While theoretical studies such as the one described in this paper can only ever be approximations of what happens in clinical practice, they can, nevertheless, yield important insights into the factors that influence the transmission of infection in hospitals. In particular, they can be useful when assessing the relative impact of various infection control measures. Provided that the data used, to a large extent, mirror the situation in a the clinical environment,, it is possible to identify general trends in the transmission dynamics. With respect to this, while the analysis presented above confirms the long held opinion that hand hygiene is an effective control measure; it also shows that the law of diminishing returns applies and that the greatest benefits are derived from the first 20% or so of compliance. Indeed, the shape of the prevalence curves presented in Figure 1 (which have the same form as those produced by Cooper et al [4]) suggests that little benefit is accrued from very high levels of hand cleansing. Above a certain threshold, which will vary depending on input data, the benefit of increased hand hygiene compliance appears to be minimal. In the case of the study reported here, the data show that, even under conditions of very high transmissibility, if an alcohol solution is used, it should be possible to ensure R
0 < 1 when compliance is in the region 55%. This appears to confirm the findings of other researchers. For example, Cooper et al [4] found that under conditions of relatively high transmissibility (i.e. p' = p = 0.13) it was possible to ensure R
0 < 1 with a hand cleansing frequency < 30%. McBryde et al [5], using a stochastic transmission model of an ICU, found that 48% hand hygiene compliance was required to ensure R
0 < 1. Investigating the transmission of vancomycin-resistant enterococci on an ICU, Austin et al [6] found that a hand cleansing frequency of 50.5% achieved an effective reproductive number, R
e
= 0.69, well below unity (R
0 = 1 when f
h
= 27.9% – extrapolated from the data of Austin). Although, these researchers assumed in their respective models a hand hygiene efficacy of 100%, their results are similar to ours, suggesting that, despite the fact that, in practise, hand cleansing is an imperfect process, it should be possible to prevent many staphylococcal outbreaks from occurring without the need to achieve excessively high hand hygiene compliance. Having said this, it is important to remember that all these researchers (including ourselves) assume the transmission of pathogens to occur exclusively via the hands of HCWs, which is unlikely to be the case [1]. Pathogens can remain viable on inanimate surfaces for long periods of time [21] and if environmental contamination in any way contributes significantly to the transmission of staphylococcal infection, then the hand hygiene compliance levels stated above may not be adequate.
The results of our study indicate that the level of hand hygiene required to ensure R
0 < 1 is greatly influenced by the rate at which HCWs and patients make contact with each other, and the transmissibility of the contacts made – as these values increase, so the hand cleansing frequency required to prevent an outbreak also increases. Again, this confirms the findings of Cooper et al [4] who found transmissibility to be the single most influential variable in their study. In our study we assumed, as they did that, for each contact event, the probability of a HCW colonizing a patient is the same as the probability that the patient will contaminate the hands of a HCW. We did this to facilitate direct comparison with the work of Cooper et al [4]. This however, may not be the case, since contamination of HCWs hands is relatively transient (lasting only for a few hours), whereas patients tend to remain colonized for the length of their stay in hospital. Therefore, each colonized patient will contaminate many HCWs (R
p
>> 1), whereas a contaminated HCW will colonize patients only infrequently (R
h
<< 1) [6]. Accordingly, Austin et al [6] used values of p' = 0.40 and p = 0.06, which equates to a combined (p' × p) value of 0.024. Similarly, Grundmann et al [8] using the same methodology, adopted values of p' = 0.152 and p = 0.01, equating to a combined (p' × p) value of 0.015. By comparison, we and Cooper et al [4] used a combined (p' × p) value of 0.010, which, although lower than that used by the other researchers, is still of the same order of magnitude. The default HCW-patient contact rate used in our model was that suggested by Cooper et al [4] (i.e. 5 contacts per patient per HCW per day); considerably greater than the value of 1.38 contacts per patient per HCW per day reported by Austin et al [6] and somewhat less than the value of 7.6 contacts per patient per HCW per day used by Grundmann et al [8]. Collectively, this gives us confidence that our analysis is valid and that the variables used are realistic.
From the foregoing it can be seen that our results are consistent with earlier studies. It can therefore be concluded that it should be possible to prevent many outbreaks of staphylococcal infection through hand hygiene measures alone, even if high compliance rates are not achieved. In the study reported here it appears that compliance rates of 40% or so, should be adequate to prevent most outbreaks occurring. If this is indeed the case, then this raises questions as to why so many outbreaks of staphylococcal infection continue to occur, despite the fact that recorded hand hygiene compliance rates are generally in the region 40% [11–13]. While the reasons for this are unclear, there appear to be four possible explanations, details of which are as outlined below:
Hawthorne Effect
While recorded hand hygiene compliance is typically in the region 40% [11–13], it may be that the Hawthorne effect is at work and that observed hand hygiene does not reflect what actually happens in reality – the implication being that general hand washing rates might be considerably lower than 40%.
Ward Management
Another reason might be the way in which wards are organized and managed. For example, if a ward is overcrowded or under-staffed, then those nurses on duty will have to attend to more patients than usual and so the HCW-patient contact rate is likely to rise and with it the hand cleansing frequency required to ensure R
0 < 1. This was graphically illustrated by Grundmann et al [8], who, in their study on an ICU, found exposure to relative staff deficit to be the only factor significantly associated with MRSA transmission. Indeed, they predicted that it would require an additional 12% improvement in adherence to hand hygiene policies to compensate for staff shortages. Given that during this study, observed hand hygiene compliance was on average 59%, the investigators concluded that under conditions of overcrowding and high workload, it would be impossible for the nursing staff to achieve the required additional compliance. The HCW-patient contact rate is also influenced by the way in which nurses are organized. Beggs et al [7] demonstrated that if nursing staff are allowed to mix freely with patients, then the number of potential transmission routes will be high, leading to increased need for hand cleansing. In order to minimise the number of transmission routes it is necessary to cohort the nursing staff so that they cannot transfer pathogens between different groups of patients. Other studies have reached similar conclusions [5, 6, 12, 20, 22] – the higher the level of cohorting, the fewer the number of contacts between patients.
Colonized Admissions
Figure 5 shows the effect of variations in the proportion of admissions already colonized with MRSA on the prevalence of infection, assuming an average hand cleansing efficacy, λ', of 83%. From this it can be seen that as the number of colonized patients entering the ward increases, so the model predicts that it is not possible to completely eradicate infection through hand hygiene measures alone. No matter the level of hand hygiene compliance, there will always be a residual level of infection which is difficult to eradicate. Therefore, if the number of colonized patients admitted to hospitals is high, then this might explain why increased hand hygiene compliance is failing to control the spread of staphylococcal infection. From Figure 6 it can be seen that when the proportion of MRSA colonized patients entering hospital is 5%, the model predicts that R
o
> 1, no matter the level of hand hygiene compliance.
Environmental Contamination
Another reason why improved hand hygiene compliance might not deliver the hoped-for results might be because environmental contamination may be important in the transmission of staphylococcal infection. A number of researchers have demonstrated that widespread environmental contamination can occur as a result of MRSA infection/colonization [22–26]. For example, Boyce et al. [23] in a study in a US hospital, found environmental contamination in 73% of the rooms of MRSA infected patients and 69% of colonized patients. Indeed, they found 27% of the surfaces sampled in rooms containing MRSA-infected patients to be contaminated with MRSA, with frequently contaminated objects including the floor, bed linen, patients' gowns, over-bed tables and sphygmomanometer cuffs. Others have cultured MRSA from the air in patient rooms [22, 24, 25], and Wilson et al [26] observed a strong correlation between the presence of MRSA patients and air samples yielding MRSA in an ICU. Although many accounts of environmental contamination have been published it has proved very difficult to determine causality, and it is not known to what extent environmental contamination contributes towards the transmission of staphylococcal infection. However, it is thought that such contamination may seed environmental reservoirs resulting in increased sporadic infection [23]. If this is indeed the case, then it might explain why staphylococcal infection has been so difficult to eradicate using hand hygiene measures alone. Accordingly we would recommend that environmental contamination with the bacterium is considered in future models of S. aureus transmission within the hospital setting.
In this paper we used a deterministic model to analyse the transmission dynamics of staphylococcal infection. While this approach has validity, it is not without drawbacks. Deterministic dynamic models predict that the persistence of infection is only possible above a certain critical threshold, R
o
= 1 (i.e. one infected or colonized patient must transmit the pathogen to, on average, at least one other patient). However in reality, outbreaks can occur even when these threshold conditions appear not to be met (i.e. when R
o
< 1). Conversely, outbreaks may die out despite R
o
> 1. This is because stochastic effects often dominate in small populations, such as those found within hospitals [27]. This means that while most outbreaks should be controlled when R
o
< 1, some will not. In a few cases, chance events will be such that outbreaks of staphylococcal infection may occur despite the presence of stringent control measures. While the use of stochastic modelling would have yielded data on the variance of the transmission dynamics, the focus of our paper is on the impact of hand hygiene on average prevalence curves, which can be predicted using deterministic methods – hence the strategy adopted in this paper.