Thursday, April 30, 2009

Swine flu, measurement error, and policy responses

Is the Mexican government overreacting to the swine flu virus outbreak? It is still to soon to know. So far, in the first week of the "contingency", most of the actions taken by the Mexican government have been backed by the WHO, and by a majority of the public opinion.

We also have witnessed a parade of numbers trying to assess the evolution and riskiness of the virus. How can we measure this? An important statistict in epidemiology is the Case Fatality Ratio (CFR = number of deaths / number of cases). The latest figure of "suspected" swine flu-deaths is 159 but only 7 of them have been confirmed as positive cases by laboratory tests. Also, the latest estimate of atypical neumonia cases in this period is about 2500 (April 29th, 2009). Thus, making the extreme assumption that all deaths become positive cases (and keeping all else equal) results in an estimated (upper bound) CFR of 159/2500 = 6.36%. Clearly, these figures will change as more positive cases and/or swine-related deaths occur (see update below).

It is important to note that measurement error on either the numerator or the denominator affect any CFR estimate, thereby making the virus to appear more or less deadly. For instance, if the virus is very contagious but not deadly, the CFR will become approximately close to zero (not too risky a ratio, but consider that seasonal influenza kills 36,000 people in the US every year). On the other hand, if the virus turns more deadly but not very contagious, the CFR will increase (for instance, SARS has a CFR of about 50%, very deadly indeed).

But then again, is the Mexican government overreacting? Do we really need to close schools and restaurants, or should we focus instead on revamping our health services? We don't know yet. Oddly enough, the more deadly this virus becomes (ie, the higher the CFR worldwide), the better our government will look in retrospective for properly reacting to a serious risk.

Alas, if the swine flu virus ends up being more deadly here than elsewhere (ie, with a CFR higher in Mexico than abroad), the worse our overall health system will look and the more off-target the Mexican reaction will be. It may be the case that Mexico's poor health care services and limited coverage turn more deadly than a new found virus. I truly hope to be wrong on this.

Related links:

Swine Flu and the Mexico Mystery: Why does the swine flu seem to be more deadly in Mexico? from Slate
Mexico's High Death Rate Poses Key Question on Virus from the Wall Street Journal
Swine flu: what do CFR, virulence and mortality rate mean? from Effect measure
Swine flu: Questions and answers from the Virology blog
Disease Outbreak News from the World Health Organization.

May 6 Update:
In Mexico: 29 deaths / 942 confirmed cases = 3.07% CFR
In the US: 2 deaths / 642 confirmed cases = 0.311 CFR

Wednesday, April 29, 2009

Risk Communication in Epidemics

Great piece on the the hazard vs. public outrage tradeoff.

"Let me tell you the basics of risk communication, and then I want to apply them, a little bit, to bird flu. The fundamental principle of risk communication can be summarized in a number, [which] is the correlation between how much harm a risk does and how upset people get about it. If you look at a long list of risks, and you rank them in order of how upset people get [about them], then you rank them again in order of how much harm they do, then you correlate the two, you get a glorious 0.2.
Those of you who remember your statistics know you can square a correlation coefficient to get the percentage of variance accounted for: If you square 0.2, you get 0.04, or 4% of the variance.
That is, the risks that kill people and the risks that upset people are completely different. If you know that a risk kills people, you have no idea whether it upsets them or not. If you know it upsets them, you have no idea whether it kills them or not."

Read the whole thing by Peter Sandman, Risk Communication Specialist