The 1893 Ganja Asylum Claim and the CUD Coding Problem

Photo by Strange Happenings via Pexels.

In 1892 a newspaper in Allahabad ran an item saying that most of the inmates in India's lunatic asylums were ganja users. It was a short piece with a large effect. It reached a member of the British Parliament, and that helped push the government into appointing a seven-member commission to examine hemp across the whole of British India.

Set that beside the recent coverage of VA records raising red flags about cannabis use disorder diagnoses, and the resemblance is hard to miss. In both cases a label has been attached to cannabis, and in both a count of labels is being read as a count of consequences.

My argument is that these are one problem separated by more than a century: an attribution problem. Someone wrote a cause or a diagnosis into a record, and we want to know who, on what basis, and against what comparison. I also think the way the Indian Hemp Drugs Commission handled its asylum records gives us a usable template for reading modern diagnosis data. The commission's verdict on hemp is the part everyone remembers. Its handling of the labelling step is the part worth borrowing.

One newspaper item and a seven-member inquiry

The documented chain is short. An 1892 item in a newspaper in Allahabad made the claim about asylum inmates. An MP took it up in Parliament, and in 1893 the government appointed the Indian Hemp Drugs Commission. The exact wording of the item and the details of the parliamentary exchange are worth pulling from the commission's report and from the Wire's reporting before you quote anyone. The sequence itself is not in dispute.

What strikes me is the shape of the claim. It was a statistic about a population: most of the people in these institutions used ganja. Nobody in that chain, as far as I can tell, stopped to ask who had written ganja into each record as a cause, or on what evidence. The number travelled, and the question behind it did not.

This is a better way into the story than the one we usually tell. When we looked at Bengal's ganja trade and the 1894 inquiry earlier, the frame was taxation, revenue and the machinery of the trade. That frame is accurate, and it explains why a colonial government cared about regulating the drug. It does not explain why an inquiry began. The asylum claim does, because it supplied the alarm.

Consider the stakes for a modern reader. A single recycled figure moved a government to commission an inquiry covering all of British India. Today's equivalents are a line in a press release, a slide in a hearing, a sentence in a ballot-measure argument. They are small, they are portable, and they are rarely traced back to the entry that produced them.

How an asylum register turned a rumour into a cause

An asylum admission register was a ledger, and one of its columns recorded the supposed cause of insanity. That column was filled in at or near admission. The key question is who supplied the entry: police, relatives, the admitting officer or medical staff, and on what evidence. The report and the scholarship should answer that, and I would confirm it there rather than assume it. What I can say is that the entry was made by people close to the moment of admission, not by anyone running a controlled study.

Here the documented and the inferred separate. It is documented that the column existed and that ganja appeared in it. It is an inference, and a large one, that the entry reflected verified causation. A name in a cause column shows that someone believed, or assumed, or was told, that ganja was involved. That is a fact about the person recording it.

The register also could not show several things a reader would want:

• How much ganja the person used.
• For how long, and in what form.
• Whether they used it at all, as opposed to coming from a region or class where ganja was common.

That last gap matters most. If a drug is widely used among poor labourers and ascetics, and those groups are overrepresented among people admitted to an asylum, ganja will turn up in the records whether or not it did anything.

The missing comparison group is the other gap. The claim needs a denominator: how many ganja users were never admitted to any asylum? The PMC paper on the commission and the Strathprints work on cannabis and colonialism both discuss how the commission and later readers argued over this point. I have not confirmed the authors' names, so I will leave them unnamed here. The core of the argument, though, needs no authority. A count of users among the admitted tells you nothing until you know the count of users among everyone else.

What the commission did with the numbers

Photo by Emin Şen via Pexels.

The commission did not simply tally the column. As I read the scholarship, it questioned where each entry came from before counting it, and it separated cases where ganja was only supposed to be the cause from cases where the link held up. This is the strongest parallel to modern coding, but I have not verified it line by line in the primary report. Check the asylum tables there, and quote exact figures and percentages only from the report itself.

I want to present this as method, not conclusion. The method is: do not count a label until you know what produced it. That step is separate from any view about hemp. A commission could apply it and still conclude that the drug was dangerous, or apply it and conclude the opposite.

On the conclusion itself, the commission's wording as quoted in the PMC paper is the thing to use. The commonly cited summary is that moderate use produced no appreciable harm, while excessive use was harmful. I would take the exact phrasing from that paper and keep it apart from my reading of the method, which is an interpretation.

What is new in this reading is the question. The usual framing asks whether the commission was right about hemp. I am asking how it treated the labelling step, because that is what transfers to a clinic or a database today. Whether its hemp verdict was correct is a separate matter, and nobody needs to settle it to borrow the technique.

The limits are real. This was a colonial body with its own incentives, and it worked within a government that had revenue and political reasons to want a particular kind of answer. Its handling of the asylum data is a template of method. It is not a model of neutrality, and I would not hold it up as one.

The same question in today's CUD diagnosis data

Our earlier piece on VA records raising red flags about cannabis use disorder diagnoses looked at what the records showed about how diagnoses were being attached. I will not restate its figures here. Pull the dataset, the period and the numbers from that article directly, and cite them as they appear there. The point I want to take from it is structural: a diagnosis is attached to cannabis in a record, and the record is then counted.

Consider how a diagnosis becomes a data point. A clinician selects a code, perhaps after a screening prompt or a positive test. The entry then sits in the record, often without a documented reason for why that code was chosen. Years later someone counts the entries and reports a number.

My inference, and it is an inference, is that the asylum column and the diagnosis code have something in common. Each is an entry made by an institution at a moment of contact with a person. So a count of entries describes the behaviour of the institution as well as the condition of the patients. Screening practices, coding incentives and what staff think a visit is for all shape the count.

The parallel breaks in one place, and it should be said plainly. Cannabis use disorder is a defined clinical diagnosis with criteria. The asylum entries were loose and unstandardised. So the question is not whether the label has a legitimate definition. It does. The question is whether the criteria were applied and written down in each case.

What would settle it is a chart-review audit. Take a sample of recorded diagnoses, check each against the criteria, and see how many hold. That is a modern version of questioning the source of each entry before counting it.

The strongest objection: heavy use does cause harm

The strongest objection is simple: noisy records do not mean there is no effect. Heavy cannabis use does cause harm for some people. Cannabis use disorder is real, and treating the whole subject as artefact would be wrong. The commission itself, as I recall and as the PMC paper's quotation should confirm, accepted that excessive use was harmful. Verify the wording there before leaning on it.

I concede the central point. A bad record-keeping process proves nothing about the true rate, in either direction. If the asylum entries were unreliable, that does not show ganja never contributed to anyone's illness. If diagnosis coding is sloppy, that does not show the disorder is rare. All it shows is that the number cannot carry the claim someone has hung on it.

That is why the thesis survives. It is a claim about what a statistic can support, not about whether harm exists. A reader can believe that heavy use harms some people and still decline to accept a particular count as evidence of how many.

The same caution applies to the psychosis literature. The PMC source revisits a nineteenth-century study of Indian hemp and insanity, and modern research continues to examine the link between cannabis and psychosis. I would keep any statement about it qualitative and tied to what that paper actually says. The relevant lesson is the same one: the strength of such a link depends on how cause, exposure and comparison were established.

If the thesis fails anywhere, it fails as an excuse. Pointing at weak records can become a way to wave away every unwelcome finding. The fix is to ask for the audit, not to assume the answer.

A three-question test for any cannabis-causes-X statistic

Photo by cottonbro studio via Pexels.

Here is the test in the form I would use it. Three questions, asked in order, of any statistic that says cannabis causes something:

1. Who recorded the cause? A clinician, a police officer, a relative, a survey respondent, or a researcher with a protocol.
2. On what basis? Direct observation, a test, self-report, or assumption.
3. What was the comparison group? Who else used cannabis and did not show up in the count, and how do the two groups differ.

Run the 1892 claim through it in one line each. The recorder was someone near admission, whose identity the register may or may not preserve. The basis is uncertain and may have been assumption. The comparison group is missing: no one counted ganja users outside the asylum.

Now run a count of CUD diagnoses through it. The recorder is documented: a clinician at a facility. The basis is the open question, because a code does not tell you whether criteria were applied. The comparison group depends on how the data are framed, and whether it exists is something you check, not something you assume. Mark which of these you can see and which you are guessing.

Growers and industry readers can use this on a lot of material. Apply it to headlines, to dispensary pamphlets, to studies cited in policy hearings and to ballot campaign literature, whichever side of the argument the material comes from. A claim that flatters your position deserves the same three questions as one that threatens it.

The question is more than a century old. Checking the labelling step was already possible in 1894, and the commission is the evidence for that. We have better tools and no excuse.

Any dataset that counts labels will be argued over, and the argument will usually be about the wrong thing. People pick a side on cannabis first and then read the number to fit. The more useful move is the one the commission made with its asylum records: ask for the audit before choosing.

So the open question is a practical one. If a sample of CUD diagnoses were reviewed against the criteria, what share would hold? And who is willing to look, given that the answer might be inconvenient to almost everyone who has an opinion?

Browse our seed collection.