During the study period (Dec 2nd 2008 - 15th Jan 2009) we identified and obtained information by questionnaire from 72 patients with ARI and 66 controls with other acute non-respiratory conditions.
The median age of the sample population was 49 years (interquartile range: 24). Table summarises the characteristics of cases and controls. Cases were more likely to be older, males, living with children and to have higher co-morbidity. Cases also appeared to be less frequent users of public transport use but this effect was not statistically significant when frequency of habitual public transport use was modelled as a continuous variable (p-trend = 0.845).
Characteristics of cases and controls (n = 138)
Table presents the crude and adjusted odds ratios for the association between public transport use and acquisition of ARI. The univariate analysis showed that cases were 10 percent more likely to have travelled by public transport in the five days before symptom onset than controls (unadjusted OR: 1.10, 95% CI: 0.55-2.21). This association was not statistically significant. After adjusting for age, gender and co-morbidity (Model 1), cases were still 10 percent more likely than controls to have travelled by public transport in the five days before symptoms, but once again, this association was not statistically significant (adjusted OR: 1.09, 95% CI: 0.50-2.38).
Odds ratios for the association between bus/tram use and ARI (n = 127)
Model 2 also adjusted for deprivation, child cohabitants, flu vaccination and habitual public transport use in addition to a priori
confounders; this time the strength of the association increased, with cases nearly six times more likely to have travelled by public transport in the five days before symptoms; this was statistically significant (adjusted OR: 5.94, 95% CI: 1.33-26.47). After taking into account age, gender, co-morbidity, child cohabitants, flu vaccination and deprivation, there was still no statistically significant association between frequency of habitual public transport use and risk of consultation for ARI (p-trend = 0.057). Nevertheless, when examining frequency of habitual public transport use as a categorical variable, there was a trend towards a decreased risk of ARI consultation in persons who were more frequent habitual users (1-3 uses/week: adjusted OR = 0.54 95% CI: 0.15-1.95; >3 uses/week: adjusted OR = 0.37 95% CI: 0.13-1.06). The results of a separate logistic regression to explore the association between frequency of habitual public transport use and recent use are presented in Additional file 1
; frequent habitual users were more likely to have been recent users but inclusion of habitual use in Model 2 nevertheless improved the overall fit compared with Model 1. We observed a significant interaction between regularity of public transport use and recent use on the risk of ARI (p = 0.019). We were unable to perform a complete stratified analysis by frequency of public transport use to explore this interaction further, because of the small sample size. However we present a limited stratified analysis in Additional file 2
, which is consistent with our key finding that infrequent habitual use of public transport poses the greatest risk of ARI.
One of the most controversial issues in pandemic preparedness is the potential benefit conferred by public health measures, including social distancing strategies such as the closure of mass transportation systems [8
]. No clear consensus has yet been reached about whether mass transportation systems should be closed during a pandemic; and the mildness of the influenza A/H1N1 pandemic in 2009 (which did not trigger transport closures) offers little prospect of shedding further light on this issue. When considering the possible policy impact, there is often confusion between the potential public health effects of closing urban transportation during a pandemic (thereby delaying population spread), versus the individual benefit of avoiding public transport (reduction in individual risk of infection). Much will also depend upon whether individuals will change their behaviour during a severe pandemic, by avoiding using public transport whilst symptomatic.
In this study we set out to explore the association between recent public transport use and the likelihood of acquiring acute respiratory infection (ARI) at an individual level. We found a significant association between ARI consultation and recent use of public transport; however, because of the small study size, the confidence intervals are wide and the magnitude of risk is not clearly defined. Furthermore, we detected a trend suggesting that the absolute risk of public transport in relation to ARI may be modified by frequency of habitual use, the risk being decreased among the most regular users. However the study lacked statistical power to explore this fully, using multiplicative modelling terms.
Our findings may suggest that whilst use of public transport in the winter potentially exposes travellers to respiratory viruses and increases the risk of ARI, the risk is offset among regular users who either acquire immunity against a range of respiratory viruses (or partial immunity sufficient to produce asymptomatic infections), presumably because they are more frequently exposed; or because they develop compensatory behaviours that reduce risk. An alternative methodological hypothesis would be that habitual users of public transport are exposed to respiratory viruses earlier in the winter season and so may have acquired protective immunity by the time this study began in early December.
If true, the implication of these findings for the 'normal' winter respiratory virus season is that occasional users of public have the greatest risk of acquiring ARI. The implications for pandemic preparedness may be somewhat different. Given a novel virus, against which there is lower background population immunity, the attenuating effect of regular public transport use would be diminished and the individual risk might be driven more by recent use of public transport and less by habitual patterns of usage. Thus the same protective effect of regular bus and tram use may not be present.
This study has a number of strengths and limitations. We used GP consultations for ARI as a proxy for 'all ARI' and acknowledge that because we selected only those ARIs, which necessitated medical attention (or selecting patients more likely to consult a GP with ARI), the findings should be interpreted carefully. Nevertheless, experienced primary care physicians identified all cases of ARI; their symptoms were cross-checked by the questionnaire tool and so misclassification bias seems highly unlikely. However without virological confirmation it is not possible to describe the range of underlying respiratory virus aetiologies in the cases in the study; thus the findings, while generalisable, are not specific to influenza. Controls were selected as the next patient with an acute non-respiratory condition, as determined by the chronological order of appointments in the practice. This may have introduced some bias if persons were seen in order of severity rather than by booking order. It is possible that if control patients with a non-respiratory complaint (e.g. worsening cardiac failure) in fact had underlying respiratory virus aetiology, this might have under-estimated the risk of public transport use. This study was performed in Nottingham where in general terms public transport use is more closely related to affluence than is the case in larger cities (e.g. London) with more extensive urban mass transportation networks. We also lacked data on subjects' normal working environments (e.g. lone working or in close contact with many people). However we adjusted for deprivation score in the multivariate analysis, which should account, to some extent, for affluence and job type. Whilst we obtained data on frequency of habitual public transport use, we did not obtain data on time of use (e.g. rush hour vs. off-peak) and passenger density may affect the likelihood of exposure to an infectious traveller. The study was somewhat underpowered due to the time and resource restrictions imposed by an undergraduate project. As participants were not asked whether they had acquired ARI symptoms earlier in the flu season, post-infection immunity in controls manifesting as sub-clinical ARI was not investigated. We were unable to address asymptomatic infection in the current study.
There are few published data available, against which this study can be meaningfully compared. Studies of transmission on board aircraft are not comparable because of differences in the number of 'passenger exchanges' per hour (very low compared with buses and trams), duration of journey, and ventilation [2
]. The study by Kar-Purthayaska et al. concludes that transmission of pandemic influenza did not occur on board a school coach. However, these data are not strictly comparable because the population was less heterogeneous than on public transport, the two school groups may have remained separated, 'passenger exchange' was far lower and the outbreak occurred in summer, albeit with a journey time of 60 minutes [4
]. In addition pandemic influenza A/H1N1 has been associated with a very high rate of asymptomatic infection in children [10
]; so many may already have been immune or may have acquired asymptomatic infection following exposure. The most relevant comparator study, by Williams and colleagues, assessed the influence of regular public transport use on the risk of serologically confirmed influenza in hospital healthcare workers, finding a risk ratio of 1.06 (95% CI 0.69-1.63) in the 2006/07 winter season [3
]. These investigators did not however report any findings about the frequency of habitual public transport use and did not investigate the influence of public transport use in the period immediately preceding illness onset. Our findings add new detail in relation to the possible role of public transport in relation to the acquisition of infection during a relevant incubation period and suggest that the frequency of habitual public transport use ought to be explored further in future studies.
Although causality can never be firmly proven on the basis of a small observational study, the effects we observed are biologically plausible including the diminution of risk with increasing habitual travel frequency. Buses and trams are generally poorly ventilated carriages of people sitting and standing in close proximity (particularly during peak travel periods) who, if travelling whilst symptomatic and exhibiting poor hand hygiene and cough etiquette, could spread respiratory virus infections via direct or indirect contact. Frequent public transport users regularly exposed to respiratory pathogens may be more likely to develop protective immunity. However, there still seems to be a risk associated with bus or tram use related to the period immediately preceding symptom onset. During a pandemic, when a novel respiratory virus is in circulation, the protective effect of habitual regular bus or tram use may be much reduced. Although small and limited in scope, this study nevertheless suggests that a larger more detailed investigation would be appropriate.