Validation of the Delirium Diagnostic Tool-Provisional in intensive care units

Abstract

Shortcomings of intensive care units (ICU) delirium screening tools include not measuring its core features, not excluding stupor/coma and not being continuous measurement instruments. We validated the Delirium Diagnostic Tool-Provisional (DDT-Pro) that assesses all 3 core symptom domains for delirium and subsyndromal delirium (SSD) in the ICU. This is a multicenter validation following STARD guideline. Delirium reference standards were DSM-5 criteria, cluster analysis (CA) of the DDT-Pro scores and clinical validators for preestablished DDT-Pro ≤ 6 and ≤ 7 cutoffs (0-9 range) for delirium and SSD. DDT-Pro dimensionality and internal consistency reliability are reported. Of 127 patients, 29 (22.8%) had DSM-5 delirium. The area under the receiver-operator DDT-Pro curve was 90% with the ≤ 7 and ≤ 6 DDT-Pro cutoffs which had 82.7% and 80.3% accuracy at the most balanced sensitivity-specificity. The ≤ 6 cutoff specificity (85.7%) was higher, while ≤ 7 cutoff sensitivity (89.7%) was higher with NPV = 96.2%. According to CA, ≤7 cutoff differentiated 100% of nondelirium patients from SSD and delirium, whereas DSM-5 misattributed SSD. Validation of an SSD group was supported by delirium severity gradients and various clinical validators. Cases in this CA also coincided exactly with nondelirium, SSD and delirium groups prespecified by DDT-Pro cutoffs from non-ICU samples. One factor explained 69.9% of the DDT-Pro variance, Cronbach α = 0.79 (cohesive delirium dimension). Our findings indicate that the DDT-Pro has very good construct validity and discriminates ICU delirium against DSM-5, performing even more cleanly using agnostic CA for SSD and delirium diagnosis. Its continuous score structure discernment of SSD was supported by clinical validators. ICU cutoffs were the same as in previous inpatient samples.

Keywords: cluster analysis; delirium screening; intensive care unit; sensitivity and specificity.

Figure 1.

Flow diagram of the participants through the study groups, distributed according to the DDT-Pro (validating tool) and the DSM-5 delirium diagnosis. DDT-Pro = Delirium Diagnostic Tool-Provisional, DSM-5 = Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition).

Figure 2.

DDT-Pro score boxplot distributions in 127 ICU patients. (A) DDT-Pro scores according to DSM-5 diagnosis which are largely overlapping. (B) The same DSM-5 delirium patients but with 2 groups resulting from CA of scores in the DSM-5 nondelirium group, delineating a nondelirium group scoring 8 to 9 and SSD/mild group ≤7 with much overlap with delirium. (C) New delirium and nondelirium distributions by a 2-group CA solution of DDT-Pro scores resulting in no score overlap between groups. (D) A 3-group CA solution of scores resulting in clear differentiation of 79 nondelirium patients (8–9 points), an intermediate group of 26 SSD (6–7 points) and delirium group of 22 patients. Bolded bars are medians (50th percentile), boxes represent the middle 50% distribution of scores and tails denotes 25^th percentile. Outliers are represented by asterisks and circles. CA without involvement of DSM-5 provides greater clarity. CA = cluster analysis, DDT-Pro = Delirium Diagnostic Tool-Provisional, DSM-5 = Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition), ICUs = intensive care units, SSD = subsyndromal delirium.

Figure 3.

Receiver-operating curve showing the AUC for delirium diagnosis using the DDT-Pro in 127 patients admitted to ICUs, diagnosed according to the reference standard DSM-5. AUC = area under the curve, DDT-Pro = Delirium Diagnostic Tool-Provisional, DSM-5 = Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition), ICUs = intensive care units.

Figure 4.

Boxplot of DDT-Pro 3 items, and commission and omission errors in vigilance, according to the 3-solution CA of the tool’s total score that grouped the sample of 127 patients in 79 nondelirium, 26 subsyndromal (SSD) and 22 delirium patients. The bars in bold are medians (50th percentile), with boxes representing middle 50% distribution of scores and tails denoting 25th percentile. Circles and asterisks are outliers. CA = cluster analysis, DDT-Pro = Delirium Diagnostic Tool-Provisional, SSD = subsyndromal delirium.