This was a pilot before-after trial in homes in the San Francisco Bay Area that had at least one school-aged child, a gas stove, and a venting range hood or over-the-range microwave and hood (OTR). The trial ran from July 2021 through June 2022 and assessed a cooking ventilation intervention consisting of (1) ensuring that the home had a working range hood that met both airflow and sound performance standards and (2) education about the hazards of cooking pollutants and the benefits of using the range hood whenever cooking occurs. The trial is registered with clinicaltrials.gov, #NCT04464720.

Because the funding for this study began (in March 2020) alongside the COVID-19 global pandemic, modifications to protocols were made to maintain as large a sample size as possible and advance the scientific goals while flexibly adapting to evolving university research guidelines, as well as the needs and comfort of study participants. These changes included (1) delaying the data collection start until the summer of 2021; (2) moving from a stepped wedge, randomized controlled trial to a before-after trial; (3) allowing some fluctuations in overall study duration and number of study visits; (4) having some study visits in the participants’ homes and others at an outdoor community site to minimize indoor contact time; and (5) expanding to including children regardless of asthma diagnosis.

Participant and study data were collected during a screening and follow-up phone call and 3-4 subsequent study visits. Written informed consent was obtained from all participants (as well as permission from the parent and assent of the child for their participation). During the first study visit, measurements of the range hood airflow and sound level were performed to assess if the existing range hood or OTR met the specified criteria (further explained in the ‘Intervention’ section below). Baseline data were collected from the air pollutant, stove, and range hood sensors (further delineated in the relevant sections below) that were installed during the initial visit through the intervention visit (either the second or third visit, depending on the protocol in place at the time of that participant’s enrollment). On average, the baseline lasted 13 days (range: 5–26 days).

At the intervention visit, all participants received a video-based educational intervention instructing the child and their guardian about cooking practices designed to decrease pollution exposure. Participants also received an associated infographic and magnet. If the range hood or OTR failed to meet the criteria at the initial visit, it was replaced either the same day or shortly before the intervention visit. The child performed spirometry and fractional exhaled nitric oxide (FeNO) measurements at the time of that intervention visit, reflecting their health status during the preceding baseline interval, and the parent completed questionnaires about both child health and cooking behaviors.

The post-intervention period then lasted from the intervention visit through the final visit (third or fourth visit overall; an average of 12 days, range 5–21 days). For each household, the post-intervention duration was usually similar to the duration of the pre-intervention study period, such that the intervention was delivered at roughly the midpoint of the entire study interval. The air pollution, stove, and range hood sensors remained in place for the entire data collection period–allowing for objective measurements of cooking intervals and range hood use both before and after the intervention–and were removed at the final visit. The child and parent completed the same questionnaires and health assessments as previously, as well as a close-out questionnaire.

We recruited children ages 6–12 living in the East Bay area of California. They were eligible for this study if the parent reported that they had both a gas stove and a venting range hood, i.e., one that extracted air from the kitchen to the outdoors. They were excluded from the study if they lived with a smoker who smoked indoors, if they knew they would not have stable housing for the period of the study, or if they were not fluent in English. In cases where more than one child per household was eligible for enrollment, all were enrolled in the study, data were collected for all children, and data from the child with the most complete lung function data were selected for analyses (see ‘Health Outcome Assessments’ below). In total, 18 participants completed the study, representing 14 distinct households with four sibling pairs.

Recruitment primarily occurred through East Bay pediatric clinics via recruitment fliers and information cards. At a few clinics, postcard mailings were also sent to potentially eligible patients. We also advertised for the study on an institutional website, through social media, and via fliers at community spaces (libraries, community centers, etc.).

Information about the child, household, and participating parent was collected in the phone interview using the ISAAC Environmental Questionnaire and demographic questions.

To decrease the likelihood that inadequate ventilation equipment would mask the effects of the educational intervention, the existing range hood or OTR were assessed by a contractor for replacement if they did not meet the target criteria. These were (1) minimum airflow of 100 cubic feet per minute (cfm; the California building code requirement as of Nov. 2020) and (2) sound pressure (loudness) no greater than 60 A-weighted decibels (dBA, a sound level often associated with annoyance) at a distance of 2 m. When feasible, the participant (and building owner, as applicable) was offered a replacement hood or OTR that met the specifications.

All study participants received education regarding use of the range hood during all cooking events. The educational intervention consisted of a 4-minute animated video featuring one of the physician researchers. The video provided background about cooking-related pollution and encouraged participants to (1) always use their range hood, (2) use the back burners when cooking on the stovetop, and (3) move other cooking appliances closer to the range hood to use it during those cooking activities as well. Local youth study assistants provided crucial feedback during the development of the video to ensure that the information was presented in a way that resonated with the local community. Research assistants played the video for the parent and child via a portable tablet computer during the intervention study visit. Participants were given a printed infographic on cardstock, as well as a small magnet to post in the kitchen, with the takeaway points from the video (both from a still frame of the video). Weekly, the participants received text messages reminding them to continue using the tips from the video. Adaptations of these educational materials have since been made publicly available at: https://wspehsu.ucsf.edu/projects/indoor-air-quality/.

Concentrations of time-resolved PM2.5, NO2, CO2, and other parameters were measured in the primary living area of each home (to represent the exposure received by the occupants) and typically logged every 1 min throughout the study interval. The air contaminants were measured along with temperature and relative humidity using an eLichens Indoor Air Quality Pro Monitor (eLichens, Grenoble, France). The eLichens has no integrated display and household members did not receive any feedback from the device other than an indicator LED to show that the device was operating. Integrated NOX and NO2 samples were collected during the intervals between study visits using Ogawa Passive Samplers (Ogawa USA [44], Pompano, Florida), placed in duplicate adjacent to the eLichens real-time monitor.

At the initial study visit, the airflow of each range hood or over-the-range microwave exhaust fan (OTR) present in the home was measured by a home performance contractor with expertise in ventilation equipment diagnostics to assess the need for intervention. Airflow was measured using a balanced-pressure flow hood method described by Walker and Wray [14] and used in several [45, 46] recent residential IAQ field studies. The measurement was made starting at the lowest fan speed setting and progressing to higher speed settings to identify the minimum setting needed to move 100 cfm. At that setting, a measurement was also made of the A-weighted sound pressure decibels (dB-A) at a 2-meter distance using a smartphone with an SPL Audio Tools app.

For the duration of the study interval, stove, oven, and range hood/OTR use were measured at 1-minute intervals using Lascar Easylog Thermocouple Data Loggers (with one thermocouple placed adjacent to each of the four cooktop burners), Onset Hobo temperature loggers above the cooktop (affixed to the range hood), and a Digisense Data Logging Vane Anemometer, affixed to the air inlet of the range hood. We used these temperature data to derive cooking intervals and the logged real-time airflow velocity data to monitor the frequency of ventilation use. Self-reported cooking data were also collected at each visit.

Baseline health information included an assessment of pre-existing medical conditions, including asthma (using the ISAAC Asthma questionnaire). Health outcomes included objective measures of respiratory health in all children at follow-up study visits.

With an EasyOne Spirometer, each child performed spirometry to complete three acceptable efforts (maximum eight attempts) in accordance with standard ATS/ERS performance criteria. Each effort was graded by two trained physicians in accordance with the 2019 ATS/ERS guidelines. If there were two or more efforts with acceptable quality and reproducible FEV1 and FVC (each ≤0.15 L), the best FEV1 and best FVC were used. If two or more efforts had usable quality and reproducible FEV1 but unacceptable FVC, only the best FEV1 was used, and FVC was considered missing.

Prior to spirometry, each child also had FeNO measured in accordance with ATS/ERS criteria by blowing a steady, sustained exhalation into a NIOX Vero device to complete two measurements with exhalation duration greater than 10 s and within 10% of each other (maximum eight attempts). If there were two measurements within 10%, this testing session was graded as acceptable, and the larger value was used. If there were one or multiple FeNO measurements of adequate duration (but none within 10%), this testing session was graded as usable, and these values were averaged to create the FeNO value for analysis. Children with asthma also completed the self-reported asthma control test with their guardian at each follow-up visit.

Because performing lung function maneuvers requires coordination of breathing in an unfamiliar manner, missing data for lung function test results is not necessarily related to underlying respiratory status and is common, especially in studies assessing children. Thus, lung function testing was performed in all interested children within the household and the child with the least missingness in the respiratory data was used as the primary study participant.

Ogawa samplers were hand-delivered from field to lab staff and kept cool on ice during transport. Twenty primary samplers were deployed, with 15 duplicates and five travel blanks. All Ogawa samples were extracted for analysis within 30 days from when the samplers were assembled. The samples were extracted and analyzed following the protocols provided by the company and as performed previously by our group (Ogawa USA, 2017) [44, 45, 47]. We computed mean temperature (T) and relative humidity (RH) for each Ogawa passive sampler deployment period based on measurements of the co-located eLichens monitors; as per the protocol, these values were then used to calculate the ambient NO2 concentrations from the measurement of the filter mass.

Data analyses were performed using Python and R in accordance with a prespecified data analysis plan posted to Open Science Framework (https://osf.io/vbm9g/?view_only=fe4ae01f9100486c9f78e93804b14583). Briefly, air pollution sensors and thermocouples were cross-calibrated before deployment, and all study data had an initial quality assurance, visual review, and alignment of timestamps for equipment deployed together. Data processing included algorithmic identification of burner use events from thermocouple data, cooking events (burner events within 30 min of each other), range hood use events from anemometer data, and pollutant events (with the associated time and pollutant statistics) from the eLichens data.

For each cooking event, we calculated several metrics, including: the duration of use for each burner (called burner-time), the number and position of burners used (front and/or back, if it could be clearly determined), the cooking event duration (from the start of the first burner event until the end of the last burner event, among burner events separated by less than 30 min), whether there were anemometer readings to indicate range hood use during the event, the number of minutes the anemometer recorded airflow through the hood, whether there were valid data for pollutants in that same time interval (PM2.5 in µg/m3; NO2 in ppb; CO2 in ppm), the maximal 10-min average concentration above the baseline for each pollutant during the cooking event (called the event peak), the event integrated concentration of the pollutant above the background level (concentration*min) during the cooking event, and a normalized event integrated concentration by dividing by the duration of the event (in burner-time).

Data were summarized within each household and study interval, which included the number of cooking events per day and their duration, as well as the number and percent of cooking events with range hood use. To capture episodes with intentional range hood use for most of the cooking event, we also determined the number and percent of events with the range hood on for ≥80% of the cooking time, as has been done previously [34]. For the pollutants, we calculated not only the means across the cooking events within each household and interval for each of the metrics calculated for the cooking events, but also the total integrated concentration (the sum of the integrated concentration across cooking events in that interval). This approximates the contribution of cooking to this pollutant in the house over the measured time interval.

Wilcoxon Signed Rank tests were used to compare the paired pre-intervention to the post-intervention data to assess the relationship between intervention status and the following groups of exposure variables: range hood use, burner use, and pollutant concentrations during cooking events (PM2.5, NO2 and CO2). The pre-specified alpha value was 0.05, so values of p below 0.05 were considered significant. Wilcoxon Signed Rank tests were also used to compare the pre-intervention data to the post-intervention data to assess the relationship between intervention status and the following health outcome variables: airways inflammation (i.e., FeNO) and lung function. Further details of the exact variables calculated are available in the supplemental material. Due to the small final sample size, the study was underpowered to assess health differences, no imputation was performed, data were analyzed in a complete case fashion, and corrections for multiple comparisons were not feasible. Median values reported in the results and tables/figures are the median values of the mean within each household for each time period.