Perhaps the Flu has us Out-matched

1/15/2013 05:24:00 PM
Tweetable
Most people treat influenza as a less-than-serious disease. This is, of course, a mistake--the flu is actually the deadliest virus in human history, and kills some 30,000 people annually. It is also completely preventable, if only people got their flu shots in time.

Or so we think. It turns out that the evidence on flu vaccine isn't so clear. You see, we don't have any placebo-controlled trials to test the actual effectiveness of the vaccine. Sure, we use placebo-controlled trials every year to test the safety of the vaccine itself (besides extra soreness at the injection sight, it has no additional side effects besides those that occur in all vaccines, if you are wondering). But these trials only test for safety, not effectiveness. Ideally, we'd like to see whether the incidence of deaths is higher in the placebo group than the vaccinated group. And that's just the problem--since the vaccine is known to be low-risk, and the flu known to be deadly, any trial that deprives a placebo group of the real vaccine is inherently unethical.

Because it would be unethical to deny the vaccine to a placebo group--which is what we would have to do to get an ideal statistic--we have to assess the effectiveness of the flu vaccine by comparing groups who voluntarily get the vaccine versus those who voluntarily don't. The summary statistics on this are quite impressive: the flu-season death rates among those who get the vaccine are half those who don't. But that is suspicious: at most, the flu is only directly or indirectly involved in about 10% of flu-season deaths among those who don't get the vaccine. Clearly, the decision to get the vaccine is correlated with other factors that affect health. Which means that this topic counts as health economics.

So here are some health economics considerations: Broadly speaking, the individual's decision on whether or not to get a vaccine is influenced by two considerations: the budget constraint and preferences. On the one hand, preferences are important, for example, if the person is especially health-conscious and seeks out the vaccine. Or maybe the person is just risk-averse and finds that the assurance against the probability of the flu is worth the cost. Both of these preference-factors, it turns out, will lead to a lower mortality rate among people who get the vaccines than don't even if the vaccine is completely ineffective. That is, people who don't get the vaccine are less health-conscious and/or less risk averse, which means that elsewhere in their life they are likely to engage in more high-risk, unhealthy behaviors that increase the mortality rates.

But then there are also budget constraint factors. Even if someone is health conscious, they may simply be too poor to get the vaccine. Better to put food on the table than spend $25 on a vaccine. Or maybe they don't get the vaccine because they have no health insurance. On the other hand, there are also other budget constraint considerations: some health-conscious employers, especially hospitals, require a flu vaccine as a condition of employment during the winters. That makes not getting the shot a hugely costly choice. These budget considerations also imply that the populations of those who do and don't get the vaccine will differ in other ways that are correlated with mortality, as well. It is well-established in the research literature that income and life-span are highly correlated. Indeed, since 1985, almost all of the increases in the average national life-expectancy has been among the wealthy, with the working class living barely any longer than they were three decades ago. And workers at companies that require vaccines differ from those who don't in two systematic ways: 1) 100% of them are employed, meaning they are less likely to be poor, and 2)they have a health-conscious employer who possibly offers more incentives to stay healthy, and better health insurance. It is well known that hospitals typically have better health insurance than other types of firms, for example.

There is, of course, another factor in the budget constraint: information. This includes both information about the prevalence of the flu, and information about the effectiveness of the vaccine. There is a lot of empirical evidence in the health economic research that people actually do change their behavior in response to changes in the prevalence of diseases in their area. For example, according to this study, an increase in the local prevalence of HIV is associated with an increase in safer sex practices in the locality. I'd guess that the same effect is true of the flu--an increase in the prevalence of the flu probably increases the propensity to get a flu shot--but I'm unfamiliar with any studies on the topic.

In addition to knowing the prevalence of the flu, individuals' decisions also require expectations about the potential costs of getting the flu--it is far deadlier than most people realize, as well as the effectiveness of the flu vaccine. Which brings us back to where we started. If people think the vaccine is ineffective, they won't get it regardless of how rich they are or how cheap we make it. The CDC responds to this by providing add campaigns to try to inform people about the estimated effectiveness--believed to be around 70% or so--but this is not data people will accept blindly. Experimental data suggests that people do, in fact, place more weight on personal experiences than official statistics, so I suspect that people who know someone who got the vaccine but still got the flu are less likely to get the vaccine themselves. In a way, this is problematic for our statistics as well. People who know someone who got the flu represent a group more likely to get the flu themselves. The fact that they are choosing not to get the vaccine further biases our estimate of the effectiveness of the vaccine.

All of this suggests that perhaps we need economists to work with the CDC to produce estimates of the effectiveness of the vaccine, if the CDC remains opposed to a true placebo-controlled trial for effectiveness.

In the mean time, let me suggest that perhaps the flu has us out-gunned. Fatalities actually have remained just as high no matter how much we innoculate the population. This could just be spurrious correlation, or it could mean that the vaccine is ineffective. There are a variety of reasons why the vaccine wouldn't work. For one, the people most likely to die from a flu-related complication may not have the immune system necessary to kill the flu virus, even if the vaccine allows them to develop the anti-bodies that would supposedly grant immunity. This would be consistent with the data showing that increasing vaccination of senior citizens hasn't decreased flu-related fatalities among them. Another possibility is that the flu vaccines actually aren't good matches for the dominant strains, or worse, that inoculating the population against one flu virus only causes another strain to take its place. It turns out that for the vast majority of flu diagnoses, not only do we not know the strain of flu it was, but we don't even know if it was the flu at all. Most viral infections cause flu-like symptoms, so that somewhere between 50% to 90% of flu diagnoses are actually something else.

The point that I think people need to realize is that inoculation against flu is a considerably more ambitious progect than the great inoculations of the past against thinks like polio. Unlike polio, flu mutates and adapts. And it does so quite quickly. And while there are thousands of potentially fatal variants of the flu, vaccines are only effective (in the sense that they generate the required threshold of anti-bodies) if we inoculate for only four or fewer strains at a time (the first ever four-strain vaccine will be in use next year).

But there is hope. Remember that with polio, the first vaccination effort--the dead-virus vaccine developed by Jonas Salk--was a complete failure. That's because Salk's vaccine didn't target the polio infection itself. Albert Sabin, working in my old office at Cincinnati Children's, figured out that polio lives in the digestive tract, and figured out how to make the first effective vaccine against polio by producing a live attenuated version of the virus that could be consumed orally rather than injected intramuscularly. This not only created immunity in recipients of the vaccine, but also increased the "herd immunity" as the attenuated version of the virus was spread from those initially inoculated to those who weren't. By understanding how Polio thrives, how it spreads, Sabin practically eradicated it from the developed world. And today we still know very little about influenza. No one is quite sure why it only stalks us in the winter, nor how it gets spread across populations. But we are getting closer. For example, new research suggests that children have a lot to do with it: when we inoculate children with the nasal spray, which contains a live-attenuated flu virus, not only do flu rates among children fall, but so do fatalities among senior citizens. This suggests that, perhaps, children are an instrumental part of the spread of the virus, and that large-scale inoculations of school children could save thousands of lives.

Oh, and one more thing. While the medical establishment believes that a placebo-controlled trial of the intramuscular dead-virus vaccine would be unethical, they did not object to such a trial for the live-attenuated nasal spray vaccine. I suppose that the reason for the differing standards has to do with the fact that the nasal spray was new and untested, unlike the traditional vaccine that dates back to the 1940s, but the distinction seems silly considering the uncertainty surrounding the effectiveness of the dead-virus vaccine. If you are curious, the live-attenuated virus decreases a child's probability of getting the flu by 92%. But--that's only true if we correctly predict the strain of flu virus, and doesn't directly tell us about the impact on fatalities, which is presumably what we care about. So, it still isn't clear how effective the vaccine is.