Intra-rater Kappa Accuracy of Prototype and ICD-10 Operational Criteria-Based Diagnoses for Mental Disorders: A Brief Report of a Cross-Sectional Study in an Outpatient Setting

Abstract

Background and Objectives The use of “operational criteria” is a solution for low reliability, contrasting with a prototypical classification that is used in clinics. We aim to measure the reliability of prototypical and ICD-10 diagnoses. Methods This is a retrospective study, with a convenience sample of subjects treated in a university clinic. Residents reviewed their diagnosis using ICD-10 criteria, and Cohen's kappa statistic was performed on operational and prototype diagnoses. Results Three out of 30 residents participated, reviewing 146 subjects under their care. Diagnoses were grouped in eight classes: organic (diagnoses from F00 to F09), substance disorders (F10–F19), schizophrenia spectrum disorders (F20–F29), bipolar affective disorder (F30, F31, F34.0, F38.1), depression (F32, F33), anxiety-related disorders (F40–F49), personality disorders (F60–F69), and neurodevelopmental disorders (F70–F99). Overall, agreement was high [K = 0.77, 95% confidence interval (CI) = 0.69–0.85], with a lower agreement related to personality disorders (K = 0.58, 95% CI = 0.38–0.76) and higher with schizophrenia spectrum disorders (K = 0.91, 95% CI = 0.82–0.99). Discussion Use of ICD-10 criteria did not significantly increase the number of diagnoses. It changed few diagnoses, implying that operational criteria were irrelevant to clinical opinion. This suggests that reliability among interviewers is more related to information gathering than diagnostic definitions. Also, it suggests an incorporation of diagnostic criteria according to training, which then became part of the clinician's prototypes. Residents should be trained in the use of diagnostic categories, but presence/absence checking is not needed to achieve operational compatible diagnoses.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 9 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.