|
Abstract
|
Aims To promote clinical research in child and adolescent psychiatry (CAP), highlighting feasibility and an accurate extraction and categorization of diagnostic formulations in chart review studies; to discuss a methodological tool for extracting diagnoses in retrospective research, its limitations and potential technological improvements. Methods A narrative review of international literature revealed limited information on specific methods for extracting and categorizing psychiatric diagnoses in CAP. A step-by-step extraction algorithm was developed based on the emergency (E) and inpatient (H) units’ files of 30 patients. The first three diagnoses were retained. Sixteen categories were framed in a specific diagnosis guide, encompassing all CAP’s DSM-5 diagnoses. Six child and adolescent psychiatrists tested the algorithm with inter-rater agreement kappa’s Cohen (k) measured at E and H in two training sessions. Results and conclusion For the principal reported diagnosis the k was of 0.8. Controversies resolution consolidated the adjustment of the guide for the specific divergent items concerning mainly the second and third diagnoses. Variability of the verbatim formulations of diagnosis hypothesis exceeding the DSM-5 frame, as well as the lack of a clear or uniform hierarchy of the diagnoses were the main challenges to the algorithm’s process. The finalized guide will be a methodological objective tool in an ongoing retrospective research project. We will discuss the role of technology, particularly natural language processing tools, which could significantly support the retrospective research in CAP.
|
