Estimating daily air quality exposure of the aging and dementia population

Introduction

A number of studies have observed an increased risk for dementia in association with worse air quality.^1–4 The conclusions of these studies are based on estimates of the environmental exposure of study populations to outdoor air quality measurements based on PM_2.5 (particulate matter or particles smaller than 2.5 microns) levels provided by community-based, public air monitoring stations. However, these studies did not consider how much time the study population was outdoors and thus truly exposed to such air quality. Healthy older adults may leave their homes on average only 3.5–4.2 h per day.^5–7 Further, with increasing age, as well as with the development of cognitive decline with aging, the time out of home further decreases.^5–7 Therefore, measuring their indoor air quality exposure may be important. Factors such as the materials that the homes are made of, home appliances and indoor activities can affect their air quality exposure.^8–10 It is also important to consider that not all time spent out of home is in the open outdoor environment. Only a small portion of time out of home may be actually spent outdoors as a person transits to another indoor setting. The specificity of location of exposure is further reduced in existing research by the fact that the public air monitors are located relatively far from one another. State and local environmental agencies run a national network of 3900 air monitoring devices covering 3.8 million square miles of U.S. land mass, which is roughly equivalent to one monitor per thousand square miles. In addition, about 120 million Americans, more than one third of the population, live in counties where there are no Environmental Protection Agency (EPA) air monitors. ¹¹ Thus, even though studies use the nearest outdoor public air monitors to the homes of study participants to estimate their air quality exposure, the estimates may not be accurate due to lengthy distances from their everyday living spaces.

Considering these potential limitations, we examined the spatiotemporal relationship between indoor air quality (using indoor environment sensing monitors) and outdoor air quality estimated by the nearest public air monitor. We compared the differences and the correlations between the two measurements in relation to the distances between the indoor and the outdoor public air monitors. The indoor air quality data was collected as part of ongoing studies on how indoor home environments may affect the behavior of persons living with dementia, their caregivers, and older adults with normal cognition. We hypothesized that indoor and outdoor air quality are concordant with each other, but the strength of the relationship may depend on the distance between the two measurements and differ by home. We also examined if the use of Hazardous Air Quality Ensemble System (HAQES) can provide better estimates for indoor air quality. The practical objective of this current work is to provide a closer examination of the fidelity of outdoor relative to indoor air quality measurements. This is particularly important since air quality is an identified risk factor for dementia.

Methods

Results

Indoor versus outdoor comparisons

Indoor air monitoring was conducted for a total of 6872 days across the 23 residences. All but one home were located in the state of Oregon in the United States. The homes were monitored for 299 days on average with a standard deviation of 184 days and a range of 33 to 641 days. Table 1 shows the summary statistics of daily indoor and outdoor air quality measured by the nearest public air monitors, the distances between the homes and the nearest public air monitors, and the Spearman correlation coefficients between the indoor and outdoor air quality readings and their 95% confidence intervals (CIs). As seen in Table 1, the mean daily indoor air PM_2.5 measurement was lower than that of the nearest public air monitors for all except 3 homes, while the variabilities of indoor air quality were larger than that of the nearest public air monitors for 8 of the homes. Figure 1 shows the scatter plots of PM_2.5 from in-home air monitors versus PM_2.5 from the nearest public air monitors, with each scatter plot representing one home. Spearman correlation coefficients range from −0.146 to 0.867, with a mean (SD) of 0.442 (0.219).

OPEN IN VIEWER

Figure 1.

Twenty-three scatter plots (one per home) of PM_2.5 measurements from in-home air monitors versus PM_2.5 measurements from the nearest EPA air monitors along with the corresponding Spearman correlation coefficients (r) and p-values. The points would fall on the red dashed lines if the indoor and outdoor measurements agree perfectly.

OPEN IN VIEWER

Table 1.

Statistics of indoor and outdoor air quality, the Spearman correlation coefficient (indoor versus nearest outdoor records) per home, and distance between the home and the outdoor public monitor.

Home	Number of days	Indoor PM2.5 daily average (µg/m3)Mean (SD)[Range]	Nearest public PM2.5daily record (µg/m3)Mean (SD)[Range]	Spearman correlation coefficient [95% CI]	p	Distance between the home and the outdoor public air monitor (kilometers)
1	103	3.42 (1.95) [0.80–11.85]	3.23 (1.45) [1.3–11]	0.45 [0.28, 0.59]	<0.001	1.013884
2	258	4.55 (5.61) [0.32–37.39]	5.27 (5.27) [0.9–51.6]	0.61 [0.53, 0.68]	<0.001	1.400126
3	641	3.42 (3.36) [0.28–37.50]	6.45 (5.74) [1.6–51.4]	0.54 [0.49, 0.60]	<0.001	4.441778
4	289	5.74 (5.16) [0.00–43.58]	6.55 (4.49) [2.4–41.3]	0.17 [0.06, 0.28]	<0.01	5.021141
5	147	1.29 (1.25) [0.00–8.72]	6.96 (5.12) [1.6–27]	0.53 [0.40, 0.64]	<0.001	5.182075
6	268	57.11 (44.16) [0.66–192.27]	4.81 (2.35) [1.6–13.7]	−0.15 [−0.26, −0.03]	<0.05	6.324706
7	117	0.67 (0.54) [0.02–4.21]	4.05 (1.66) [1.8–10.3]	0.45 [0.29, 0.58]	<0.001	6.550014
8	203	1.96 (4.19) [0.00–44.47]	7.16 (8.29) [1–74.4]	0.87 [0.83, 0.90]	<0.001	6.646574
9	278	1.23 (1.20) [0.37–16.93]	5.70 (4.44) [1.6–41.8]	0.54 [0.45, 0.62]	<0.001	6.807508
10	169	7.18 (12.31) [0.015–94.6]	4.53 (2.37) [0.6–14.7]	0.30 [0.16, 0.43]	<0.001	7.065003
11	54	3.40 (1.91) [1.22–8.14]	8.64 (5.52) [2.3–27]	0.81 [0.70, 0.89]	<0.001	7.065003
12	33	3.60 (4.08) [0.83–20.87]	4.47 (1.91) [1.9–9.7]	0.27 [−0.08, 0.56]	0.12483	8.111074
13	605	2.35 (5.26) [0.03–122.25]	6.62 (4.07) [2.3–41.3]	0.52 [0.46, 0.58]	<0.001	8.127167
14	522	2.45 (4.86) [0.00–67.23]	5.60 (5.91) [1–74.4]	0.65 [0.59, 0.69]	<0.001	8.416848
15	180	1.82 (2.70) [0.00–30.07]	8.00 (7.50) [1.4–46]	0.42 [0.29, 0.53]	<0.001	9.301985
16	265	3.40 (3.66) [0.37–33.76]	5.79 (4.16) [1.6–27]	0.48 [0.39, 0.57]	<0.001	9.800881
17	214	3.54 (4.43) [0.03–39.43]	5.53 (4.26) [1.4–40.7]	0.47 [0.36, 0.57]	<0.001	10.86305
18	545	2.76 (2.71) [0.36–45.22]	6.45 (6.50) [1.2–75.4]	0.55 [0.49, 0.60]	<0.001	11.97349
19	317	2.85 (2.60) [0.43–26.69]	10.00 (4.21) [3.5–35.6]	0.47 [0.38, 0.55]	<0.001	13.01956
20	286	5.74 (6.52) [0.05–54.27]	8.05 (7.06) [2.4–65.1]	0.14 [0.03, 0.26]	<0.05	13.92079
21	472	4.84 (6.03) [0.46–51.51]	6.86 (6.13) [1.5–61.5]	0.28 [0.19, 0.36]	<0.001	22.62732
22	641	3.49 (4.11) [0.00–50.47]	6.98 (7.38) [1.3–79.8]	0.28 [0.21, 0.35]	<0.001	26.71504
23	265	6.08 (5.33) [0.45–39.55]	7.13 (7.85) [1.6–70.6]	0.51 [0.42, 0.59]	<0.001	33.42599

Figure 2 shows the plots of the outdoor PM_2.5 measurements from the nearest public air monitors and indoor PM_2.5 measurements versus calendar date, along with the EPA's 24-h PM_2.5 level recommendation. As seen in both Figures 1 and 2, indoor PM_2.5 were generally lower. However, there were days when the indoor air quality was much worse (i.e., higher PM_2.5) than the outdoor air quality with frequent large spikes in the indoor PM_2.5 measurements. Also, the air qualities were much worse during the month of October 2022 across all homes because of the wildfires that occurred in Oregon during that time. The average indoor PM_2.5 measurement from all homes before the month October 2022 was 3.62 µg/m³ while in the month of October 2022 the average indoor PM_2.5 measurement was 8.81 µg/m³.

OPEN IN VIEWER

Figure 2.

Twenty-three plots (one per home) of outdoor air quality from the nearest public EPA air monitor (red) and indoor air quality (blue) versus date, along with the EPA's 24-h PM_2.5 recommendation (dashed black line). Only days with indoor air quality data are shown. Participants living at these homes enrolled into the study on different days and therefore the start dates of data collection are different for these homes.

Correlations versus distances

Figure 3 shows a scatter plot of Spearman correlation coefficients between the indoor and outdoor PM_2.5 measurements versus the distances between the homes and the nearest outdoor air monitors for the 23 homes and their linear regression line. As seen from the regression line, the longer the distance between the home and the outdoor air monitor was, the lower the correlation between indoor and outdoor PM_2.5 measurements was, even though the association was not significant at the 0.05 level (p-value = 0.42).

OPEN IN VIEWER

Figure 3.

Scatter plot of Spearman correlation coefficient of PM_2.5 measurements obtained in home correlated with outdoors versus distance between home and the nearest outdoor air monitor for the 23 homes, regression line and confidence interval (shaded area around the regression line).

Three-way comparison among indoor, HAQES air quality forecast data, and outdoor air quality data

Figure 4 shows a visual of the HAQES data for the estimated average daily PM_2.5 level data for December 18–31, 2022. The air quality was worse for the more populated areas. Figure 5 shows the daily average PM_2.5 levels for the 16 homes from which data were collected during December 18–31, 2022 along with estimates from the gridded HAQES data and nearest outdoor air monitors. The indoor PM_2.5 levels were generally lower but occasionally could spike and even exceeded the US EPA 24-h PM_2.5 recommendation (< 35 µgm/m³). Table 2 shows the distances between the homes and the outdoor public air monitors and between the homes and the HAQES gridded data. If the homes and the public air monitors were very far apart, the HAQES provided gridded data that were much closer to the homes.

OPEN IN VIEWER

Figure 4.

Average daily modeled PM_2.5 levels at the grid points from HAQES. Beneath these points is an Inverse Distance Weighted Surface Map based off of the average daily modeled PM_2.5 levels.

OPEN IN VIEWER

Figure 5.

Average daily PM_2.5 level measured indoors (blue), estimated through the HAQES data (yellow), and measured through nearest outdoor air monitor (green) for the 16 homes from which data were collected during the last 2 weeks of 2022.

OPEN IN VIEWER

Table 2.

The distances between the homes and the nearest outdoor monitors, and the distances between the homes and the nearest HAQES gridded data points.

Home	Distance from Home to Nearest Outdoor Monitor (kilometers)	Distance from Home to Nearest HAQES Gridded Data Points (kilometers)
3	4.44	6.29
4	5.02	5.45
5	5.17	4.17
8	6.65	5.79
9	6.81	4.74
11	7.07	1.59
13	8.13	8.34
14	8.41	5.39
15	9.30	4.98
16	9.80	4.61
17	10.87	5.86
18	11.97	6.11
20	13.92	4.95
21	22.62	5.78
22	26.71	3.09
23	33.43	5.25

As seen from Figure 6, indoor mean PM_2.5 levels for the 16 homes were significantly different from both the PM_2.5 levels measured by the nearest public outdoor air monitors and forecasted HAQES air quality data. This shows that even though the HAQES ensemble model provides estimates for PM_2.5 levels at a higher spatial resolution, they still may not represent the PM_2.5 levels indoors.

OPEN IN VIEWER

Figure 6.

Boxplot showing the mean PM_2.5 level measured indoors by the Awair Omni, calculated by HAQES and the nearest outdoor public air monitors for the 16 homes which had data for December 18–31, 2022, along with the significance from Wilcoxon signed-rank tests. ***p < 0.001.

Daily PM_2.5 level profile analysis

Figure 7 shows the daily mean profile of PM_2.5 at hourly intervals for the 5 homes with least variability and the 5 homes with most variability. For Home 6, the night-time PM_2.5 levels were hazardous while for Home 10, both mean PM_2.5 level and its variability were higher during the daytime than the night-time. These may be due to use of home appliance during the night and indoor activities during the day.

OPEN IN VIEWER

Figure 7.

The daily mean profile of PM_2.5 at hourly intervals for the 5 homes with the smallest variability of PM_2.5 (top row) and the 5 homes with the largest variability of PM_2.5 (bottom row). The vertical lines along the profiles are showing the standard deviations for those hours.

Discussion

In this analysis of the relationship of outdoor PM_2.5 and estimates from HAQES to indoor, home-based PM_2.5 measurements, we found that there was a wide range in the strength of this correlation and the strength of correlation may depend on the distances between the homes and the public air quality monitors though not statistically significant in our data. For each home there was typically high variability in day-to-day readings over time and within this variability, the indoor PM_2.5 measurements were usually lower than outdoors. Nevertheless, occasionally indoor PM_2.5 was much higher than outdoor measurements. Fourteen of the 23 homes during the monitoring period had levels suggested by the Indoor Air Hygiene Institute to be concerning (spikes in daily PM_2.5 levels above 35 μg/m³). ¹³ Notably the spikes in PM_2.5 detected at home were not reflected in most outdoor measures. Such “spike” periods may have been affected by indoor activities and may not be detected by an EPA station located even close to a home such as use of home appliances, burning wood in a fireplace, burning candles, cooking, smoking, or vigorous house cleaning. ¹⁴ The HAQES data did not provide a better correlation or proxy for indoor air qualities. This study underscores the importance of indoor air quality monitoring to accurately estimate an older adult's air quality exposure over time.

Despite the fact that the older population tends to spend most of their time indoors in their home and there are imperfect correlations between indoor and outdoor air quality as currently assessed, a majority of epidemiologic studies^1–4 have reported a relationship between outdoor air quality and dementia incidence with few exceptions. ¹⁵ This may reflect several factors in play. The overall air quality of a region can certainly affect the indoor air quality experienced over time. Thus, for example, during wildfire periods indoor air quality may be uniformly degraded across a large population. ¹⁶ However, experienced at the individual level, the degree of this exposure may vary depending on a dwelling's ventilation and air-tightness, ¹⁷ as well as each individual may have other personal behaviors that can affect their local home environment over time ranging from pet exposures, use of home appliances, to cleaning and cooking activities. To improve our understanding of the relationship of dementia to air quality and the environment, estimates of air quality exposure will need to improve in several ways. First, at the macro level, providing denser and more evenly spaced air monitor coverage of local areas will allow a more direct tying of the true everyday outdoor ambient environment that a person may experience when outdoors. Second, at the micro or personal home level, assessment of indoor residential sensors will be important as we have shown in this study that the indoor environment could often be hazardous. There were days when indoor air quality reached a concerning range not reflected in the outdoor measurements. Indoor activities and use of for example home appliances such as a humidifier could significantly influence the indoor air quality. ¹⁸ Finally, connecting all the air quality data to time of actual exposure by location whether outdoors or within a dwelling, as well as more detailed accounting of the person-level environment (e.g., household size, dwelling type, insulation and air infiltration), and occupants’ behaviors and activities (e.g., periods of cooking, cleaning, smoking), will help provide the most precise data to be able to gain insight into the mechanisms that drive the relationship of air quality to cognitive impairment and dementia over time.

This study had limitations. First, we have air quality data from only 23 homes, with all except one located in the state of Oregon in the United States. Therefore, the results may be limited in their generalizability. Another study limitation is that the devices used for collecting the air quality data are different between outdoor and indoor conditions. The Awair Omnis, which are relatively low-cost commercial-grade indoor air quality monitors were used to collect indoor air quality, while the data sources of outdoor air quality were from publicly available outdoor measurements generated from the EPA's Air Quality System (AQS) networks which use a variety of sensor systems dispersed in the field which are frequently calibrated. ¹⁹ It would be ideal to use the same sensors for indoor and outdoor measurements. However, that would not be affordable nor scalable given the cost of the sensors used in the AQS networks. The Awair Omni device has been independently tested by RESET which provides their accreditation to help ensure trusted reliable and scalable indoor air quality assessments to primarily the building industry. As such, within the limitations of inherent variability of these devices and real-world measurements, these data can be considered consistent for the purposes of demonstrating a proof of principle relationship of indoor residential PM_2.5 measurements to available public outdoor monitors. Future studies suggesting a relationship of the ambient environment to dementia will need to take these real-world measurement conditions into careful consideration. Another limitation of this study is that limited information was collected regarding the interior characteristics of these homes such as the exact location of ventilation ducts or whether windows were left open or closed. Therefore, we were not able to identify room-specific contributing factors in the discrepancies between indoor and outdoor air quality for most homes. Future studies should collect such data. Last, the indoor air quality monitors were placed in where participants slept as participants most likely spent the most amount of time there. However, this may not capture high PM_2.5 levels caused by activities such as cooking in the kitchen if they were far apart. Future studies should deploy air monitors in multiple spaces indoors, combine such air quality data with people's locations within the home, to get the most accurate air quality exposure.

This work represents an essential step toward determining the extent to which existing measures of air quality can be used for understanding cognitive health and dementia risks. While we examine the fidelity of air-quality measures at both regional and individual levels, ideally these measures can be combined with other measures of individual behavior such as time out of home, in-home activity, and home-ventilation status, to provide a more comprehensive picture of environmental- and individual-level risks and associations.

Supplemental Material