Sunday, October 25, 2026

The AI Psych Brief

🔬 Section 1

Psychiatric Research Article

Long-term health, criminal justice, and socioeconomic outcomes among people subject to compulsory mental health care in Norway: a nationwide registry cohort study

Hofstad T, Nyttingnes O, Blindheim AA, Håheim ES, Aarre TF, Rugkåsa J, Bjelland I, Mykletun A · The Lancet Psychiatry · November 2026

Norwegian investigators linked national registries for every adult aged 18–65 with specialist mental health contact in 2015–16 and followed them to 2021–23, comparing five mutually exclusive groups: community treatment orders (CTO, n=3,331), compulsory hospitalisation, compulsory observation only, voluntary care (n=177,396), and an age- and sex-matched general-population sample (total 321,708). All three compulsory groups had substantially worse long-term outcomes across health, justice, and socioeconomic domains than voluntary patients or the general population.

The surprise was which compulsory group fared worst on acute risk: people held only for compulsory observation — assessed involuntarily but never formally committed — had the highest adjusted probabilities of self-harm or suicide (12% of women, 5.5% of men) and of substance use disorder (39% of women, 53% of men). Suicide and other external causes accounted for 56% of deaths in that group, versus 38% in voluntary care and 19% in the general population. The CTO group instead stood out socioeconomically, with roughly 90% receiving disability benefits.

💡 Why it matters

A patient released after an involuntary evaluation without commitment is not a low-risk patient — this group carried the heaviest burden of later suicide, self-harm and substance use, so discharge from an observation hold deserves the same safety planning and follow-up you would give a committed patient. Legal status is a marker of distinct prognoses, not one homogeneous 'involuntary' population.

Read the paper → doi:10.1016/S2215-0366(26)00275-0

🩺 Section 2

General Medicine Article

Benzodiazepine Use Early in Pregnancy and Risk for Congenital Malformations

Fung K, Sujan AC, Bateman BT, Straub L, Hernandez-Diaz S, Brill G, Zhu Y, Cohen LS, Gray KJ, Huybrechts KF · Annals of Internal Medicine · October 2026

Using two large US claims databases — Medicaid (2000–2018) and MarketScan (2003–2020) — this cohort compared pregnancies with at least one first-trimester benzodiazepine fill (51,339 of 2.48 million publicly insured and 27,779 of 1.75 million commercially insured pregnancies) against unexposed pregnancies, using propensity-score weighting to handle confounding by the underlying anxiety, insomnia, and co-prescribing. A relative risk of 1.3 or higher was prespecified as clinically meaningful.

Adjusted absolute risks of any major malformation were essentially identical: 4.0% vs 4.1% in Medicaid and 4.9% vs 4.8% in MarketScan, for a pooled RR of 0.99 (95% CI 0.96–1.03). Organ-specific signals, including the historically cited oral clefts, did not hold up consistently across sensitivity analyses, there was no dose-response, and individual benzodiazepines looked similar. Some co-exposure analyses (benzodiazepines plus other psychotropics) showed scattered elevations without a consistent pattern.

💡 Why it matters

When a patient discovers she is pregnant while taking a benzodiazepine, this large dataset argues against abrupt discontinuation driven by malformation fear — first-trimester exposure did not raise major-malformation risk. The decision should instead rest on the indication, dependence and withdrawal risk, and late-pregnancy concerns such as neonatal sedation and withdrawal, which this study did not address.

Read the paper → doi:10.7326/ANNALS-25-05148

💊 Section 3

Psychiatric Fact

Desvenlafaxine sidesteps CYP2D6 but hands the dosing problem to the kidney

Venlafaxine depends on CYP2D6 to become O-desmethylvenlafaxine, so poor metabolizers and patients on fluoxetine, paroxetine or bupropion carry more parent drug. Desvenlafaxine is that metabolite: the label states the CYP2D6 pathway is not involved, and its pharmacokinetics are similar in poor and extensive metabolizers. It is cleared mostly by UGT conjugation, with a minor CYP3A4 route, and only 30% protein-bound — a cleaner choice when a 2D6 inhibitor has to stay on board.

The trade-off is renal: about 45% is excreted unchanged in urine. The label caps the dose at 50 mg/day for CrCl 30–50 mL/min and at 25 mg daily or 50 mg every other day below 30 mL/min or on dialysis, with no supplemental dose after dialysis. And because doses above 50 mg/day showed no added benefit in trials while adverse effects and dropouts rose, 50 mg is the dose, not a starting point. It is still a weak 2D6 inhibitor at high doses: the label suggests halving 2D6 substrates such as metoprolol or atomoxetine only at 400 mg.

🛋️ Section 4

Psychotherapy Teaching Pearl

In MBT, when mentalizing collapses, stop and rewind to the moment before it broke

Mentalization-based treatment assumes that in high arousal a patient with borderline organization loses the capacity to hold minds in mind, and further interpretation then lands on nothing — it is heard in psychic equivalence as accusation or confirmation. The technical response is stop, rewind, explore: halt the current exchange, return to the last point in the session where both people were still thinking, and examine, slowly and in detail, what shifted between then and now. The therapist's own contribution is part of the material; acknowledging it ("I think I said something there that landed hard") lowers arousal and models curiosity about minds, including the therapist's. The aim is not insight into content but restoring the capacity to think at all, so the therapist stays in the present, keeps interventions short, and resists explaining. Rewinding treats the rupture as an event that happened between two people in the room, which is exactly the frame a hypomentalizing patient cannot yet hold alone.

🗒️ Vignette

A 24-year-old woman is describing a weekend argument with her sister when the resident asks, "What did you think she was feeling?" She goes flat, then sharp: "So it's my fault. You're on her side like everyone else." The resident's impulse is to clarify that he was not blaming her — a defense of his intention that would keep the argument going on her terms. Instead he says, "Let's stop for a second. A minute ago we were figuring out what happened on Saturday together. Then I asked about your sister, and something changed. Can we go back to just before my question?" She says that when he asked, she pictured him siding with her sister. He replies, "That makes sense of why it felt like an attack — I may have jumped to her side of it too quickly." Her shoulders drop. Only then does he return, more slowly, to the sister. The move was to explore the break in the room before resuming the story outside it.