Aim: A serious syndrome for cancer in-patients, delirium risk increases with age and medical acuity. Screening tools exist but detection is frequently delayed or missed. We test the 'Single Question in Delirium' (SQiD), in comparison to psychiatrist clinical interview.

Methods: Inpatients in two comprehensive cancer centres were prospectively screened. Clinical staff asked informants to respond to the SQiD: "Do you feel that [patient's name] has been more confused lately?". The primary endpoint was negative predictive value (NPV) of the SQiD versus psychiatrist diagnosis (Diagnostic and Statistics Manual criteria). Secondary endpoints included: NPV of the Confusion Assessment Method (CAM), sensitivity, specificity and Cohen's Kappa coefficient.

Results: Between May 2012 and July 2015, the SQiD plus CAM was applied to 122 patients; 73 had the SQiD and psychiatrist interview. Median age was 65 yrs. (interquartile range 54-74), 46% were female; median length of hospital stay was 12 days (5-18 days). Major cancer types were lung (19%), gastric or other upper GI (15%) and breast (14%). 70% of participants had stage 4 cancer. Diagnostic values were similar between the SQiD (NPV = 74, 95% CI 67-81; kappa = 0.32) and CAM (NPV = 72, 95% CI 67-77, kappa = 0.32), compared with psychiatrist interview. Overall the CAM identified only a small number of delirious cases but all were true positives. The specificity of the SQiD was 87% (74-95) The SQiD had higher sensitivity than CAM (44% [95% CI 41-80] vs 26% [10-48]).

Conclusion: The SQiD, administered by bedside clinical staff, was feasible and its psychometric properties are now better understood. The SQiD can contribute to delirium detection and clinical care for hospitalised cancer patients.