A Modified Case-Based Reasoning Approach for Triaging Psychiatric Patients Using a Similarity Measure Derived from Orthogonal Vector Projection

摘要

A modified case-based reasoning method is introduced aimed to fulfill the need for a triage tool that differentiates likely psychiatric diagnoses and associated risk level. Clinical cases are represented as a set of clinical features rated on a numerical scale according to level of severity. One standard case is used for each diagnostic category, represented as a vector denoting the expected severity of each clinical feature. A new case represented as another vector denoting the severity of observed clinical features in a patient is assessed against the standard cases. Measurement based on orthogonal vector projection was used as a clinically intuitive measurement of similarity. Using thirty different test cases representing six different diagnostic categories, this measure and alternative similarity measures consisting of cosine similarity and Euclidean distance were evaluated. Results indicated that orthogonal vector projection was superior to the other two methods in differentiating diagnoses and predicting severity.