Study: Reliability of mental health diagnoses is only moderate

An international study shows that diagnosing mental health disorders using standardised interviews is not as reliable as previously thought. The accuracy of diagnoses varies significantly depending on the condition and the clinician, calling into question current diagnostic methods.

Results from an international study challenge one of psychiatry's cornerstones: standardised diagnostic interviews, which doctors worldwide use to identify mental health disorders. The study found that these methods have only moderate reliability, and the agreement between clinicians on diagnoses varies considerably depending on which disorder is being assessed.

Why do results differ?

Standardised interviews were developed to reduce subjectivity and ensure that different clinicians would reach the same conclusion for the same patient. In practice, however, how the clinician interprets the questions, how the patient answers them, and how clearly defined the diagnostic criteria for a specific disorder are all play an important role. For some disorders, the level of agreement between clinicians is significantly higher than for others.

Impact on patients and treatment

Diagnostic inaccuracy can have a direct impact on patients' treatment journey. When two clinicians give the same person different diagnoses, it can lead to inappropriate treatment, which in turn may worsen the patient's condition. Researchers emphasised that existing tools should be improved and clinician training enhanced to reduce interpretative differences.

According to experts, this does not mean that standardised interviews are useless, they remain the best tool currently available. However, their results should be interpreted cautiously and supported by other clinical data, not solely on the basis of the interview score. The study calls for the development of additional diagnostic tools, including biomarkers and digital assessment instruments, which could help make diagnoses more accurate in the future.