Wednesday, November 1, 2017

Is It a Public Health Emergency If We Drive More?

Suppose you are the head of your nation's public health bureaucracy. (Say, the CDC, FDA, EPA or similar agencies in the United States.) You notice a disturbing recent trend. Auto accident fatalities are three times higher than they were a few years ago! You do a little digging and find an incredibly mundane explanation for the rise. People are simply driving more. It's not like some new, dangerous behavior has emerged. There was some risk of a fatal accident per mile driven, and people are simply driving three times as many miles. You think for a moment about other societal costs, like pollution or pedestrian fatalities (people who didn't necessarily consent to this added risk). But many of the additional miles driven are highway miles (thus more fuel efficient), and besides a move toward hybrid and electric cars has canceled any pollution effects. And the pedestrians are drivers, too. They aren't some separate non-driving population that isn't consenting to the risk; they look pretty much exactly like the population that is driving more miles. When they're not walking, they're driving and occasionally causing accidents, including hitting pedestrians.

It would be silly to declare a public health emergency. Everyone basically knows there is some risk of a fatal crash for each car trip. This risk is unchanged. It's just that people are willing to indulge this risk three times as much. Who knows why? Some kind of wanderlust? More frequent trips home from work over the lunch break? More frequent drives to hobbies and social clubs in the evening? Long, unplanned drives on the highway just for the sake of driving? It doesn't matter. An enlightened public understands this risk and is indulging it three times as much. Public health emergency? Was the status quo from a few years ago acceptable? Should you have declared a public health emergency then in an attempt to cut the fatality rate by 66%? If "No," then what is the rationale for acting now? Why fetishise the mortality rate from a few years ago as "normal"? By what objective criterion is the new rate "too high"?

Now, suppose you recognize the same disturbing trend of increasing auto accident fatalities. This time, you find that new risky behaviors are responsible for the trend. People are driving the same amount of miles as before, but they are distracted by their smartphones. You find through surveys, data from smart-phone-based driver-tracking apps (I recommend trying one of these, by the way!), and post-mortem reports of fatal crashes that most of the rise in auto fatalities is coming from smart-phone related distraction. (That new augmented reality app that fills the road with bad guys and places bonus items on the sidewalk is particularly to blame.) In this case, it's not that an enlightened public has simply chosen to indulge a known risk a lot more than it used to. A new risk has surprised them. In this latter case, it is far less silly to declare a public health emergency. It is important to inform the public that the risk has actually changed. Plausibly it's worth taking some action to reduce the risky new behaviors (laws against smartphone use while driving? increase discretion of police officers to issue tickets?).

If you buy this distinction between two different kinds of risk trends, there is a lesson here for the recent opioid epidemic. The rise of prescription opioid overdoses is a lot like driving more often. The number of prescriptions roughly tripled and the number of overdose deaths from these substances also roughly tripled. Society made an enlightened decision to treat pain more often. A known, quantifiable risk-per-prescription did not change; it was simply applied to a larger population. The rise in heroin and "other synthetic opioid" overdoses is more like smart-phone distraction.*  It is a new kind of risk, different in character than what existed before. The mortality per user has increased dramatically; it's not simply that the number of heroin users increased. I don't think it ever made sense to freak out about the rise in prescription painkiller overdoses. Indeed, it's quite plausible that cracking down on prescriptions led people to the cheaper-but-deadlier alternatives of heroin.

It drives me nuts when people write about "the opioid epidemic" as if it's one thing wrapped up in a neat, tidy narrative. It's not. There are some very different things going on here. One is a tiny risk applied to a very large number of cases: about six overdose deaths per 100,000 legal prescriptions. (Roughly speaking, a little more than 13,000 overdoses divided by 200 million legal prescriptions, both annual figures.) Not bad odds if you're in a lot of pain. And this number is pretty much unchanged since a decade and a half ago. The other trend is a huge risk applied to a very small population. There are roughly half a million heroin users and just about 18.5 thousand overdose deaths related to heroin/fentanyl (2015 figures; keep in mind these are rapidly changing in recent years).** This yields a 3.7% mortality rate per year, which is just enormous. And this is up from ~1.5% just a few years ago. (Annual numbers are volatile, but it bounced around 1.5% until 2011, then started rising in 2012 or 2013. BTW, a 1.5% mortality per year is still enormous. This was a very risky population, even before the recent dramatic rise.) As I said in a previous post, I think we have prohibition to blame for the spike in heroin deaths. Some people are blaming loose opioid prescribing practices for the rise in heroin deaths, but I find this story implausible. Really these deaths would probably not be happening if we gave people safe, legal opportunities to get high and issued helpful, specific warnings about the risks of drug use.

_________________________________________________________

*ICD-10 has a code for "other synthetic opioids", which includes fentanyl (and a separate code for heroin). Really these are people taking heroin that's been adulterated with fentanyl, something that would probably never happen if people could legally buy pharmaceutical grade heroin. So I think it makes sense to count these along with deaths coded as heroin overdoses when tallying up death totals. The medical examiner doesn't always know which substance is actually responsible, so surely there's some miscoding going on.

**Some relevant number in this post, which is really making a different point. Note I use a different denominator to calculate the mortality risk for opioid prescriptions (legal users, not # of legal prescriptions). But I argue that the risk is small, even with this smaller denominator/higher risk. It is quite possible that the number of heroin users is understated due to survey bias, so feel free to use a different denominator if you think it changes any of my conclusions.

No comments:

Post a Comment