A New Severity Scoring Scale for the 3‐Minute Confusion Assessment Method ( 3D‐CAM )

The 3D-Confusion Assessment Method (CAM) is a short structured instrument to determine the presence or absence of CAM-defined delirium that can be completed in about 3 minutes.1 Recent literature emphasized the importance of measuring delirium severity in addition to determining its presence.2 Some of us previously reported a method for measuring delirium severity using the 3D-CAM with scores ranging from 0 to 7 (with 7 the most severe).3 This method has numerous strengths including that it quantifies the severity of each of the four features measured by the 3D-CAM, and it assigns non-zero severity scores only to those patients who are highly likely to have delirium by gold standard clinical assessments. For some applications, it is also important to have a delirium severity scoring scale that captures the severity of subsyndromal delirium symptoms and has a wider range for quantifying the severity of delirium signs and symptoms. This latter feature is particularly important in delirium biomarker studies,4-8 in which delirium severity scores with a wider range would provide greater statistical power for detecting associations with biomarkers. A delirium severity score with a wider range may also be useful for tracking the severity of delirium over time within individual patients, in both clinical practice and research studies. Thus we developed and validated a new delirium severity scoring method based on the 3D-CAM instrument that yields an expanded raw severity score ranging from 0 to 20 points. This raw score is the sum of positive items present on 20 questions of the original 3D-CAM instrument, where positivity is defined as an incorrect response to a cognitive test item (3D-CAM items 1-7), patient endorsement of a symptom probe (items 8-10), or interviewer endorsement of an observational feature (items 11-20).1 To describe and validate the new raw 3D-CAM severity measure, we calculated 3D-CAM severity scores using both the old and new method in the original 3D-CAM validation cohort stratified by the presence or absence of delirium determined by an independent clinical reference standard assessment and also by the 3D-CAM itself.1 We then calculated the discriminant measure area under the receiver operating characteristic curve (AUC) and the 95% Wald confidence interval (CI), assuming asymptotic normality of the estimated area for both severity measures against the presence or absence of delirium measured both ways. The first provides a measure of external validity against an independent reference standard; the second provides a measure of internal consistency between 3D-CAM delirium severity and diagnosis. Informed consent was obtained for all study procedures, and analyses were performed using SAS v.9.4. The distribution of severity scores in patients with vs without delirium from a previously described cohort of 201 older adults (mean age = 84 years; 62% women; 28% with dementia)1 is shown in Supplementary Tables S1 and S2, for both the original 3D-CAM severity scoring scale and for this new raw 3D-CAM severity scoring scale, respectively. As expected, the new raw severity scale has a wider range than the original severity scale (17 vs 6 points) including among those with delirium (16 vs 6 points) and those without delirium (8 vs 5 points; Supplementary Tables S1 and S2). The AUC (95% CI) was .950(.915-.984) for the original 3D-CAM severity score and .950(.914-.986) for the new raw 3D-CAM severity score for discriminating the presence or absence of delirium as determined by an independent reference standard clinical interview (Figure 1A). For discriminating the presence or absence of delirium determined by the 3D-CAM, the AUC (95% CI) was .979(.965-.993) for the original 3D-CAM severity score and .985(.973-.996) for the raw 3D-CAM severity score (Figure 1B). These data suggest that this new raw 3D-CAM delirium severity scoring method has a wide range, demonstrates high validity against an independent clinical reference standard, and strong internal consistency with the 3D-CAM delirium diagnosis. In the last instance, it works similarly well as the previously described original 3D-CAM severity scoring method.3 This new raw 3D-CAM delirium severity scoring method provides two clear advantages. First, it has a nearly 3-fold wider range for assessing delirium severity than the original 3D-CAM severity scoring method,3 which may be useful for both delirium biomarker studies and for tracking the clinical course of individual patients' delirium over time. Second, the new method's wider range among the non-delirious allows the severity of subsyndromal delirium to be quantitated, that is, the severity of delirium-related signs and symptoms that may be present yet insufficient to meet full diagnostic criteria for delirium. Future work should evaluate the relationship between this new raw 3D-CAM delirium severity score and clinical outcomes such as hospital length of stay, readmission rate, and mortality. This research was supported by the National Institute on Aging grants (K01AG057836 [SMV], R03AG061582 [SMV], R01030618 [ERM], K24AG035075 [ERM], K76AG057022 [MB], P30AG028716 [MB], UH3AG056925 [MB]), the Alzheimer's Association (AARF-18-560786 [SMV]), a PACT grant from the Alzheimer's Drug Discovery Foundation [MB], and Duke Anesthesiology Department funds [MB]. All authors have declared no conflicts of interest for this letter. All the authors conceived of the idea for this project. Michel J. Devinney, Leah Acker, and Miles Berger drafted the letter. Edward R. Marcantonio acquired the data. Yoojin Jung and Long Ngo analyzed the data. All the authors interpreted the data and critically revised and approved the final submission for important intellectual content. The sponsors did not participate in the design or data analysis of any aspect of the study or in the preparation of this letter. Supplementary Appendix S1: Supplementary Material. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.