A key question in contingency planning for a possible influenza pandemic is the value of restricting international travel to and from affected countries or regions, or imposing entry or exit screening of passengers at airports1. The potential for infectious diseases to spread rapidly through an increasingly well-connected, steadily growing world population2,3,4,5 was brought into sharp focus during the 2003 epidemic of severe acute respiratory syndrome (SARS). By 12 March, 150 suspected cases had been reported in seven countries. By the end of the outbreak, 27 countries had reported 8,096 suspected cases6 and five incidents of transmission on commercial aircraft7. The World Health Organization (WHO) issued several travel advisories in an attempt to slow the international spread of the disease8. These were controversial because of their adverse economic impact and the uncertainty about their effectiveness. Here we analyze the impact of WHO travel advisories in 2003, and evaluate the likely effect of similar interventions in the case of an influenza pandemic.

The direct effect of the 2003 WHO travel advisories is difficult to quantify. There were small reductions in airline passenger numbers at the start of the epidemic, followed by larger reductions as the epidemic grew (Fig. 1a). Reduced international travel continued well beyond the time when WHO travel advisories were lifted. Despite markedly fewer airline passengers, cases continued to be exported throughout the epidemic, although in decreasing numbers (Fig. 1b).

Figure 1: International travel and epidemic spread.
figure 1

(a) Daily SARS case numbers in 2003 (thin black bars)13 and monthly percentage change (over previous year) in airline passenger volume (thicker bars). Four air passenger datasets are shown: global passengers (black)14, Asia-Pacific passengers (light gray)14, Hong Kong airport passengers (dark gray)15 and Beijing International Airport passengers (white)16. (b) Timeseries of the average number of countries importing cases from a source country within a SARS epidemic, as calculated from a mathematical model of epidemic spread within a set of 100 countries connected by airline travel (Supplementary Note); 1,000 model simulations were performed. p represents the proportion by which travel is reduced at the start of the epidemic. Assuming source control fails (that is, uncontrolled spread in source country), the impact of 0% (dashed line), 80% (dash-dot line) and 99% (dash-dot-dot line) reductions in travel are shown. Control of the source epidemic (represented by the reproduction number, R, falling to 0.8 on day 50) is shown for p = 0% (heavy solid line) and p = 80% (heavy dashed line). The number of countries having reported SARS cases is plotted by onset date of first case (dots; day 50 was 12 March 2003)6. The predicted rate of spread of an influenza pandemic (with no interventions) is also shown (gray line)13.

Case isolation was highly effective in reducing onward transmission of SARS, and most imported cases were contained at their destination6 because infectiousness peaked well after the onset of clinical symptoms. In contrast, considerable infectiousness can be associated with presymptomatic or mildly symptomatic influenza infection5,9. Secondary influenza infections would be more likely to arise on international flights and from imported cases, accelerating international spread well beyond that seen for SARS. Conversely, the incubation period of influenza is short compared with SARS (1.5 d10,11 rather than 4 d12), which reduces the probability an asymptomatic person will travel. This effect, however, is more than offset by the much faster epidemic growth rate for pandemic influenza—potentially an average doubling time as short as 2.3 d (compared with 5.8 d for SARS; Fig. 1b and Supplementary Note online).

To examine the impact of travel restrictions more rigorously, we constructed a simple mathematical model of an epidemic in a source country with cases exported to any of 100 other countries with equal probability. As a best-case scenario, we assume exported cases do not seed new epidemics and exit screening is 100% effective, so only asymptomatic cases are exported (Supplementary Note). Travel restrictions reduce the probability of any individual leaving an outbreak area, and so reduce the rate at which non-source countries import cases (Fig. 1b). However, travel reductions of the order of 80%, for example, only increase the interval between exports by days (Supplementary Note). Travel restrictions with >99% effectiveness are needed to increase the time between exports to the order of weeks (Supplementary Note). Even at this level, travel restrictions only slow the exportation of cases rather than halting spread (Fig. 1b). Compliance with travel advisories and effectiveness of screening are major issues in implementing such a stringent policy.

Key to the impact of travel reductions is the rate of growth of the epidemic in the source country and its eventual final scale. If the source epidemic is controlled before there are thousands of cases (bringing the number of secondary cases per infected individual, R, to below 1) travel restrictions during the containment phase may have a large impact on the probability that an infected individual travels out of the source area and potentially seeds a new outbreak (Fig. 1b). Early intervention in the source region is crucial: containment of a pandemic influenza strain is probably only feasible when there are less than 50 cases11.

The effect of airline travel on international spread of infection is complicated by heterogeneities in the airline network3,4,5. If only a partial restriction of air travel is possible, closure of highly connected hub airports has the potential to slow (but not halt) the epidemic more effectively than a homogenous reduction in global air travel4. But these heterogeneities only have a major role when global case numbers are low: once there are tens or hundreds of thousands of cases and multiple epidemics, travel restrictions have little impact even if optimally targeted.

Governments will use several methods to try to reduce the risk to their own population in the event of an emerging epidemic in another country, including travel advisories, passenger screening or even rigorously enforced border restrictions. Our analysis of SARS in 2003 strongly suggests that the issued travel advisories (along with individuals' perception of risk) induced large reductions in travel, but that these were too late (after the effective implementation of internal control strategies) and of too small a magnitude to impact the global spread of SARS had there not been such effective control of the epidemics within affected areas. Our analysis also indicates that restrictions on travel will be of limited benefit in slowing global spread of a pandemic influenza outbreak that is not contained at its source. There may be a role for travel restrictions applied to the source country while containment efforts are underway—minimizing the chance that one of the first few hundred infections of an outbreak might be exported to a region where containment would be less feasible.

Country-based contingency plans for the next influenza A pandemic should therefore largely focus resources on facilitating treatment, monitoring and control of new cases at home. Once case numbers exceed a few hundred in source areas, only rapidly implemented and almost total restriction of international travel can prevent the export of cases and the triggering of new epidemics in unaffected areas. The efforts of the WHO and the international community should be targeted at bringing any emerging influenza epidemic under control as rapidly as possible in the country in which the new strain first emerges. Success in this task is likely to be the dominant factor in restricting international spread.

Note: Supplementary information is available on the Nature Medicine website.