Our study demonstrates that the SQiD has good performance and is comparable with the CAM in terms of NPV against psychiatric interview, the diagnostic reference standard. This is an important attribute of the study as without a robust standard applied systematically across a validation study, in keeping with recommended delirium diagnosis methodology [3], the validity of a detection tool cannot be established. Our study found fair correlation of both SQiD and CAM in relation to the reference standard. The CAM had a better PPV but a low sensitivity, whereas the SQiD had better sensitivity. The CAM however, found much lower number of patients with delirium than the reference standard, but with the caveat that all those identified as being delirious or non-delirious by the CAM were correctly identified, hence the contrast between Sensitivity (26%) and Specificity (100%) of the CAM.
Methodological strengths of our study include a clear description of patient characteristics and setting that mitigated case ascertainment bias and the method which prevented training bias due to study flow.
The use of an accurate and reproducible reference standard with a clear description of reference-rater training and characteristics is of central importance. This was a pragmatic, clinically embedded, ‘real world’ study. The SQiD is simple and provides a pragmatic approach to support uptake and also enabled families to be involved in discussions with bedside clinicians regarding delirium. Informal feedback suggests it has good face validity to engage staff in actively seeking delirium. A better understanding of what happens after the SQiD question is asked may benefit future iterations and aid an understanding of the role for the SQiD in staff education and the promotion of helpful discussion between carers and staff.
Current guidelines do not support routine delirium screening in Oncology settings due to lack of evidence [26], they do however provide the following advice “any changes in cognitive or emotional behaviour or psychomotor activity suggestive of delirium are present, a trained healthcare professional with expertise in evaluating delirium should carry out a clinical assessment to confirm the diagnosis of delirium” which lends support to engaging staff and family in detecting this change. While the SQiD does not replace clinical assessment its NPV may support clinicians in identifying patients for further review.
Of the 16 patients with hypoactive delirium, 6 were identified on SQiD, the CAM identified only one. The detection of hypoactive delirium is an important consideration in delirium detection tools as typically hypoactive delirium poses more difficulties for clinical identification. Given the small numbers in each motoric sub-group, this observation needs to be interpreted with caution, however this is one aspect of the SQiD which may support further investigation.
As the SQiD is not directly asked of the patient, it is not important that patients speak English. This may even be an advantage but at this point it is a theoretical advantage not tested by this study which was constrained by exclusion of patients with insufficient English fluency for consent and psychiatrist interview.
Limitations of our study relate to a non-consecutive sample and exclusion of some patient groups. The study population had some statistically-significant differences to the source population – they were older and had longer length of stay. Further it is of note that delirium detection rates in our sample are not presented as representative of the base population and cannot therefore be used to estimate delirium incidence or prevalence. The rates detected in the study population were however not dissimilar to incidence or prevalence established in several other medical in-patient cohorts [13, 14, 24, 27,28,29]. A further limitation was that SQiD administration appeared to be dependent on prompting by senior staff, making recruitment subject to variation due to competing demands and staffing. The time between SQiD, CAM and psychiatric interview was also generally longer than anticipated in the protocol. This can have implications for diagnosis as delirium symptoms fluctuate over time. There is a possibility of recruitment centre bias however in terms of Australia-modified Performance Status (AKPS) [25] no bias was detected. Lastly, the study did not reach the planned sample size, which resulted in lower power and wider confidence intervals.
Although CAM training for study staff was conducted it fell short of that recommended for diagnostic purposes and we did not achieve the sensitivity and specificity in using the CAM that has been reported elsewhere [9]. One of the reasons the CAM was included as a comparator to the SQiD was to determine if it was an option for routine use in our clinical setting. With the training possible in the cancer clinical context this was not the case, and we note that other recent studies have reported similar short-comings in training staff so that optimum sensitivity and specificity of the CAM has not been achieved in those clinical and clinical research settings [12, 13].
A further limitation is that the only information gathered about the SQiD informant was co-habitation with the patient. Cognitive testing to understand how family and carers interpret the SQiD question, the subsequent discussion, and understanding of the meaning of question prior to the study may be of relevance, and is recommended for consideration in future studies.
Finally, a number of patients in our study were excluded due to lack of English fluency (n = 43). Therefore, conclusions about the utility of the SQiD in patients across cultural and linguistic backgrounds other than that of English speakers remains untested. Similarly, the SQiD may be of lower utility for patients who live alone or have limited contact with other people. Nevertheless, of 142 patients in our study, only five were excluded due to not having a person available to answer the SQiD.
Implications for training and practice
The SQiD may engage nurses, medical staff, carers and/or patients in discussion regarding the components of syndromic delirium. The use of the SQiD by clinical staff may reinforce the concept that delirium is an important medical issue for staff, patients and family/carers.
Implications for research
Other studies have compared short tests of cognition or attention [4, 8, 12, 30, 31] however, to our knowledge, having a clinician directly approach the main carer with a conversational approach and a simple question regarding recent changes in cognition, although it makes common sense/good clinical practice, has not been previously evaluated. Future research questions may focus on the content of the discussion that ensues following administration of the SQiD, and the ability of the tool to engage staff and family/carers in looking for changes that would indicate a need for further clinical delirium assessment.
The combination of a detection tool with high sensitivity followed by one with high specificity may be a useful approach to screening and could be addressed in future studies. Since inception of this study other short and brief tools have appeared in the peer review literature, including the 4AT [30] and more recently the delirium RADAR [31]. These tools are promising either alone or in combination, however they may not have the same qualities of engaging carers and staff that is possible with the SQiD, which again raises hypotheses relating to combined administration of short tests.