Influenza viruses are believed to be spread between humans through a number of modes of transmission, including primarily through inhalation of respiratory droplets containing infectious virus, and possible contact of respiratory secretions containing infectious virus with mucous membranes. A distinction is sometimes drawn between larger versus smaller respiratory droplets, as large droplets quickly fall to the ground [1], [2], while droplet nuclei can remain suspended in the air for prolonged periods because of their low settling velocity [3]. However, aerosols are easily removed from the environment through ventilation, and infectious virus suspended in aerosols could be fragile and easily lose infectivity. The threshold for small particles is typically drawn in the range 5 µm to 20 µm [3]–[5]. Only a small number of pathogens are thought to transmit via aerosols, including varicella virus, M. tuberculosis and rubeola virus (measles) [6]. The potential for influenza virus to spread by aerosols remains controversial [3]–[5], [7], [8]. There is growing evidence that influenza A virus can spread by aerosols [3]–[5], [8]–[10], but less discussion over the potential role of aerosols in influenza B virus transmission with limited published literature. Infectious influenza B virus can be detected in the aerosol fraction (particles <5 µm) of exhaled breath of subjects with influenza B virus infection [11].

Influenza B viruses can infect all age groups. Compared to influenza A viruses, infections with influenza B virus are more commonly identified in children compared to adults [12], perhaps because of slower evolutionary rates [13] leading to greater herd immunity among adults. Influenza B virus infections can cause severe illness in all ages [14], and the mortality impact of influenza B epidemics in populations is generally estimated to be comparable to the impact of influenza A(H1N1) epidemics but somewhat less than influenza A(H3N2) epidemics, with the majority of excess deaths occurring in the very young and very old [15]–[18].

Historical volunteer challenge studies reported a difference in clinical presentation of influenza A virus infections depending on the mode of infection [9]. In one classic study, 23 people were experimentally inoculated with aerosols, 7 subsequently had serologic evidence of infection and virus was recovered from one additional volunteer without serologic evidence of infection, and 4 of those 8 had typical ILI with fever [19].

In another study, 24 people were inoculated intranasally and had milder illness than people with naturally-acquired illness [20]. In some infectious diseases (e.g. smallpox, plague), the clinical severity is known to depend on the mode of acquisition, and this property has recently been termed ‘anisotropic’ infection [21]. We previously assumed that influenza A virus also has the anisotropic property, and based on that property, further assuming that hand hygiene and face masks act primarily against contact and large droplet transmission respectively, we estimated that up to 50% of influenza A virus transmission within households in Hong Kong and Bangkok occur via the aerosol route [9]. Here, we propose that the same anisotropic nature may hold for influenza B virus infections, specifically that the mode of exposure leading to an infection may affect the pattern in subsequent signs and symptoms [21], and we use the same modeling framework to infer the proportion of household transmission of influenza B virus that occurs via the aerosol route.

In Hong Kong and Bangkok there were 104 and 113 households, respectively, with an index case with RT-PCR-confirmed influenza B virus infection. The characteristics of index cases and their household contacts are shown in Tables 2 and 3 for Hong Kong and Bangkok respectively. We examined the cumulative hazard of RT-PCR-confirmed influenza B virus infections for household contacts, and found increases in the risk of infection with fever plus cough, and decreases in the risk of infection without fever plus cough, in the intervention arms compared to the control arm. The change was particularly apparent in the households in Bangkok (Figure 1). To be more specific, we found a statistically significant decrease in the risk of infection without fever plus cough, in the hand hygiene plus face masks arm compared to the control arm in the households in Bangkok.

Under the scenario where randomization to the hand hygiene intervention reduced contact transmission by 50% while randomization to face mask and hand hygiene interventions reduced both contact and droplet transmission by 50%, we fitted the transmission model to the Hong Kong and Bangkok data. We estimated that in the absence of interventions, aerosol transmission was responsible for 37% and 26% of secondary infections in Hong Kong and Bangkok, respectively (Table 4). We also varied the assumed efficacy of hand hygiene and face masks from 0% to 100% and estimated the relative importance of aerosol transmission in the absence of interventions, which ranged from approximately 20% to 80% in Hong Kong and 20% to 32% in Bangkok (Figure 2).

We compared the cause-specific probabilities of each mode of transmission as well as the associated illnesses in the control arm for influenza A and B virus infections, in Hong Kong and Bangkok respectively (Figure 3). Data for influenza A were extracted from a previous report [9]. Both influenza A and B virus infections attributed to aerosol transmission were associated with a higher risk of fever plus cough, compared with the other two modes of transmission. The point estimates of aerosol transmission were lower for influenza B compared to influenza A in both Hong Kong and Bangkok.

We propose that the mode of spread associated with an influenza B virus infection affects the probability of experiencing fever plus cough for that infection. Based on that hypothesis, we estimated that approximately 37% and 26% of transmission was via the aerosol mode in households in Hong Kong and Bangkok, respectively. However, we should exercise caution in interpreting these findings because we have not been able to find literature supporting the anisotropic nature of influenza B virus infection, whereas we previously described literature supporting this property for influenza A virus infections [9]. Nevertheless, patterns in secondary infections and disease in the controlled trials in Hong Kong and Bangkok were consistent with this hypothesis (Figure 1). This also implicitly suggested that though hand hygiene and face masks could reduce the risk of transmission through contact or large droplets, but meanwhile increase the risk of aerosol transmission, which was associated with a greater risk of illness with fever plus cough.

Whereas we previously estimated that approximately half of within-household transmission of influenza A virus could be associated with aerosols [9], here we estimated a slightly reduced importance of aerosols for influenza B virus (Figure 3). One explanation for such a difference could be the age mix of cases of influenza A versus B, if aerosol transmission were more important among adults than children. We did not have sufficient sample size in the present study to examine whether modes of transmission might vary by age, but this would be an interesting area for further exploration.

If aerosol transmission is indeed an important mode of spread of influenza B virus, this may have important implications for control efforts. In particular measures targeting contact transmission, such as hand hygiene, and measures targeting large respiratory droplet transmission, such as surgical face masks, may not be sufficient to substantially reduce the risk of transmission. Control measures that might reduce aerosol transmission indoors include improvement in ventilation [26], modification of humidity [27], or the use of personal protective equipment that is more effective against aerosols than surgical masks. While the use of N95 respirators may not be practical in community settings and fit-testing is unlikely although required for optimal performance, other types of face masks with improved filtration compared to standard surgical masks or procedure masks may be available in the future.

There are a number of limitations to our analysis. First, our model did not include the possibility of variability in infectiousness between index cases, variability in immunity to different modes of transmission, or variability in within-household transmissibility associated with physical dimensions of the home, ventilation rates etc, and inclusion of these or other factors potentially affecting transmission dynamics could be natural extensions to our model. Because interventions were allocated randomly among households, the possibility of confounding should be minimized. Second, our model implicitly assumes that only the first infectious exposure is relevant to susceptible contacts, and once infected by that first exposure, further exposures are unimportant. Our model could be modified to allow for multiple simultaneous exposures by one or more modes, if it were understood how this might affect the course of disease. Third, while we assumed that all infections of household contacts during the 7-day follow-up were acquired within the household, it is possible that some infections were acquired outside. However in a separate study with a similar design in Hong Kong we used molecular epidemiology analyses of virus sequence data to demonstrate that most secondary influenza cases acquired infection from within the household [28], and a similar observation was reported in a household transmission study in Canada [29]. Fourth, it is possible that some secondary influenza virus infections were not confirmed due to poor quality specimens collected during home visits, or if peak influenza B viral shedding in the respiratory tract occurred between home visits at 3-day intervals. We did include serological data although this could have provided additional information on infections among household contacts. Fifth, by recruiting in outpatient clinics and using a rapid test to screen index cases, we may have introduced selection bias towards index cases with more serious illness or higher levels of virus shedding, affecting the relative importance of different modes of transmission. Finally, we did not explicitly account for imperfect adherence to the interventions, although the parameters in our model account for moderate efficacy of interventions against specific modes of transmission. Further improvements in the model might be obtained by incorporating limited data on adherence that was mainly self-reported by participants.

In conclusion, we propose that the aerosol route may be an important mode of transmission of influenza B virus in households. Further studies of non-pharmaceutical interventions in households would be improved by more careful monitoring of viral contamination on surfaces [30], [31] and in the air, and inclusion of this information in transmission models.