The transmission of airborne disease is essentially a stochastic process, strongly influenced by chance events and system variance. It is therefore important to take this into account when modelling the airborne transmission of infection. The Gammaitoni-Nucci and Wells-Riley equations have frequently been applied in a deterministic manner [5, 6, 8]. Although this approach has merit, with a deterministic methodology it is only possible to determine the mean probability of a susceptible person becoming infected - it is not possible to predict those chance 'high risk' events that are most likely to result in the transmission of infection. Given that it is the 'high risk' events, rather than the mean probability that are critical, we adopted a stochastic methodology in our study to overcome some of the shortcomings associated with the deterministic approach.
From Table 3 it can be seen that as the quanta production rate rises, so the standard deviation of the results greatly increases. Consequently, the number of 'high risk' events associated with infections such as measles, are much greater than, say, for TB. Indeed, our analysis suggests that under normal conditions (i.e. without other complicating factors which might increase the production of droplet nuclei [5]) the chances of contracting tuberculosis in the waiting room under the default conditions are minimal. By comparison, if an infectious individual with measles is present in the waiting room, then under the same default conditions, the risk of transmission will be >10% on 59% of occasions, with the result that transmission of this disease is very likely to occur. This is reflected in the differences in observed attack rates for viral infections such as measles [6] and influenza [18], compared with those for open pulmonary TB. Indeed, our finding reinforces current guidance on TB prevention in the United Kingdom which recommends that periods of exposure to an infectious person in a confined space of less than eight hours should not be considered a significant risk [19].
Beggs et al. [5] demonstrated that length of exposure time plays a critical role in the transmission of TB. In most situations (i.e. situations where artificial aerosols are not generated) TB is not easily transmitted and long periods of exposure are generally required in order to contract an infection [2, 5]. In the 1950s, Riley et al. undertook a study in a well-ventilated TB ward in a Baltimore, MD Hospital, from which they estimated that 'one unit of infection' was suspended in every 15 000-20 000 ft3 (500-667 m3) of air in the ward [2]. From this they concluded that airborne M. tuberculosis bacilli were "not very numerous even in the vicinity of patients", and that it would take the "better part of a year" for a nurse working on a TB ward to breathe in 500-667 m3 of air. The results of our study concur with this opinion and suggest that, although sporadic transmission may occur, the duration spent by most patients in waiting is too short for the risk of contracting TB to be anything other than minimal. Having said this, it should be noted that our study assumed: (i) the presence of only one infectious individual and (ii) a waiting area typical of that found in a UK hospital (i.e. not overcrowded). In many developing countries these assumptions may not be valid. In such countries hospital waiting areas are often very overcrowded, waiting times may be long (i.e. several hours), and more than one infectious individual might be present. Under such circumstances the risk of TB transmission is likely to be greater.
One factor that can vary greatly from day-to-day is the number of susceptible patients present in a waiting room at any given time. This, fact (often overlooked) is important because it has a profound effect on the number of people who are likely to contract an infection. If for example, the personal risk of contracting a given infection is constant, then equation 6 indicates that the likelihood of a person within the waiting area becoming infected is ten times greater if there are 20 susceptible people in the room, rather than just two. Although, the personal risk of contracting an infection is independent of the occupancy density (i.e. the probability of transmission is the same for each person in the waiting room), the fact that more people are present means that the chance that someone will become infected is much greater. In our model we allowed for this by varying the number of susceptible individuals present in each simulation. Figure 4 presents results of analysis undertaken for TB and influenza under the default conditions, assuming that the mean number of susceptible individuals varies. From this it can be seen, that for both infections, as the number of patients present increases, so the number of new infections occurring also increases, despite the fact that the personal risk to each individual in the space is nominally constant. Furthermore, it can be seen from Figure 4 that the standard deviation increases as the number of susceptible individuals present increases. Consequently, 'high risk' transmission events are more likely to occur when many individuals are present.
From the analysis above it becomes clear that the increased number of infection cases experienced in overcrowded spaces is simply a reflection of the fact that there are more susceptible individuals present to become infected. However, while this is undoubtedly true in part, in reality the whole picture is more complex, as proximity to an index case also influences personal risk. For example, if a person walks into a room containing one individual with open pulmonary tuberculosis, according to the Gammaitoni-Nucci model their personal risk would be the same whether there were six uninfected people in the room or 60. If, however, as a result of having 60 people in the room space an individual was forced into close proximity with the index case, then their risk of contracting an infection would probably increase. However, because of the logistical and computational difficulties associated with simulating proximity, it was not possible in our study to assess this issue.
In order to prevent the transmission of airborne infection many researchers have advocated the use of improved building ventilation [8, 12, 20] or air disinfection devices [21–23]. While such strategies undoubtedly have potential in applications where exposure times are long, their use in the waiting areas of healthcare facilities appears to be much more questionable. Despite this ultraviolet lamps were reported to be installed in 12/144 (8%) and HEPA (high efficiency particulate air) filtered air provided in 22/138 (16%) of Emergency Department waiting areas in facilities which recorded >1 case of TB per month [24]. From Figure 3 it can be seen that little impact is made on the transmission of TB and influenza by increasing the ventilation rate to 12 air changes per hour. This is because the exposure times are generally too short for these infections to have much impact. Any benefit derived from increased ventilation is significantly outweighed by factors such as the exposure time and the numbers of susceptible individuals present. Even with high quanta producing diseases, such as measles, where improved ventilation might be beneficial, reducing: (i) waiting times; and (ii) the number of susceptible individuals present, appear to be as important as installing expensive ventilation/air cleaning equipment. It is therefore important when seeking to prevent the transmission of airborne viral disease to first minimize waiting times and the number of susceptible individuals present before turning to expensive technological solutions.
While the Gammaitoni-Nucci model calculates the risk that an airborne disease might be transmitted in a confined space, it is important to remember that the results it produces rely wholly on the quality of the data used. With respect to this, the room volume, room ventilation rate, average pulmonary ventilation rate, and average occupancy time, are all variables that can be estimated with some degree of accuracy. However by comparison, the quanta production rate is much harder to estimate. This is because there is a paucity of good quality data regarding quanta production rates. Values are generally calculated retrospectively, using either the Gammaitoni-Nucci or Wells-Riley equations after an outbreak has occurred. Furthermore, published quanta values can vary greatly for the same disease, making comparisons difficult. For example, while Riley et al. [4] estimated that the average TB patient in hospital produced only 1.25 quanta per hour, Nardell et al. [8], investigating a TB outbreak in a Massachusetts office building, calculated the infectious dose to be 12.7 quanta per hour. Given the variability in the published quanta production data, the results generated by the Gammaitoni-Nucci model should only be considered as indicative of trends rather than absolute values. Furthermore, because infectious individuals are likely to generate differing quanta production rates, it is important when modelling the risk of transmission to use a range of values, rather than use a single fixed quanta production rate.