Chicago, Illinois, USA Published online: 30 Nov 2010.

This article was downloaded by: [148.251.235.206] On: 17 August 2015, At: 11:03 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick Place, London, SW1P 1WG Aerosol Science and Technology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/uast20 Time and Space Uniformity of Indoor Bacteria Concentrations in Chicago Area Residences D. J. Moschandreas a, K. R. Pagilla a & L. V. Storino a a Department of Chemical and Environmental Engineering, Illinois Institute of Technology, Chicago, Illinois, USA Published online: 30 Nov 2010. To cite this article: D. J. Moschandreas, K. R. Pagilla & L. V. Storino (2003) Time and Space Uniformity of Indoor Bacteria Concentrations in Chicago Area Residences, Aerosol Science and Technology, 37:11, 899-906, DOI: 10.1080/02786820300935 To link to this article: http://dx.doi.org/10.1080/02786820300935 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content ) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Aerosol Science and Technology, 37:899 906, 2003 Copyright c American Association for Aerosol Research ISSN: 0278-6826 print / 1521-7388 online DOI: 10.1080/02786820390229039 Time and Space Uniformity of Indoor Bacteria Concentrations in Chicago Area Residences D. J. Moschandreas, K. R. Pagilla, and L. V. Storino Department of Chemical and Environmental Engineering, Illinois Institute of Technology, Chicago, Illinois, USA The objective of this article is to assess spatial and temporal variation of indoor gram-positive bacteria and Staphylococcus sp. in 20 urban residences. At each residence, air was sampled at one outdoor site and four indoor sites (rooms) to assess spatial variation and once each season for five consecutive seasons to assess temporal variation. All temporal and spatial comparisons were performed using data obtained by the Andersen sampling technique. A secondary objective of this study is to evaluate relationships among several sampling methods used to measure bacteria levels; since differences in the measured concentrations are expected, the focus is to discern if corresponding measurements relate to each other. Not surprisingly, comparisons among the four sampling systems revealed statistically significant differences, although levels correlated relatively well. Using data only from the Andersen samplers, we conclude that seasonal variation is residence dependent with typically higher summer levels, but a clear pattern of variation could not be established. Room-to-room difference is not statistically significant, but the basement levels render the basement a distinct microenvironment. Indoor concentrations exceeded corresponding outdoor concentrations 75% of the time. Staphylococcus accounts for approximately 27% of the average indoor bacteria levels. The highest levels of gram-positive bacteria are found in the kitchen. The presence of multiple indoor sources with variable emission rates in multiple indoor locations results in bacteria levels that vary with time and space. INTRODUCTION Biological agents, such as fungi, bacteria, endotoxins, protozoa, and antigens, are natural and ubiquitous components of the earth s ecosystems. Many of these organisms and their byproducts are harmless to humans. Even those that are associated with human disease are usually only problematic when they Received 17 December 2001; accepted 5 May 2003. Address correspondence to D. J. Moschandreas, Department of Chemical and Environmental Engineering, Illinois Institute of Technology, Perlstein Hall, Suite 127, 10 West 33rd Street, Chicago, IL 60616, USA. E-mail: moschandreas@iit.edu become amplified under specific conditions (Cutter Information Corp. 1994). It is well established that bacteria are agents of important infectious diseases (Burge 1995). Occasionally such bacteria in droplet nuclei from individuals shedding pathogenic agents can be rapidly and evenly dispersed throughout a closed indoor environment by means of the Heating, Ventilation, and Air-Conditioning (HVAC) system (Riley 1979). Some medically important bacteria may also be transmitted via aerosols and can cause respiratory problems (Verhoeff 1993). Building construction has become more energy efficient, and HVAC systems are designed to use less outdoor air. Therefore, mixing with outdoor air does not readily dilute indoor biocontaminants. Additionally, building envelopes have become tighter and contribute to the buildup of moisture in indoor environments, which facilitates growth of micro-organisms. The buildup of dirt and debris in poorly maintained HVAC systems is a potential source of nutrients for micro-organisms. Thus indoor environments may provide the appropriate nutrients, water, and temperature for the amplification and aerosolization of micro-organisms (Cutter Information Corp. 1994). The field study reported in this article sampled twenty residences in the city of Chicago to measure biological pollutant levels at four indoor locations and one outdoor location by each residence. The objective of the study was to examine spatial and temporal distributions of indoor bacteria levels. A second objective was to investigate relationships among several methods of field monitoring levels of bacteria. This objective was designed to determine the type of measurement method and the type of sampling system to be used in this research based on literature-reported methods. Previous investigations by several researchers have extensively researched measurement methods and their comparative advantages and disadvantages, and hence the objective of comparing measurement and sampling methods was a objective of this study (Hernandez et al. 1999; Henningson et al. 1997; Terzieva et al. 1996). Determinants of indoor biological pollutants that were investigated include relative humidity, temperature, number of 899

900 D. J. MOSCHANDREAS ET AL. people inhabiting the residence, and the presence of pets. Waterdamaged versus nonwater-damaged residences were measured within six months of water damage to investigate the potential impact of accidental water release indoors. Major gram-positive bacteria and Staphylococcus were cultured to identify their indoor levels. Indoor levels were compared with outdoor levels to investigate the presence of indoor sources. Correlation between the Biotest RCS centrifugal air sampler and the Andersen six-stage cascade impactor was investigated for any difference in measurement techniques of total culturable bacteria levels using agar as the collection medium. Similarly, correlation coefficients were established for measurements by the Modified Andersen and the liquid impinger samplers using water as the collection medium and were investigated to assess the difference in measurement of total culturable and nonculturable bacteria levels. The analytical measurements were made by UV fluorescence microscopy using actinide orange dye stain (Hernandez et al. 1999; Moschandreas et al. 1996). MATERIAL AND METHODS In this study we used an inertial sampler known as the Andersen six-stage cascade impactor (Andersen 1976), the Biotest Reuter Centrifugal Sampler (RCS) (Biotest Diagnostics Corporation 1980), liquid impingement sampler (AGI-30 impinger), and Modified Andersen sampler with UV fluorescence microscopy. Each of these methods is explained briefly in the following paragraphs. The Andersen six-stage sampler separates particles into six aerodynamic ranges. The particle size ranges sampled by the six-stage Andersen sampler are from 0.65 µm for stage 6 to greater than 7 µm for stage 1, as illustrated in the horizontal axis of Figure 3 of this paper. Each stage contains a plate with 400 orifices. Directly below each stage is a petri dish containing a selective or nonselective nutrient agar medium, and air is drawn in at a rate of one cubic foot per minute (cfm) for 20 min. After the samples are collected, the petri dishes are incubated and the colonies are counted using microscope or the positive-hole correction method, which counts colonies under a magnifying lens. The total count per plate is then adjusted using the positive hole correction table, which accounts for the jets that deliver culturable particles to the petri plates. The observed culturable airborne bacteria concentrations are reported in colony-forming units per meter cubed of sampled air (CFU/m 3 ). The RCS sampler is a multibladed impeller that draws air into the sampler from up to 40 cm away. The microbes are impacted on a plastic strip containing an agar medium inside of the impeller housing. The air volume relevant for calculating the number of biological organisms per air volume is 40 liters/min. The sampling time employed in this study was 8 min. The conventional steps of adjusting the impeller angles before each sampling and changing the sampler batteries on daily basis were implemented. After the samples are collected, the strips were incubated and the colonies are counted under a magnifying lens, and the levels are denoted in CFU/m 3. The RCS is used in many field studies because it is a compact field sampler that is battery-operated and samples a large volume of air in a short time. Sampling by liquid impingement (AGI-30 sampler) collects microbes directly into a liquid, which provides some protection to the micro-organisms and allows immediate initiation of repair of any damage that may have been caused by the collection process itself. The sample liquid can be processed to detect a wide range of aerosol concentrations. In this method, 30 ml of distilled water is bubbled with the sampled air at 12.5 L/min flow rate for 60 min. It was shown by Hernandez et al. (1999) that the evaporative losses from the liquid in the impingers is less than 2% for AGI-4 impinger. Although distilled water may induce osmotic stress on the organisms, we minimized the time period between sampling and filtration and analysis. If nutrients are added to the distilled, it is possible that a culturing effect could be introduced. Hence distilled water was used as the collection media. Then the sampled water is filtered thorough a 0.2 µm black Nucleopore membrane filter. The filter is stained with 3 ml of 5 mg/l acridine orange solution for 3 min. Each filter is counted at 10 random locations using an Olympus BH2 UV fluorescence microscope at 1000 magnification. The bacterial count is represented as total cells per m 3 of air sampled. A filter in a plastic cassette attached to a personal sampling pump can be used to collect culturable dust and inactive cells such as spores. Culturable cells, however, will most likely be dehydrated and killed by the large volumes of air that pass over the filter in the course of sampling. Therefore, filtering is not recommended for culturable cells (Ness 1991). To measure both culturable and nonculturable bioaerosols, the Modified Andersen impactor-ffdc method may be used (Moschandreas et al. 1996; Hernandez et al. 1999). The Modified Andersen impactor-ffdc method uses a six-stage Andersen sampler with filtered distilled water instead of an agar medium (Modified Andersen sampler). After collection, the water is filtered on a 0.2 µm black filter and stained with an acridine orange dye as in the impingement method. The filter is then mounted on a slide and viewed under an ultraviolet fluorescence microscope at 1000 magnification to get total counts for culturable and nonculturable bacteria. Liquid Impinger and Modified Andersen sampling were carried out in a side-by-side approach to compare results by the two methods that measure both culturable and nonculturable aerosols. All samplers were operated with the optimum air flow rates for that particular sampler, and the bacterial count was normalized to per cubic meter of air. A Quality Assurance and Quality Control (QA/QC) plan was followed during the preparation of the agar medium, sterilization of equipment, and handling, shipping, and documentation of samples throughout the field study. All sampling equipment and glassware were sterilized using an autoclave prior to use whenever possible. At each sampling site, equipment was sterilized using isopropyl alcohol inbetween sampling. One of every ten agar plates was used as a media blank, and one agar plate per sampling date was used as a field blank. Similarly, one agar strip

UNIFORMITY OF INDOOR BACTERIA CONCENTRATIONS 901 per sampling date was used as a field blank. Media blanks refer to prepared petri dishes, which were incubated immediately after preparation to check for contamination. Even if one colony appeared on blank agar plates after a 48 h period, sample plates made from the same batch were discarded. Field blanks refer to petri dishes that were subjected to the same transportation and handling conditions each sampling day, but no air was sampled by the sampling system using field blanks. Field blanks were incubated for 48 h; if colonies appeared on the field blanks, the sample total counts were adjusted to reflect this finding. Equal amounts of filtered water used for the Modified Andersen and impinger methods were stained with acridine orange each sampling date and counted for bacteria by UV flourescence microscopy. Background bacteria counts from each method were subtracted from the measured total counts. For documentation purposes, an identification form was prepared for each sampling method and was used to keep record of each sampling event. The information contained in the documentation forms included date, time, location, flow rates, duration of sampling, identification numbers, and counts. Finally, a logbook was maintained to describe the site and conditions that may affect sampling counts. Experimental Design Volunteers were sought in Chicago by placing an advertisement in newspapers and on the radio. Twenty residences were selected randomly among those that volunteered to allow the team to sample their residences. A two-stage sampling design was carried out. Using data from the first stage, n = 6, and assuming that log-transformed bacteria concentrations are normally distributed, we determined that a sample of 20 residences would detect differences of 300 CFU/m 3 for a population variance of 16 10 4, with α = 0.05 and β = 0.10. For the second sampling stage we selected an additional 14 residences randomly from our volunteers. The first objective of the study, determining spatial and temporal variation of indoor biological pollutants, was investigated with data from sampling with the Andersen sampler, which was used for all sampling sites and all seasons. Correlation coefficients relating the various sampling methods were established by setting side-by-side samplers over short periods of time, one or two seasonal sampling campaigns. Sampling was carried out over five seasons: winter sampling took place twice to ascertain the difference between biological pollutant levels over the same season. The bathroom, kitchen, bedroom, basement and one outdoor location were sampled at each residence to determine the levels of total airborne bacteria, culturable bacteria, and Staphylococcus. Five of the twenty residences had experienced water damage from either seepage of water through cracks in the basement or from sewer problems within the last six months of the initial sampling. The impact of such water damage was investigated using this rather small database. Descriptive statistics were estimated to summarize observations from all sites and times, and spatial and temporal differences were assessed with t-tests for two populations and with one-way ANOVA for several populations. Additionally, χ 2 tests are used to test difference for proportions, and two-way ANOVA to test for two factors. Bonferroni post-hoc tests were used to determine the source of statistical difference in multiple mean comparisons (Wagner 1969). RESULTS Results from this study are presented in four sections. The first section deals with descriptive statistics summarizing results of indoor air sampling for total culturable bacteria, and Staphylococcus sp. The second section clarifies relationships between the Andersen sampler and the RCS sampler, and the Modified Andersen-FFDC and liquid impingement-ffdc methods. The third section assesses determinants of indoor bacteria. The last section addresses spatial and temporal relations. In the balance of this article, natural log transformed data and significance level α = 0.05 are used for all inferences. Data Summaries Descriptive statistics of indoor airborne bacteria measured by the Andersen sampler are given in Table 1. Table 2 illustrates the levels of gram-positive bacteria and Staphylococcus as a percentage of the total culturable bacteria. Figures 1 and 2 illustrate seasonal and spatial variation of total culturable bacteria, respectively, with box-plots. Both the spatial and temporal distributions are skewed to the high end, as illustrated by Table 1 Average and standard deviation of indoor airborne bacteria levels Bathroom Kitchen Outdoor Bedroom Basement CFU/m 3 x avg SD x avg SD x avg SD x avg SD x avg SD Winter 640 503 455 443 332 211 506 365 422 297 Spring 568 478 498 511 245 169 791 1009 299 247 Summer 1003 753 1026 814 444 371 818 660 1207 1966 Fall 1061 1089 758 621 302 169 672 323 457 453 Winter II 491 417 350 203 101 107 394 262 403 461 Average 753 719 620 600 286 249 636 604 558 975

902 D. J. MOSCHANDREAS ET AL. Table 2 Gram-positive bacteria and Staphylococcus sp. levels as a percentage of total culturable bacteria Bedroom Bathroom Kitchen Basement Outdoors (%) (%) (%) (%) (%) Gram-positive bacteria 58 62 75 68 50 Staphylococcus 37 26 30 16 10 the number of measurements denoted by the circles in each figure. Typically, summer levels and bathroom levels of culturable bacteria appear to be the highest, with winter and basement levels being the lowest among indoor locations. The highest levels of gram-positive bacteria are found in the kitchen. This can be attributed to increased occupant activity patterns in the kitchen. Staphylococcus was found to account for approximately 37% of bedroom bacteria and for 27% of the average indoor bacteria levels. This may be attributed to the sloughing or shedding of human skin or to occupant activities such as changing bed sheets and other cleaning activities. Indoor concentrations exceed corresponding outdoor concentrations 75% of the time, indicating the presence of indoor sources. The size distribution of bacteria by sample site was obtained for both total culturable bacteria and Staphylococcus. (Figures 3 and 4). The size distribution is uniform at all sampling sites. About 25% of the total bacteria fall in the largest size group of 7 µm or more, while approximately 5% of the bacteria are at the smallest size range of less than 1.1 µm. Single bacterial cells are smaller than the observed sizes. One may speculate that either aggregation or higher survival rate of aggregated bacteria may explain the observed size distribution. The distribution is divided almost equally in the remaining four size categories. Approximately 40% of indoor bacteria levels have size that enables them to penetrate the lung (diameter < 3.3 µm). Similarly, roughly 50% of Staphylococcus levels was in the size range capable of penetrating the lungs. These may pose a human health concern since exposure to Staphylococcus are known to cause toxic shock syndrome, scalded skin syndrome, and soft tissue infections. As indicated, the distribution of bacteria levels in all sites is skewed right for all locations. The outdoor bacterial levels have a peak value of 200 CFU/m 3, with a high probability of occurrence of about 30% and a small dispersion around the peak. The indoor locations have larger peak values with a 15 20% probability of occurrence and large dispersion around the peak. These characteristics demonstrate that indoor sources cause elevated indoor concentrations with large variation possibly due to variable indoor sources. Measured indoor pollutant concentrations higher than corresponding outdoor concentrations indicate the presence of indoor sources emitting pollutants. The probability distribution of observed indoor to outdoor ratios of all data collected over the five seasons (n = 368) at all indoor locations indicates that over 75% of the ratio values are greater than one, Figure 1. bacteria. Box plot of seasonal variation of total viable Figure 2. Box plot of spatial variation of total viable bacteria.

UNIFORMITY OF INDOOR BACTERIA CONCENTRATIONS 903 Figure 3. Size distribution of total viable bacteria over all seasons in different rooms. Figure 4. Size distribution of Staphylococcus sp. over all seasons in different rooms.

904 D. J. MOSCHANDREAS ET AL. further suggesting the presence of indoor sources. A few ratio values were found to be as high as 14. Comparison of Sampling Methods Several methods were used to investigate the presence of indoor culturable and total bacteria. The Andersen six-stage impactor was used as the benchmark for comparison of culturable bacteria and for the spatial and temporal variations investigated in this article. The Biotest RCS sampler was used simultaneously with the Andersen six-stage impactor during the first season (winter) to measure bacteria levels in the bathroom of each residence, and sample size for comparisons was n = 100 for each sampling device. While this was not the main objective of the study, bacteria levels found with the RCS sampler were compared with those found with the Andersen six-stage sampling method. Agreeing with several other studies comparing bacteria levels obtained by different samplers, the null hypothesis of no difference between RCS and Andersen samples is rejected in favor of higher levels using the RCS sampling method. It is interesting to note that the side-by-side sampling leads to a relatively strong correlation as indicated by the r 2 = 0.62 value of the coefficient of determination. Thus, while measurements from the two methods correlate relatively well, depending on the site of measurement the r 2 value varied from 48 to 62 percent, which means the difference of the measured levels is statistically significant. Using the bathroom as the experimental unit and during winter sampling, the only period with complete simultaneous sampling with both methods, the average CFU/m 3 in the case of RCS sampler was 2,103 with a standard deviation of 1,275 compared to the average CFU/m 3 and standard deviation of 640 and 503, respectively, for the Andersen sampler. Jensen et al. (1992) observed similar behavior and reported that the difference may be due to the sampling efficiency of each instrument and possibly due to particle size effects. The RCS sampler has a higher sampling efficiency than the Andersen sampler, especially for larger particles, and this may explain the higher bacteria levels measured by the RCS sampler. Furthermore, this could be due to the fact that the RCS sampler is single stage and large particles do not have the bounce effect seen with smaller particles. It is possible that if the RCS sampler is used for larger particles ranges, it is more efficient than the Andersen sampler, and vice versa for the Andersen sampler. Griffiths and Stewart (1999) indicated the differences are due to sampling and collection efficiency differences of the sampling systems. It is possible that the overall efficiency is also influenced by the impacts of the sampler on the culturability of the organisms such as bounce effects and particle shear. During the fall and the second winter sampling seasons, the liquid impinger method was used for sampling side-by-side with the Modified Andersen sampling method. The fall season Modified Andersen and impinger sampling mean (and standard deviation) levels in the bathroom site are 1,538 (1,106) and 9,982 (6291), respectively; the corresponding values for the second winter season are 1,391 (1,411) and 9,869 (9031). The sample size for each season was n = 100. Each method measures culturable and total bioaerosol particles per cubic meter of air sampled. A one-sided t-test rejected the null hypothesis of no difference, indicating that impinger sampling method measurements are higher than those obtained by the Modified Andersen method. This statistically significant difference is attributed to the different collection techniques. The liquid impinger uses a gas-liquid interface contact as the collection method, where as the Modified Andersen sampler uses an impaction method for collection. The air-liquid contact time is much greater in the liquid impinger, which may contribute to greater collection efficiency. The measured airborne concentrations of culturable bacteria are influenced by many factors including aerosol sampling efficiency, sampler collection efficiency, survival of the bacteria in the sampler, culture conditions, sampling time, and others. Thus the statistical differences noted in the side-by-side measurements are not surprising, they simply confirm findings of studies found in the literature (Griffiths and Stewart 1999; Jensen et al. 1992). This study did not attempt to isolate the cause(s) of the difference in the measured levels. Potential sources include different sampling efficiencies, different particle cut-off size, and possibly different bounce effect. Yet the relatively high value of the coefficient of determination estimated in the comparison between the Andersen and RCS samplers indicates that the relative findings, but not the absolute levels, would be similar. Since the coefficient of determination of the relationship between the Modified Andersen sampler and the liquid impinger is smaller, this conclusion would not be valid for these two sampling methods. Determinants of Indoor Air Bacteria It is suggested in the literature that there are many indoor sources and factors for culturable airborne bacteria such as relative humidity, temperature, water damage, number of home residents, and the presence of pets (Flannigan et al. 2001). These factors were investigated for any association with the bacteria levels observed from sampling by the Andersen samplers. Relative humidity and temperature measurements were taken at the time of air sampling. Correlation coefficients were calculated to relate bacteria levels to these parameters. A maximum of 27% of the variation in kitchen bacteria levels can be attributed to the variation in temperature during the winter sampling. Increased relative humidity levels in the home are thought to cause elevated bacteria levels within the home. Data were characterized into dry locations, those having relative humidity levels less than 35%, and wet locations, those having relative humidity levels greater than 35%. Since relative humidity under controlled residential conditions did not vary much among homes, this classification of dry and wet sites was applicable only during the two winter seasons. Analysis of this portion of our database suggests that the relative humidity is not a major cause of elevated bacteria levels in Chicago residential environments. Five of the twenty residences sampled had reported water damage within six months of the initial sampling. The water

UNIFORMITY OF INDOOR BACTERIA CONCENTRATIONS 905 damage occurred primarily in the basement of the five residences. The basement levels of the first sampling season, winter, were grouped into those greater than the outdoor level, and those less than the outdoor level. Ratios of bacteria levels in water-damaged homes over the outdoor levels were compared with bacteria levels in nonwater-damaged homes over the outdoor level. The outdoor level served as the background concentration. Statistical tests reveal no significant difference of the proportions of houses with elevated bacteria levels due to water damage between the two house groups. This may be attributed to the following two facts: (1) in no case was the period between sampling and the water incident shorter than three weeks, and (2) all occupants were aware of the potential impact and had taken specific steps to dry out their homes. It is suspected that an increased number of residents in a home results in increased bacteria levels. Residences were grouped into those that had fewer than three occupants three occupants, and more than three occupants. The null hypothesis cannot be rejected; that is, there is no significant difference in bacteria levels as a function of the number of occupants. This may be attributed to the large variation of bacteria levels and the activity patterns of the residents. Having pets in a home is also thought to cause elevated bacteria levels within the home. Data was grouped into those residences with a pet and those without a pet. The pets included cats and dogs only. A t-test of no difference between the two classes of houses was rejected. Further investigation of the database demonstrates a surprising result: homes with a pet in the dwelling have lower indoor bacteria counts. This conclusion contradicts the nearly universal finding that pets indoors are sources of biological contamination, One may attribute this finding to the small sample size, n = 5, and to owner awareness that a pet may be a source of indoor air pollution. The pet owners stated that they take preventive steps to avoid pet-induced contamination in their homes. Spatial and Temporal Analysis Andersen samplers generated all data used in this analysis. To investigate the difference between indoor and outdoor bacteria levels, a one-sided-paired t-test was used to compare indoor concentrations with outdoor concentrations. The null hypothesis of no difference was tested against the alternative of indoor population mean value greater than the outdoor population mean. The null hypothesis is rejected at every location except in the kitchen during the winter and the basement during the winter, spring, and fall seasons. Room-to-room variation was tested with ANOVA for sample means, and the outcome is season dependent. The null hypothesis cannot be rejected in the winter, summer, and winter II, and can be rejected in the spring and fall, indicating room-to-room variation. Post-hoc analyses indicate that the basement bacteria levels are significantly lower than all other indoor locations for the spring and fall. Seasonal variation was investigated using a two-sided paired t-test by location. The null hypothesis of no difference by season at each site cannot be rejected for roughly half of the paired tests, while for the remaining tests the hypothesis of no difference is rejected. Importantly, no clear patterns are observed. When comparing the first winter with the second consecutive winter, the null hypothesis cannot be rejected for the bathroom, kitchen, bedroom, and basement sites. However, a statistically significant difference is established between outdoor levels during winter I and winter II sampling. The results taken jointly substantiate the importance of indoor sources. Interaction between seasonal and spatial variations was investigated using a two-way ANOVA with replications. Each cell in the two-way ANOVA table represents the twenty measurements (one per residence) per indoor location per season. The five seasons were used as treatments with a hypothesis that there are no seasonal effects. The four indoor locations were used as the blocks to test the null hypothesis of no difference in levels among the four sampling sites. The null hypothesis of no interaction asserts that there are no interaction effects among seasons and indoor locations. All null hypotheses are rejected, confirming that there is seasonal and spatial (room-to-room) variation and interaction at the α = 0.05 level. The significant interaction implies that both factors are important and that the level of each factor must be known to characterize the effect of either factor on the response. Post-hoc analysis, a Bonferroni test, was performed to identify the source(s) of variation detected. Results of the Bonferroni test (Figure 5) indicate that outdoor bacteria levels are significantly different from corresponding levels at all indoor locations and that bathroom levels are statistically different from the basement. However, the basement, kitchen, and bedroom measurements are not significantly different. The Bonferroni test of seasons indicates that winter II is statistically different from the other four seasons measured. The results also indicate that spring, winter, and fall bacteria levels are not statistically different. Similarly, fall and summer bacteria concentrations are not statistically different. Figure 5. Bonferroni test results for variations in mean levels of total viable bacteria in (a) different rooms and (b) different seasons.

906 D. J. MOSCHANDREAS ET AL. CONCLUSIONS Conclusions reached by this study should be viewed with three qualifying factors: 1. The sample residences of this study were selected randomly from a self-selecting population that responded to a public announcement for volunteers interested in knowing the biological pollutant levels in their residences. 2. As indicated in the section on experimental design, the sample size is sufficient to detect the desired differences, yet any generalization is limited by the location of the sampling. The conclusions refer to climatic conditions in Chicago, Illinois, USA, and any generalization should be carefully applied only for similar conditions. 3. While we have measured Staphylococcus for a specific species of biological pollutants, several of the conclusions are generalized for all bacteria. We are cognizant of the fact that seasonal and spatial variation of biological pollutants should be performed on individual flora, and therefore this study should be considered as the first step in a series of additional studies that would investigate individual flora. These qualifying factors not withstanding, the results verified some conclusions reached by other researchers and posed questions that should motivate additional research. Indoor bacteria levels were found to be higher than outdoor levels for 75% of indoor locations. This supports conclusions reached by other studies and indicates the presence of indoor sources of total culturable airborne bacteria. On the average, Staphylococcus concentrations were about 27% of total culturable bacteria levels and up to a level of 37%. Measurements using side-by-side sampling methods correlated relatively well, but the levels measured were statistically different. Relative humidity, temperature, number of occupants, and in-house pets are suspected sources of indoor air bacteria. However, when these variables were examined, their association with total bacteria counts was minimal. These variables were measured and noted as found, and they were not controlled for in this study. A control room study varying each of these variables may show the suspected dependence of bacteria levels and the independent variables tested. One should also consider that the subject residences were conditioned and thus the relative humidity and temperature were nearly uniform while the bacteria levels varied considerably. No significant association was found between bacteria levels in water-damaged versus nonwater-damaged homes. This lack of statistical and indeed practical difference may be attributed to steps implemented by concerned home owners of waterdamaged residences and specifically to the extensive cleaning that was carried out to reduce the perceived high levels of indoor bacteria. Residential occupied spaces appear to have uniform levels because the difference of bacteria concentrations among occupied rooms was not statistically different, but the difference between basement levels and those of the room sampling sites is statistically significant. Frequently, but not always, those who attempt to characterize the concentration of biological pollutants in a residence measure once at one site assuming that bacteria levels in residences are uniform in space and time. Agreeing with a large segment of the research community, this study found that this assumption is not valid. Basement bacteria levels were significantly lower than other indoor locations during the spring and fall. This signifies that the rarely occupied basements are distinct indoor microenvironments of bacteria levels. Seasonal variation was also observed, but a clear pattern of variation could not be established. This points to the importance of indoor sources and implies corresponding variation of indoor source emissions. REFERENCES Andersen Samplers Inc. (1976). Operating Manual for Andersen Sampler Inc. Particle Sizing Sampler, Andersen Samplers Inc., Atlanta, GA. Biotest Diagnostics Corporation. (1980). Manual for Reuter Centrifugal Sampler, Danville, NJ. Burge, H. A. (1995). Bioaerosols. CRC Press, Inc., Boca Raton, FL. Cutter Information Corp. (1994). Biocontaminants in Indoor Environments. Cutter Information Corp., Arlington, MA. Flannigan, B., Sampson, R. A., and Miller, J. D. (Eds). (2001). Microorganisms in Home and Work Indoor Environments. Taylor and Francis, London. Jensen, P. A., Todd, W. F., Davis, G. N., and Scarpino, P. V. (1992). Evaluation of Eight Bioaerosol Samplers Challenged with Aerosols Free Bacteria, American Industrial Hygiene Association Journal 53(10):660 667. Griffiths, W. D., and Stewart, I. W. (1999). Performance of Bioaerosol Samplers Used by the UK Biotechnology Industry, J. Aerosol Sci. 30(8):1029 1040. Henningson, E. W., Lundquist, M., Larsson, E., Sandstrom, G., and Forsman, M. (1997). A Comparative Study of Different Methods to Determine the Total Number and the Survival Ratio of Bacteria in Aerobiological Samples, J. Aerosol Sci. 28(3):459 469. Hernandez, M., Miller, S. L., Landfear, D. W., and Macher, J. M. (1999). A Combined Flourochrome Method for Quantitation of Metabolically Active and Inactive Airborne Bacteria, Aerosol Sci. Technol. 30:145 160. Moschandreas, D. J., Cha, D. K., and Quian, J. (1996). Measurement of Indoor Bioaerosol Levels by a Direct Counting Method, J. Environmental Engineering 122(5):374 378. Ness, S. A. (1991). Air Monitoring of Toxic Exposures. Van Nostrand Reinhold, New York. Riley, R. L. (1979). Indoor Spread of Respiratory Infection by Recirculation of Air, Bulletin of Physiopathological Respiration 15:699 705. Terzieva, S., Donnelly, J., Ulevicius, V., Grinshpun, S. A., Willeke, K., Stelma, G. N., and Brenner, K. P. (1996). Comparison of Methods for Detection and Enumeration of Airborne Microorganisms Collected by Liquid Impingement, Appl. Environ. Microbiol. 62(7):2264 2272. Verhoeff, A. (Ed.). (1993). Indoor Air Quality and Its Impact on Man. Report No. 12. Office for Official Publications of the European Communities, Luxembourg. Wagner, S. F. (1969). Introduction to Statistics. Harper Perennial, Harper Collins Publishers, Inc., New York.