Reducing Exposure With Air Cleaners and Technology

Historically and persistently divested communities like Bayview Hunters Point (BVHP) in California's San Francisco Bay Area experience disproportionate air pollution and health burden. Local sources of PM2.5 include heavy industrial activity, commuter traffic, heavy-duty diesel trucks, freight trains, and port and maritime activities. Health-protective guidance in these highly impacted communities would improve if we better understood indoor infiltration of outdoor air pollution, especially during extreme events like Spare the Air Days and wildfire smoke periods.

It has been well documented that air contaminants and pollution have adverse health effects, including increased hospitalizations for respiratory illnesses such as asthma, pneumonia and cardiovascular effects. During climate related events such as wildfires, both indoor and outdoor PM2.5 increase, posing additional adverse health effects. Previous studies have noted that low-cost sensors, such as PurpleAir sensors, are effective at measuring airborne particles. Additionally, previous studies have noted that the use of air filters can help mitigate health effects from wildfire smoke and other causes of air pollution. However, use of air cleaners can often be inconsistent. The present study will build on previous literature to examine whether use of air quality monitors in conjunction with air cleaners can increase air cleaner usage and effectiveness.

This study will be a collaborative partnership between UC Berkeley, UC San Francisco, and Brightline Defense. PurpleAir monitors will be used to quantify PM2.5 concentrations before and after air cleaners are distributed. Using a multistage randomized controlled trial design, we will also evaluate how additional information and training can impact participant use of the air cleaners. We will recruit 100 households in BVHP and monitor indoor PM2.5 concentrations continuously and collect in-the-moment health and behavior data twice weekly for a total of 90 days in each participating household. Once consented, households will be randomized to receive sensors with visual indoor air quality (IAQ) cues (n = 50, PurpleAir Zen light ON; Intervention Arm 1) or sensors without IAQ indicators (n = 50, PurpleAir Zen light OFF; Intervention Arm 2). Additionally, household will be randomized to receive one, two, or all three of the technical assistance methods (texts, short-form videos, or video calls). Participants will be randomized 1:1 using blocks into the control arm or intervention arm. Our randomization schema will attempt to minimize differences in reported respiratory symptoms across the two groups. The use of block randomization also reduces the potential for unbalances with studies that have small sample sizes. At the end of study, all participants will keep the air cleaner (including a replacement filter) and PurpleAir monitor. As compensation for hosting the monitors and taking the questionnaires at the beginning and end of the monitoring period, each participant will receive a stipend at the start of the study and at the end. Additionally, each participant can earn an additional stipend for taking the interval surveys over the course of the study.

Before study recruitment and procedures begin, we will recruit 10 community members to serve on the Community Advisory Board. The CAB will be formed and will follow a hybrid model that includes both key stakeholders and organizations who work in the community as well as BVHP residents. The CAB will be used for internal quality improvement of the research study (i.e., helping develop surveys and outreach materials) and thus does not fall within the realm of research. The CAB's formation and activities are separate from other study procedures. CAB members will be recruited separately from study participants. CAB members will not participate in the study, and study participants will not participate in the CAB. The CAB will meet up to 10 times over the 2-year study period. The CAB will provide feedback on the literature review, provide input on study questionnaires and in-the-moment SMS surveys for the study, advise on the planned study protocol, including eligibility criteria and outreach plan, and co-develop a dissemination plan to ensure the study is incorporated into broader community goals.

Health and behavior data will be collected at baseline, during in-the-moment surveys, and at study completion. The questionnaires and surveys will allow the research team to evaluate potentially confounding factors that may influence the intervention benefits and self-reported health impacts, as well as capture behavior data to assist the research team in interpretation of collected data.

Questionnaires will be distributed at baseline and completion of the study, and will include questions about demographics, household characteristics, baseline health status, and baseline knowledge of IAQ and air cleaners. For demographics, questions will include income, race/ethnicity, gender, education, whether they rent or own their home, and how long they have lived at their current residence. For household characteristics, questions will include address, unit size (number of rooms, approximate square footage), building type (single or multi-level, single- vs multi-family home, etc.), stove type (gas or electric), and presence of heating, ventilation, and air conditioning (HVAC) system or personal air cleaners. The number and age of all household members to assess crowding will also be asked. For baseline health status, we will ask about smoking status, underlying health conditions, and daily symptom burden. Last, we will ask questions about the participant's familiarity with IAQ and air cleaners, including any previous experience with air quality information and using air cleaners.

In-the-moment surveys will be sent by SMS twice per week throughout the 90-day study period and ask questions on occupancy (number of people currently home), recent indoor activities (within the last hour), and health and IAQ status. Recent indoor activities include whether there have been open windows, air cleaner in use, running of HVAC system, cooking using a stove or other appliance, burning candles or incense, and smoking or vaping of any kind. For health and IAQ status, questions will ask about any current respiratory symptoms and the participant's impression of their current IAQ and air cleaner use.

Health outcome will be measured at baseline, during in-the-moment surveys, and at study completion using the Breathlessness, Cough, and Sputum Scale (BCSS) for participants with COPD and the Asthma Control Test for participants with asthma.

The interrupted time series intervention study will be conducted as follows:

Phase 1 (Days 1-30): After consent and enrollment, participants will complete the intake questionnaire. During Home Visit 1, a PurpleAir monitor will be installed in each household, with the IAQ indicator turned OFF. All participants will receive an educational handout on the PurpleAir monitor. We will collect baseline indoor PM2.5 concentrations from the PurpleAir monitors and health and behavior data from in-the-moment surveys over the first 30 days. Our prior work informs this monitoring duration as sufficient for establishing daily average PM2.5 and typical behaviors.

Phase 2 (Days 31-60): During Home Visit 2, an air cleaner will be installed in each household with a data logger to capture use. For the 50 participants in Intervention Arm 1, the visual IAQ indicator on the PurpleAir monitor will be turned ON. For the 50 participants in Arm 2, the visual IAQ indicator on the PurpleAir monitor will remain OFF. All participants will receive an educational handout on the air cleaner and best practices for use. All participants will also receive a fact sheet about the Air Quality Index (AQI). For those in Arm 1, the second half of this fact sheet will describe the IAQ indicator on their PurpleAir monitor. For those in Arm 2, the second half of the fact sheet will share general information about where local air quality information can be found. We will continue to collect baseline indoor PM2.5 concentrations from the PurpleAir monitors and health and behavior data from in-the-moment surveys over this next set of 30 days.

Phase 3 (Days 61-90): All participants will receive additional information and training to explore if more targeted technical assistance increases air cleaner use. Participants will receive technical assistance on air cleaner usage in three different forms: (1) text messages that reinforce air cleaner use through reminders and tips and targets spatial barriers and in-the-moment behaviors; (2) short-form videos that show the participants how to use their air cleaner and share community member perspectives on benefits, safety concerns, and household dynamics; and (3) video calls with study staff to build skills and confidence and reinforce value of using the air cleaner. Participants will be randomized to receive one, two, or all three of these forms of technical assistance. We will continue to collect baseline indoor PM2.5 concentrations from the PurpleAir monitors and health and behavior data from in-the-moment surveys over this next set of 30 days.

At the end of the 90-day study period, participants will complete the outtake questionnaire. During Home Visit 3, researchers will download PurpleAir and air cleaner logger data. The power logger will be removed. Participants keep the PurpleAir monitor and air cleaner, with a spare replacement filter. For those in Arm 2, the visual IAQ indicator will be turned ON and the corresponding educational material will be shared.

Perceived benefits, including perceived respiratory improvements, will be qualitatively assessed by semi-structured interviews with participants at the end of the study. Interviews will last approximately 30 minutes and ask about use of air cleaners, including factors that may contribute to increase (e.g., IAQ indicators, extra support) and decrease (e.g., motivation, belief about efficacy) and perceived IAQ and health benefits. Interviews will be audio recorded, transcribed, and coded using an inductive approach (i.e., examine for emerging patterns and themes from interviews). Final thematic analysis will be completed in partnership with CAB.

The installation of the PurpleAir monitors and air cleaners will follow a standardized method. Both will be placed in the same room, either the participant's bedroom or in a central, communal area like the living room. PurpleAir sensors will be installed on a non-floor surface with at least one foot of clearance on all sides, away from any vents, PM2.5 sources like candles, and windows that may be opened. PurpleAir monitor readings will be stored locally to an SD card that is retrieved during Home Visits 2 and 3. When installed, the air cleaner will be installed in the same room as the PurpleAir and placed at least three feet away from the monitor. The air cleaner will be installed with an energy data logger to determine usage and speed setting; data will be downloaded during Home Visit 3 when the energy data logger is uninstalled. The air cleaner is Energy Star rated and certified by the California Air Resources Board, meaning these devices use less energy and do not produce significant ozone. Collected data from the PurpleAir and cleaner use loggers will be assigned a unique identification and will not include any identifiable information of the participant. During installation, technicians will survey the home and make a rough sketch of the layout with dimensions using a laser distance finder to estimate the total volume of the household. No personal or identifiable information will be stored with study data.

Each household and its installed PurpleAir monitor, air cleaner, and cleaner use logger will have a unique ID assigned to them. This unique household ID will be assigned to study participants upon enrollment and linked to unique participant ID. The linked datasets for PM2.5 concentrations, air cleaner use, responses from the questionnaires and surveys, and reported health outcomes will not include any identifying information like participant name, address, or date of birth. The PurpleAir monitor will not be registered with any identifying information. The PurpleAir Monitors will be registered with an assigned number and with a study email address and a proxy physical address. The data stored on PurpleAir will be linked only to the monitor number and a proxy address but will not be linked to any participant identifying information. Data will only be used for the purpose of this study. Information collected through the monitors is used only for aggregate analysis. No personal or identifiable information will be stored with study data; it will be stored on an encrypted HIPAA-compliant, password-protected cloud server at UCSF.

We will leverage the existing publicly accessible networks of low-cost PM2.5 monitors that have been deployed in BVHP. Indoor monitors will be paired to the nearest outdoor sensor to characterize infiltration, indoor/outdoor ratios, and identify the indoor emissions signal relative to the outdoor background.

We will assess the changes in PM2.5 and self-reported health outcomes over the intervention periods relative to the baseline. Air cleaner usage and speed settings will be verified with the recorded energy consumption data. To characterize the relationships between indoor and outdoor PM2.5, we will determine indoor/outdoor ratios and infiltration factors (Finf). We will utilize morphological algorithms to identify and separate indoor episodic PM2.5 peaks from the indoor baseline, which will be further decomposed into persistent indoor-generated and infiltrating outdoor-generated PM2.5. We will then use modified random component super-composition to estimate Finf. To evaluate the efficacy of air cleaners, we will determine decay rates for each identified indoor episodic event. Household-specific total average PM2.5, indoor episodic PM2.5, and PM2.5 decay rates comprise our PM2.5 exposure metrics of interest. We will employ segmented regression models to assess the changes in these exposure metrics over the study course. Each metric will be modeled as a function of the intervention and covariates, e.g., outdoor PM2.5 concentration, Finf, time of day, season, and self-reported behaviors like opening windows. These segmented models will account for the clearly defined interruptions in the PM2.5 time series data resulting from the phased interventions. Inclusion of random effects will account for autocorrelation at the household level, and sub-analysis by housing type will be explored for effect modification. Self-reported health outcomes will also be modeled using segmented regression but will include sub-analysis for potential effect modification by individual characteristics like age, gender, and pre-existing and co-occurring diseases. We will isolate and quantify the immediate outcome changes following each intervention, as well as any changes in the trends.