Aims & Objectives:
Delirium diagnosis in critically ill children is multifactorial and fraught with challenges. Current evidence suggests that medical diagnosis of delirium is unreliable. A sub-study of a larger delirium project, the primary aim of this research was to examine the psychometric properties of medical decision-making for delirium screening.
Participants were attending physicians (n = 16) at an Australian tertiary referral mixed-PICU. Eligible children (> 6 months, SBS > -2) were reviewed for delirium once daily, up to five days, by an attending physician. Psychometric properties were compared against the gold-standard using sensitivity, specificity and receiver operator characteristic curves. Further comparisons were made between two doctors who acted as study ‘champions’ and all other attending physicians.
A total of 80 children were recruited to the study and 101 paired assessments were undertaken. The incidence of delirium in this sample was 13.9% and a total of 39 (38.6%) of assessments were undertaken by one of the study champions. Table 1 summarises the main findings. Overall, medical staff were able to detect delirium with reasonable accuracy (AUC = 78.6, 95% CI = 59.5–97.7), however diagnostic accuracy was significantly better when patients were reviewed by one of the two study champion doctors (AUC = 88.5; 95% CI = 67.3–100.0).
Historically, medical decision-making alone has been viewed as an unreliable method of detecting delirium. These findings indicate that, while overall, physicians’ assessments of paediatric delirium can be accurate, exposure to delirium and delirium research can improve this accuracy.