The schedule for the assessment of drug-induced movement disorders (SADIMoD): inter-rater reliability and construct validity.

Abstract

The Schedule for the Assessment of Drug-Induced Movement Disorders (SADIMoD) is a newly developed instrument, consisting of a compilation of rating scales, to measure the severity of drug-induced movement disorders: dystonia, dyskinesia, Parkinsonism, akathisia, ataxia, and several types of tremors. The inter-rater reliability and the construct validity of this scale were investigated. Six investigator teams were trained by means of a standard package of instruction material to such an extent that a single member of the team could represent the entire team. Thirty-one patients [20 male/11 female; 57.1+/-6.5 yr (mean+/-S.D.)] with a variety of movement disorders were recorded on videotape according to the SADIMoD Schedule. Single representatives of all six teams scored these video recordings. To this set the existing SADIMoD ratings of 82 patients were added to form the so-called 'total data set'. These patients were examined by 6 different researchers, who rated 4, 8, 10, 14, 18 and 28 patients, respectively, mostly in the context of a research protocol. A specific subset consisted of 12 patients that were examined three times with a two-weekly interval without any change of their medical condition or treatment. The 6 ratings of the 31 individual patients correlated to a highly significant degree, with Kendall's Coefficients of Concordance of 0.436 to 0.891 (median 0.717). The same was true for the 6 ratings of the 7 SADIMoD subscales (median 0.578, range 0.462-0.715) Considering the total data set, the homogeneity of the various subscales was good (Cronbach's alpha = 0.81-0.94, median: 0.87). The SADIMoD dyskinesia and dystonia subscales showed a highly significant mutual correlation. The Parkinsonism subscale correlated highly significantly with the rest and postural tremor subscales and to a lesser extent with the akathisia and ataxia subscales. An analysis of variance showed that the three ratings in the subset of 12 patients were not significantly different for any scale. Also Scheffé tests for homogenous subsets did not reveal any significant differences. When investigated under 'real world' circumstances, the inter-rater reliability of the SADIMoD was found to be satisfying. The instruction material, that was developed and used in this study, fully comes up to the requirements. The construct validity of the SADIMoD is more than sufficient.