Classification Algorithm for the Determination of Suicide Attempt and Suicide (CAD-SAS): Development and Psychometric Properties

Izabela E. Fedyszyn, Meredith G. Harris, Jo Robinson, Susan J. Paxton

One methodological difficulty in research into suicide attempts and suicide is distinguishing these phenomena from nonsuicidal self-harming behaviors and accidents. This is problematic because a reliable assessment of the presence or absence of the outcome variable is fundamental for the validity of the findings. To develop a standardized rating system, the Classification Algorithm for the Determination of Suicide Attempt and Suicide (CAD-SAS), and to investigate its psychometric properties. To examine the test-retest reliability, one investigator rated 217 narratives of real-life self-harming incidents at initial assessment and 4 weeks later. To establish the interrater reliability, three independent raters assessed a random sample of 70 narratives using the CAD-SAS. To examine the validity, one investigator using the CAD-SAS compared ratings to clinical judgments made by a consultant psychiatrist without the CAD-SAS on the same random set of 70 narratives. Test-retest reliability was excellent (97.2% agreement) and interrater reliability was substantial (70.0% agreement, κ = 0.70). Agreement in the classification of incidents with the "real-world" clinical judgments supports the validity of the CAD-SAS (64.3% agreement, κ = 0.46). The reliability and validity of future studies can be enhanced through the standardized assessment and classification of incidents.