Some Quantitative Relations between Indoor Environmental Quality and Work Performance or Health

Poor indoor environmental quality (IEQ) has been related to increases in sick building syndrome symptoms, respiratory illnesses, sick leave, and losses in productivity. Calculations indicate that the cost of poor IEQ can be higher than energy costs space conditioning and ventilation, and that many measures taken to improve indoor IEQ will be highly cost-effective when accounting for the monetary savings resulting from an improved health or productivity. To enable building professionals to make selections of building designs and operating practices that account for effects on health and productivity, we need models for quantifying the health and productivity benefits of better indoor environments. Therefore, we have reviewed the literature on the effects of indoor environment on health and performance and used existing data, when possible, to develop some initial models. Based on the best-available evidence we present quantitative relationships between ventilation rate and short term sick leave, ventilation rate and work performance, perceived air quality and performance, temperature and performance, and temperature and sick building syndrome symptoms. We show also that a relationship exists between SBS symptoms (sick building syndrome symptoms) and work performance. INDEX TERMS Ventilation, Temperature, Performance, Modeling, Perceived air quality, SBS symptoms, Cost benefit calculations INTRODUCTION There is increasing evidence that indoor environmental conditions substantially influence health and performance. Building professionals are interested in improving indoor environments and quantifying the effects. Macro-economic estimates of nationwide financial gains have been developed. They show that the potential benefits from indoor environmental improvements for the society are high (Fisk 2000, 2001). Some calculations show that the estimated cost of poor indoor environment is higher than energy costs of heating and ventilation of the same buildings (Seppänen 1999). A few sample calculations have also shown that many measures to improve indoor air environment are cost-effective when the health and productivity benefits resulting from an improved indoor climate are included into the calculations (Djukanovic et al. 2002, Fisk 2000, Fisk et al. 2003, Hansen 1997, Seppänen and Vuolle 2000, Tuomainen et al. 2002, Wargocki, 2003). There is an obvious need to develop tools and models so that economic outcomes of health and performance can be integrated in cost benefit calculations with initial, energy and maintenance costs. The use of such models would be expected to lead to improved indoor environments, health and productivity. To systemize these building level calculations we have earlier developed a conceptual model (Seppänen and Fisk 2003) to estimate the cost-effectiveness of retrofits of indoor environment. In this paper we derive and present estimates of some quantitative linkages in the model for cost benefit calculations namely between ventilation rate and sick leave, ventilation rate and performance, perceived air quality and performance, temperature and performance, and temperature and SBS symptoms (sick building syndrome symptoms). We also suggest that a link between SBS symptoms and performance exists, and that a linkage from building factors to SBS symptoms and further to performance and health outcomes will be an attractive way to evaluate the financial benefits of indoor environmental improvements. VENTILATION RATES AND SHORT TERM SICK LEAVE Ventilation reduces the indoor concentration of indoor-generated airborne pollutants. The effects of ventilation rates on human responses has been summarized by Seppänen at al. (1999), Fisk (2000) and Wargocki et al. (2002). These summaries show that the prevalence of some types of communicable respiratory diseases is higher under conditions with lower ventilation rates. In our earlier paper (Fisk et al. 2003) a quantitative relationship between ventilation rate and sick leave was estimated combining published field data and a theoretical model of airborne transmission of respiratory infections. The model (Figure 1) accounts for the effects of ventilation, filtration, and particle deposition on airborne concentrations of infectious particles and for the feedback process by which more disease transmission in a building leads to more sick occupants who are sources of infectious particles. The theoretical model is calibrated, i.e., fit to several sets of empirical data, resulting in different curves relating ventilation rates with illness prevalence. Figure 1. Predicted trends in illness of sick leave versus ventilation rate (from Fisk et al. 2003) The relationships of sick leave or absence with air change rate that are depicted in Figure 1 are only applicable for the levels of occupant density encountered in the studies. To illustrate how the illness or absence rate is predicted to vary with ventilation rate per person in an office building, Figure 2 provides a re-plot of two of the curves in Figure 1, assuming an occupant density of 2900 ft (83 m) per person, which was derived using data from a survey of 100 U.S. office buildings (Burton et al. 2000). From the data of Milton et al. (2000), one can derive a baseline short-term sick leave rate of 2% for an office building with a ventilation rate of 12 L s per person , enabling a calculation of the annual average sick leave rate, for higher or lower ventilation rates. Applying the curve in Figure 2 based on the particle concentration model, which corresponds to a mid-range among the results depicted in Figure 1, , one can estimate that doubling the average ventilation rate to 24 L s per person , would decrease the sick leave prevalence in an office from 2% (5 days per year) to 1.5% (3.8 days per year). Figure 2. Predicted trends in illness or sick leave versus ventilation rate per person. There are many sources of uncertainty in the model used to relate ventilation rates to sick leave. Most important is the limited empirical data available to calibrate and evaluate the model. In addition, there are uncertainties in the size, filtration rate, and deposition rate of infectious particles in typical buildings. Also, the natural loss of viability of airborne infectious particles has not been accounted for in the model due to a lack of information on the survival times of the airborne virus and bacteria that cause respiratory diseases. If suitable information were available, viability loss could be incorporated in the model as filtration and depositional losses were incorporated. The rate at which an infector disseminates infectious particles will likely vary among illnesses. The susceptibility to infection will vary with the age, health status, and immunizations of the occupants of the building. It is likely that these and other factors, including different amounts of time spent in different types of buildings, partially explain the different curves shown in Figure 1. Despite these large sources of uncertainty, a rough accounting of the influence of ventilation rates on sick leave may lead to better decisions about building design and operation than totally neglecting this issue. VENTILATION RATES AND PERFORMANCE Ventilation affects productivity indirectly through its impact on short-term sick leave due to infectious diseases, but also directly. To establish the relation between ventilation rate and performance we identified five relevant workplace studies (Heschong group 2003, Federspiel et al. 2004, Tham 2004, Tham and Willem 2004, Wargocki et al. 2004), and two studies with data collected in controlled laboratory environment (Bako-Biro 2004, Wargocki et al. 2000a). All workplace studies were performed in call centres where the time required to talk with customers and the processing time between calls with customers, and other relevant information was automatically recorded in computer files. In these studies, the speed of work, i.e., time per call, was used as a measure of work performance. Laboratory studies assessed work performance by having subjects perform one or more computer-administered tasks that simulate aspects of actual work and by subsequent evaluation of the speed and/or accuracy of task performance. We also included a study made in schools using Swedish performance evaluation system with reaction times (Myhrvold and Olesen 1997). We used adjusted results when possible and unadjusted results when the authors made no adjustments. Some of the studies have compared only two ventilation rates some provide the data comparing several ventilation rates. We included in the summary all reported data points regardless of the level of the statistical significance, which actually was not reported in all studies We normalised the data from the studies by calculating the change in performance per increase of 10 L/s-person in ventilation rate. Thus, the relative performance increase was calculated by subtracting the performance with the lower ventilation rate from the performance with the higher ventilation rate and dividing the difference, by performance by lower ventilation rate. This relative change in performance was further divided by the difference between the two ventilation rates in L/s-person, and multiplied by 10 L/s-person, and converted to percentages. The number represents thus the change in performance of a specific task per increase in ventilation rate of 10 L/s-person. The included studies also varied greatly in sample size and method. In the regression we weighted the studies by adjusted number of subjects. We also applied a weighting factor based on the authors’ judgement of the relative relevance of the performance outcome to real work. We used the following relative weighting factors: overall work performance (1), single tasks (0.5) and reaction time (0.25). The sample size weight and outcome relevance weight were then combined to get a final set of weights (Seppänen et al. 2006). Normalized Adjusted Change in Productivity (%) vs. ventilation rate, unweighted, weighted by sample size, and weighted by combined final weight are plotted in Figure 3. The very large (21.9%) improvement in performance reported by Tham (2004) at a ventilation rate of 10 L/s-person compared to 5 L/s-person (when the temperature was 24.5 C) was a clear outlier among the data and was excluded from the final analysis. Figure 3 shows also the 90% confidence limits for the model with composite weights. Figure 3. change in performance per 10 L/s-person increase in ventilation rate versus average ventilation rate in the experiment (data points), and regression models (curves). One outlier data point (43.8% at 7.5 L/s-person) is excluded. Dashed line: no weighting factors; broken line: data points weighted by sample size as described in text; solid line: data points weighted by sample size and relevance of out come (composite weighted) as described in the text. The shaded area represents 95% confidence interval and the dashed-dot line represents the 90% interval for the curve for with composite weights (solid line). Figure 3 shows that the trend of increasing performance with increased ventilation rate is statistically significant ventilation rates up to approximately 16 L/s-person with 90% CI and up to 14 L/s-person with 95% CI. In practise the equipment and energy cost also limit the ventilation rates. Based on the estimated polynomial models, the performance at all ventilation rates relative to the performance at a reference ventilation rates of 6.5 L/s-person and 10 L/s-person were calculated and plotted in Figure 4.