Wednesday, October 14, 2026

The AI Psych Brief

🔬 Section 1

Psychiatric Research Article

Attention-Deficit/Hyperactivity Disorder and Late-Life Cognitive Impairment

Bauer-Negrini G, Leffa DT, Ferreira PCL, Bellaver B, Povala G, Medeiros MS, Lussier FZ, Tudorascu DL, Cortese S, Rohde LA, Molina BSG, Pascoal TA · JAMA Psychiatry · October 7, 2026 (online)

This retrospective cohort study used the NIH All of Us Research Program to follow 187,341 adults aged 50 and older, linking electronic health records to genotype data. The investigators asked two questions: does a clinical ADHD diagnosis predict new mild cognitive impairment or dementia, and does a higher ADHD polygenic risk score predict it even in people never diagnosed?

A clinical ADHD diagnosis carried a hazard ratio of 4.60 (95% CI, 3.99–5.29) for incident cognitive impairment, which fell to 3.19 (2.76–3.68) after adjustment for depression, anxiety, and other neuropsychiatric comorbidity. Participants with the highest versus lowest ADHD polygenic risk had a hazard ratio of 1.35 (1.25–1.46). Among people who developed cognitive impairment, those with ADHD reported worse mental health, quality of life, and social satisfaction, and used about 20% more health care.

The design cannot establish causation: reverse causation through a dementia prodrome, diagnostic misclassification in EHR data, and a volunteer cohort are all real limits, and ADHD medication exposure was not examined.

💡 Why it matters

When an older adult with lifelong ADHD presents with cognitive complaints, do not reflexively attribute them to the ADHD — this cohort supports a lower threshold for formal cognitive screening and longitudinal monitoring in this group. Equally, taking a careful childhood history matters when a 'new' attention problem appears after 50, because the distinction changes both workup and treatment.

Read the paper → doi:10.1001/jamapsychiatry.2026.3165

📖 Section 2

Psychotherapy Research Article

Mapping negative experiences in psychotherapy: A network analysis

Bobei T, Flückiger C, Podina IR · Psychotherapy Research · October 6, 2026 (online)

Most outcome research asks whether therapy helped; this study asks how the ways it can hurt hang together. The authors gave the Negative Effects Questionnaire to 439 adult outpatients currently or previously in psychotherapy and estimated a Gaussian graphical network linking its domains — symptom deterioration, hopelessness, negative self-evaluation, dissatisfaction with the treatment process, dependency on the therapist, and stigma.

The network organized around a single hub: hopelessness was the most central node, directly connected to symptom worsening, negative self-evaluation, and negative appraisals of the therapy itself. Dependency and stigma sat at the periphery. The core cluster held its shape across exploratory subgroup comparisons.

The data are cross-sectional and self-reported, so centrality shows connection, not causal direction.

💡 Why it matters

Hopelessness about the treatment is the node that ties a patient's sense of getting worse to their sense that therapy is failing them, which makes it worth asking about directly rather than inferring from symptom scores. A patient who says 'this isn't going anywhere' is giving you the most connected piece of a negative-effects picture, and it belongs on the table in the session, not just in your notes.

Read the paper → doi:10.1080/10503307.2026.2743645

💊 Section 3

Psychiatric Fact

Ziprasidone and lurasidone are dosed with a meal, not just at bedtime

Two antipsychotics have food requirements that work like a dose change. Ziprasidone needs a meal of roughly 500 kcal or more: without it absorption falls by about half, and Lincoln, Stewart and Preskorn (2010) showed that calories, not fat content, drive the effect, so a light snack does not count. Lurasidone needs only 350 kcal: in the label's food-effect study, Cmax rose about threefold and AUC about twofold with food, and exposure did not climb further between 350 and 1,000 kcal or with higher fat.

The practical point is that 'nonresponse' on either drug is often an exposure problem. Before calling ziprasidone or lurasidone a failed trial or raising the dose, ask what the patient eats with it — a bedtime pill taken on an empty stomach, a skipped breakfast, or an inpatient whose evening dose lands after dinner trays are cleared. Schedule the dose with the main meal of the day, and document it, because a patient who starts eating with the dose has effectively had their dose doubled.

🛋️ Section 4

Psychotherapy Teaching Pearl

The doorknob disclosure belongs to the next session, not the next ten minutes

Material raised as the patient stands to leave — a suicidal thought, an affair, 'I've been drinking again' — is not a scheduling accident. The timing is part of the communication: the patient has found a way to say the thing while guaranteeing there is no room to explore it, and often to test whether you will break the frame to rescue them. The two errors are mirror images. Extending the session gratifies the wish to be held without examining it, and teaches that the boundary bends under pressure. Ignoring it leaves the patient alone with something they could only half-say. The technical move is to acknowledge, assess only what safety requires, and hold the frame — then make both the content and its timing the first item of the next session. Safety is the one exception: if the disclosure carries acute risk, that assessment happens now.

🗒️ Vignette

A 34-year-old man in weekly psychodynamic therapy says, hand on the door, 'Oh — I started looking at apartments without telling my wife.' The resident feels the pull to sit back down. Instead: 'That sounds important, and I don't want to rush it in the doorway. Let's start there next week.' After a brief check that nothing is unsafe, the session ends on time. The following week the patient doesn't raise it. The resident does: 'Last week you told me about the apartments just as you were leaving. I wonder what made the doorway the safest place to say it.' The patient laughs, then goes quiet: 'If I say it in here, it's real.' The timing becomes the doorway into ambivalence — the content alone would have produced only a discussion of logistics.