung on two self-closing hinges to keep them always closed (Supplementary Fig. 4). The windows were also screened as already described above. The eaves were screened by introducing a piece of timber at the edge of the wall just before the eave space and another piece of timber on the roof directly above the wall. Fiberglass insect mesh was then attached to the two pieces of timber, hence covering the eave space. An overlap of the insect mesh was tacked into the grooves between the timber and the corrugated iron sheet to block gaps between the corrugated iron sheet and the timber.

Indoor thermal environment

Daily temperature and humidity were collected every 15 minutes in both modified and control houses from March to July 2023 using Onset HOBO UX100-003 data loggers. The data loggers were placed indoors, hanging at approximately 1.5 m from the floor, in the sleeping zone, away from walkways and out of reach of children to avoid interference with the daily activities of household members. Data download was conducted twice during the study period. We used temperature and humidity to calculate the HI^10,48 based on an algorithm from the US National Weather Service, which is implemented in the R package weathermetrics⁴⁹. HI is also known as the apparent temperature. It is the heat felt by the human body when relative humidity is combined with the air temperature. More than a dozen indices have been developed in the recent past to describe the complex interaction among ambient air temperature, relative humidity, wind speed and radiation and their impact on human health and performance⁹. The Wet-Bulb Globe Temperature (WBGT) is a measure of heat stress that combines air temperature, humidity, wind speed, sun angle and cloud cover to indicate how hot it feels in direct sunlight⁵⁰. It is the most widely used index to assess heat stress in humans and to recommend rest/work cycles at different physical work intensities, especially under hot and humid conditions⁵¹. However, the WBGT is difficult and costly to measure directly, and most observations are sparse and based on estimations. In addition, the WBGT is more informative for workplaces. For these reasons, we opted for the HI, which combines temperature and humidity, the most crucial factors in this specific indoor environment, especially at night, as ventilation can be dramatically reduced by the absence of openings or windows closed due to safety reasons. To verify our approach, we conducted a small side test to compare HI measured with a HOBO device and a portable hygrometer designed to measure WBGT (model PCE-WB 20SD; PCE Instruments) directly. The results showed complete agreement; in addition, we made a direct comparison between the HI from HOBO and the WBGT index directly measured from such a hygrometer and observed the same pattern (see Extended Data Figs. 4 and 5 for comparison description).

To provide a more comprehensive overview of the impact of the house modification on thermal comfort, we provided psychrometric charts and predicted mean vote (PMV), metrics used to assess thermal comfort in building design and other applications. Both are established models for standards such as ASHRAE 55 and ISO. Input variables for the psychrometric chart were temperature and relative humidity from the data logger (tPost data) and humidity ratio calculated using the R package PsychroLib⁵². For these charts, we also added a TCZ based on the estimated effects of clothing, metabolism and air speed. Data points falling outside this zone suggest that occupants may be experiencing thermal discomfort. Boundaries of the TCZ were determined using the CBE Thermal Comfort tool, a free online tool that implements thermal comfort calculations from standards⁵³. We used the so-called ‘psychrometric (air temperature)’ (https://comfort.cbe.berkeley.edu) method and entered the following into the calculation mask: a mean radiant temperature of 24 °C (because ASHRAE Standard 55-2017 recommends temperature between 19.4 °C and 27.8 °C, mean 23.6 °C), a metabolic rate of 1 MET (because 0.8 MET corresponds to sleep and 1.2 MET corresponds to standing relaxed), an air speed of 0.1 m s⁻¹ and clothing varying between 0.5 clo and 0.6 clo (because these corresponds to typical summer indoor clothing). Based on this input, the limits of the TCZ at 0% relative humidity are 28 °C and 34.8 °C and, at 100% relative humidity, 23.4 °C and 28.0 °C⁵³.

The PMV is a point thermal sensation scale to quantify the degree of thermal comfort. It generally ranges from −3 to +3 (that is, −3 cold, −2 cool, −1 slightly cool, +1 slightly warm, +2 warm and +3 hot), with 0 representing neutral comfort. The farther the values are from zero, the greater the perceived thermal discomfort. The PMV was calculated using the method by Fanger implemented in the R package comf⁴⁹. The input consisted of six parameters: air temperature from the indoor data logger, radiant temperature (=air temperature), relative humidity from the indoor data logger, air speed of 0.1 m s⁻¹, clothing level of 0.5 clo and metabolic rate of 1 MET. Here, fixed values for air speed, clothing level and metabolic rate were the same as used for TCZ in the psychrometric chart (see above). It is important to note that, in the frame of this study, we focused specifically on day and nighttime measurements separately, because residents spend the longest time at home especially during the night, whereas, during daytime, they might be mostly outdoors, due to the different occupational activities.

Mosquito collection

Mosquito collection was performed indoors using the CDC Miniature Light Trap (model 512; John W. Hock Company). Collections were conducted twice before modification (baseline) and three times after modification in each house. The light traps were set in the sleeping area, next to an occupied bed net, at approximately 1.5 m from the floor. The traps were run from 18:00 to 7:00 the next morning. During the mosquito collection period, the collector administered a brief questionnaire to collect information on household characteristics, including roof type, wall type, presence of eaves, presence and use of malaria control products such as bed nets, presence of cattle and number of people who slept in the house the previous night. The location of each house was recorded using the Global Positioning System. The collected mosquitoes were identified to genus levels as either Anopheles or Culex and in each genus as either male or female based on morphological features. The anophelines were further identified morphologically^54,55 to species level as either A. gambiae s.l. or A. funestus s.l. All the collected mosquitoes were counted for each day, trap and house, and the counts were recorded based on genus, species and sex. All the mosquitoes that were morphologically identified as A. funestus s.l. were further identified as species by polymerase chain reaction (PCR)⁵⁶, and all were confirmed to be A. funestus s.s. The mosquitoes identified as A. gambiae s.l. were not subjected to further analysis because the numbers were so low. Data on household characteristics and mosquito information were collected on a CommCare (Dimagi, Inc.) application run on an Android tablet and transmitted to a project cloud server.

Collection of data on community knowledge, attitude and perception of house modification

A structured questionnaire assessing community perception, knowledge and attitude toward house modification for vector control and temperature reduction was administered to intervention households before and after house modification. Data were collected on the community’s building practices, including reasons for the inclusion or exclusion of certain building elements such as windows, eaves spaces, ceiling, wall and roof types. The questionnaire further assessed the community’s understanding of the relationship between the various building elements and the entry of mosquitoes into houses and indoor heat levels. Additional questions were administered to gauge the community’s perception of changes made to their houses, their potential contribution to house cooling and reduction of mosquito entry as well as their willingness to use their resources to modify their houses. In the post-modification survey, the perceived benefits and risks of house modification and the community’s willingness to continue using the modifications beyond the lifetime of the study were assessed.

Thermal imaging

Thermal images of the houses were taken using a FLIR T450sc camera (Teledyne). The images were taken by the same operator with the same camera on different days but within the same time slots. In addition, a camera built-in laser pointer was used to ensure a precise distance and pointing for subsequent images of different houses. In this way, confounders were minimized, ensuring a standardization of measurements in the photographed houses. Images were taken early in the morning before sunrise, at midday and in the evening after sunset to assess the source of heating in the houses at different times of the day. Images were taken centered on a specific reference point, identified by using a built-in laser pointer, to ensure reproducibility and to compare different times of the day. A set of n = 4 houses was selected, as easily accessible for photographs at all times of the day and representing the different structural modifications—mCR, mCV and mMC—as well as control (Extended Data Fig. 6).

Economic cost

The costs for housing modification were based on material and labor costs in cases where the raw materials were used directly in the modification process—for instance, in the application of cool-roof paint and installation of mat-ceiling and eave screens. Screened windows and doors were procured as complete units including the costs of installation. Material, labor and item costs for windows and doors were provided within the market rates.

Data management and analysis

Field data were collected using CommCare software run on Android tablets. Every participating house was identified by a unique code, and a collection code was generated by the tablet for every mosquito sampling effort. These codes were used to track data generated from the different study components for ease of management. Individual mosquitoes from each collection were placed in microcentrifuge tubes labeled with pre-printed barcodes and linked to the field data using a house code and a collection code. Results of species identification by PCR were linked to individual mosquitoes by the unique barcode label.

Statistics

Sample size consideration

Because this was an external pilot study, a sample size calculation was not performed. The objectives of the pilot study were to assess the feasibility of combining housing modification for passive cooling with malaria vector control, community acceptance and engagement, identification of implementation costs and challenges and development of data collection tools for the main trial. A sample of 10 structures per treatment arm was considered sufficient based on recommendations by Whitehead et al.⁵⁷ for an external pilot with a continuous outcome variable. However, the outcome of the pilot was not used to inform the sample size for the main trial.

Indoor environmental data and thermal comfort

To assess the indoor environment of houses before and after their modification (mCL, mCR, mCV and mMC), each temperature and relative humidity recording was divided into tPre and tPost. The tPre set includes data from 4 days at the end of March, and the tPost set data comprises data from the entire month of May. Each recording was then grouped into three different subsets: 24hTime, data covering a full day; dayTime, data covering periods from 7:00 to 19:00; and nightTime, data covering periods from 19:00 to 7:00. To ensure data quality, each was checked for completeness. Only datasets with a completeness of at least 95% were considered for further analysis. In addition to indoor temperature and relative humidity, the HI was calculated using the R package weathermetrics¹⁰. Means and 95% confidence intervals were calculated for temperature, relative humidity and HI at different timepoints (tPre: 24−27 March; tPost1: 1−7 May; tPost2: 8−15 May; tPost3: 16−23 May; tPost4: 24−31 May). Variability in indoor environmental data between the different groups of houses was assessed separately for tPre and tPost using linear mixed-effect models. The tPre model consists of time (days 1−4), modification (mCL, mCR, mCV and mMC) and their interactions as fixed effects and house as a random effect with random intercepts for house. The tPost model consists of time (tPost1, tPost2, tPost3 and tPost4) and modification (mCL, mCR, mCV and mMC) and their interactions as fixed effects and house as a random effect with random intercepts for house. Linear mixed-effect models were run using the R package afex. Analyses and graphical illustrations were performed using R version 4.3.3 (ref. ⁵⁸). Linear mixed-effect models were run using the R package afex. Estimated marginal means and contrasts were calculated using the R package emmeans. Figures were created using the R packages ggplot2 and cowplot.

Mosquito data

Vector abundance was assessed using descriptive statistics (means, proportions and 95% confidence intervals). For inferential analysis, we used rate ratios as a measure of intervention effect⁵⁹. Generalized linear mixed models (GLMMs) using Template Model Builder (glmmTMB) were fitted using negative binomial distribution for analysis of mosquito counts among different study arms. Study arm (treatment allocation), presence of deterrence (use of mosquito coil indoors, fire burning indoors and/or use of insecticide spray can) and presence of domestic animals indoors with their interaction were considered as fixed effects. Models were adjusted for repeated measures using the structure ID as a random effect. The rate ratios were obtained by exponentiating the model coefficient on the log rate ratio scale.

All data analyses were performed using R statistical software version 4.4.1, and the significance level was set at α = 0.05.

Confounding

The following confounders were accounted for in the analysis of mosquito data: use of mosquito coils, use of insecticide spray cans, open cook fires and the presence of cattle indoors. No confounders were considered in the heat stress analysis.

Ethics approval

The study received ethical review and approval from the KEMRI Scientific Ethics Review Unit (SERU 4796). Before being included in the study, written informed consent was obtained from every household for modification of houses and mosquito collection.

Inclusion and ethics statement

This project was designed through a partnership among KEMRI, Habitat for Humanity International and Charité – Universitätsmedizin Berlin. A section of Kadenge Ratuoro village in Siaya County was selected as a study site by KEMRI in consultation with the village elder. It is within the KEMRI/CDC HDSS and has high mosquito numbers and malaria burden given its proximity to Yala swamp. Allocation of interventions to the study structures was conducted randomly by lottery with the participation of structure representatives.

In study implementation, the KEMRI team led the intervention design, community engagement and data collection, whereas Habitat for Humanity International led the sourcing of funds and provided technical oversight for housing modifications. All team members collaborated in data ownership, intellectual property and authorship of the publication relating to the work.

As housing modification involved major alteration in the house design, measures were put in place to guard the safety and privacy of study participants. The housing modification workers were trained on safeguarding and good clinical practice. At the end of the data collection, the control houses were modified, and the study results were deiminated to all participants.

Previous work on housing modification in the region guided the design of this study. Additionally, findings from similar investigations in the regions were taken into account in the citations for this paper.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.