Diagnostic interviews – the commonest method to diagnose substance use and psychological problems together with melancholy, nervousness, bipolar and character problems – differ in reliability from situation to situation, in line with a brand new examine in Jama Community Open.
Laura Duncan, a psychiatry professor at McMaster College in Ontario, Canada, and one of many examine’s authors, mentioned diagnostic interviews are “typically handled as a ‘gold normal’ for assessing psychological problems in each medical settings and analysis”, however identified that these interviews fall in need of offering a “definitive benchmark that demonstrates wonderful validity and reliability”.
Although proof on the reliability of those interviews has lengthy been blended, “they proceed to be extensively seen as the very best accessible strategy, presumably as a result of lack of higher alternate options,” Duncan mentioned. The overview examine brings collectively proof from research on “test-retest reliability” of diagnostic interviews reviewed from February 2024 to September 2025.
The examine’s authors used Cohen’s kappa coefficient to estimate how dependable diagnostic interviews have been for various psychological well being circumstances; this allowed them to see how typically sufferers would obtain the identical prognosis when given the identical diagnostic interview twice, whereas accounting for the truth that generally this could occur by luck.
The typical reliability was typically higher for substance use problems, and highest general for opioid use dysfunction. Duncan mentioned this was as a result of substance use dysfunction standards are largely primarily based on conduct. For example, it’s typically simpler to estimate what number of drinks you had in every week, than the variety of days you felt unhappy or anxious.
Dr Michael First, a psychiatrist and professor at Columbia College who authored the Structured Medical Interview for DSM 5 (SCID), was annoyed with components of the examine. Whereas he agreed that diagnostic interviews differ in reliability and too typically fail to appropriately diagnose folks, he wished to see extra details about which particular devices have been most dependable.
“It’d be good to have the ability to take a look at this and say: ‘Oh, primarily based upon this paper, I ought to choose this one due to this.’ That may be doing the sphere an actual service,” he mentioned. “However there’s merely not sufficient data right here.” Duncan mentioned that the data within the examine was primarily based on the restricted quantity of related analysis accessible throughout the examine interval.
The overview included papers on diagnostic instruments together with the SCID, which First authored, in addition to Mini Worldwide Neuropsychiatric Interview (Mini), each which display screen for a number of psychological well being circumstances – in addition to instruments meant for particular problems, just like the Clinically Administered PTSD Scale (Caps.)
First additionally took subject with how the examine lumped “totally structured”, and “semi-structured” interviews collectively. Totally structured interviews usually tend to yield the identical end result when administered greater than as soon as, “since you stick with the script and can’t deviate from it in any respect”, First famous.
“If the individual says one thing contradictory, you’re not allowed to even level out that it’s contradictory,” First mentioned. The sort of interview is usually used for epidemiological analysis on giant populations, and is due to this fact designed for folks with little coaching to manage.
Semi-structured interviews, however, are designed for educated clinicians to diagnose sufferers. With any such interview, clinicians have the liberty to “ad-lib their questions as wanted”, First mentioned. This implies if a affected person’s reply is imprecise or contradictory, their supplier is ready to ask follow-up inquiries to make clear. That permits for extra correct prognosis, however the affected person’s solutions additionally may differ extra from session to session.
Whereas Duncan famous that it could be helpful to handle all of First’s considerations, she mentioned the information she would wish to take action merely doesn’t exist but. Within the papers her examine included, Duncan mentioned they “tried to extract data on interview format, however this was typically unclear or not reported”. The shortage of accessible data needed to match completely different interview designs one after the other is one other signal of the necessity for extra rigor in relation to psychiatric prognosis.
Although he helps design them, First readily admits structured interviews are lower than ideally suited instruments. For many years, psychiatrists have been hoping at some point extra goal laboratory exams will grow to be accessible for psychological circumstances, he mentioned.
“We’ve been saying that for 50 years,” First mentioned. Duncan pointed to another future strategy the place clinicians “transfer away from strict diagnostic classes, the place a situation is both current or absent, and take into consideration signs on a spectrum or continuum”.
