Contribution of Indoor-and Outdoor-Generated Fine and Coarse Particles to Indoor Air in Taiwanese Hospitals

This study quantified the contributions of both indoorand outdoor-generated particles to the air inside hospitals and examined whether air conditioning type, working area, working hours, and ambient pollution affect these contributions. Indoor and outdoor fine and coarse particles were measured at 33 hospitals, and the building characteristics were recorded. The infiltration factor (Finf) was calculated, and the contributions of both indoor and outdoor particles to indoor air were assessed. Additionally, their influencing factors on the indoor air were evaluated. The Finf values of fine and coarse particles were higher in hospitals with window and signal split type air conditioning than in hospitals with other types of air conditioning. No significant differences in the Finf values between working areas were observed. Outdoor-generated fine and coarse particles were major contributors to the indoor air, regardless of air conditioning type and working area. Higher contributions from indoor-generated fine and coarse particles to the indoor air were recorded in clinic waiting areas and lobbies during working hours than nonworking hours. Ambient air pollutant emissions and air conditioning characteristics influenced the contributions of indoorand outdoor-generated particles to indoor air according to a regression model. In summary, the contribution of outdoor-generated particles to indoor air must be reduced to improve occupants’ health in hospitals.


INTRODUCTION
Particulate matter (PM) is widely studied because of its effects on human health.Previous studies have identified a link between exposure to PM with aerodynamic diameters less than 2.5 µm (PM 2.5 ) and hospital admissions for respiratory diseases (Dominici et al., 2006;Tsai et al., 2014) and cardiovascular diseases (Dominici et al., 2006;Pope et al., 2008).Studies have also determined that PM 2.5 exposure increases the risk of diabetes mellitus (Pearson et al., 2010;Chen et al., 2013).Particle sizes between 2.5 and 10 µm (PM 2.5-10 ) have also a noticeable effect on human health.Cheng's study found that high PM 2.5-10 exposure increases the risk of hospital admission for respiratory diseases on cool days (Cheng et al., 2015).Heart rate variability was associated with PM 2.5-10 exposure among older adults with coronary artery disease (Lipsett et al., 2006).Therefore, investigating the characteristics and sources of PM is vital for reducing exposure.
People spend 80-90% of their time indoors; therefore, investigating the characteristics and sources of indoor PM is necessary, especially in hospitals.Hospitals are complex environments and differ from other indoor environments.Many patients and employees stay and work in hospitals.Studies indicated that PM exposure in hospitals was associated with sick building syndrome (Chang et al., 2015).Thus, understanding the characteristics and sources of PM in hospitals is necessary.Studies in Asia and Europe have determined that PM 2.5 (10-215 µg m -3 ) and PM 10 (58-250 µg m -3 ) levels differ greatly between hospitals (Wang et al., 2006a, b;Slezakova et al., 2012;Lomboy et al., 2015).Additionally, indoor activities (Wang et al., 2006b), carpets, and curtains (Verma and Taneja, 2011) are major sources of indoor PM in hospitals.Other major sources of indoor PM are outdoor PM 2.5 and PM 10 (Wang et al., 2006a, b;Slezakova et al., 2012;Lomboy et al., 2015).Studies have shown that both indoor air and outdoor air contribute to the PM inside hospitals.However, no study measured the contributions of indoor and outdoor air to PM in hospitals for assessing health risks and reducing exposure levels.
Studies have determined that the level of PM varies in different working areas in hospitals.PM 10 levels were higher in outpatient departments than in inpatient departments (Li et al., 2016) and the level of PM 2.5 was higher in clinics than in intensive care units (Lomboy et al., 2015).Verma's study found the level of PM 2.5 or PM 10 varied in different types of wards (Verma and Taneja, 2011).Air conditioning types also affect the distribution of PM levels in hospitals.A previous study found the levels of PM 2.5 and PM 10 were higher in hospitals with window and signal split type than those with central air conditioning (Jung et al., 2015).PM 2.5 levels were higher in hospitals with natural ventilation than in those with central air conditioning (Lomboy et al., 2015).However, the differences between indoor and outdoor air contributions to indoor PM in hospitals for different air conditioning types and working areas are inadequately understood.
This study investigated the contributions of indoor-and outdoor-generated PM 2.5 and PM 2.5-10 to the air in hospitals and assessed whether the air conditioning types and working areas affected the aforementioned contribution.Because hospitals are not accessible on a 7-d, 24-h basis to those who are not staff or patients, this study calculated these contributions at various working hours.Additionally, we used a regression model to analyze the effects of temperature, relative humidity, ambient pollution sources, gaseous pollutants, and air conditioning characteristics on the contributions of indoor-and outdoor-generated particles to air in hospitals.

Hospital Selection and Building Characteristic Questionnaire
We used a simple random sampling to select 50 hospitals from the government registry; 33 (average hospital age: 22 years) of the 50 hospitals agreed to participate in the study.From November 2007 to January 2008, we measured indoor air quality (IAQ) at two to four sampling sites within each hospital.We measured IAQ at nurse stations, clinics, clinic waiting areas, lobbies, and wards; the sampling site numbers were 7, 20, 21, 23, and 4, respectively.Outdoor sampling sites where outdoor air entered the hospital were included.Building characteristics, including air conditioning type, working hours, and ambient pollution sources, were also surveyed for each sampling site by the researchers or hospital staffs during the sampling period in accordance with a standardized checklist.Table 1 summarizes the characteristics of the 75 sampling sites from 33 participating hospitals.

Measurement of Indoor and Outdoor Air Pollutants
Indoor samplers were placed in the center of the lobby; at the other sampling sites (viz., nurse stations, clinics, clinic waiting areas, and wards), the indoor samplers were placed near walls (Fig. S1).At all sampling sites, instrument inlets were located 1.2-1.5 m above floor level.For the outdoor sampling sites, the samplers were located on balconies in proximity to the air inlet of the air conditioner.When a balcony of the hospital was unavailable, the samplers were attached to the windows, and their inlets were used to connect the tubing for outdoor sampling.Indoor and outdoor PM 2.5 and PM 10 levels were measured in the study hospitals over 24 h (starting at 09:00), and the PM 2.5-10 level was calculated by subtracting the PM 2.5 from the PM 10 .Moreover, indoor and outdoor carbon monoxide (CO), carbon dioxide (CO 2 ), ozone (O 3 ), temperature, and relative humidity (RH) levels were determined over 24 hours (starting at 09:00).
PM 2.5 and PM 10 levels were measured using DUST-TRAK Aerosol Monitors (Model 8520, TSI Corporation, Shoreview, MN, USA).A Q-TRAK Indoor Air Quality Monitor (Model 7575, TSI Corporation, Shoreview, MN, USA) was used to monitor CO, CO 2 , temperature, and RH levels.The O 3 level was measured using an Ozone Monitor (Model 202, 2B Technologies, Boulder, CO, USA).

Infiltration Factor Estimation
Some studies have used physics models or tracer elements to estimate the contributions of indoor and outdoor air to indoor PM (Meng et al., 2007;Allen et al., 2012;Ji and Zhao, 2015).However, these methods required sampling and analysis of the compositions of PM and thus were relatively expensive and inconclusive (Samek et al., 2016).The infiltration factor (F inf ) was employed to ascertain the penetration of outdoor PM into indoor air (Chen and Zhao, 2011).The method only measured the indoor and outdoor PM levels.Some studies have also used the F inf to estimate the contributions of indoor and outdoor air to indoor PM (Kearney et al., 2011;MacNeill et al., 2012).
The F inf can be calculated using Eqs.(1)-( 4).The level of indoor PM (C in ) is the sum of C out × F inf and C ins (1), where C out is the outdoor PM level, and C ins is the PM level generated by indoor sources.
Certain sampling sites were inaccessible to the staff or patients at night, such as the clinic, clinic waiting area, and lobby.At nurse stations and wards, human activity decreased at night.This study assumed that indoor sources could be ignored at night (Long et al., 2001); therefore, Eq. ( 1) can be simplified as Eq. ( 2).We used indoor and outdoor PM levels between 02:00 and 06:00 to calculate the F inf .
The contributions of the indoor-and outdoor-generated particles to indoor air were estimated according to Eqs. ( 3) and ( 4), where C outs is an indoor particle level that is contributed by the outdoor source.

Statistical Analysis
This study used one-way analysis of variance (one-way ANOVA) to analyze the differences in F inf values of PM 2.5 or PM 2.5-10 for different air conditioning types or working areas.Additionally, ANOVA was also used to test the differences in the contribution levels of indoor-or outdoorgenerated particles to indoor air according to different air conditioning types or working areas.Furthermore, the t-test was used to analyze the contribution of indoor-generated particles to indoor air during various periods.
A predictive model of indoor-and outdoor-generated PM 2.5 and PM 2.5-10 contributions to indoor air was analyzed using a step-wise regression model (backward elimination), and variables were required to be p < 0.05.We analyzed the associations between all variables and indoor-generated and outdoor-generated particles and removed variables that were statistically nonsignificant in the regression model.We repeated this process until no more variables could be removed without a statistically significant change in the regression model.SAS (v 9.3) was used for data analysis.

Hospital Characteristics and IAQ
Hospital characteristics are shown in Table 1.Most sampling sites from study hospitals near main roads (81%), parking lot entrances (68%), or with plant growth (55%) exhibited substantial emission of particles.Over two-fifths of the sampling sites were located near loading docks (43%) or restaurants (40%).The other characteristics can be found in Table 1.
The indoor PM 2.5 and PM 2.5-10 levels for various air conditioning types and working areas are presented in Tables S1 and S2, respectively.Indoor PM 2.5-10 levels differed with air conditioning type (p < 0.05), indicating that air conditioning type affects the characteristics of indoor PM 2.5-10 in hospitals.No significant differences were observed in the indoor PM 2.5 levels between air conditioning types or the levels of PM 2.5 and PM 2.5-10 between working areas.Tables S1 and S2 indicate that temperature and RH did not significantly differ between air conditioning types or working areas.These results imply that air conditioning usage controls the temperature and RH to within a stable range in Taiwanese hospitals.The characteristics of gaseous pollutants for different air conditioning types and working areas were ascertained in our previous study (Jung et al., 2015).

Infiltration Factor Estimates
The average F inf values of PM 2.5 and PM 2.5-10 were 0.63 and 0.54, respectively.The F inf values of PM 2.5 and PM 2.5-10 for air conditioning types and working areas are reported in Table 2. Hospitals with window and signal split type have higher F inf values of PM 2.5 and PM 2.5-10 (p < 0.05) than those with other air conditioning types.No statistical significance was observed for the F inf values of PM 2.5 and PM 2.5-10 between working areas.* One-Way ANOVA was used to examine the differences in F inf values of PM 2.5 or PM 2.5-10 in different air conditioning types or working areas (statistical significant was set at p < 0.05).

Indoor-and Outdoor-Generated Particles
Table 3 presents the contributions of indoor-and outdoorgenerated particle levels to indoor air.The contributions of indoor-and outdoor-generated PM 2.5 to indoor air were 2.0 µg m -3 (14%) and 12.0 µg m -3 (86%), respectively, and the outdoor contribution was significantly higher (p < 0.05).A similar result was found for indoor-and outdoorgenerated PM 2.5-10 (indoor: 2.6 µg m -3 (18%); outdoor: 12.1 µg m -3 (82%), p < 0.05).The contributions of indoorand outdoor-generated PM 2.5 and PM 2.5-10 to indoor air were calculated for air conditioning types and working areas in Table 3.Our data shows that outdoor-generated PM 2.5 and PM 2.5-10 were major contributors to indoor air for different air conditioning types and working areas.
Table 4 reveals higher contributions of indoor-generated PM 2.5 and PM 2.5-10 to indoor air in both the clinic waiting areas and lobbies during working hours than during nonworking hours (p < 0.05).Additionally, higher contributions of indoor-generated PM 2.5 and PM 2.5-10 to indoor air were recorded during working hours than during nonworking hours at nurse stations, clinics, and wards; however, statistical significance was not attained for this result.Fig. 1 reveals that the average contribution of indoor-generated and outdoor-generated PM 2.5 and PM 2.5-10 in all study spaces (N = 75) increased gradually from 08:00 and steadily declined from 16:00.

Effect Factors of Indoor-and Outdoor-Generated Particles
A step-wise regression model was used to analyze the factors (ambient pollution, air conditioning characteristics, and weather) affecting the contributions of indoor-generated and outdoor-generated PM to indoor air (Table 5).Obstacles in the air outlet (β = 2.82) and lack of a return air pathway (β = 2.15) were major factors affecting the contribution of indoor-generated PM 2.5 (42% explained).The lack of a * One-Way ANOVA was used to examine the differences in the contribution levels of PM 2.5 or PM 2.5-10 in different air conditioning types or working areas (statistical significant was set at p < 0.05).** T-test was used to examine the differences in the contribution levels of indoor-and outdoor-generated PM 2.5 or PM 2.5-10 according to air conditioning types or working areas (statistical significant was set at p < 0.05).5) 0.73 1.9 (13) 0.7 (8) 0.44 * T-test was used to examine the differences in the contribution levels of indoor-generated PM 2.5 or PM 2.5-10 in different working hours according to working areas (statistical significant was set at p < 0.05).

DISCUSSION
This study determined that the F inf is higher in hospitals with window and signal split type air conditioning than those with central air conditioning types.Moreover, outdoor-generated PM 2.5 and PM 2.5-10 were the principal contributors to indoor air, regardless of air conditioning type and working area.During working hours, higher contributions of indoor-generated PM 2.5 and PM 2.5-10 were recorded in both clinic waiting areas and lobbies than during nonworking hours.Air conditioning characteristics affected the contribution of indoor-generated particles to indoor air.The outdoor CO level, ambient pollution sources, and air conditioning characteristics influenced the contributions of outdoor-generated particles to indoor air.
The mean PM 2.5 level recorded in the present study was lower than that recorded in other studies (Wang et al., 2006a, b;Li et al., 2016); however, the PM 2.5-10 levels recorded in the present study were similar to those recorded in the aforementioned studies.In those studies (Wang et al., 2006a, b;Li et al., 2016), the hospitals were closed to the main roads, restaurants, and industrial areas, and the mean outdoor PM 2.5 levels were higher (86-105 µg m -3 ).Moreover, those studies also found that the ratios of indoor and outdoor (I/O ratio) PM levels were below 1.0 and suggested that outdoor-generated PM was an important source of indoor PM.This was the reason for the higher level of indoor PM in the previous studies.
In a previous study, the F inf (nighttime I/O) of PM 2.5 and PM 2.5-10 ranged from 0.1 to 1.2 and from < 0.1 to 0.9, respectively (Long et al., 2001).The daily median F inf values of censored I/O and sulfur I/O for PM 2.5 were 0.55 and 0.49, respectively, in winter, and 0.80 and 0.83, respectively, in summer (MacNeill et al., 2012).Our estimation was similar to the previous studies.Therefore, our results reflected that the F inf estimation was reasonable and can be used to calculate the contributions of indoor-and outdoor-generated PM to indoor air.
Table 2 shows that the F inf values of PM 2.5 and PM 2.5-10 were higher in hospitals with window and signal split type air conditioning than in those with other types of air conditioning.We speculate that more window openings were present in hospitals with this type of air conditioning, increasing the F inf values.Studies have indicated that window opening increased the F inf values (Long et al., 2001;MacNeill et al., 2012).Moreover, in this study, the correlation between the indoor and outdoor PM 2.5 level in hospitals with window and signal split type air conditioning was superior to that of central air conditioning (Supplementary Table 3).Hospitals with central air conditioning install air filters to remove PM from the outdoor air.One study also found that the level of aerosol was lower in indoor spaces with central air conditioning with high-efficiency particulate air filters than in those with signal split type air conditioning (Chuaybamroong et al., 2008).Thus, the F inf value was higher in hospitals with window and signal split type air conditioning than in those with other types of air conditioning because outdoor air was more directly combined with indoor air.
In this study, most of the hospitals used central air conditioning, and less window opening occurred.Therefore, we hypothesized that the contributions of indoor-generated particles to indoor air were dominant.However, our results reveal that outdoor-generated PM 2.5 and PM 2.5-10 contributions to indoor air were the principal contributions in terms of contribution level and percentage (Table 3).This result contradicted our hypothesis and the findings of a previous study (MacNeill et al., 2012).In MacNeill's study, they conducted the measurement in homes and found that oven use, candle burning, and wood fireplaces are associated with the contribution of indoor-generated fine particles.However, no indoor burning occurs in our study hospitals; therefore, the contribution of indoor-generated particles was lower.Our data indicated that the I/O ratios of PM 2.5 and PM 2.5-10 were below 1.0 (PM 2.5 : 0.70; PM 2.5-10 : 0.61); these results were similar to those of previous studies (Wang et al., 2006a, b;Slezakova et al., 2012;Lomboy et al., 2015).Therefore, outdoor air is a key source of indoor PM in Taiwanese hospitals.However, the correlation between indoor PM and outdoor PM in hospitals with central air conditioning was low (Table S3).A time delay may affect this association.Similar results were found for the contributions of outdoor-generated particles to indoor air between different working areas (Table 3).These results suggest that the effect of outdoor PM on indoor air in hospitals should not be ignored, regardless of the air conditioning types or working areas.
To the best of our knowledge, this is the first study to investigate the different contributions of indoor-generated particles to indoor air for various working hours in hospitals (Table 4).The contributions of the indoor-generated PM 2.5 and PM 2.5-10 to indoor air were high during working hours, and statistically significant differences were observed in both the clinic waiting areas and lobbies.The concentrations of indoor-generated PM 2.5 and PM 2.5-10 gradually increased after 08:00 and gradually declined after 18:00 (Fig. 1).Human activities, such as walking and cleaning, may be influencing factors.Ferro's study used a mathematical model to calculate the effect of human activity on the level of PM (Ferro et al., 2004).They found PM levels gradually increased during human activity, such as walking, dancing on carpeted surfaces, and dancing on noncarpeted surfaces.One study indicated that cleaning behavior also increased the particle emission rate using a chamber test (Géhin et al., 2008).Mašková's study also found that the movement of visitors during visiting hours was a major source of particulate matter in library (Mašková1 et al., 2016).Thus, human activities increase PM level during working hours.
Ambient pollution (Massey et al., 2012;Chithra and Nagendra, 2013), air conditioning characteristics (Chithra and Nagendra, 2012), and weather (Chithra and Nagendra, 2012;Massey et al., 2012) have been demonstrated to affect indoor PM concentration.A regression model was used and indicated that obstacles in the air outlet and the lack of a return air pathway increased the contribution of indoor-generated PM 2.5 , and the lack of a return air pathway increased the contribution of indoor-generated PM 2.5-10 (Table 5).The PM level should increase in air conditioning systems without a return air pathway.Moreover, obstacles in the air outlet prevented the air flow from effectively removing PM (Hu et al., 2014); therefore, these obstacles may influence variations in indoor PM concentrations in hospitals.
Hospitals near restaurants were associated with the contribution of outdoor-generated PM 2.5 to indoor air.A previous study indicated that particulate matter emission from restaurants affected the air quality (Lung et al., 2011).Moreover, our results indicated that the outdoor CO level and lack of a return air pathway were positively associated with the contribution of outdoor-generated PM 2.5-10 .Studies have indicated that CO is a marker for traffic-related air pollution (Jo and Lee, 2006;Both et al., 2013) and that traffic pollution produces particulate air pollutants (Charron and Harrison, 2005;Both et al., 2013;Oakes et al., 2016).Moreover, 81% of the sampling sites from the study hospitals in this study were near a main road.Therefore, traffic pollution could be an important factor contributing to the outdoor-generated PM and the resulting indoor air.In hospitals, particulate air pollutants from the outdoor air can infiltrate indoor air through air conditioning.The lack of a return air pathway may hinder particulate air pollutant removal in indoor air and lead to a higher contribution of outdoor-generated PM 2.5-10 to indoor air.Several limitations affected this study.First, the air exchange rate was not measured in calculating the F inf value.Calculation of the F inf value used the levels of indoor and outdoor PM between 02:00 and 06:00; we assumed that indoor sources of PM were negligible in these periods.Therefore, measuring the air exchange rate to calculate the F inf value was unnecessary.Moreover, the previous study successfully used the same equations to calculate the F inf value (Long et al., 2001), and our value was also similar to those of previous studies (Long et al., 2001;MacNeill et al., 2012).Therefore, the value was reasonable for calculating the contribution of indoor-generated or outdoor-generated PM to indoor air.Second, we did not complete the sampling over four seasons in characterizing seasonal variations in the F inf value.In this study, 88% of the sampling sites used central air conditioning to control the temperature and air flow.The temperature and air flow are generally stable when central air conditioning is used; therefore, the F inf estimation may remain stable across seasons.

CONCLUSIONS
This is the first study to calculate the infiltration factors and contributions of indoor-and outdoor-generated fine and coarse particles to the air inside hospitals.Our study found that air conditioning type affects the infiltration factor.Outdoor air was a primary source of indoor PM.The air conditioning characteristics and ambient pollution influence the contributions of indoor-and outdoor-generated fine and coarse particles to indoor air.This paper provides valuable suggestions for reducing PM exposure and effectively assessing the health risks of occupants in hospitals.
Diurnal patterns of average hourly indoor-and outdoor-generated particles in all study spaces (n = 75).(a) PM 2.5 and (b) PM 2.5-10 .

Table 2 .
Infiltration factor of particles for different air conditioning type and working area (mean).

Table 3 .
Contribution levels of indoor-and outdoor-generated particles to indoor air for different types of air conditioning and working areas (µg m -3 (%)).

Table 4 .
Contribution level of indoor-generated particles to indoor air for different periods according to various working areas (µg m -3 (%))

Table 5 .
Predictive model for the contribution levels of indoor-and outdoor-generated PM The effect was selected into the final model when the p < 0.05 by a step-wise regression model. *