Premium
Predictive power of the DASA ‐ IV : Variations in rating method and timescales
Author(s) -
Nqwaku Mphindisi,
Draycott Simon,
AldridgeWaddon Luke,
Bush EmmaLouise,
Tsirimokou Alexandra,
Jones Dominic,
Puzzo Ignazio
Publication year - 2018
Publication title -
international journal of mental health nursing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.911
H-Index - 54
eISSN - 1447-0349
pISSN - 1445-8330
DOI - 10.1111/inm.12464
Subject(s) - predictive power , predictive validity , receiver operating characteristic , medicine , predictive value , clinical psychology , philosophy , epistemology
This project evaluated the predictive validity of the Dynamic Appraisal of Situational Aggression – Inpatient Version ( DASA ‐ IV ) in a high‐secure psychiatric hospital in the UK over 24 hours and over a single nursing shift. DASA ‐ IV scores from three sequential nursing shifts over a 24‐hour period were compared with the mean (average of three scores across the 24‐hour period) and peak (highest of the three scores across the 24‐hour period) scores across these shifts. In addition, scores from a single nursing shift were used to predict aggressive incidents over each of the following three shifts. The DASA ‐ IV was completed by nursing staff during handover meetings, rating 43 male psychiatric inpatients over a period of 6 months. Data were compared to incident reports recorded over the same period. Receiver operating characteristic ( ROC ) curves and generalized estimating equations assessed the predictive ability of various DASA ‐ IV scores over 24‐hour and single‐shift timescales. Scores from the DASA ‐ IV based on a single shift had moderate predictive ability for aggressive incidents occurring the next calendar day, whereas scores based on all three shifts had excellent predictive ability. DASA ‐ IV scores from a single shift showed moderate predictive ability for each of the following three shifts. The DASA ‐ IV has excellent predictive ability for aggressive incidents within a secure setting when data are summarized over a 24‐hour period, as opposed to when a single rating is taken. In addition, it has moderate value for predicting incidents over even shorter timescales.