Monday, 26 February 2024

Statistics of a sort

My mail this morning included one from Medscape which caught my eye, covering an advertisement infested article - it is free, after all - the top of which is snapped above. And I took a look in the intervals of bread batch No.711, now brought to a satisfactory conclusion.

The story seeming to be that methamphetamines and cocaine have overtaken opioids as what is taken with fentanyl. Also that these supplements are increasingly being taken by smoking rather than by injection, possibly because this last is seen as high risk. A story covering the reports at references 1 and 2.

All this being in the US, a country of some 335m people and 3.5m deaths annually, that is to say around 1 per 100 population. Comparable figures in the UK being some 70m people and 0.7m deaths annually. Making the US between four and five times bigger than we are. Not that that stops them having ten times more carrier task forces than we do, once the winners of the Battle of Trafalgar. Which, incidentally, was a battle that we won without the help of the Germans, unlike Waterloo.

Reference 1 starts by telling us that there are around 110,000 drug overdose deaths in the US every year, with 70% of them involving illegally manufactured fentanyl, which rather goes against them being more like suicides than accidents, which was my first thought. Not that the reports says anything about that. But it does say that the proportion associated with smoking has increased a lot, while the proportion associated with injecting has decreased somewhat. On the other hand, the number of adults injecting increased from under 1m in 2011 to near 4m in 2018, this being put down to a shift from prescription opioid misuse to the use of heroin and IMFs (illegally manufactured fentanyl and fentanyl analogs).

A lot of the detail is based on the 139,740 overdose deaths 2020-2022 which occurred in the 28 jurisdictions participating CDC’s State Unintentional Drug Overdose Reporting System (SUDORS), which looks like rather less than half the total. The message being that drug users need to understand that smoking is not risk free.

Reference 2 draws on the large Oracle-Cerner database of urine tests maintained by Millennium Healthcare, a database involving millions of patients and even more millions of tests. For which see references 3 and 4. Tests requested by doctors, for one reason or another. Tests which can detect all kinds of substances, markers and derivatives. The snap above is alleged to be the data model for this database - but it does not zoom and so only suggests complexity, rather than give anything more substantial away.

Which is a pity, as I am a great believer in data diagrams of this particular variety. If it is a good database, they can tell us a lot both about the database and the subject, the problem that it addresses. All the thing in my days in the world of work.

According to reference 2, this database runs from 2013 to present, and the analysis in question there used one specimen each from more than 4m adults, drawn from all 50 states.

While reference 5 tells us something about the makeup of this database, on the basis of a large sample. The table above having been slightly awkwardly trimmed on its way here for some reason. The second block down (hopefully visible if you click on it) tells us something about why the sample was requested in the first place, with one of the answers being pain management.

However, among all this stuff, I failed to find any kind of statistical review of the sample, of this data gathered for administrative rather than statistical purposes, the sort of thing that here in the UK one would hope to find lurking somewhere in the depths of ONS. But I suppose the US is too big a country for that highly centralised approach to work. 

Nor did I find anything which explained when a substance was recorded as being present. Was it a binary choice? Did it have to reach a certain concentration? Could it have got there by contamination - either in the home of the consumer or in the slummy back room of the supplier?

But there was a note to the effect that samples were excluded from the analysis where the doctor supplying the sample had said that the patient concerned had been prescribed the drug in question - a note carrying the implicit suggestion that some doctors were more careful about this than others. Which would mean that rates of unprescribed use of controlled drugs derived from these samples would be overstated to that extent.

It also occurs to me that the fact that doctors are probably incentivised to order tests, probably helps rather than hinders in this case. It makes the database of tests bigger than it might otherwise be, quite possibly without introducing further biases of interest.

One of the more striking graphics in reference 2. One vaguely knew that lock-down likely meant a lot more drug use, but I had not seen it brought out in quite this way before. While the point here was more that UDT (urine drug testing) data was a pretty good indicator of what was going on out in the world.

So all in all, I am not much the wiser after this little excursion. Nor has it disturbed my belief that we would do better to treat abuse of controlled drugs as a health problem rather than as a crime problem. Spend your money on therapists rather than jailors.

PS 1: I believe that the rather flamboyant, long-time boss of Oracle UK used to be very keen on racing yachts, and made regular appearances on that scene. He must have given his US colleagues the bug to judge by the snap above.

PS 2: I noticed a few negative references to benzodiazepines in reference 5, once very widely used, perhaps still very widely used worldwide, and including the well known Valium. Presumably to do with the fact that while GPs here in Epsom had, say ten years ago, been fairly relaxed about prescribing the stuff for mild medical or dental anxiety, they are no longer.

References

Reference 1: Routes of Drug Use Among Drug Overdose Deaths — United States, 2020–2022 - Lauren J. Tanz, R. Matt Gladden, Amanda T. Dinwiddie, Kimberly D. Miller, Dita Broz, Eliot Spector, Julie O’Donnell - 2024. CDC.

Reference 2: The “Fourth Wave”: The Rise of Stimulants and the Evolution of Polysubstance Use in America’s Fentanyl Crisis - Millennium Health - 2024. Signals Volume 5.

Reference 3: https://www.millenniumhealth.com/. 'Empowering Clinicians with Comprehensive Drug Testing and Data Insights: Millennium Health is an accredited specialty laboratory with over 15 years of experience in drug testing services helping clinicians monitor use of prescription medications and illicit drugs and effectiveness of treatment plans'. Very hot on urine samples.

Reference 4: https://www.cerner.com/.

Reference 5: Millennium Health Signals Report: National Drug Use Trends - Millennium Health - 2020. Signals Volume 2.


No comments:

Post a Comment