Does quarantine work for pandemic flu?

We've written here frequently about the ineffectiveness of quarantine for stopping the spread of influenza, but now a piece comes out in the Journal of the American Medical Association (JAMA) that claims quarantine was an effective mitigation method for influenza in 1918. Time Magazine, for example, had an article with the headline: "Study: Quarantines Work Against Pandemics":

To plan for the future, researchers in Michigan went straight to the past. Led by Dr. Howard Markel, director of the University of Michigan Medical School's Center for the History of Medicine, a team of public-health experts evaluated the U.S. response to the world's last great pandemic -- the Spanish flu in 1918. The new report, published in the Aug. 8 issue of the Journal of the American Medical Association, analyzed the public-health measures taken by 43 U.S. cities, all with populations greater than 100,000, during the six months between Sept. 1918 and Feb. 1919. Markel found that cities that early on adopted "old-fashioned," non-pharmaceutical interventions -- such as school closures, social-distancing in the community and workplace and quarantine -- and "layered" multiple interventions at once for a long period of time fared better than other cities, with slower rates of infection and lower rates of death. (Time)

Naturally I was curious about the basis for this, so I retrieved the article (subscription only, alas) by Markel and his colleagues. Examination of mortality data, week by week, in 43 cities showed great variation. Markel et al. wondered if the measures taken to stem the epidemic could account for the differences. They classified the non pharmaceutical interventions (NPIs) into three broad categories: school closures; bans on public gatherings; isolation and quarantine and looked at timing and timeliness as well as combinations of NPI categories for effects on excess mortality, height of epidemic peak and timing of epidemic peak, using multivariate analysis. One problem is that isolation and quarantine were grouped together, although they are quite different. Isolation is segregation of those who are already sick, while quarantine is segregation of those who are well but have had exposure to those who are sick. The only example given of true quarantine was in New York City where the sick were isolated and their contacts confined to their homes, which were placarded. New York City was one of the earliest to act and one most vigorous enforcers of NPI and was said by Markel et al. to have suffered less than other cities, at least by official records. John Barry in his book The Great Influenza (p. 276) suggests NYC suffered severely and many cases were just not recorded. In any event, in this case, too, it is not possible to disentangle the separate effects of isolation and of quarantine.

What is quite clear from the analysis, however, is that information about when, how long and in what combination NPIs were used in relation to the onset of the epidemic in a city explains a great deal of the variation in epidemic experience. Every city did something (i.e., adopted a measure from at least one of the three categories). 15 out of 43 used all three concurrently. The most common combination was school closure and bans on public gatherings (79%) for a median duration of 4 weeks (range 1 - 10 weeks). The most common single intervention was school closure (93%), but New York City, New Haven and Chicago never closed their schools, even in the face of widespread absenteeism (more than 45% at one point in Chicago). Several cities closed their schools more than once. The median time of school closure was 6 weeks (range 0 - 15 weeks).

The analysis showed that combinations were more effective than single interventions. This may be a correlate of how seriously the city was taking the epidemic or a real effect of the interventions themselves. But it also showed that there were features of individual cities that made a difference (the NPI x city interaction term was significant), although what about the cities modified the effect is not known. None of population size, density, sex distribution or age distribution accounted for differences in mortality. Many cities suffered several waves of the pandemic, but there was no association between severity in one wave with that of another.

The bottom line is that the earlier a city acts and the more coordinated and multifaceted its response the better off it seemed to be -- in general. The data certainly do not demonstrate that quarantine itself is effective -- they are not able to make that statement. Cities that acted using isolation and quarantine did seem to do better, especially if in combination with other measures, but we don't know the effectiveness of either separately or in combination since they were not reported or analyzed separately in the paper.

What these data do seem to establish is that the better prepared and organized a community is, the better off it will be. And the more a community ignores and denies a problem, the worse off it will be. This is the real message, not that "quarantines work."

But you knew that.