Hacker News new | past | comments | ask | show | jobs | submit login

For me, the most useful page of the linked PDF is page 22, the tables of 1999-2015 stats, without any projections. I see numbers from the CDC for general population like:

    Alcohol deaths per 100,000
    1999  2015  Pct Increase
    7.0   10.3  47%

    Drug deaths per 100,000
    1999  2015  Pct Increase
    6.9   16.3  136%

    Suicide deaths per 100,000
    1999  2015  Pct Increase
    10.5  13.8  31%
So, devil's avocado -- anecdotes aside, is this all just due to the opioid epidemic?

Ie, the US is in the middle of a widespread opioid problem right now, and suicide and alcohol use may correlate.

I would love to see some tables that compare this specific data across multiple countries over the same time frame, as that could help tease apart how much of this is really fallout from one single problem in the US, the opioid crisis -- and, if not, what factors might be involved ('are there other countries on a similar track? What do they have in common with the US?').

I wonder if OurWorldInData might be interested ...

Edit: others have posted the link below, with what might be 1 of the 6 sources of data that would be helpful for this comparison: recent global suicide stats by country. (Ideally we'd have this for all 3 causes, and then also for 1999).

https://en.wikipedia.org/wiki/List_of_countries_by_suicide_r...




There is the Werther-effect[1] aka copycat suicide, which could explain the problem. But probably compounded by other factors such as rising uncertainty (job security, financial struggle) many face today. E.g. for rich nations the problem might be that losing something (job/credit) is harder on people than never having had that "thing" (whatever is "lost") in the first place. (the latter is probably a philosophical point but having gone through rough times after good times myself it always felt harder digging myself out than it felt working hard initially to reach the "good life")

Speaking of philosophy I also have seen claims that the damage that is done to the environment would feed back into the social fabric - so people would kind of feel the tension from the damage that is done to nature. I was unable to find anything that could corroborate these claims nor do I subscribe to it, but fwiw the thought has stuck with me. (my interpretation of it is that many people today are hooked to some form of news on social media which is mostly negative. When I grew up news seem to be something only old people read now everyone is engaged in the problems of far away issues. Or maybe it's because I'm old myself now in comparison and am more prone to consuming news with all its negativity ...)

https://www.ncbi.nlm.nih.gov/pubmed/18082110


Mass/cluster suicide is, unfortunately, a poorly understood and researched phenomenon. The Werther-effect seems to be a good first pass at describing what is likely to be a manifestation of many underlying causes [0]. Though this may sound like a broken record: more funding is needed.

Suicide can be described like any other disease [1]. It has a coefficient of communicability/spread. Though I can't find the citation, it is thought that social media has increased this coefficient by ~22x ; meaning that suicide has become much easier to 'catch'. In the US, a fair few suicide 'clusters' have sprung up and many more are likely being hidden by empathetic, yet mis-guided, coroners. One famous one was in at Palo Alto HS in California[2], where a lot of Stanford's Professors sent their kids. Another, still ongoing, is in Colorado Springs' religious High School network[3]. I'll let Newsweek speak for the true horror that is possible via new technology:

>“This is the part that kills me—I know she was texting other kids at the time and letting them know,” he says. She wrote, “My feet are off the floor,” and “Everything is getting hazy and dark.” None of the kids intervened; one responded by suggesting she “unhang.”

Suicide is also prevalent in east Africa currently [1] and there have been spats of it throughout time in a wide variety of place and for a large range of reasons[4].

Even talking about suicide at all, like I am currently doing, can increase it's reach. It's a very tricky thing to research and combat, as a consequence.

PLEASE, if you are feeling suicidal, reach out! Call a friend, a family member, or someone you feel safe with. CALL THIS NUMBER NOW:

1-800-273-8255

[0] Like describing a stuffy nose; it has many causes.

[1] https://en.wikipedia.org/wiki/Epidemiology_of_suicide

[2] https://www.theatlantic.com/magazine/archive/2015/12/the-sil...

[3] https://www.newsweek.com/2016/10/28/teen-suicide-contagious-...

[4] https://en.wikipedia.org/wiki/Suicide_epidemic


Relevant question: If someone dies and they have both alcohol and an opioid in their bloodstream, do they count against alcohol, opioid, both, or do they try to work out which one killed them?


They try to work out which one killed them. If they can't, it should to my knowledge be filed as an unknown-substance drug death, although the toxicology report would have the details.

Forensic toxicology screens are pretty thorough, and benefit from being able to take invasive samples like heart blood, brain, liver, and even eye tissue samples. That provides a lot of flexibility to find any substance that's present and look at relative concentrations and metabolites. Toxicology is usually sufficient in combination with the coroner's report on proximate cause of death; even if multiple substances are found, it's only a problem if they're all able to kill by the relevant mechanism.

In this example, someone with a high alcohol tolerance might die with a potentially-lethal level of alcohol in their blood, and also test positive for opioids. A tox screen should then be able to estimate that they had stopped drinking a few hours before (e.g. high acetaldehyde levels in blood and urine, no alcohol in stomach fluids), then freshly used opioids just before the time of death.

This gets harder if you have drugs with additive effects (e.g. death by respiratory depression, testing positive for heroin and ketamine); the autopsy would list both, but I don't know how the stats are handled. And it seems like a nightmare in the case of drug interactions; I have no idea what happens if somebody dies of serotonin syndrome triggered by mixing alcohol with MAOIs.


Funnily enough, I happened to be on the page that answers this.

Page 160: "Alcohol-induced deaths and drug-induced deaths are mutually-exclusive. However, these deaths may also be considered suicide deaths."

Also, I can't find any more info on what is included here. They just say it doesn't use ICD-10 codes:

>"NCHS has defined selected causes of death groups for analysis of all ages mortality data: Drug-Induced causes, Alcohol-Induced Causes, All Other Causes. The group code values are not actual ICD codes published in the International Classification of Diseases, but are "recodes" defined to support analysis by the Selected Causes of Death groups." https://wonder.cdc.gov/wonder/help/mcd.html#Drug/Alcohol%20I...

EDIT:

If I go to CDC WONDER (https://wonder.cdc.gov/controller/datarequest/D76) and click "Drug/alcohol induced causes" a box populates showing (presumably) ICD-10 codes: X40-44, X60-64, X85, Y10-Y14

That includes stuff you wouldn't expect:

X44:

  agents primarily acting on smooth and skeletal muscles and the respiratory system
  anaesthetics (general)(local)
  drugs affecting the:
  · cardiovascular system
  · gastrointestinal system
  hormones and synthetic substitutes
  systemic and haematological agents
  systemic antibiotics and other anti-infectives
  therapeutic gases
  topical preparations
  vaccines
  water-balance agents and drugs affecting mineral and uric acid metabolism 
http://apps.who.int/classifications/apps/icd/icd10online2004...

So this could be partially due to "better access to healthcare". More people are being put on blood pressure, etc medications and are dying from over/under-dosing on them. I don't see any reason for them to include X44 in the current study besides trying to mess with the numbers...

EDIT 2:

Indeed, when I selected only x44 on CDC WONDER, here are the results:

  Year Deaths per 100k Pop
  1999  1.5
  2000  1.7
  2001  1.9
  2002  2.4
  2003  2.6
  2004  2.9
  2005  3.3
  2006  3.8
  2007  4.2
  2008  4.5
  2009  4.7
  2010  5.0
  2011  5.6
  2012  5.3
  2013  5.5
  2014  5.8
  2015  6.3
  2016  7.4
  2017  8.1
So the death rate from normal (not addictive or recreational) medications is 5.4x higher in 2017 than in 1999.


The full information is available on this via WONDER, it's just harder to find.

ICD coding features both an underlying cause of death, the X40-X44 range you're checking, and a "contributing cause" code which attempts to specify the substance involved. In this case, that's the T40.0-T40.6 range. The government guidelines on using WONDER to study opioids include "Please note: X and Y codes must be used in combination with T codes to identify opioid-related deaths."

That's because X42 only covers certain narcotic opioids like heroin, so an X44 opioid overdose is a possibility. Since that's "Accidental poisoning by and exposure to other and unspecified drugs", it could accompany T40.2 ("other opioids") or T40.6 ("unspecified narcotics"). And if the coroner doesn't record anything beyond "accidental drug overdose", then even heroin overdoses will result in X44, T50.9 ("unspecified drugs"). That last case alone apparently describes 25% of all US overdoses.

There's no slight of hand here, just a confusing two-layered classification system.

https://www.samhsa.gov/capt/sites/default/files/capt_resourc...

edit: here's a CDC page listing which primary and contributing causes can go together. Both prescription and illicit opioid deaths can be be filed as X44. https://www.cdc.gov/drugoverdose/pdf/pdo_guide_to_icd-9-cm_a...


Thanks. Any insight into how they justify grouping in unknown poisonings with those due to common medications?


...poorly? The ICD-10 has been criticized quite a lot for where it splits categories; arcane problems get specific codes, while common issues are both fuzzily spread across buckets and conflated within overlarge buckets.

That said, I think the specific rationale is that X44 is meant to hold all types of low-frequency poisoning by drugs, and 'unknown drug' usually means "not one of the notable drugs that gets its own code and is tested for". This usually works alright. X45-X49 handle cases like "alcohol poisoning", "tainted food", "inhaled chlorine fumes", and "other/unspecified chemicals", so X44 is only representing drugs. And "poisoning" is separate from "adverse effects" (Y40-Y84), so an allergic reaction or a standard risk like bleeding from warfarin won't get mixed in here. Given that, X44 is just a reasonably narrow category to assign a group of T## codes to. Those go into enormous detail, and have secondary "unknown" categories to handle the case where you know the class of drug but not the specific drug. Totally unknown poisonings are restricted to X44/T50.9 and should be extremely rare.

But this was all designed by 1990; it looked reasonable to divide both X and T categories by "non-opioid analgesic", "anesthetics", and "narcotics". It wasn't until we started widely prescribing strong opioids for pain that "painkillers plus heroin" became a key grouping you'd want to research. That, plus the international nature of the system, also left us with "heroin", "opium", and "other opioids" as the only 3 T codes for this topic. The listed examples for 'other' are morphine and its prodrug codeine; as a result we've got fentanyl and morphine in the same category despite a 100x difference in strength.

Compounding that, the narcotic poisoning (X42/T40) grouping was basically doomed from the beginning through no fault of the designers. It tries to replicate the legal/treaty category of narcotics, which has no biological coherence; what on earth do cocaine and morphine overdose have in common? Why is "cannabis derivative poisoning" singled out in the 'narcotics' section of a document created when the only synthetic cannabinoid in use was a rare antiemetic no one abused?

tl;dr: It was supposed to be a low-frequency grouping with subcategories clarifying about specific drugs, but that fell apart for opioids.


Muscle relaxers, certain types of steroids, GHB, NO2, and even cocaine are all technically part of that list.

Your not wrong, but it's not that super clear cut either.


Yes, they are grouping together the (lets call them) "DARE drugs" with "medical drugs".

Presumably this is strategic to hide whatever issue is going on with the "medical drugs" and inflate the apparent problem with "DARE drugs".


This is a good point, but it's not just that; X44 includes opioid deaths!

X42 governs accidental narcotic poisoning, but only in cases where the drug is specified, and isn't necessarily applied for prescription opioids. The list above for X44 is the 'including' list of suggestions, but the category includes any death from "unspecified drugs". And contributing causes of death are handled under a separate system making it possible to list more drugs in more detail, so the CDC says that both prescription and illicit opioid deaths can crop up under X44.

https://www.cdc.gov/drugoverdose/pdf/pdo_guide_to_icd-9-cm_a...


>or do they try to work out which one killed them?

Yes, they put their best medical experts to the task...

https://www.youtube.com/watch?v=v7acD4q0lp0


Is this per 100,000 deaths? In that case, wouldn't lower rates of traffic accidents and lung cancer contribute to higher relative deaths of drug-related deaths?


This is an excellent question, but my first impression is that they aren't doing anything sneaky.

I see that some of the data is labeled "per 100,000 individuals," not deaths.

Obviously, there is a specific policy push and point of view outlined in the paper, which is probably why most such papers are published.

But a lot of this data is from the CDC and should be independently verifiable and from Appendix B of the paper, I see that the ultimate source is something called "CDC Wonder".

https://wonder.cdc.gov/

And it looks like one can make a request for that data: https://wonder.cdc.gov/controller/datarequest/D77


It's "crude rate per 100,000", which is the number of people in a sample of 100,000 who would die of that cause in a given year. It's basically a death rate multiplied by 100,000 to be more readable. If you check the CDC WONDER stats, the total annual death rate is ~840/100,000.

I suppose there is some replacement effect in here, where if fewer people die in car crashes they're more likely to live long enough to die of Alzheimers. That's part of why the death rate has been climbing; the population is aging overall. To handle that, there are also "age-adjusted" death rates available, which standardize years of life lost. That can be a big difference; for instance the UK suicide rate is much lower than the US one, but also skews much younger so the age-adjusted rates are closer.


Have a read off https://www.economist.com/international/2018/11/24/suicide-i.... The causes are wide and varied and can be specific to certain countries and cultures.


Did they control for age? We have an unprecedented number of older people now, too.


They do break it out by age as well.

snipped to imgur: https://i.imgur.com/yLTwL3J.png

Edit: my original comment may have implied that I was minimizing the importance of the issue, which I hope wasn't the impression I gave -- it's striking that the only good news whatsoever is a small reduction in alcohol deaths among black people, which was canceled out and then some in drug deaths.


> across multiple countries

It's really difficult to compare suicide rates across countries, especially across time, because how they count suicide is different in different countries and changes across time.


Yeah, but I'll bet it's high-quality and comparable enough among some countries to be useful.

Tracking these three causes together would probably help as well; if one culture is biased towards over-counting overdoses, for instance, and under-counting suicides (which makes me think of the Ray Bradbury line, that someone had "jumped off a pill bottle"), tracking stats for a few proximate causes probably helps get a better overall picture.


Yep, suicide comparisons between developed countries are very possible - you just can't use headline reported numbers.

I had occasion to look into this one recently, actually. In the UK, for example, undetermined-intent drug and injury deaths among adults are included in suicide statistics by default. (That's based largely on the logic that intent is hard to show and false positives are quite painful, while determination of an accident is comparatively easy and harmless.) The US doesn't include those deaths as suicides, but with standardized ICD codes we can easily recover the data. Adding them in raises the US suicide rate about 11%.

Demography is also available, and pretty significant. The US suicide spike is overwhelmingly in adults, so while the US suicide rate is 66% higher than the UK (without definition correction), the age-adjusted death rate is 40% higher.

And you're right about the mixed causes, too. There are probably a nontrivial number of suicides which are explicitly ruled 'accidental', but stats like "overdoses plus suicides" do a lot of work correcting for that. Even for laypeople it's enough to start with, and then actual researchers of course go into specific data on traffic deaths, drownings, and so on.


But even in the UK you have variation across years.

For one example - the burden of proof used by coroners changed in 2018. This means it's going to be a bit tricky to compare deaths before and after this date, and there's going to be variation as this information is applied by coroners as they get used to the new level.

https://www.bailii.org/ew/cases/EWHC/Admin/2018/1955.html

And the coroners numbers do affect the ONS data.

The ONS describe the difficulties of comparing data across nations here: https://www.ons.gov.uk/peoplepopulationandcommunity/birthsde...




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: