Director Frieden of the CDC likes phone surveys. He relies on them for determining how many people have been infected with the new H1N1 despite clear-cut evidence that phone surveys are worthless for this purpose. The number infected is the denominator used when calculating the case fatality rate. The numerator is the number dead. Turns out, Dr. Frieden has found a way to calculate this number, also from phone surveys:
From the CDC Weekly 2009 H1N1 Flu Media Briefing, November 3, 2009:
Joanne Silberner: Thanks. You mentioned before the idea that half the people with asthma once they’ve gotten sick have not come in. Where does that number come from?
Tom Frieden: this is from telephone survey day that that we have. We call randomly households. We identify a subset of people who have had influenza-like illness in the previous 30 days. Then for the subset we ask a series of questions about what conditions they have and what they did to seek care. We don’t know what the baseline is for that. We haven’t asked that survey, that question on that widespread basis before. We know we would like more people to seek care if they have an underlying condition with flulike illness.
Joanne Silberner: From the survey you don’t have a sense of what happened to them because you didn’t seek care. You know that from other information.
Tom Frieden: They were still answering the phone and answering our questions a few weeks later. In that regard, we think nothing terrible happened.
In other words, because they were able to answer the phone, they weren’t dead. Really? Did they all answer the phone? Every single one? Beause that would be really incredible. No-one went on vacation. No-one went out to dinner. No-one went to a movie. They all just sat home every night waiting for the CDC to call them again. Dr. Frieden must have found the most cooperative bunch of phone survey respondents, ever.
There are lots and lots of problems with phone surveys. Anyone who follows politics knows that such surveys can yield widely varying results. Just ask Matt Drudge. Anyone who uses phone surveys should know what they can and cannot be used for. I’d really like Nate Silver’s opinion on this, but I’m pretty sure that phone surveys are not a valid way to determine how many people have died of pandemic flu.
But then, I’m not an epidemiologist, so what do I know.