Evaluation of alcohol intoxication and withdrawal syndromes based on analysis of tremor signals

Abstract In this paper, we analyze the relationship between the clinical assessment of tremors caused by alcohol withdrawal (AW) syndrome and the underlying tremor signal as recorded by an accelerometer. The ultimate goal of the study is to find out the relationship between energy of the tremor signal in patients in AW and tremor rating provided by physicians. We have developed an iOS application that calculates the tremor component of the Clinical Institute Withdrawal Assessment, revised (CIWA-Ar) score using accelerometer data. We report on the characteristics of AW tremor, the accuracy of electronic tremor assessment compared to expert clinician assessment. It should be noted that this application is not a replacement for physician's assessment, but it will help physicians to quantify the severity of the tremor which is one of the components of CIWA-Ar protocol. Our first and key finding is that there is a logarithmic relationship between the CIWA-Ar score and the tremor energy in the [5,15] Hz range. Based on training on 84 recordings from 50 patients with the clinical diagnosis of AWS, and testing on a separate set of 33 patients with AWS, the proposed logarithmic relationship provides a CIWA-Ar Root Mean Square Error (RMSE) of 0.91 (with respect to a consensus rating from 3 expert physicians on a 7 point scale). This compares to a RMSE of 1.04 for junior physicians, 0.94 for junior nurses, 0.94 for senior nurses, and 0.95 for expert physicians.