This study will be conducted in the residential modules at the Well Living Lab (WLL) in Rochester, MN. The WLL creates realistic indoor environments that allow researchers to capture realistic, long-term human health, behavioral and environmental data in a highly controlled environment. For this study, two apartments built at the lab will each house one participant at a time for one month.

Each residential module simulates one-bedroom apartments with the floor area of 32.6 m² (Figure 1). Both modules are equipped with electric stove and stove hood. The stove hood is installed above the stove and have ducted exhaust to the outside. In standard control mode, the stove hood will be turned ON manually, and in advanced control mode, the stove hood will be automated and triggered based on the environmental readings. Details of the operation are discussed in Section 2.5. Modules have 2 PACs (Blueair 480i, Blueair, Stockholm) located in the bedroom and between the kitchen and living room. In each module, participants were able to control air temperature. Windows in both modules are not operable. Ventilation is provided through the central HVAC system. IoT CO₂, TVOC, and PM_2.5 sensors are placed in the kitchen, living room, bedroom, and bathroom (Figure 2).

The study is designed to simulate occupant daily exposure with the focus on the indoor sources in the residential environment, specifically PM emitted during cooking and cleaning. Study is designed as a combination of exposures in controlled environment of simulated residential space and uncontrolled daily exposures outside the residential setup. Although exposure outside of the module is uncontrolled it is accounted for by personal exposure measurements. Each participant will be exposed to 2 sets of conditions in simulated residential environment in the crossover format. Half of the participants will spend first two weeks in the residential module operated with standard control and then we will change control mode to advanced and participants will spend another two weeks in those conditions. The second half of the participants will first start with advanced control conditions for two weeks and then we will change operation to standard control for the remaining two weeks. The main study will take place over 28 weeks, with a total of 14 participants that will spend 4 weeks each in the simulated residential environment. The pilot study was a shorter version of the main study, taking place over the period of 1 week per participants during the month of Jun, with a total of 3 participants. In the pilot we tested how participants carried out tasks defined in the study protocol (Figure 1). Schedule of daily activities is presented in Figure 1. Pilot provided insights on how study participants felt about the quantity and difficulty of the tasks they are asked to perform, surveys they need to take daily, equipment they needed to carry. The pilot study also serves to test data collection inside and outside the residential modules, collected and manually stored and collected using IoT and automatically stored. We tested how participants respond to the frequent blood and urine sample draws and how Mayo Clinic analyzes our samples.

The same as in the pilot, in the main study, participants will spend most of their non-work hours in the residential module. During their time in the module, they will cook breakfast and dinner each weekday and clean twice a week (Figure 2). During the workday and for a short period after the workday, participants can leave residential modules for a short period of time. The requirement for participants will be to stay in the module at least 1 h and 30 min after they have cooked dinner. A full schedule is detailed in the Appendix.

The residences simulate one-bedroom residential units with typical amenities, including a full kitchen, bathroom, bedroom, living room/office space, and laundry, as shown in Figure 2. In the residential modules, we measure T, RH, PM_2.5 CO₂, VOC, noise, and lux levels in the kitchen, bedroom, living room, and toilet. Alongside these environmental measurements, we will use Real Time Location System (RTLS) to map out participant locations in the residential modules. RTLS data will be paired with environmental data to accurately calculate indoor exposure to cooking emitted PM_2.5. Previous findings by Lie et al. (Liu et al., 2022) from experiments conducted in the WLL residential modules suggest that the environment is not well mixed from the cooking emitted PM_2.5. perspective and that pairing of the location to the sensor in that particular zone is necessary for accurate exposure calculation. Participants will also wear a PM2.5 personal sampler (SidePak, TSI, Shoreview, MN) while inside and outside the module and iPhone with GPS. GPS and personal sampler information will be paired for the correct apportionment of cooking emitted particles compared to the total daily exposure.

Table 1 outlines the experimental conditions and building system set points for the standard and advanced environmental control scenarios.

Cohort size estimates are based on a literature review of previous trials with similar aims, outcomes, and design. Detail description of statistical power and size effect is in section 2.12. At least 14 participants will individually reside in a residential module for four weeks, with a two-week observation period per condition. Participants aged 18 years and over will be recruited in Rochester, MN. Exclusion criteria includes smoking or having quit smoking <1 year prior; respiratory infection; signs and symptoms of obstructive pulmonary disease, asthma, heart failure or cardiac arrhythmia; use of corticosteroids anti-arrhythmic medication, beta-blockers, anti-inflammatory drugs or aspirin; women who are pregnant or intend to become pregnant during the duration of the study; and shift workers. Additional exclusion criteria include individuals who have not been fully vaccinated against SARS-CoV-2. Participants included in this study will be able to relocate to the Well Living Lab for 20 nights (4 weeks), fulfill scripted tasks (Figure 1), including cooking, cleaning, sample collection, and provide informed, written consent.

Participants will be recruited through five different approaches (Burnett et al., 2018): a mass e-mail recruitment (Steinle et al., 2013), word of mouth (Tran et al., 2021), a recruitment announcement posted on the Mayo Clinic Classifieds (Assimakopoulos et al., 2018), placing recruitment flyers on boards in Mayo Clinic buildings, and/or (Saraga et al., 2014) using Mayo Clinic’s social media webpage. All forms of recruitment will guide participants to complete a Qualtrics screening survey for the initial eligibility assessment.

A screening survey using Qualtrics will be used as the first step in determining eligibility among potential participants. The survey will have questions regarding demographics information (e.g., age, sex), contact information (e.g. name, email, and phone number) and preferred method of contact, occupational factors (e.g., current work department, job description), environmental characteristics of current home environment, health factors (e.g., use of medications known to impact the outcomes being observed), and exclusion criteria assessment.

Once a participant is successfully recruited and before the study participant moves in, biometric baseline measures will be collected and the participant will be trained on the schedule of activities, devices to be used during the study, module rules, and emergency procedures.

Participants scheduled activities include morning and evening tasks, Monday through Friday. Participants may leave the module during the day. They will be asked to cook breakfast and dinner using lab-provided recipes that are expected to emit significant concentrations of PM_2.5 and TVOC, including oil-based frying (Yu et al., 2015; Gao et al., 2015; See and Balasubramanian, 2008; Buonanno et al., 2009; Olson and Burke, 2006; Torkmahalleh et al., 2012; Wallace et al., 2004; Abdullahi et al., 2013; Dennekamp, 2001; Poon et al., 2016; Li S. et al., 2017; Liu et al., 2018a; Dobbin et al., 2018; Liu et al., 2018b; Militello-Hourigan and Miller, 2018; Singer et al., 2017b). Food will be delivered when they move in and on a weekly basis. Participants will be able to cook additional food if desired but cooking methods will be restricted to boiling and using the microwave. Boiling is associated with low PM emission rates (See and Balasubramanian, 2008) and will result in lower unaccounted exposures. A microwave can only be used for food heating, because studies of emission rates suggest that this food preparation method can emit a lot of particles (Zhang et al., 2014). This restriction will minimize variability in cooking-related air pollutant exposures between participants. During Tuesday and Thursday evenings, participants will be asked to conduct cleaning activities known to generate particles like vacuuming (Sercombe et al., 2007; Knibbs et al., 2012; Avershina et al., 2015; Vicendese et al., 2015; Nazaroff, 2016) and TVOCs like wiping down surfaces after using a spray cleaning solution (Singer et al., 2006a; Singer et al., 2006b; Carslaw, 2013; Nørgaard et al., 2014; Carslaw et al., 2017; Shin et al., 2017; Fang et al., 2019). Some recent studies have also shown that several categories of cleaning products (e.g. bleach, botanical disinfectants, terpene-based cleaning agents) can initiate the formation of particles (Mattila et al., 2020; Jiang et al., 2021). Participants will be able to run the dishwasher and do laundry whenever necessary. They will be allowed to leave the residential module with the restriction that they have to stay in the module for 1 h and 30 min after they cook dinner. On Wednesdays and Fridays prior to cooking breakfast, the participant’s blood, urine, and blood pressure will be collected in the module by a Clinical Research and Trials Unit staff member. Sleeping and wake-up time will be flexible on all days, but activities that occur after waking up (e.g., HRV measurement, cooking breakfast, nurse visit) and prior to sleeping (e.g., HRV measurement, surveys) should occur in the scheduled order during the indicated periods. During the four-week study, participants are not required to stay in the module on Friday when they complete all their scheduled activities. Condition transitions will involve installing or removing PACs, initiating and ceasing filtration/ventilation control algorithms, and changing the air handling unit filter. Following the participant move-out after the fourth week, a professional cleaning service will clean the residential modules following COVID-19 disinfection protocol. More detailed information on participant schedule can be found in Supplementary Table A1.

Participants will be provided with a pre-programmed tablet using an WLL designed mobile application which will include their schedule of tasks for the week and an electronic checklist for the participant to indicate when a task is completed. The application was developed for participants to keep track of and check off all the scheduled activities they need to perform during each day. In addition, participants will be able to report if they engage in any additional cooking or cleaning activities that may not be part of the study protocol. This information will help us understand the frequency of unplanned events that may impact the study outcomes.

The air supply system servicing the residences consists of a single air handling unit where outdoor air mixes with return air after which mixture goes through the MERV14 filter cooling and heating coil before it is supplied to the residential modules. Each residence has a supply air variable air volume (VAV) terminal unit with reheat. The flow rates used in the pilot and main study are described in Table 1. MERV14 filter is installed in the AHU to reduce the risk of cross-contamination of particulate emissions and airborne pathogens (e.g., SARS-CoV-2).

Residential module with Standard Control operates as VAV system maintaining air temperature set point and has manual ON/OFF control of stove hood and bathroom exhaust. Advanced Control mode introduces automatically controlled air quality intervention strategies consisting of a stove hood, two PACs, bathroom exhaust, and an increased supply airflow rate. These interventions contain and remove PM_2.5 (Liu et al., 2022) emitted during cooking. When PM_2.5 concentrations of 15 μg/m³ or higher are detected in the kitchen, bathroom, bedroom, or living room, then the stove hood, bathroom exhaust, bedroom PAC, and living room PAC will turn ON, respectively (Figure 1). When the PurpleAir sensors detect an average of 3 consecutive measurements of PM_2.5 < 6 μg/m³, then the corresponding devices will turn OFF. If any of the PurpleAir sensors detect PM_2.5 concentration beyond 50 μg/m³, the supply airflow rate increases to 510 m³/h (“air flush” mode) to dilute the air pollution level. Setting the air flush to trigger at a higher threshold allows for stove hood, exhaust fans, and air cleaners to address emission events without altering the standard HVAC operation. Sensor details are in Supplementary Table A2. The outdoor air ratio (∼20%) will be maintained during air flush, rather than being set to fully open to prevent ducts from freezing during severe winter weather. The air supply system provides occupants the ability to control temperature setpoint within a set range ±3 °C using a slider located on the wall thermostat (BAPI 418MHZ). Humidity was maintained at 35% during the pilot and will be maintained at 35% during the main study. Details about the IoT connection of different components to the system is described in the Appendix.

In Figure 3 and Figure 4, we present results from the pilot study with standard and advanced control modes. Hourly box plots in Figure 3 show that the residential module with advanced control maintained lower PM_2.5 levels in the module throughout the day and especially during scheduled cooking hours. Results in Figure 4 show that when meals with the same recipe were cooked in both residential modules, the module with the advanced control had lower peak concentration and shorter exposure time. Results in Figure 3 and Figure 4 suggest that participants were exposed to different levels of PM_2.5 in the residential modules with the standard and advanced control. Results of the stove hood use show that stove hood was not used in the manual mode, while it was turned on during all the cooking activities when advanced controls were used.

Environmental monitoring consists of room-level and personal PM_2.5 monitoring (worn outside the module). Besides environmental monitoring we use RTLS and GPS to include participant location information in the exposure analysis. With the exposure analysis we would like to 1) accurately estimate exposure in the residential module using sensors placed in four locations in the module, and 2) provide insights into the relationship between cooking exposure at the residential module to other exposures that participants experience during the day (total exposure). Table 1 describes the environmental data collected in the pilot. The same data will be collected in the main experiment. Module-level sensor locations are included in Figure 2. Collocation calibration was performed for low-cost sensors for PM_2.5, and TVOC. All CO₂ sensors were calibrated using 3-point concentration collocation against research grade sensor (Li-Cor LI-840A, LI-COR Inc., Lincoln, NE).

During the pilot several surveys were administered to the participants once or multiple times. Baseline surveys regarding participants’ general background, home environment, and office environment were administered once before moving in and an exit survey was administered once after participants completed the study. During the participants’ stay, participants answered a set of surveys multiple times, including the COVID-19 screening, health symptoms and concerns associated with cooking and cleaning activity, and experience with the residential module at the end of the day (EoD).

Baseline surveys collect information about demographics, medications, sleep quality (Snyder et al., 2018), perceived stress (Cohen et al., 1983), wellbeing, general cooking habits, home characteristics assessment, work background, and office environment (Andersson, 1998). The exit survey collects feedback on participants’ experience on performing assigned tasks, satisfaction as a study participant, and the quality of instructions and schedule of tasks through the tablet. Health symptoms and concerns associated with sick building syndrome (SBS) were assessed daily after cooking. In the pilot the health symptoms and concerns questions were derived from the Emory Indoor Air Quality questionnaire. Health symptoms were assessed additionally in the morning and evening. In the symptoms question, if participants reported the presence of symptoms, they were subsequently asked to report the perceived intensity. After each cooking and cleaning event, participants were asked to fill out questions related to perceived air quality during and if the activity was performed according to the protocol. For example, if during cooking they followed a recipe or if during cleaning they only vacuumed or also performed surface cleaning.

At the end of the day on Monday through Thursday during the pilot, we collected data on participants’ general satisfaction with the residential module, perception, satisfaction, and sensation of indoor environment, environmental concerns, stress level, and caffeine consumption. The perceived quality of the indoor environment questions includes assessments of air quality, thermal comfort, light, and acoustic quality. Environmental concerns encompass questions regarding noise, fumes, temperature, and humidity. Thermal comfort was assessed following ASHRAE 55 methods and included a thermal sensation scale, a thermal comfort scale, and a question intended to estimate the amount of clothing being worn for estimating thermal insulation (Nørgaard et al., 2014). A subset of these questions was obtained from a recently developed residential comfort scale (Singer et al., 2006a). In addition, prior to visit of staff for blood draw on Tuesday and Thursday morning, participants completed the COVID-19 screening survey as a part of COVID-19 safety protocols.

Results presented in Table 2 show that during the pilot participants completed 100% of the repeating surveys. 2 out of 3 participants completed the exit survey, however, due to the technical issues, baseline surveys were not distributed. This result shows that during the pilot survey fatigue did not play an important role. The survey output results illustrated in Figure 10A depict IAQ perception differences between the 3 participants. Participant 3 was less satisfied with the IAQ in the residential module compared to other two participants. Results also show that Participant 3 experianced stronger negative impacted by the cooking fumes (Figure 10B).

In the main study, we will adjust distribution of the surveys and survey questions. To minimize technical issues, baseline surveys and exit survey are integrated into the scheduled tasks. We simplified baseline, thermal comfort, and SBS-related symptoms and concerns questions so that we can focus more on perception of cooking exposure and daily activities as well as evaluation of each environment. After observing behaviors related to cooking activities, we included more questions about participants’ cooking habits and environments at home in the baseline survey. In addition, to understand possible differences in the perception of the environment, especially between the home and residential modules, the same questions regarding indoor environmental perception for the home, office, and residential modules are also included. Furthermore, we include questions related to commute and physical activities at the end of day and sleep quality in the morning to better assess each outcome of interest.

Several studies have shown that PM_2.5 can play a role in oxidative stress, resulting in a change in 8-hydroxy-2-deoxyguanosine (8-OHdG) levels (Pan et al., 2011; Lin et al., 2013; Chuang et al., 2017). PM and TVOCs have also been seen to influence systemic inflammation through biomarkers interleukin-6 (IL-6), C-reactive protein (CRP), and fibrinogen (Lin et al., 2013; Karottki et al., 2014; Hassanvand et al., 2017). These physiological measures are important indicators to determine the impact of IAQ on cardiorespiratory health.

Two recent literature reviews identified various biomarkers in blood, urine, and saliva impacted by outdoor and indoor air pollution exposure (Senerat et al., 2021). Both reviews found that air quality affected multiple biomarkers in various body systems, including inflammation and oxidative stress. The IAQ review found four main biomarkers that consistently showed association with IAQ through various sources, including cooking (Pan et al., 2011; Lin et al., 2013; Chuang et al., 2017), (Pan et al., 2011; Lin et al., 2013; Karottki et al., 2014; Hassanvand et al., 2017). While numerous research studies show an association between IAQ and various human biomarkers, many of these studies range in settings, including restaurant kitchens, workplaces, or residential homes. These study limitations do not allow for a controlled environment to factor in confounding variables of the effects of IAQ on human biomarkers.

During the pilot, blood and urine were collected 2-3 times per participant. A mobile nursing unit collected these samples on Wednesday and Friday mornings for all three participants, and one participant provided baseline samples before move-in due to time constraints. Up to 20 ml of blood and a minimum of 30 ml of urine will be sampled per collection and stored in a fridge for subsequent analysis. Analysis was performed for seven biomarkers. Biomarkers tested in the pilot include 8-hydroxy-2-deoxyguanosine (8-OHdG, urine), vonWillebrand Factor (vWF, blood plasma), high sensitivity C-Reactive Protein (hsCRP, blood serum), Interleukin-6 (blood plasma), CD11b (blood), Fibrinogen (blood plasma), and Myeloperoxidase (blood serum).

In the main study, the same amount of blood and urine collected in the pilot will be collected up to 9 times per participant (four times per condition and one time as baseline). The same mobile nursing unit used in the pilot will collect these samples on Wednesday and Friday mornings during the participants’ four-week stay. In the main study, one blood and urine sample each will be collected one morning of the week before participants move into the residential modules as a baseline sample. Only six of the seven blood and urine biomarkers will be quantified per sample to detect changes in inflammation, coagulation, and oxidative stress pathways. We removed CD11b due to the lack of quantitative results provided by laboratory analysis. Biomarker levels will be tested at the Mayo Clinic Laboratories.

To assess the effect of indoor air pollution exposure on cardiovascular health, research studies have monitored changes in several physiological outcomes relevant to IAQ exposures, including diastolic and systolic blood pressure, heart rate, and HRV under different exposure conditions (Brook et al., 2010; Delfino et al., 2011; Kelly and Fussell, 2017).

During the pilot, sphygmomanometer-based blood pressure measurements (SphygBP) were collected 2 times per participant. In the main study, SphygBP will be collected in triplicate up to 8 times per participant (four times per condition) during the morning by a mobile nursing unit following each two-day exposure sequence. The average of the three closest measures at each time point will be recorded. A baseline measure of blood pressure was not collected during the pilot but will be collected in the main study on the Tuesday morning of the week prior to occupancy during the baseline blood and urine collection. Additionally, in the main study, semi-continuous measurements of blood pressure will be collected during periods of interest using an ambulatory blood pressure monitor (OnTrak 90227) on Tuesdays and Thursdays over a 4-week period. In the pilot and in the main study during mornings, ambulatory blood pressure will be measured every 15 min during a 2-h period, from 1 h before cooking to 1 h after cooking breakfast. During evenings, ambulatory blood pressure will also be measured during a 3-h period, from 30 min before cleaning or cooking to 1.5 h after cooking dinner.

In the pilot and during the main study, the BioNomadix Electrocardiogram (ECG) system (BIOPAC, United States) will measure continuous HR and HRV (ConHRV) two days per week (Monday and Wednesday) during waking hours (about 7:00 AM to 10:00 PM). Additionally, a Photoplethysmography (PPG) based sensor will be used to measure HRV and HR at four specific time points daily (Monday-Friday) to collect resting HRV (RestHRV) measurements: 1. Immediately after the participant wakes up, 2. After cooking breakfast, 3. Before cooking dinner, and 4. After cooking dinner. Resting HRV measurements will be collected after the participant has rested for 5 min (stabilization period) for approximately 10 min with the subject seated and as motionless as possible using spaced breathing. For HRV measurements taken throughout the day when participants are away from the modules, participants will be asked to report in an activity log the times they exercise and stress and anxiety levels. In the main study resting HRV readings will be collected on the same schedule from Tuesday to Friday before occupancy as a baseline measure. During the pilot, occupants performed most of the scheduled measurements. The pilot results showed that the participant completed all required biomatrix measurements. This was one of our major concerns, considering the number and strict schedule when they need to be performed. We will keep the same biomatrix measurement protocol during the main study. We included BP, HR, and HRV baseline measurements up to a week before moving into the module in the main study in an attempt to provide better base for observation of exposure effects.

To evaluate the impact of indoor air pollutants on respiratory health, researchers have used lung function measures and respiratory health questionnaires. The Force Oscillation Technique (FOT) has emerged as an alternative to spirometry for measuring lung function (Maugeri et al., 2017). FOT is a technically simple, minimally time-consuming, and accurate technique for characterizing respiratory mechanics (Faria et al., 2009). In addition to FOT and spirometry, measurements of concentrations of exhaled nitric oxide (FeNO) has been used to detect airway inflammation. Nitric oxide is a gas produced by the endothelial cells involved in the inflammation response, and the concentration of FeNO can be easily measured in an exhaled sample of breath. Measurements of inflammation of lung airways complements conventional lung function testing in the assessment of patients with non‐specific respiratory symptoms (Taylor et al., 2006).

During the pilot, lung function and inflammation measurements were be collected in triplicate via the forced oscillation technique (FOT) and exhaled nitric oxide (FeNO), respectively, up to 32 times per participant. Participant collected collected these measurements during four scheduled periods. On Mondays, FOT and FeNO measurements were collected before and after cooking dinner. On Wednesdays, FOT and FeNO measurements were collected before and after cooking breakfast, as well as before and after cooking dinner. On Fridays, FOT and FeNO measurements were collected before and after cooking breakfast. The main study will use the same protocol. In the main study additional baseline lung function and inflammation measures will be collected on Tuesday morning prior to occupancy.

Although environmental air quality factors such as noise, temperature, and humidity have been shown to impact the sleep (Okamoto-Mizuno and Mizuno, 2012; Harding et al., 2019; Akiyama et al., 2021), there is limited research on the effect of ambient and indoor air pollution exposures and sleep health measures, including sleep duration, sleep quality, and the frequency of periods of wakefulness after sleep onset (Strøm-Tejsen et al., 2016; Canha et al., 2021). Strom-Tejsen et al. (Strøm-Tejsen et al., 2016), assessed the impact of bedroom CO₂ levels on objectively measured sleep quality and sleepiness (Strøm-Tejsen et al., 2016). Improved sleep quality and a reduction in next-day sleepiness levels were observed in bedrooms with ventilation and, thus, a lower concentration of CO₂ (∼835 ppm) compared to bedrooms with no ventilation and an average CO₂ concentration of 2395 ppm. Similarly, Yu et al. (Yu et al., 2019), reported reductions in sleep duration (0.68 h of sleep), measured using subjective sleep questionnaires with increases in air pollution (PM_2.5, PM₁₀ and NO₂) among students living in Beijing, China (Yu et al., 2019). To our knowledge, limited studies have investigated the impact on sleep following exposure to particles from cooking emissions.

A mattress-based sleep monitor, EMFIT QS (EMFIT, Finland), was used in the pilot to evaluate the association between exposure and individual sleep quality every night. The main study will include Dreem headband (Dreem, France) alongside EMFIT. We will assess metrics that includes total sleep time, time in bed, sleep onset, sleep onset latency, wake after sleep onset, sleep quality and HRV. As HRV is monitored with this device, in combination with the daytime wearable HRV biometric device, nearly 24 h of continuous HRV measurements will collected on certain days during the study. Measures of sleep will be averaged across each condition and compared using statistical analysis.

Figure 11 details the data that were collected in the pilot and will be collected in the main study, associated planned analyses for the main study data, and questions the study tries to answer. Linear mixed-effects models or similar modeling approaches will be used to assess differences between health and survey outcomes in two residential modules. ANOVA or similar will be used to assess differences between environmental conditions and occupants' behavior related to pollution control. Air Pollution Exposure models, which include occupant tracking, estimates of inhalation rate based on occupant activity, and environmental measurements will be used to assess differences in personal exposure between experimental conditions for health effect modeling.

All power calculations are performed assuming the use of a linear mixed effects model with a random intercept for participant and no random slope. We also assume that there are a total of 14 participants engaging with both ventilation conditions. Variables for which there are few measurements per participant per condition will have the lowest statistical power, whereas more frequently collected variables will have greater power.

We will collect blood and urine biomarkers four times per condition for each participant. This sample collection represents the lower limit in terms of collection frequency. For such variables with four measurements per participant per condition, we have 84% power to detect a difference of one standard deviation between the two conditions. This power decreases to 32% to detect a difference of one-half standard deviations.

Resting heart rate variability is collected four times per day, five days each week per condition. This gives us forty measurements per condition for each participant. With this collection frequency, we have >99% power to detect a difference of one standard deviation, and 94% power to detect a difference of one-half standard deviations.

In general, we will be well-powered to detect differences as small as one standard deviation. However, for smaller differences between conditions, only certain variables with a relatively high degree of collection frequency will be well-powered.

All data captured from environmental sensors, wearable devices, surveys/questionnaires, and information documented from the subject’s chart will be stored and processed in our secure data management and machine learning ecosystem residing on Mayo Clinic’s secure instance of Microsoft Azure Cloud. Data from environmental sensors, surveys, and other devices will be collected using data pipelines residing on Azure Cloud. The raw data being ingested is stored in Azure Data Lake Gen 2. The processed data will be stored in a secure Microsoft SQL database for analytical purposes. Streaming data will be available immediately for analytical and visualization purposes. The data from instruments and devices which do not stream data or do not have the ability to send data to Azure Cloud will be uploaded via a data management web application or scripts hosted on Microsoft Azure Cloud. Survey data will be uploaded via automated data pipeline to ingest the data from Qualtrics into Well Living Lab’s data repository residing on Microsoft Azure Cloud. Data visualization dashboards are created using the data stored in the SQL database for streaming and non-streaming data. The study dashboards ensure data collection is occurring from sensors and other various data sources.

The data storages are secure requiring two factor authentication for access. The data will not contain subject identifying information. Survey responses will be stored in a secure database (with restricted access limited to a few individuals listed on this protocol) once the subject has finished their period of active participation during relocation. Survey and questionnaire responses will be coded and de-identified upon collection to protect the subject’s privacy. Before being manually stored to the HIPAA compliant Mayo Server (to which our broader team has access), survey data will undergo a second layer of de-identification such that they will be reassigned a second subject ID number that further distances them from any remote possibility of being re-identified. Only the study coordinator will have access to the first and second set of subject ID numbers that can be used to re-identify participants. Any data that may be abstracted from the subject’s medical chart or that contains any identifiers will be stored in one of our on-site HIPAA compliant secure databases. Data management architecture is depicted in Supplementary Figure A2.

The Well Living Lab data storages are HIPAA compliant and secured by passwords and user restrictions. All subject information will be de-identified and the identifying key will not be shared with collaborators. All de-identified data will be backed up, and manual entry of data will be limited to minimize data entry errors. All subject data collected during the study will be stored indefinitely and may be used for future analysis.