Abstract
The Clinical Dementia Rating (CDR) is a common rating system used in clinical trials and longitudinal research projects to rate the presence and severity of cognitive problems in Alzheimer disease and related disorders. The interview process requires training and can be time consuming. Here we describe the validity, reliability and discriminative ability of a computer-generated CDR using a personal digital assistant (PDA) format. This project utilized clinical data from 138 archival and live evaluations (patient and informant interviews) collected for research purposes at Washington University to develop and test a software-based system for the administration and automatic scoring of the CDR. The system was programmed for use on a hand-held computer via the Palm
™
Operating System. We developed domain-specific algorithms to quantify and translate clinical scoring decisions for the three cognitive (Memory, Orientation, Judgment and Problem Solving) and the three functional (Community Affairs, Home & Hobbies, Personal Care) domains of the CDR. An acceptable set of algorithms were developed using data from 104 research cases, reflecting a range of impairment levels (CDR 0 – 3) and expert scoring decisions. These algorithms were then tested for accuracy in a validation sample of 34 cases. The computer-generated CDR has excellent internal consistency (Cronbach’s alpha ranging from .94–.98) and inter-rater reliability (intraclass correlation coefficient ranging from .88 – .96). The computer-generated CDR showed excellent discrimination between demented and nondemented cases (Area under the curve = 0.95; 95%CI:.84–1.1). The computer-generated CDR using a Palm
™
Operating System is easy to use, valid and reliable. The level of agreement compares favorably to published inter-rater reliability data for the CDR. Software-based administration and automatic scoring of the CDR is a viable alternative to paper-based methods and may be useful in research and clinical settings, especially where electronic data management and reliability in scoring are critical.
