Our results demonstrate that the impact of social distancing measures on the number of Canadians requiring antiviral treatment during a severe pandemic depends on the transmissibility of the virus, as well as the willingness, magnitude and duration of distancing measures adopted by individuals in the population. We have shown that although informal social distancing measures can reduce the need for antiviral treatment dispensed from a centrally held stockpile, if the pandemic strain were highly transmissible, stockpiles in excess of 20% population coverage could still be required even in the presence of significant behavioural change. Public health planners struggle with the costs and logistics of stockpiling large amounts of treatment. The idea that stockpiling fewer doses and relying on informal social distancing measures in the case of a severe pandemic to decrease demand for the stockpile until vaccine becomes available seems reasonable. However, our results indicate that informal distancing does not actually decrease demand enough to warrant only a small stockpile unless the duration and magnitude of the intervention was sustained for long periods of time with high compliance. Although the main outcome of the model is the quantity of antiviral drugs required from the stockpile to treat pandemic cases, it is also a measure of the use of any health care resources that occur proportionally to the rate at which individuals seek medical attention and could be considered a proxy measure for a variety of healthcare resources that may be in limited supply during a pandemic.
If the duration of behaviour change adopted by the population is maintained for 12 weeks, the impact on the consumption of antivirals is relatively small with a mean of 25.6% (IQR = 21.7 – 28.7%) of the population requiring antivirals from the stockpile. This might be expected in a pandemic where initial media reports highlight the unknown nature of the emerging threat causing fear and panic in the population. However, significant uncertainty exists around the likelihood that compliance would remain high over the course of the 12 weeks. As the situation evolves, people may relax the behavioural changes they had previously implemented if they feel that their risk of infection has decreased [2, 57]. Behavioural changes lasting significantly less than 12 weeks are unlikely to have enough of an impact to reduce the need for antivirals from the stockpile which is a finding that is in line with the results of Maharaj and Kleczkowski . Maharaj and Kleczkowski demonstrated that if the disease is highly infectious, social distancing cannot be relied upon to control the epidemic and any social distancing that does occur would have to be extreme in order to have any effect .
The duration of time that distancing measures are in place has important implications for the economic stability of a country. In some cases, individuals may be able to change their contact patterns easily for example, by arranging to work from home. However, for individuals who work in an industry that provides critical infrastructure, significantly altering work related contact patterns is likely not a possibility especially if the behaviour change needed to be implemented for up to 6 months. Compliance fatigue would likely play a significant role regardless of if the social distancing were to be maintained for 12 or 24 weeks.
Even though we have assumed high levels of vaccine uptake in the population, vaccination has only a small impact on the overall antiviral stockpile need in the population because of the time at which we assume vaccine becomes available (Table 1). This is especially relevant during a severe pandemic as a virus with a higher R0 value will spread very rapidly and infect many individuals before vaccine becomes available even if vaccine could be produced more quickly. For both in-season and out of season emergence, the 6 month time frame for vaccine availability means that a large number of Canadians will be infected before vaccine is available to them. Reducing the timeframe from viral emergence to vaccine availability would further decrease the demand for antiviral treatment.
Our results are in contrast with a UK study examining antiviral stockpiling. A 2005 study by Gani et al.  which modelled antiviral treatment requirements in the UK population suggested that a stockpile size of 12% would be sufficient to treat pandemic cases even if the overall attack rate in the absence of intervention was 25%. However, in this study, the authors assume that treated individuals have a shortened duration of infectiousness which significantly reduces the effective reproductive number as more and more individuals are treated. However, empirical evidence suggests that this is likely only the case if treatment is started within 48 hours of symptom onset . Recent analyses of the 2009 influenza pandemic indicates that in at least one geographic region of Canada, the vast majority of patients receiving antivirals received them > 48 hours after symptom onset . In this case, the effect on subsequent transmission would be negligible and the proposed 12% recommendation would be an underestimate because of suboptimal antiviral treatment timing.
We have not explicitly considered school closure in any of these scenarios nor have we considered behavioural changes such as hand hygiene or coughing into your sleeve as the impact of these at a population level is difficult to quantify. We assume a constant level of behaviour change over some duration of time which is likely unrealistic. Some individuals may fatigue such that the magnitude of their behaviour change wanes over time. Others may react to external factors such as media reports and may increase or decrease the magnitude of their social distancing measures over time based on their perception of risk. A variety of studies have examined the intention of people to engage in social distancing measures should a severe pandemic occur and those are the studies we have used to inform this work. The results of these studies vary significantly especially by geography. For example, Asian populations that experienced SARS tend to respond at higher levels than European populations . Also, there is evidence that the way in which people respond in practice often differs from their response if asked about their “intention” (e.g. healthcare workers intention to vaccinate for seasonal influenza vs. actual vaccination rates) [60, 61]. Our model does not provide specific advice for how to distribute the antiviral stockpile. It is possible that in the case of a severe pandemic, stockpiles could be used up relatively quickly. In this case, some of the stockpile should be reserved for treatment of hospitalized cases and other vulnerable groups. Targeted antiviral use in this context would likely be much more effective at preventing deaths per dose however; our model has not specifically examined strategies for optimal deployment or management of a stockpile. The model does not consider the possibility of reduced transmission due to antiviral treatment. There is evidence that individuals who receive treatment within 48 hours of symptom onset are less infectious to others . However, it is unclear what proportion of cases would actually begin treatment within this 48 hour window. During the 2009 H1N1 influenza pandemic in Canada, many cases received antiviral treatment outside of this window . If antiviral treatment does decrease influenza transmission by making treated individuals less infectious to others this could lead to a lower population attack rate, and in turn to lower antiviral usage than we project here [25, 62]. We have also not considered antiviral wastage due to the presence of other co-circulating respiratory pathogens although this is an area that has important implications for stockpiling and is an area of ongoing research.