Informant single screening questions for delirium and dementia in acute care – a cross-sectional test accuracy pilot study

Cognitive impairment is a term covering a range of disorders representing a clinical deficit in cognitive ability with a significant deterioration from the person’s previous level of function [1]. It can present acutely with rapid onset and a fluctuating short-term course such as delirium, or as a chronic illness with a gradual, progressive course, for example mild cognitive impairment (MCI) and dementia.

Cognitive impairment is a strong predictor of negative outcomes such as increased length of hospitalisation, increased 6 month mortality and increased risk of readmission in older hospitalised individuals [2,3]. Cognitive impairment is commonly seen in the acute care setting, with between 40% and 70% elderly patients in acute care in UK hospitals having dementia with with less than half of these patients having a known previous diagnosis [4,5].

In-depth psychiatric assessment is not usually available as a first means of detecting cognitive impairment due to time and opportunity constraints. Instead, cognitive screening tests that are quick and easy to administer are used to identify individuals at high risk of cognitive impairment, suitable for further assessment [6]. Single question screening tests have been described for dementia [7] and delirium [8]. There is debate surrounding which screening test is the best to use due to the wide range available and lack of validation in hospital settings [9].

We aimed to pilot the performance of two single screening questions (SSQ), one for delirium and one for dementia, in hospitalised, elderly individuals as part of a secondary analysis of a previously collected data set. The SSQs were compared to a validated informant assessment of dementia; the 16 item Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) [10] and delirium diagnosis based on the Confusion Assessment Method (CAM) [11]. We hypothesised that single informant-based screening questions would be sensitive and specific in detecting dementia and delirium in hospitalised older people.

The characteristics of the 161 participants recruited for this study are shown in Table 1, as well as separate analyses for patients with complete and incomplete single screening question data. SSQ’s were completed for 70 patients. There was found to be no significant difference between characteristics of respondents and non-respondents except in terms of age (p = 0.049).

Of the 70 patients who had data for both the dementia and delirium SSQ, 26 (37.1%) had a positive screen on both single questions, 25 (35.7%) had a negative screen on both single questions, 9 (12.9%) had a positive screen only for delirium and 10 (14.3%) had a positive screen only for dementia.

Kruskal-Wallis H Test analysis revealed a statistically significant difference in MMSE scores between the different SSQ-delirium outcomes (H(2) = 21.4 , p < 0.001) (see Table 2).

A statistically significant difference in MMSE scores was also found between the different SSQ-dementia outcomes (H(2) = 16.8, p < 0.001), with a median of 3.06 for patients identified as “no change or better”, 3.94 for patients identified as “a bit worse” and 4.88 for patients identified as “much worse”.

Kruskal-Wallis H Test analysis also revealed a statistically significant difference in IQCODE scores between the different SSQ-delirium outcomes (H(2) = 27.3, p < 0.001), with a median of 3.07 for those identified as “no change or better”, 3.75 for those identified as “a bit worse” and 4.85 for patients identified as “much worse”.

A statistically significant difference in IQCODE scores was also found between the different SSQ-dementia (H(2) = 41.2, p < 0.001), with a median of 3.06 for those identified as “no change or better”, 3.94 for patients identified as “a bit worse” and 4.88 for patients identified as “much worse”.

ROC analyses were carried out to calculate the sensitivity and specificity values of the SSQ-dementia and the SSQ -delirium, when comparing screening accuracy to a directly comparable routinely used screening instrument (IQCODE with a cut score of <3.38 and CAM positive diagnosis, respectively). Analysis revealed the SSQ-dementia (AUC = 0.882) had good diagnostic accuracy and SSQ-delirium (AUC = 0.665) had fair diagnostic accuracy. Sensitivity and specificity data are reported in Table 3.

The most prominent finding from this pilot study was the high sensitivity and specificity of the single screening question for dementia. This one-item screen performed at a similar level to the routinely used 16-item IQCODE. However, while the single screening question for delirium showed a similar sensitivity, it had low specificity. Almost half of individuals with normal cognitive functioning, as classified by CAM diagnosis, were identified by the single screening question having suspected delirium.

While the SSQ-delirium appeared not to perform well as a first step in delirium detection, this may be better explained by methodological issues with difficulties in screening for delirium, in general. A defining feature of delirium is fluctuation in presence of symptoms, and as such it is very difficult to have exact concurrence between a screening tool for delirium and the reference standard; in this study the SSQ-delirium and CAM, respectively. Thus, it is possible that delirium can be present at one testing point but not the other. Interpretation of the performance of the SSQ-delirium in this study is limited as data on the time lag between the SSQ-delirium and CAM were not collected. It would also be of interest for future studies to test the SSQ-delirium blinded from CAM diagnosis.

Undiagnosed dementia may be able to account for the high number of false positive results identified by the single screening question for delirium. The majority of patients identified as positive by the single screening question for delirium also had a positive results on the single screening question for dementia. It is possible that an informant based question is not suitable to accurately differentiate those at high risk of having delirium from those at high risk of having dementia.

Direct cognitive testing of patients is the most commonly used screening method for cognitive impairment [14]. However, informant testing shows promise in improving detection of at-risk individuals. It has been demonstrated that an informant questionnaire shows the same performance as direct cognitive testing, despite the fact that these different screening tools measure different patient attribute [15]. A major advantage of informant based assessments is that they do not suffer the same problems as cognitive testing of being independent of patient’s education level and less susceptible to ceiling effects [16]. Single question screening for cognitive impairment is a hot topic at the moment and there is a paid incentive being rolled out across hospitals in England in an attempt to improve diagnosis of patients with cognitive impairment known as the Commission for Quality and Innovation (CQUIN) framework. Of particular interest to this study is the first stage, ‘Find’ whereby the patient or informant is asked the question, ‘Has the patient been more forgetful in the last 12 months to the extent that it has significantly affected their daily life?’ This suggests that single question methods of screening for cognitive impairment are beginning to be used on a large scale despite not being fully validated.

The advantage of using an informant-based screening tool for delirium is less clear than for dementia. Due to the fluctuating nature of delirium, it can easily be missed and the need for a relative to be present to provide information introduces further timing challenges as access to informants is only possible at discrete selected times. However, evidence suggests that delirium is most prevalent in older patients when they are at their most sick, usually soon after admission or in those patients who have prolonged hospitalisation. Thus, those may be particularly important times to use the SSQ-delirium. It is clear from the literature that some form of delirium screening is needed rather than subjective clinical judgement which has been shown to perform poorly at detecting prevalent delirium [17]. This supports the need for a more structured approach to be implemented as a brief first step to identify those with suspected delirium who would then be assessed using diagnostic tools such as the CAM.

Caution must be taken when interpreting these results as the reference standards were screening tests rather than a formal clinical diagnosis. lack of comprehensive clinical evaluation for dementia was a further major limitation . The recruitment and consent strategy could have lead to biases within this study; those at higher risk of cognitive impairment may have had more visits from family and hence more likely to return the informant questionnaire, especially in cases where the patient was unable to provide consent and hence a relative or carer had to provide consent. As this study obtained single question informant report in less than half of patients this raises issues of feasibility as there is strong potential for many individuals with suspected cognitive impairment not being assessed. From this data set, it is unclear whether such low response rates were due to patients not having a suitable relative or carer available to answer the single screening questions or whether it was due to a lack of appropriate measures taken by the researcher to insure return of informant responses.

Presenting informants with the IQCODE prior to answering the single screening questions may also have influenced the results, possibly enhancing how well the single screening questions appeared to perform.

However, within these limitations we believe our results still provide useful information on how simple responses from informants may perform. This pilot study was strengthened by broad inclusion criteria thus providing a sample relatively representative of older adult acute care admissions. We based our SSQ’s on the format of a validated informant questionnaire, the IQCODE.

These preliminary findings show promise for use of a single question screening tool as the first step in the detection of cognitive impairment and prompt more thorough investigation. As yet there is no one cognitive screening tool that has achieved widespread consensus, thus the comparisons made to the IQCODE and MMSE only provide evidence for the value of carrying out a more in-depth study. Future research should compare the single question screen to a gold standard clinical evaluation of dementia and delirium to determine more reliable diagnostic accuracy figures. The use of an informant-based single screening question may be particularly useful in combination with a direct cognitive testing method in helping to distinguish patients who fall within the middle, grey area of scores on cognitive test based screening tools. Furthermore, it is apparent that there is need for investigation to determine how to get higher uptake of relatives/carers to provide informant report.

Routine screening for cognitive impairment is necessary in older hospitalised patients and effective screening is beneficial to clinical staff as well as the outcomes of the patients. This task is made more difficult by the lack of uniformity regarding the various screening tools available [18]. This pilot study has provided evidence that a single-item informant based screen may perform in a comparable way to much longer screens for dementia by effectively differentiating those with suspected cognitive impairment from normally functioning patients. Further validation is warranted. However, if a screening tool as straightforward as asking a single question, which has low time cost and requires little to no training to administer, can perform as well as more complex screens such as the IQCODE, then it would seem intuitive that this is a preferable option. However, results showed that a single question screening tool may not have the same potential when identifying individuals with suspected delirium.