Rationale: Delirium is a serious, morbid condition affecting 2.6 million older Americans annually. A major problem plaguing delirium research is difficulty in identification, given a plethora of existing tools. The lack of consensus on key features and approaches has stymied progress in delirium research. The goal of this project was to use advanced measurement methods to improve delirium’s identification.

Aims and Findings:

(1) Determine the 4 most commonly used and well-validated instruments for delirium identification. Through a rigorous systematic review, I identified the Confusion Assessment Method (CAM), Delirium Observation Screening Scale (DOSS), Delirium Rating Scale-Revised-98 (DRS-R-98), and Memorial Delirium Assessment Scale (MDAS).

(2) Harmonize the 4 instruments to generate a delirium item bank (DEL-IB), a dataset containing items and estimates of their population level parameters. In a secondary analysis of 3 datasets, I equated instruments on a common metric and created crosswalks.

(3) Explore applications of the harmonized item bank through several approaches. First, identifying different cut-points that will optimize: (a) balanced high accuracy (Youden’s J-Statistic), (b) screening (sensitivity), and (c) confirmation of diagnosis (specificity) in identification of delirium. Second, comparing performance characteristics of example forms developed from the DEL-IB.

Impact: The knowledge gained includes harmonization of 4 instruments for identification of delirium, with crosswalks on a common metric. This will pave the way for combining studies, such as meta-analyses of new treatments, essential for developing guidelines and advancing clinical care. Additionally, the DEL-IB will facilitate creating big datasets, such as for omics studies to advance pathophysiologic understanding of delirium.