DSM: who wrote it, and with what ties

The 69% — and why it does not mean an increase

Lisa Cosgrove and Sheldon Krimsky reported that 69% of the 29 members of the DSM-5 task force — about 20 people — disclosed financial ties to the pharmaceutical industry under the APA's new mandatory disclosure policy. That policy did not prohibit ties: it capped them at US$10,000 a year in industry income and US$50,000 in pharmaceutical stock.

By work group: Mood Disorders 67% (N=12), Psychotic Disorders 83% (N=12), Sleep/Wake Disorders 100% (N=7).

What it does not establish. Because DSM-IV carried no mandatory disclosure and its figure was reconstructed by outside screening, the comparison between the two editions cannot establish that ties increased. APA President John Oldham raised exactly this objection in March 2012 and stated that more than 70% of DSM-5 work-group members disclosed no industry ties.

[22] Cosgrove L, Krimsky S. PLoS Med 2012;9(3):e1001190, DOI 10.1371/journal.pmed.1001190

Before that: 95 of 170, and two panels in their entirety

In 2006, Cosgrove, Krimsky, Vijayaraghavan and Schneider screened publicly available records and found that 95 of 170 members of the DSM-IV and DSM-IV-TR panels (56%) had at least one financial association with industry. Research funding 42%, consultancies 22%, speakers' bureaux 16%.

Every member of the two panels covering mood disorders and psychotic disorders had at least one such tie. Both panels were small — we will not publish that as a percentage until the exact membership counts are read from the paper's own table.

The study establishes association, not causation. Its authors could not determine whether any tie preceded or followed an appointment. The question has not closed: in 2024 the BMJ published an analysis of undisclosed conflicts of interest in DSM-5-TR.

[21] Cosgrove L et al., Psychother Psychosom 2006;75(3):154-160, DOI 10.1159/000091772 · [23] Davis LC et al., BMJ 2024;384:e076902

How often two clinicians agree

The DSM-5 field trials tested how often two clinicians, interviewing the same patient independently on separate occasions in ordinary clinical settings, reached the same diagnosis. Of 23 diagnoses with adequate samples — 15 adult and 8 child/adolescent — the investigators rated, on their own published scale: 5 very good (kappa 0.60–0.79), 9 good (0.40–0.59), 6 questionable (0.20–0.39), 3 unacceptable (below 0.20). Eight further diagnoses had samples too small for a precise estimate.

We do not quote specific kappa values: the table containing them sits behind a barrier we could not pass, and we do not publish numbers we have not read.

And the standard those numbers were judged against was written a year earlier. In January 2012, Kraemer, Kupfer, Clarke, Narrow and Regier published "DSM-5: How Reliable Is Reliable Enough?" in the same journal. Verbatim: «to see kappa for a DSM-5 diagnosis above 0.8 would be almost miraculous; to see kappa between 0.6 and 0.8 would be cause for celebration… a realistic goal is kappa between 0.4 and 0.6, while kappa between 0.2 and 0.4 would be acceptable».

That is a lowering of the conventional threshold, on which kappa below 0.4 counted as poor. Three of its five authors — Clarke, Narrow and Regier — went on to co-author the results paper judged by that standard. That it appeared before the results means it was a pre-specified criterion, not a post-hoc apology. Both things are true at once.

And "kappa 0.28 means a coin flip" is wrong. Kappa is not a percentage of agreement; it does not translate into "50%". Anyone who says it that way has not read the definition.

[24] Regier DA et al., Am J Psychiatry 2013;170(1):59-70 · [25] Kraemer HC et al., editorial, January 2012

The diagnosis that was deleted by a vote

In December 1973 the Board of Trustees of the American Psychiatric Association voted to delete homosexuality from DSM-II and substitute a new category, "sexual orientation disturbance". The proposal was written by Robert L. Spitzer of the Task Force on Nomenclature and Statistics.

Opponents, largely psychoanalysts, petitioned for a referendum of the membership. It upheld the decision: 5,854 in favour, 3,810 against, 367 abstentions — out of 17,910 eligible voters. About ten thousand voted, and the winning side amounted to roughly a third of the eligible membership.

Where these numbers come from, and what is missing. From two contemporaneous New York Times reports, of 9 April and 26 May 1974, quoting the APA's own announcement. The APA's own ballot record was not obtained. On the month, the APA's own materials disagree — the seventh printing of DSM-II says May 1974, its foundation's archive says February; the 9 April report describes the vote as already concluded, which places the count in April.

And what it does not prove. It does not show that psychiatric classification is generally decided by ballot: the referendum was an exceptional measure, forced by petition after a Board decision. It does show that a diagnosis existed and then stopped existing without new data, and that the profession itself split roughly 60/40 among those who voted.

[61] "Psychiatrists Approve Change on Homosexuals", The New York Times, 09.04.1974 · [62] "8 Psychiatrists Are Seeking New Vote on Homosexuality as Mental Illness", The New York Times, 26.05.1974 — contemporaneous journalism about the APA's announcement, not the ballot record

When grief became diagnosable

DSM-IV barred a diagnosis of major depressive episode where symptoms lasted less than two months after the death of a loved one. DSM-5 dropped that rule and replaced it with a footnote directing clinicians to distinguish: in grief, painful feelings arrive in waves intermixed with positive memories and self-esteem is usually preserved; in depression, mood is near-constantly negative, accompanied by worthlessness.

The APA's stated reasons — quoted because they are reasonable: two months wrongly implied that bereavement lasts two months, when the usual duration is one to two years; and bereavement-related depression resembles other depression in family history, genetic influence, comorbidity, chronicity and treatment response.

The change remains among the most disputed in DSM-5. It is presented here as a decision with a stated rationale, not as a scandal.

[26] American Psychiatric Association, "Highlights of Changes from DSM-IV-TR to DSM-5", Bereavement Exclusion section

And a correction: most countries do not code on the DSM

Neither Greece nor Italy uses the DSM as the official instrument for diagnostic coding or reimbursement. Both use the ICD family, as does essentially every national health system.

Italy still codes hospital discharges and DRG reimbursement on the Italian translation of ICD-9-CM, 2007 version, in force since 1 January 2009. Greece established ICD-10-GrM by ministerial decision in 2018, initially only for hospitals piloting the Sy.K.N.Y. costing system.

The Seventy-second World Health Assembly adopted ICD-11 in May 2019, to come into effect on 1 January 2022 — but the same resolution provides transitional arrangements for at least five years, so that member states can keep reporting statistics under previous revisions.

What this means in practice: the diagnosis recorded in your medical notes is written as an ICD code, not a DSM one, in almost every country. Criticism of how the DSM was compiled — who sat on the panels, with what ties — does not automatically carry over to it.

[29] WHA72.15, 28.05.2019 · [30] Greek ministerial decision establishing ICD-10-GrM (2018)