What is the relationship between indoor air quality parameters and airborne microorganisms in hospital environments? A systematic review and meta- analysis

Airborne microorganisms in hospitals have been associated with several hospital-acquired infections (HAIs), and various measures of indoor air quality (IAQ) parameters such as temperature, relative humidity, carbon dioxide (CO 2 ), particle mass concentration, and particle size have been linked to pathogen survival or mitigation of pathogen spread. To investigate whether there are quantitative relationships between the concentration of airborne microorganisms and the IAQ in the hospital environment. Web of Science, Scopus and PubMed databases were searched for studies reporting airborne microbial levels and any IAQ parameter(s) in hospital environments, from database inception to October 2020. Pooled effect estimates were determined via random- effects models. Seventeen of 654 studies were eligible for the meta-analysis. The concentration of airborne microbial measured as aerobic colony count (ACC) was significantly correlated with temperature ( r = 0.25 [95% CI = 0.06– 0.42], p = 0.01), CO 2 concentration ( r = 0.53 [95% CI = 0.40– 0.64], p ˂ 0.001),


| INTRODUC TI ON
Hospital-acquired infections (HAIs) are a globally significant problem and their treatment can be costly. [1][2][3] In the UK, HAIs are estimated to cost up to a billion pounds per year as of 2017 and the hospital environment is thought to play a role in approximately 20% of all HAIs by influencing the survival and spread of pathogens in the environment. 4,5 The hospital environment is subject to workplace design and layout, operation and maintenance, and hosts multiple interactions between environment and people. Studies investigating microbial contamination of the environment have suggested that a wide range of factors may influence the presence of microorganisms including IAQ parameters such as temperature, relative humidity, and ventilation; staff activities, patient status, and visitor numbers; and surface types, including how and when they are cleaned. 1,3,6-12 A very small number of studies correlated virus concentrations to these factors, hence the focus of this study on the investigation of relationships between bacteria and fungi in the air and IAQ parameters. Surfaces, air, and indoor structure including ventilation systems have all been shown to act as reservoirs for pathogens, and in some cases, these pathogens can persist for months in a hospital environment. [13][14][15] Previous studies have used the information from environmental sampling to link bioburden levels in the air, on surfaces bioburden and HAI rate. [16][17][18][19][20] Microbial sampling of the air can be used to evaluate the likely concentration of airborne microorganisms present in the hospital environment. The majority of studies apply culture-based methods to assess viable microorganisms. Airborne microbial load can be quantified by using either active or passive sampling methods. 8,16 IAQ parameters such as temperature, relative humidity, CO 2 level (which reflects the ventilation rate), particle mass concentration, and particle size are important for the health and well-being of those in hospitals and may also influence the bioburden in the environment.
Ambient air temperature and relative humidity are usually measured in indoor environments to understand the thermal comfort and well-being of occupants. However, both parameters are also linked with the survival of microorganisms, with humidity a particular concern. Many bacteria and fungi favor more humid conditions. 21,22 However, there is evidence that virus survival increases at humidity below 40% RH. 23 Guidance varies around the world, but temperatures within 16-25°C and humidity in the range 40%-60% RH are commonly recommended. 24 CO 2 is related to the exhaled breath of occupants and is frequently measured in indoor environments as an indicator of ventilation rates. A number of studies have also shown that ventilation rates expressed through CO 2 concentrations can be used to evaluate airborne infection risk. 25 Airborne particles provide a general measure of indoor air quality (IAQ) and can be related to indoor sources and activity or outdoor conditions. 8,26 Some studies suggest using airborne particles as a proxy for cleanliness of the air, including to commissioning of specialized hospital ventilation systems. 27 The directed acyclic graph (DAG) approach is a good way to investigate causality with variables with respect to confounding.
Although the correlation between IAQ parameters and microorganism prevalence and survival has been studied for decades, there are conflicting results, 2,8,28-32 and it is not clear which parameters may be significant and how they interact together. If there are significant and consistent relationships between the microbial load in the air and IAQ parameters, this could allow IAQ to be used as a proxy for evaluating the likelihood of microorganisms being present in the air.
The aim of this study is to carry out a systematic review and meta-analysis to investigate the relationships between the level of airborne microorganisms and IAQ parameters in a hospital environment. By bringing together data from multiple studies, the paper aims to formally assess the strength of relationships between parameters and to determine where there are gaps in data that could inform future experimental studies in healthcare settings. This study can also inform new predictive models that provide an improved method for monitoring the concentration of airborne microorganisms in real time through measurement of IAQ parameters. hospital environment. These data would inform models to improve the understanding of the likely concentration of airborne microorganisms and provide an alternative approach for real-time monitoring of the healthcare environment.

K E Y W O R D S
airborne microorganisms, hospital environment, hospital-acquired infection, indoor air quality, meta-analysis

Practical Implications
• This study encourages the introduction of new predictive models that provide a better method for monitoring the concentration of airborne microorganisms in real time from knowing the IAQ parameters.
• Controlling IAQ parameters could lead to reducing airborne bioburden which might reduce the infection risk from airborne microorganisms.
• This study provides a basis for designing further studies with improved data reporting and makes it easier for such studies to maximize the outcome and ensure a more unified approach.

| Search and inclusion criteria
A systematic review was performed to identify relevant studies. For the identification phase, three electronic databases (Web of Science, Scopus and PubMed) were searched systematically from inception to October 2020 using keywords "air, sampling, hospital, environment, AND contamination." Full-text articles published in English that include air sampling data for microorganisms and IAQ parameters in patient areas of hospitals were selected for inclusion. The reference lists of all selected studies were screened to identify other likely eligible studies. We excluded papers conducting air sampling in other types of healthcare buildings (eg, G.P. surgery, clinic), in hospital rooms with specialist ventilation ≥10 air changes per hour (e.g., isolation rooms, operating theaters) or in areas undergoing construction or renovation. Studies with relevant data were included for the meta-analysis ( Figure 1); studies had to present quantitative data on the airborne microbial concentration measured as aerobic colony count (ACC) or airborne total fungi (TF) with at least one IAQ factor: temperature, relative humidity, CO 2 , particle mass concentration (≤5 or >5 µg/m 3 ), or particulate matter of size (≤5 or >5 µm) measured at the same time point.
The DAGitty and statistical software R 4.0.0 (package "ggdag" version 0.2.3) were used to build a DAG ( Figure 2) to describe how potential confounders and the air quality parameters relate to microbial measures. 33

| Data extraction and quality appraisal
All corresponding authors for included studies were contacted for raw data where the data available within the paper were not sufficient to conduct analysis. Correlation coefficient and sample size F I G U R E 1 Flowchart of the systematic review and meta-analysis phases, search strategy, and exclusion criteria. ACC, airborne microbial concentration measured as aerobic colony count and TF, airborne total fungi were extracted directly from the study, derived from graphed points, obtained from tabulated values, or calculated from raw data, which were provided by the corresponding author via private correspondence. Equation (1) was used to compute the correlation coefficient from multiple regression and the general linear model for taking covariates into account. 34 where df is the degrees of freedom used for a corresponding t value in a linear model. Outliers and influential observations are very likely to weaken the validity and robustness of the conclusions from a metaanalysis. 35 Sensitivity analysis of the meta-analyses to detect potentially outlying studies was performed using visual approaches including (1) externally standardized residuals, (2) difference in fits (DFFITS) values, (3) Cook's distances, (4) covariance ratios, (5) leave-one-out estimates of the amount of heterogeneity, (6) leave-one-out values of the test statistics for heterogeneity, (7) hat values, and (8) weights. 36 If observations were beyond the lower and upper limit of DFFITS, they were excluded from the meta-analysis, as their inclusion could lead to notable changes in the pooled (overall) estimate effect size of metaanalysis. To test heterogeneity between studies, the Q statistic was used to examine the null hypothesis that all studies had the same true effect: τ2 = 0. 37 The 95% CI around the I 2 statistic was also calculated to determine the level of heterogeneity present.
The meta-analysis was based on a Fisher Z transformation of the correlation coefficient to obtain weightings for each study. Fishertransformed correlations are always less biased than when untransformed correlations are used. 38 A random-effect meta-analysis model is used since the studies came from different populations and included design-related heterogeneity. Random-effects models are more appropriate since the aim is to generalize beyond the studies included in the meta-analysis. 39,40 Forest plots were used to visualize the overall estimates of the study effects with corresponding confidence intervals. 41 This systematic review and meta-analysis was performed according to the Preferred Reporting Items for the Systematic Reviews and Meta-Analysis (PRISMA) guidance. 42,43 The statistical software R 4.0.0 (package "meta" version 4.12-0 and package "metacor" version 1.0-2.1) was used to perform the metaanalysis (Appendix S1).

| RE SULTS
A total of 1173 studies were retrieved and 654 studies screened after duplicates were removed. After screening through titles and abstracts, 197 studies remained for full text assessed for eligibility. Seventeen studies were included in the final meta-analysis ( Figure 1). These presented quantitative airborne microbial concentration measured as ACC or airborne TF concentration (TF) with at least one quantitative factor of the IAQ parameters at the same time point in a hospital setting and the correlation coefficient values and sample size for the relationships are given for each study (Table 1). 1,2,7,8,10-12,28-32,4 4-48 The forest plots prepared were for the Fisher Z-transformed correlation which was used to test the hypotheses about the value of the correlation coefficient. In order to interpret the results, the transformed values of pooled correlations were converted back to the original metric in the text. The studies were checked for the presence of outliers and influential observations that might bias the results, but none was detected. The heterogeneity was not statistically significant, and I 2 was very low between most studies. The correlations between ACC or TF and IAQ are as shown below. A DAG approach was used to identify possible confounders within the data structure and which variables need to be included ( Figure 2). 49 Uncorrelated measurement error cannot be elucidated from the articles, so we assume similar bands of error and thus do not include it in the statistical analysis.

| Correlation between airborne microorganisms and ambient air temperature
Six studies provided quantitative data to assess the relationships between the concentration of airborne microorganisms and temperature within the hospital environment. Temperatures recorded within the studies ranged from 17.4°C to 27°C, for measured microbial

| Correlation between airborne microorganisms and airborne particles
Three eligible studies considered the correlation with particle mass concentration, with values reported only for ACC and not TF. Across these studies, the sample size ranged from 11 to 70, with a total of 141 measurements of airborne microorganism concentration (ACC ranged from 378 to 3000 cfu/m 3 , particle mass ≤5 µg [5-61 µg/

| Correlation between ACC and TF level in the air
The final analysis considered the correlation between ACC and TF, and this was measured by more studies. The sample size ranged from 4 to 96 with a total of 305 values across the ten studies. 7,10,11,28,32,44,45,47,48,50 The pooled estimated was moderately pos-

| DISCUSS ION
To our knowledge, this is the first systematic review and metaanalysis to quantitatively examine the relationships between microbes in air and IAQ parameters in hospital environments. Although the importance of ensuring good IAQ to minimize airborne microorganism transmission is recognized, 51 we found that there are a very small number of studies that carry out sufficient quantitative measurement to reliably assess relationships between airborne microorganisms and environmental parameters. The majority of studies considered bacteria and/or fungi, and no studies had sufficient data to assess correlations between virus in air and the IAQ parameters in a hospital setting.

| Sampling approaches
There are two main approaches used in air sampling, active sampling and passive sampling. Most of the studies in the literature use active sampling (see  The summary of results in Table 2 shows that most of the studies present their findings based on a snapshot air sampling rather than intensively performing multiple samples over a long time. This leads to misleading conclusions as the results are too few to reflect the accurate correlation.

| Temperature and Humidity
The meta-analysis suggests there is a significant positive relationship between airborne bacteria concentration and temperature, while there was no statistically significant relationship between airborne microbial concentration and relative humidity. For the airborne fungi concentration, the correlation with both temperature and relative humidity was not found to be significant. dations are given for humidity, and it is rare that humidity is controlled. 27 In the United States, ASHRAE recommends 21-24°C in patient rooms and also does not specify humidity control and however in clinical areas, they typically recommend 30%-60% RH. 56 Recommendations for patient rooms in Japan vary by season with temperature (24-27°C) and humidity (50%-60% RH) recommended for summer compared to winter (20-24°C, 40%-50% RH). 55 The lack of clear correlation between microbial load in the air and the tem-

| Ventilation rate
The moderate and positive significant relationship between airborne bacteria concentration and CO 2 using a pooled estimate (2) the room occupancy which will contribute to bacterial generation through respiratory sources, natural skin shedding, and activities such as bed making that may resuspend microorganisms. 7,10,57 Conversely, there was no relationship between TF concentration and CO 2 level. This result can be interpreted according to previous work that found people shed half the number of bacteria as fungi. 58 It is also likely that in many settings, TF is influenced by the fungi in outdoor air and hence would only be influenced by ventilation is there is effective filtration in place. 26 Studies have shown that the level of CO 2 level has a positive correlation with occupied rooms, room temperature, and relative humidity. 7,10,59 Although it is possible to estimate ventilation rates using exhaled CO 2 levels as a proxy, measuring the ventilation rate is not straightforward. Recommended ventilation rates in hospital wards vary worldwide and depend on the climate and ventilation approach. In the United States, ASHRAE recommends 6 ACH and however only 2 ACH is required to be fresh F I G U R E 6 Forest plot for the studies reporting relationships between the particle mass concentration and airborne microbial concentration using Fisher's transformed correlation. (A) Correlation with particle mass concentration ≤5 µg/ m 3 . (B) Correlation with particle mass concentration ˃5 µg/m 3

F I G U R E 7
Forest plot for the studies reporting relationships between particulate matter size and airborne microbial concentration using Fisher's transformed correlation. (A) Correlation with particulate matter of size ≤5 µm. (B) Correlation with particulate matter of size >5 µm F I G U R E 8 Forest plot for the studies reporting relationships between the airborne microbial concentration and airborne total fungi concentration using Fisher's transformed correlation air and the remaining 4 ACH can be recirculated with appropriate filtration. UK hospitals recommend 6 ACH full fresh air, but do permit natural ventilation which will be variable. 27 Ventilation rates in many hospitals do not necessarily meet these standards and reflect the standards at the time of construction and the maintenance of the ventilation systems.

| Particulates
Airborne particulate matter may be indicative of the transport and deposition of a microorganism in air, and where microorganisms are released alongside other particle generating activities, it is important to understand whether particle measurement is a useful proxy for microorganisms. It is evident that particulate matter of size <5 µm is likely to be of greatest importance as they fall within the size range of bioaerosols that can remain airborne for long periods of time (between 100 and 1000 s). 60 This study result shows that there is significantly moderately positively correlation between airborne microorganisms, particle mass concentration (≤5) µg/m 3 , and diameter particle concentration (≤5 and ˃5 µm) particle/m 3 , while not significantly correlated with particle mass concentration of >5 µg. It is hard to determine whether these relationships between microorganisms in the air and particles are directly or indirectly a result of the hospital environment. Previous studies illustrate that increased activity in hospital wards (eg, patient bathing or wound toilet behind closed curtains) is correlated with increased concentrations of bioaerosols and particles; wards are generally full of patients, healthcare workers, and visitors leading to contamination and re-contamination of the environment. 8,13 Additionally, human occupancy has a strong link with indoor particle mass concentration. 61

| CON CLUS ION
We have systematically reviewed studies that sampled airborne microorganisms in hospital wards and presented quantitative data with one or more IAQ parameters (temperature, relative humidity, CO 2 , particle mass concentration, and particulate matter of size).
We found that there are only a small number of studies that provide quantitative data to assess relationships between airborne microorganisms and IAQ parameters from measurements made in hospitals outside of settings with specialist ventilation (eg, operating rooms). Overall, we can conclude the following from the meta-analysis: 1. There are likely to be positive correlations between airborne bacteria and other types of microorganism, particularly fungi.
2. There are positive correlations between airborne bacteria and fungi, measured as ACC, and several IAQ parameters (temperature, CO 2 , particulate matter of size of ≤5 and ˃5 µm, and particle mass concentration ≤5 µg/m 3 ). However, the data did not demonstrate a clear correlation with relative humidity, and correlations between TF and IAQ parameters were weak.
3. There are only a very small number of studies that present quantitative data while measuring the environmental and activity factors that affect the presence and quantity of airborne microorganisms.
Our conclusions lead to the following recommendations: 1. There is a need for more detailed sampling studies including air sampling (active and passive), measurement of air quality parameters, and observation of level of healthcare worker activity to understand the spatial and temporal fluctuation in microbial bioburden in hospitals.
2. Reporting of data should be quantitative as far as possible to enable comparison between studies and future meta-analysis.
It is difficult to compare studies that present microbial samples in terms of percentage positive or in a semi-quantitative way.
Studies that carry out statistical analysis should provide the correlation coefficient and the sample size.
3. Instead of referring to the season or geographic location of the study, seasonal factors need to be reported quantitatively in terms of temperature and relative humidity to ensure they are consistent and comparable between different locations around the world.
4. It is important that data are reported at the time that each sample is taken rather than as an average for the whole study. Studies that simply present IAQ or ventilation parameters as a mean and standard deviation across all the samples do not provide sufficient data for further analysis.

ACK N OWLED G EM ENTS
We are very grateful to the EPSRC for funding this project. The lead author Hiwar is funded through an EPSRC DTP studentship project reference: 1955605. Noakes, King, Dancer, and Fletcher are funded through the EPSRC HECOIRA project EP/P023312/1.

CO N FLI C T O F I NTE R E S T
None to declare.

PEER R E V I E W
The peer review history for this article is available at https://publo