Study design
Two experiments were conducted in rural Gambia using two houses of an identical size and shape. The first experiment was designed to determine what methods could be used for measuring indoor airflow in two houses with metal roofs, closed eaves and badly-fitting doors. The second experiment used these methods to compare indoor airflow in the reference house in comparison with an alternative design. The study was carried out from 1st July 2017 to 18th August 2017, during the beginning of the rainy season, when high numbers of mosquitoes are present [8].
Study area
The study site was at the Medical Research Council Unit The Gambia’s field station at Wali Kunda (13° 34.440′ N, 14° 55.471′ W) on the south bank of the River Gambia in Lower Fulladu West, Central River Region, in The Gambia (Additional file 1: Figure S1). This is an area of open Sudanian savanna with extensive rice irrigation nearby. There is an intense rainy season from June to November, followed by a long dry season. Most clinical malaria occurs between October and December [22].
Experimental houses
Two experimental houses, positioned on a north-west to south-east axis, were constructed 10 m apart and 10 m clear of other constructions and vegetation on all sides. Houses were the average size of a single-roomed house in rural Gambia, obtained from a survey of 400 randomly selected houses in the Upper River Region of The Gambia [8]. The external base of each house was 4.20 m × 4.20 m in area and the 2.20 m high walls were constructed from sun-baked mud-blocks (each 16 cm high × 20 cm wide × 32 cm long), with a front and back door on opposite sides, perpendicular to the line of houses, each 175 cm high and 75 cm wide. The only non-traditional building components were reinforced concrete ring-beams (20 cm high) on top of the wall, which were added to prevent the mud-blocks cracking when the heavy roofs were moved, and the metal profiles used to construct the roof frame.
The study was explained in the local language to male villagers and healthy volunteers, aged over 15 years who provided signed-witnessed consent, were recruited to the study. Two volunteers in each house slept in separate beds under an intact long-lasting insecticidal net (Olyset, Sumitomo Chemical, Japan) from 21:00 h to 06:00 h.
House typologies
There were five typologies of single-roomed houses used in the study (Fig. 1). Firstly, the most common typology of housing in rural Gambia is a metal-roofed house with closed eaves and two badly-fitting doors served as the reference (MCB). Secondly, the traditional Gambian house with a thatched-roof, open eaves and two badly-fitting doors (TOB). Thirdly, a metal-roofed house with closed eaves, two badly-fitting doors and four eaves tubes (MCBE). Fourthly, a metal-roofed house with closed eaves and two screened doors (MCS). Lastly, a design based on a ventilated house used in the RooPfs trial [18], which had a metal-roof, closed eaves, two screened doors and screened windows (MCSG).
Houses with open eaves had a 3 cm gap between the top of the wall and the roof and those with closed eaves were blocked with a mixture of broken mud blocks and clay mortar. Traditional doors were constructed from a single panel of corrugate galvanized steel pinned to a wooden frame (2 cm × 2 cm). To simulate poorly-fitting doors, which are common in villages, we made a 2 cm gap along the top and bottom of each door. Screened doors were made of 25 mm square steel profiles treated with anti-corrosion paint. Two screened panels made from polyester netting (2 × 2 mm mesh) (Additional file 1: Figure S3) were placed at the top and bottom of the door, each 75 cm wide and 60 cm high and fixed to the steel frame with flat bars and bolts. Both houses had a third door with an extra tightly-fitting steel door mounted with rubbers seals in order to prevent any ventilation around the door. Eave tubes were locally made from 15 cm diameter polyvinyl chloride pipes with polyester netting (2 × 2 mm mesh). Four eave tubes were placed at a height of 180 cm on each of the two facades without gables. The ventilated metal-roofed house had two screened doors and two triangular screened windows (200 cm wide and 45 cm high at the apex), constructed with wooden frames (50 mm x 50 mm) and mosquito screening, and positioned in the gable ends of the building. Thatched roofs were pyramidal in shape and 3.6 m high and metal roofs, saddle shaped and 3.1 m high. The houses with thatched roofs had an internal volume of 37 m3 and the metal roof houses had 38 m3. Floors were beaten mud.
Experiment 1. Ventilation tests
Four methods for measuring ACH and one method for measuring IAS were tested.
Method A. Fan pressurization (blower door test)
ACH was estimated using a standard blower door (Model DB B, BlowerDoor, Serial #DB-CE1475, calibrated 17.10.2016, Certificate 8-DB-CE1475-10-17-16, The Energy Conservatory, Minneapolis, USA), based on a standard methodology [16, 17, 23]. The tightly-fitting steel door was opened for this experiment only and the blower door fitted in the door-opening (Additional file 1: Figure S3). In an empty house, the blower door was activated in order to create and maintain a pressure across the building shell, and the flow through the fan measured. Air leakage through the building was calculated and evaluated using the American Society for Testing and Materials method E779-19 [16]. The building leakage is usually defined by ACH50, the amount of air that escapes a building in m3/h when exposed to a positive or negative pressure of 50 Pa. In order to get a precise curve for the determination of the ACH50 a minimum of five target building pressures of 5 Pa, 10 Pa, 15 Pa, 20 Pa and 50 Pa was used. The test procedure requires both a depressurization and a pressurization test to be performed, giving a total of 10 measurement points for each final ACH50 value (Additional file 1: Figure S4). During an automated process a large fan in the blower door is either pulling or pushing air in or out of the building while measuring the positive or negative pressure via a pressure gauge. The two measurements are often similar but might differ if openings in the building are working as unidirectional valves. The experiment was carried out six times through a 24-hour cycle; at 02.00 h, 06.00 h, 10.00 h, 14.00 h, 18.00 h and 22.00 h. Each test took approximately 15 min including moving the blower door from one house to the other. The experiment was made on the same day as method B.
Method B. Artificial CO2 elevation & decay
Here CO2 levels were elevated using compressed gas and the rate of decay used to estimate ACH [12, 24, 25]. All ventilation openings including windows, open eaves and gaps in doors were sealed and all people vacated the room. A 10 mm diameter plastic hose hung from the ceiling, 180 cm from the ground, was connected to a CO2 gas canister kept outside the house. CO2 was released at a rate of at 500 ml/min for approximately 60 s while CO2 levels inside the house was measured from the outside using the software HOBOmobile connecting the CO2 data loggers to an Iphone via Bluetooth. In this experiment and all those measuring CO2 (Methods B, C and D), CO2 data loggers (HOBO® MX CO2 Logger #MX1102, Onset Computer Corporation. 470 MacArthur Blvd., Bourne, MA 02532, USA) were placed inside in the middle of the room, one metre above the ground. When a steady state of CO2 of 2000–3000 ppm was reached indoors, the CO2 was manually mixed by waving a 30 cm diameter plastic lid for 60 s. While this was done a second person outside the building ensured the safety of the person in the room. After manually mixing the CO2, the room was left empty for 20 min for the air to equilibrate before opening all sealed ventilation openings and starting the measurements. Temperature, relative humidity and CO2 were measured every 5 s for approximately 15 min (approximately 180 measurements) on both data loggers or until the CO2 concentration indoors has declined to values near to that outdoors. The experiment was carried out six times through a 24-h cycle; at 04.00 h, 08.00 h, 12.00 h, 16.00 h, 20.00 h and 00.00 h. Each test took approximately 30 min.
Method C. Natural CO2 elevation and decay
This experiment was similar to the proceeding one, but with CO2 levels raised through human activity, based on previous methodology [12, 25, 26]. All ventilation openings including windows, open eaves and gaps around doors were sealed. Three people walked around inside each house for 20 min or until the expected baseline CO2 level of 400–600 ppm was raised to 700–800 ppm, whereupon all people left the house and all sealed ventilation openings opened. Temperature, relative humidity and CO2 were measured every 3 min for 8 h (approximately 160 measurements) on both data loggers or until the indoor CO2 concentration decreased to values similar to those outdoors, whichever occurred first. The experiment started at 09.00 h and ended at 17.00 h.
Method D. Natural CO2 rise
This experiment measured the rate at which CO2 increased indoors due to human activity and was based on methods developed previously [25, 26]. At the start of the experiment the doors were opened for 30 min to ventilate the house. Two people slept indoors under a LLIN inside the house from 21.00 h to 06.00 h the following morning. Sleepers only left the room under exceptional circumstances. All ventilation openings remained open and temperature, relative humidity and CO2 were measured on both data loggers every 3 min for approximately 9 h (approximately 180 measurements). The experiment was carried out daily for 6 days and made on the same day as method C and E.
Method E. Hot wire thermo-anemometer
Here a hot wire thermo-anemometer was used to measure indoor air speed (IAS) based on an established methodology [27]. Two hot wire thermo-anemometers (ATP Instrumentation AAVM-8880 Hot Wire USB Logging Thermo-Anemometer, ATP Instrumentation Ltd, Ashby-de-la-Zouch, Leicestershire, United Kingdom) were placed inside the building on a tripod in the middle of the room, one metre above the ground. A CO2 data logger, which also record temperature and relative humidity, was placed just below the anemometer. All doors were opened for 30 min, ventilating the house. The internal air volume and surface area of all openings were calculated. Two adult men slept under a LLIN in each house from 21.00 h to 06.00 h. From 22.00 to 06.00 h they were instructed to rest or sleep under a LLIN. All ventilation openings were open. IAS, CO2, relative humidity and temperature were measured every 3 min for approximately 9 h (approximately 180 measurements). The experiment was repeated for six consecutive days and made on the same day as method C and D.
Experiment 2. Comparisons of ACH between housing typologies
Firstly, ventilation was compared in a reference metal-roofed house with closed eaves and badly-fitting doors (MCB) with a similar house (MCB). Secondly the reference metal-roofed house with closed eaves and badly-fitting doors (MCB) was compared with: (1) a house with thatched roof and open eaves and badly-fitting doors (TOB), (2) a metal-roofed house with closed eaves and badly-fitting doors plus eaves tubes (MCBE), (3) a metal-roofed house with closed eaves and screened doors (MCS) and (4) a metal-roofed house with closed eaves, screened doors and screened gable windows (MCSG). All five test methods described under experiment 1 were repeated on each pair of houses for six consecutive days. The indoor mean temperature and humidity was measured over six nights in each house typology.
Statistical analysis
For method A, mean ACH was determined using the six ACH50 values generated by the blower door software. For method B, C and D, ACH was determined using the natural logarithm CO2 concentration plotted over time. The slope of the line represented the ACH. Each test was done 5 times (method B) or 6 times (method C and D) and the mean calculated. For method E, ACH was calculated knowing the IAS (m/s), the volume of the building (m3) and the area of openings (m2). Six repeated tests were carried out over a 24-h cycle for each house and the mean value calculated. Comparisons between ACH values determined for each house each night were made using paired t tests to adjust for variation between nights. All analyses were performed using IBM SPSS v24 software.
Ethical considerations
The study was approved by the Gambia Government/Medical Research Council’s joint ethics committee (16th May 2016 and 16th March 2017) and the Department of Biosciences ethics committee, Durham University, UK (13th May 2016 and 29th June 2017).