The AI Psych Brief

GLP-1 receptor agonist initiation and risk of psychiatric hospitalisation and emergency department presentation in adults with schizophrenia and type 2 diabetes in the USA: an observational cohort study using electronic health records data

Davoodian N, Williams LJ, Berk M, Arnautovska U, Kim JH, McLure J, Panizzutti B, Rossell SL, Siskind D, Taipale H, Taquet M, Upthegrove R, Walder K, Yung AR, Mohebbi M · The Lancet Psychiatry · November 2026

Using US psychiatric electronic health records (NeuroBlu), the authors identified 70,257 adults with schizophrenia and type 2 diabetes. They compared people who started a GLP-1 receptor agonist with non-initiators and with people starting a DPP-4 or SGLT2 inhibitor. Inverse probability weighting balanced demographic, clinical, and medication covariates, and analysis was intention-to-treat over up to 2 years.

GLP-1 RA initiation was associated with about 24 percent lower risk of psychiatric hospitalisation than non-initiation (HR 0.76), and about 17–18 percent lower risk than DPP-4 or SGLT2 initiation. Psychiatric emergency department presentations were also lower (HR 0.71 against non-initiation). The active-comparator design matters: the benefit held against other diabetes drugs, not just against no treatment.

Read the paper →  doi:10.1016/S2215-0366(26)00276-2

Earlier Disclosure of Blood-Based Biomarkers, Diagnostic Certainty, and Clinical Management: A Randomized Clinical Trial

Estragués-Gázquez I, Contador J, Martínez LD, et al; Suárez-Calvet M · JAMA Neurology · October 2026

At a Barcelona memory clinic, 220 new outpatients with subjective cognitive decline, mild cognitive impairment, or mild dementia were randomized to learn their plasma p-tau217 and neurofilament light results at 3 months or at 9 months. Both patients and neurologists received the results.

At 3 months, 50 percent of the early-disclosure group had a very high-confidence etiologic diagnosis, compared with 4.6 percent of the delayed group. Early disclosure also led to more symptomatic Alzheimer treatment, fewer repeat neuropsychological evaluations, and more discharges back to primary care. Most relevant for us: anxiety, depression, stress, and quality of life were no worse with earlier disclosure.

Read the paper →  doi:10.1001/jamaneurol.2026.3497

A once-daily lithium dose reads higher at 12 hours than the same dose split

The 0.6–1.2 mmol/L reference range was built largely on 12-hour trough levels from divided dosing. With twice-daily dosing, a level drawn 12 hours after the evening dose really is a trough. With a single bedtime dose, a level drawn 12 hours later still sits on the descending limb, closer to the peak, so the same total daily dose gives a 12-hour level roughly 10 to 25 percent higher.

That has two practical effects. First, when consolidating a patient from twice-daily to once-daily dosing, expect the reported level to rise even though exposure has not, and do not cut the dose just to get back to the old number. Second, a once-daily patient at 0.7 mmol/L may have similar effective exposure to a divided-dose patient at a slightly lower trough. Once-daily dosing is generally preferred anyway, because the daily trough gives the kidney time each day without high lithium exposure, and it is linked to less polyuria. Always document the dosing schedule next to the level so whoever reads it next interprets it correctly.

In supportive therapy, leave a defense alone if it is still working

Expressive therapy treats defenses as obstacles to insight. Supportive therapy asks a different question: is this defense adaptive right now? Denial, suppression, humor, and intellectualization can protect a patient who is facing an overwhelming reality. They become a target only when they block necessary action, such as refusing treatment, driving unsafely, or skipping planning that cannot wait. Confronting a defense that is still doing its job removes the patient's protection without giving them anything else to hold onto.

So the supportive clinician separates the patient's stated belief from their behavior. If the patient says "it's nothing" but keeps every appointment, the defense is regulating affect and is not interfering with care. Support it, and use anticipatory guidance to prepare for what is coming without forcing the patient to agree with the diagnosis first.