• Chris Brunker and Ruth Harris.
• St George's Healthcare NHS Trust, Neuro-ICU, St George's Hospital, London SW17 0QT, United Kingdom. Electronic address: chris.brunker@stgeorges.nhs.uk.
• Intensive Crit Care Nurs. 2015 Apr 1; 31 (2): 69-75.

Objectivesto evaluate accuracy, sensitivity, specificity and inter-rater agreement of AVPU (Alert, Voice, Pain, Unresponsive) when used by non-specialist nursing staff assessing consciousness, and to investigate users' views.MethodsVideo-recorded simulations of assessments of consciousness were developed and verified by an expert panel. Participants scored simulations using AVPU and completed questionnaires eliciting views on the scale. AVPU scores were compared with functional levels agreed by the panel.SettingA large urban teaching hospital.ResultsFifty-one participants scored 255 simulations. Overall accuracy was 82.4% (95% CI=77.7-87.1%), sensitivity 0.94 (95% CI=0.90-0.98), specificity 0.74 (95% CI=0.66-0.82) and inter-rater agreement (un-weighted kappa) 0.782. Accuracy was low for simulations depicting an orientated patient whose eyes open to speech (49% correct) and a confused patient with spontaneous eye opening (61.5% correct). Sensitivity and agreement for levels corresponding to "Alert" and "Voice" were 0.81 (95% CI=0.69-0.93) and kappa=0.506. Participants expressed uncertainty about aspects of AVPU's use.ConclusionsAVPU had low rates of accuracy, sensitivity and agreement in distinguishing between "Alert" and "Voice", and low specificity overall, suggesting it may be unsuitable for early warning scoring. Participants expressed doubts about the use of AVPU.Copyright © 2014 Elsevier Ltd. All rights reserved.

19 keywords...

hide…