Personal NO2 sensor demonstrates feasibility of in-home exposure measurements for pediatric asthma research and management.

BACKGROUND: One of the most common pollutants in residences due to gas appliances, NO2 has been shown to increase the risk of asthma attacks after small increases in short term exposure. However, standard environmental sampling methods taken at the regional level overlook chronic intermittent exposure due to lack of temporal and spatial granularity. Further, the EPA and WHO do not currently provide exposure recommendations to at-risk populations. AIMS: A pilot study with pediatric asthma patients was conducted to investigate potential deployment challenges as well as benefits of home-based NO2 sensors and, when combined with a subject's hospital records and self-reported symptoms, the richness of data available for larger-scale epidemiological studies.
METHODS: We developed a compact personal NO2 sensor with one minute temporal resolution and sensitivity down to 15 ppb to monitor exposure levels in the home. Patient hospital records were collected along with self-reported symptom diaries, and two example hypotheses were created to further demonstrate how data of this detail may enable study of the impact of NO2 in this sensitive population.
RESULTS: 17 patients (55%) had at least 1 h each day with average NO2 exposure >21 ppb. Frequency of acute NO2 exposure >21 ppb was higher in the group with gas stoves (U = 27, p ≤ 0.001), and showed a positive correlation (rs = 0.662, p = 0.037, 95% CI 0.36-0.84) with hospital admissions. SIGNIFICANCE: Similar studies are needed to evaluate the true impact of NO2 in the home environment on at-risk populations, and to provide further data to regulatory bodies when developing updated recommendations.

Introduction

As many as 300 million people across the globe suffer from asthma, making it one of the most common chronic diseases in the world (1). In 2007 alone there were 1.75 million emergency hospital visits and 456,000 hospitalizations directly attributed to asthma in the United States (2). This prevalence had an estimated annual cost of $20B in 2009 (3). Alarmingly, over the last three decades the prevalence of asthma has also increased worldwide (1,4,5).

Although asthma affects persons in all age groups, it is more prevalent among children (9.3%) than adults (7.3%) (6). Studies have also shown that children are more susceptible to air pollution and at greater risk of asthma exacerbations due to developmental differences such as higher air intake per kilogram of body weight (7,8,9). However, while correlations between ambient air pollution and risk of asthma attacks have been shown through regional studies (10,11,12,13,14,15) and spatial associations (16,17,18,19), little is known about the direct impact of Nitrogen Dioxide (NO2) exposure in the indoor micro-environment on the progression of the condition (4,5).

Based on over 40 years of evidence, a recent review indicates that gas stoves in the home are more harmful than previously assumed due to pollutants including nitrogen dioxide and carbon monoxide, with documented health effects for sensitive populations observed at levels well below the US Environmental Protection Agency (EPA) outdoor standard (20). Meanwhile, gas stoves used without an exhaust hood have been shown to generate NO2 levels above the outdoor standard (21). This is particularly alarming given a recent survey showing that over a third of US households cook primarily with gas (22). These gas appliances are then partially responsible for indoor air pollution which is typically two to five times higher than outdoor air pollution levels (23).

Short- and long-term exposure to NO2, one of the most common pollutants due to gas stoves, has been demonstrated to have many negative health effects. For children, even small increases in short-term exposure can increase the risk of asthma attacks. This sensitivity was highlighted in a 2013 study examining urban and suburban household exposure. The study found that for every 5-parts-per-billion (ppb) increase in NO2 above a threshold of 6ppb, the risk of wheezing and need for medication increased (24). This may indicate that children in an urban environment have an increased risk of asthma symptoms due to more frequent exposure to elevated NO2 levels.

However, to date, a consensus threshold for safe levels of indoor NO2 has not been determined. The World Health Organization (WHO) has set limits for NO2 exposure based on hourly (104ppb) and annual averages (21ppb) (25), however, these levels lack temporal resolution which may have meaningful correlations with the at-risk population (26). Meanwhile, the EPA has yet to establish an acceptable indoor limit of NO2 exposure (20), despite the fact that individuals may spend up to 90 percent of their time indoors (27).

Working towards an NO2 monitoring solution with high temporal and spatial resolution, we have developed a compact and easy to deploy air quality sensor to monitor indoor exposure down to a resolution of 15ppb with samples taken every minute (See Figure 1) (28). Combining this high-resolution data with hospital records, a more in-depth profile of the patient’s condition can be provided to their physicians for analysis, and correlative studies can be more impactful for sensitive populations. Demonstrating the feasibility of such studies based on these sensing modalities, a pilot observational study was conducted to observe correlations between gas appliances in the home, frequent elevated acute NO2 levels, and indicators of asthma exacerbation and severity pulled from hospital records. Despite limitations in sample size and challenges typical of small-scale deployments, by performing this pilot study, the quality and range of data available for future studies is demonstrated.

Figure 1.

System-level overview and data collection strategy. Custom stationary air quality sensor records in-home pollutant data which is imported to a custom Amazon-Web-Services cloud database along with patient diary information for symptom tracking and hospital records. Data can be viewed in real time using Grafana or downloaded for post-processing and analysis.

Methods

Stationary Air Quality Sensor for Data Collection, Cloud-Based Informatics System

As shown in Figure 2, a custom stationary sensor was developed for the study which is able to monitor NO2, O3, humidity and temperature within the subject’s home. The stationary sensor is comprised of a Raspberry Pi 3B for wireless communication and data processing, NO2 (NO2-A43F, Alphasense, New York, NY, USA), O3 (Alphasense OX-A431), temperature, and humidity sensors (HTU21D, Measurement Specialties, Schaffhausen, Switzerland) and a custom interconnect PCB used to pass data from the sensors to the Raspberry Pi. As reported by the manufacturers and confirmed through in-laboratory calibration routines, NO2 and O3 sensor sensitivity are 10ppb and 15ppb, respectively, temperature accuracy is +/−0.3 degrees C, and relative humidity accuracy is 0.3%. Though not used in present study, the additional sensors were included as they may provide additional insights for future work.

Figure 2.

Exploded view of custom stationary air quality sensor assembly.

The unit takes a recording every minute and stores it to local memory on an SD card. If Wi-Fi connectivity is available, the unit is also able to transmit the data securely to a custom Amazon-Web-Services (AWS) backend for remote analysis. From this point, users provided a secure login (including approved clinicians and researchers) are able to access the data remotely with an expected delay up to a few minutes allowing for data transmission. For users without Wi-Fi access, the sensor is sent back to the approved clinician or researcher for data access through the SD card.

Study Design

Participants were identified in the greater Washington, D.C. region, including Maryland and Northern Virginia, through the Severe Asthma Clinic at Children’s National Hospital. As asthma is more prevalent in the pediatric community (6), subjects recruited were 5–21 years old and presenting with chronic asthma symptoms. Subjects presenting with documented diagnoses of rheumatologic, cardiac, or immunologic conditions were excluded from the study.

Clinical data was collected through Children’s National Hospital in accordance with a plan approved by the Institutional Review Board (IRB; #Pro00009593). Physicians and trained healthcare professionals in contact with potential subjects described the pilot studies, including information about the nature of the study, expectations of their participation, any potential risks and benefits of the study, and their voluntary participation. To participate, each subject (or parent, as applicable) was required to sign a consent form approved by the IRB. As the study fundamentally relied on proper sensor placement within the home, families were selected based on assumed reliability from prior medication compliance and frequency of visits to the clinic. Primary subject recruitment occurred during regular clinic hours on Mondays, with secondary recruitment occurring on Wednesdays, Thursdays, and Fridays. Deployments were not performed over federal holidays to minimize the risk of data loss due to delayed device returns.

Protocols were established to ensure inclusion of women and minorities in the subject groups, as physiological manifestations of reduced pulmonary function may differ amongst women and men, as well as in minority groups based on potential differences in lung composition. The need for this inclusion is highlighted by Subbarao et al., showing the relative difference in prevalence and severity between the sexes at different ages, with the potential for environmental risk factors to be modified by sex (5). As shown in Figure 1, a custom stationary air quality sensor was developed to take NO2 readings in the home every minute and record the values to on-board memory. Each consenting family received a sensor unit, a diary, and a peak flow meter (optional) to collect symptoms and medication usage data. The sensor unit was placed in the kitchen in the patient’s home for six days, and families/subjects recorded symptoms in the diary during that time period. The devices and diary were then mailed back to Children’s National Hospital for data retrieval and analysis.

Clinical data was collected through electronic medical records (Cerner Millennium, Cerner Corp., Kansas City, MO, USA) and included age, gender, body mass index (BMI), ethnicity, associated medical or surgical problems, allergies, medications, family and social histories. Other data collected included physical examination, laboratory studies (if performed), pulmonary function testing (PFT), history of asthma related emergency department (ED) visits and admissions, and courses of oral steroids in the past 12 months. National Asthma Education and Prevention Program (NAEPP) guidelines were used for diagnosis, classification and assessment of asthma control. Further assessment of asthma control included the validated Asthma Control Test (ACT) (29).

Data Analysis

Between the WHO and EPA, the WHO has a lower recommended threshold for NO2 exposure. To minimize risk for affected populations, the WHO recommends annual average indoor NO2 exposure levels below 21ppb and 1-hour average NO2 exposure levels below 104ppb (25). This recommended hourly exposure limit was initially based on a meta-analysis of indoor studies. These studies assumed a baseline in-home annual average NO2 exposure of 8ppb, with an expected increase of 15ppb for homes with gas appliances (or 23ppb total). More recent studies have indicated that these baseline levels may actually be much higher, while NO2 exposure profiles may have an impact on the asthma community at much lower exposure levels (20).

Based on this body of knowledge, data analysis was performed on exposure levels down to 21ppb average per hour. Correlations observed at this level may indicate that a value lower than 21ppb for short-term exposure may need to be selected for future guidance recommendations. This would agree with previous work which has observed sensitivities in pediatric asthma patients starting below this level (24). Using standard guidance language as a guideline, one-hour averages were selected under the assumption that this language would be easier to reflect and monitor in future recommendations by regulatory bodies.

Hospital records were gathered up to 12-months prior to the study for analysis of trends of asthma exacerbations. The complex nature of the exposure-response mechanism for NO2 makes it difficult to monitor short-term exposure, as the inflammatory response may be cumulative in nature. Additionally, it has been reported that while NO2 may not be a direct cause of inflammation within the lungs, it may lead to exacerbation when other allergens are present (26). In such cases, it would be difficult to capture a response unless the other allergen was present, which may be seasonal in nature. By making an underlying assumption that NO2 exposure patterns in the home are relatively consistent across time, collecting lagging 12-month hospital records may provide an additional indicator of NO2 exposure impact on the subject. Hospital and urgent care admissions, lab records for immune system responses, asthma control test results, forced expiratory volume / forced vital capacity, and prescribed medication levels were all included for response metrics pulled from hospital records.

Statistical Methods

For the two demonstrative hypotheses, the average number of hours per day with NO2 levels >21ppb was used as an indicator for frequency of acute exposure. Data collected from the 31 test subjects displayed one outlier with NO2 values far out from the rest of the data in the study (30).

For hypothesis one, to determine the impact of the presence of a gas appliance on the frequency of acute NO2 exposure, a Mann-Whitney U test was performed with subjects separated into “gas stove” and “no gas stove” groups.

For the second hypothesis, the number of times the subject was admitted to a hospital for asthma symptoms during the 12 months leading up to the study was used as a proxy for severe asthma exacerbations/status asthmaticus, compared to less severe exacerbations identified as urgent care or emergency department visits frequently cited in other studies.

Due to the small sample size, Spearman’s rank-sum correlation for nonparametric data was performed between these two metrics. Using statistical methods developed by Bonnet and Wright, a confidence interval for this correlation was found by transforming the correlation to a Fisher Z, calculating the Standard Error for that Z, calculating the confidence interval for the Z, then translating the upper and lower bounds of the Z back to correlations (31). This method, more commonly used for Pearson correlations, has been argued to be valid for Spearman Rank correlations (32).

Results

Deployment Results

One of the primary goals of the present work is to demonstrate the feasibility of an in-home deployment of pollutant sensors with high temporal resolution for the purpose of monitoring the home environment of at-risk populations. Results below highlight subjects recruited, deployment challenges experienced which may be common to real-world deployments in similar populations, an overview of data that was able to be collected, and a brief discussion of one observed outlier.

Subject Characteristics

40 total high-risk asthma subjects were recruited, of which 30 deployments successfully collected exposure data (accounting for the outlier), comprising 16 males and 14 females with an average age of 12 years. 9 subjects were obese, 6 were overweight, and 15 were of healthy Body-Mass-Index (BMI). 3 reported with moderate persistent asthma, and 27 with severe persistent asthma, with all subjects having a previous diagnosis of rhinitis. Of the 26 subjects tested, 22 showed a positive skin prick test result, with an average positive result of 3.6 per subject. 17 reported having gas appliances. Participant characteristics are broken down in Table 1.

Table 1.

Subject Demographics and NO2 Exposure Overview.

Characteristics	Overall(n=30)	Gas-Stove Group(n=17)	No Gas-Stove Group(n=10)
Demographics
Age at recruitment, years	12±5	12±5	11±4
Male, n	16	11	5
BMI
Healthy, n	15	9	4
Overweight, n	6	3	3
Obese, n	9	5	3
Clinical characteristics
Skin Prick Testing, n	26	14	9
Positive Result, n	22	11	9
Average Positive Result/Person	3.6±2.7	3.8±3.3	3.6±1.7
Rhinitis, n	30	17	10
Inhaled corticosteroid use, n	30	17	10
NO2 Exposure
Hours/Day with NO2>21ppb	0.95±1.42	1.45±1.57	0
Peak NO2 Level Observed (ppb)	64.8±66.3	99.6±60.5	5.7±5.1
Daily Average NO2 Level (ppb)	6.59±5.75	9.4±5.3	1.8±2.4
Maximum 1-Day Average (ppb)	36.8±40.3	56.2±39.4	3.7±4.1

Definition of abbreviations: BMI = body-mass index; SD = standard deviation.

Data are presented as mean ± SD unless otherwise noted.

Deployment Data

Shown in Supplements 1–2 are sample reports generated as examples of common data received. Supplement 1 shows the full range of data that was recorded for one subject; Supplement 2 demonstrates a sample diary entry from this subject. Average hours per day with NO2 exposure > 21ppb correlating with hospital admissions over the past 12 months had the highest level of statistical significance after analysis; other clinical measurements had no observed significance. Broken down by starting day-of-the-week, 19 deployments began on Monday, 6 on Wednesday, 3 on Thursday, and 2 on Friday. This provided 2 overlapping weekdays and one overlapping weekend-day for all deployments. No differences in metrics for asthma exacerbation or exposure levels were discernable between days based on this limited deployment; however, standardizing the day of the week for initial deployment or moving to a 7-day deployment schedule may help reduce the impact of any day-to-day variability.

Deployment Challenges

Several challenges were observed during the course of the study. First, a significant lack patient compliance in the deployment was observed, as may be expected in real-world environments. These hurdles include unreturned equipment, lack of diary information, late equipment returns, and equipment left unplugged. In an attempt to reduce these issues, patients were offered a gift card for full compliance; however, this incentive did not appear to have a major impact. It was found that simple phone call reminders and consistent communication lines was more successful in improved patient compliance. These communication lines still proved difficult due to changing phone numbers and primary caregivers. Grouping these factors together, user error accounted for approximately 22.5% of all deployments and 90% of all failed deployments. Second, WiFi connectivity was not available in many of the homes of the study. To circumvent this issue, all data from the sensor is stored locally to be downloaded upon return to the clinic. Lastly, a factor to consider in the success of deployments is the potential for data loss or broken devices. In the current study, a single device deployment resulted in a data fault; with a small sample size, this accounted for roughly 3% of available deployments (one of 40 total). A breakdown of all deployments highlighting data available for investigation is shown in Figure 3.

Figure 3.

Deployment results breakdown

Observed Outlier

During data analysis, one strong outlier was observed. This subject had the highest overall NO2 exposure levels despite no reported gas appliance in the home. With the primary focus of this study being inner-city populations, there is a high likelihood that despite indicating no known gas appliances, there is another significant source of NO2 in or near the subject’s home. This could be a neighboring unit (this particular subject lives in an apartment complex, for example) a bus stop, or other known source of combustion. As the purpose of this study is to observe the impact of indoor gas appliances on the health of severe pediatric asthma patients, this subject was omitted from data analysis; however, it is important to note the potential impact of outside influences for future studies which may need to include geographic information and regional air quality data to account for these sources of exposure.

Correlation Analysis Results

As a demonstration of correlative analyses that can be conducted through the rich data available by combining high resolution environmental sensors with individual hospital records, data collected was run through SPSS to observe correlations between gas appliances, frequent acute NO2 exposure, and indicators of asthma exacerbation observed in hospital records. Following this high-level observation, two example hypotheses were formulated as shown below. An overview of clinical findings correlated with NO2 exposure is provided in Supplement 3, and complete results from statistical testing are shown in Supplement 4.

Gas Appliances and Frequent Repeated Acute NO2 Exposure

Hypothesis 1: Gas appliances lead to frequent acute NO2 levels >21 parts-per-billion (ppb) in the home.

The gas appliance group (n=17) showed an average of 1.5 hours per day with NO2>21ppb (s=1.6 hours), while the group without a gas stove (n=11) showed an average of 0 hours per day. Performing a Mann-Whitney U test on this data, we concluded that frequency of acute NO2 exposure >21ppb in the home is statistically significantly higher in the group with gas stoves than the group without gas stoves (U = 27, p ≤ 0.001), as shown in Figure 4.

Figure 4.

Average Hours per Day with NO2 Exposure>21ppb vs. Gas Appliance Reported in Home.

Acute Exposure of NO2 and Hospital Admissions

Hypothesis 2: Frequency of elevated acute exposure of NO2>21ppb in the home correlates with the number of hospital admissions due to asthma.

Spearman’s correlation was performed to determine the relationship between the frequency of elevated acute NO2 exposure in the home and hospital admissions due to asthma in the 12 months prior to the sensor deployment. The test revealed a positive correlation between the frequency of elevated acute NO2 exposure and hospital admissions due to asthma (r=0.662, n=30, p=0.037, 95% CI 0.36–0.84). Sample calculations are shown in Supplement 5.

Discussion

While the WHO and EPA attempt to provide broad guidelines to advise the general public in their exposure to NO2, further research may identify lower levels of NO2 as being associated with negative outcomes in specific disease populations, such as the pediatric asthma community. Original standards created by each organization were limited by availability of technology and feasibility of deployments. Sensors such as the one utilized in this study may enable a better understanding of the indoor environment and NO2 exposure’s direct impact on sensitive populations to inform future standards.

It is known that indoor exposure is often significantly higher than outdoors (20,23), and NO2 can worsen asthma even at ambient concentrations (33). For this reason, higher temporal resolution of NO2 exposure within the home may provide better insights into an asthmatic’s environment, beyond the recommendations of the WHO or EPA, which does not provide recommendations for an indoor environment or impacted population.

In testing the hypotheses, two significant findings were observed. First, 3 of the 31 subjects recruited had at least one hour where the one-hour average exceeded the WHO’s recommended hourly 104ppb limit.

Second, one subject experienced a daily average exposure of 11ppb and a maximum hourly average of 91 ppb. These values lie within recommended guidelines for indoor exposure from the WHO. However, this subject experienced a peak NO2 exposure of 188ppb, the second highest observed in our study. With standard pollutant observation technology only providing outdoor monitoring of hourly and annual average exposure limits, a full comprehension of a subject’s potential asthma triggers is impossible.

NO2 Health Impact

Augmented allergic reactions in asthmatics have been observed after brief exposures to NO2 by mimicking real-life exposure conditions. In one 3-day study, 266ppb NO2 was introduced to subjects for just 15 minutes followed by allergen exposure 3–4 hours later, with symptoms, pulmonary function, and inflammatory response in sputum and blood being measured daily. The study found that two to three brief exposures to NO2 was enough to enhance eosinophilic activity in the sputum to inhaled allergens (26). With our allergic dominant study cohort, NO2 may exacerbate underlying immune responses, contributing to severe exacerbations leading to hospital admissions.

Studies have identified the potential impact of NO2 as a trigger for asthma exacerbations (34,35,36), increased hospital visits (37,38), enhanced asthmatic reactions to inhaled allergens (39), and as a cause of respiratory system conditions in the healthy population, including reduced lung function (40,41). However, a majority of this work has been performed in population-wide epidemiological studies (42), with very few linking well quantified indoor exposure values to clinical cardiopulmonary effects, even though indoor concentrations are often much higher than outdoors (43,20).

Additionally, NO2 is known to cause a variety of clinical responses depending on the concentration and duration of exposure (42). At low concentrations, inhaled NO2 interacts with moisture in the respiratory tract to form nitric acid (HNO3), which further dissociates into nitrates and nitrites. Higher concentrations of NO2 will reach further into the respiratory tract and into the lungs, where it will remain or be transported via the bloodstream to extrapulmonary sites. Once it has reached these sites, NO2 or its derivates can interact with hemoglobin, forming methemoglobin, an ineffective carrier of oxygen (44). This inefficiency creates a significant health risk for vulnerable populations such as asthmatics and children, already faced with inefficient oxygen transportation.

Exposure Threshold and Temporal Resolution

For outdoor environments, the EPA recommends a 1-hour maximum average NO2 exposure of 100ppb and an annual maximum average of 53ppb. The annual limit was first provided in 1971 as part of the original Clean Air Act, and the 1-hour standard was issued in 2010 as part of the National Ambient Air Quality Standard (NAAQS) addendum to the Clean Air Act (45). This hourly exposure limit is intended to limit an individual’s exposure to short-term peak concentrations of NO2 which can largely be attributed to highly congested areas of road traffic. While these standards are intended to represent a guideline for the general public based on common outdoor sources of NO2, they do not take into account occurrences of indoor exposure which are often significantly higher than outdoors (43).

Meanwhile, as discussed above, the WHO has lower recommended exposure levels than the EPA, with the lowest absolute value at 21ppb annual average. It has been estimated that annual NO2 exposure less than 21ppb may contribute to 4 million new cases of asthma globally (46). One study demonstrated that repeated exposure to an ambient level of NO2 can enhance an asthmatic’s response to an asymptomatic allergen dose (47). Another study found asthma exacerbations in children occurring at exposure levels significantly lower than those recommend by the EPA and WHO, with a dose-dependent increase in the risk of higher asthma severity scores for every 30ppb increase in NO2 exposure over 6ppb (24).

These studies demonstrate the need for real-world sensing modalities capable of monitoring indoor low-level exposure with high temporal resolution to further assess the impact of NO2 exposure on the asthmatic community. However, due to conflicting findings and inconsistent testing methods, there remains a lack of consensus in the scientific and health communities as to what an appropriate indoor NO2 exposure recommendation should be. As shown in Figure 5, using devices similar to the one developed for this study, minute-level, hourly, daily, and annual temporal resolution is all feasible to study indoor NO2 exposure, with sensitivity down to 15ppb, enabling the necessary research to reach this consensus.

Figure 5.

Sample NO2 Record from Custom-Developed Stationary Air Quality Sensor.

Confounding Variables

While a correlation between hospital admissions due to asthma and the frequency of elevated acute exposure to NO2 for individual pediatric asthma subjects was observed, the present work is meant to demonstrate a deployment strategy that may be utilized when using sensing techniques similar to those discussed. Although correlative strength of the two hypotheses tested is relatively high, it should be noted that the present study with n=30 may not represent the population due to the small sample size and single-site recruitment strategy. Hospital admissions could be for several reasons. For example, it may be possible that the impact observed was due to a cumulative effect of exposure to NO2 in the home. Another possible explanation may be that frequent acute exposure to NO2 leads to enhanced sensitivity to other allergens, as has been demonstrated in other work (47,26). In this case, it should be noted that all subjects reported more than one known asthma trigger, such as pets or seasonal allergies. A brief overview of additional known confounding variables is outlined in below, though this list is not considered exhaustive.

The current 6-day study period may be too short for a direct comparison to previous work of physiological responses due to pollutant exposure. Other similar studies which show the strongest correlation between pollutant exposure and biological response typically involve epidemiology studies over a larger study period (20,42,46). Additionally, the metropolitan region of the present study may display different pollution exposure levels than the greater regional area. In this regard, epidemiological studies which consider larger regions may demonstrate alternative findings. Differences at the population level may lead to study results contrasting to larger regional studies; the DC area pediatric asthma subjects may have different pollutant responses than rural Kentucky, for example, as it has been shown that children with asthma in inner-city communities may be particularly susceptible to air pollution due to exposure to motor vehicle emissions from a young age (48).

As part of the study design, the sensor was placed within the subject’s home, however specific times that the subject was present or near the sensor were not always recorded. Thus, the data collected can only be used as a representation of typical exposure profiles within the home, without a definitive recording of subject-specific exposure. Additionally, as shown through the outlier observed in this study, external influences may impact the NO2 exposure levels observed in the home.

Application, Power Analysis, and Study Methods for Future Work

While this study focuses on pediatric asthma subjects due to higher rates of asthma in children, older adults are more likely to have undiagnosed lung conditions. Many other populations are known to be sensitive to NO2 exposure, including people with preexisting lung conditions such as emphysema, chronic obstructive pulmonary disease, and chronic bronchitis (49). These groups may also benefit from additional air pollution studies with high temporal resolution and sensitivity.

These studies could inform groups such as The National Institute for Occupational Safety and Health (NIOSH), which still provides a recommended exposure limit below 1ppm over an 8 hour average (50). Similarly, improved quality of data is needed to update EPA guidelines to provide indoor NO2 exposure recommendations, and the WHO to update recommendations for effected populations.

The work provided in this study is intended to represent feasibility of improved personal exposure data through deployment of air quality monitors indoors, where an individual may spend around 90% of their time (27). However, implications of the study are limited due to the relatively small sample size. Based on results from this study and statistical methods of Bonett and Wright (31), 152 subjects will be needed to achieve a Fisher confidence interval of 0.2 in future work correlating acute NO2 exposure at this level to hospital admissions due to hospital exacerbations. Calculations used to derive this sample size are shown in Supplement 6.

Due to deployment challenges which may be difficult to overcome, other methods of data analysis may be considered in future studies as well. By aggregating data to person-level on an hourly basis to closer match existing standards, a smaller relative sample size may have been created as compared to a theoretical 174 person-days which could be used for a repeated measures design. Additionally, while correlations based on Spearman’s statistical methods were not observed between many of the input variables, Body-Mass-Index (BMI), Asthma Control Test (ACT), and other individual characteristics could be used to adjust for between-person differences in a mixed effects model testing for associations between indoor NO2 concentrations and health outcomes.

Another potential improvement for future studies is highlighted through the limitations of the physical symptom monitor diaries which were provided to subjects. Though these are standard protocol in asthma clinics, they rely heavily on subject compliance, and lack sensitivity and temporal granularity that may prove meaningful when observing correlations between exposure and symptom. Improvements to this symptom monitor system, which may come through wearable technologies or simplified, automated symptom diaries, may enable correlative studies with higher impact.

Conclusion

In this paper, we have demonstrated a deployment strategy which combines subject hospital records with a compact pollutant sensor with high temporal resolution that may be utilized to monitor the home environment of asthma subjects for potential triggers. Following a small pilot study, and in agreement with recent data indicating NO2 exposure limits set by the EPA and WHO are insufficient (20), we designed a demonstrative observational study to test two hypotheses: one linking gas stoves to increased frequency of acute NO2 exposure in the home, and the other linking the frequency of acute indoor elevated NO2 exposure to hospital admissions due to asthma. 31 subjects were recruited in the Severe Asthma Clinic of Children’s National Hospital to deploy a stationary air quality sensor to monitor NO2 levels in the home every minute for a period of 6 days.

Homes with gas appliances had increased frequency of elevated hourly NO2 (U=27, p=0.001), and the frequency of acute elevated NO2 exposure correlated with hospital admissions due to asthma (r=0.662, α=0.037, 95% CI 0.36–0.84). While the strength of the study is limited due a small sample size (n=30), the results agree with other research indicating that current recommendations may not be adequate for the effected population.

Despite challenges of real-world deployments described herein, similar deployment methods and in-home pollutant monitors with fine temporal resolution may be used to monitor an individual’s microenvironment and further identify asthma triggers. Long-term outcomes of such studies may help change subject behavior, provide clinically relevant data to improve subject outcomes, and inform agencies to update recommended safe indoor exposure limits to NO2 to be more appropriate for effected populations.