BACKGROUND: The lack of universally agreed criteria has hampered population studies of the prevalence and causation of soft-tissue disorders of the upper limb. OBJECTIVES: To establish core variables for classification of the commonest disorders seen in population samples. METHODS: Consecutive new cases seen in clinical practice in five different centres were evaluated with respect to 30 variables shown to have discriminatory value in univariate analysis. Multivariate analysis using logistic regression modelling was carried out with these as the independent variables and with the clinical diagnosis as the dependent variable. RESULTS: A total of 1382 cases of soft-tissue disorder were recorded and only those diagnostic groups with 50 or more cases were included. In multivariate logistic regression, significant variables positively discriminating for each disorder were identified for carpal tunnel syndrome (n = 56), lateral epicondylitis (n = 87), tenosynovitis (n = 63), shoulder tendonitis (n = 157), non-specific upper limb disorder (n = 458), fibromyalgia (n = 124) and inflammatory arthritis (n = 100), which was used for comparison purposes. Significant discrimination for each model was demonstrated by the construction of receiver operating characteristic (ROC) curves and appropriate area under the curve statistics. CONCLUSIONS: This approach to classification criteria is based on multivariate modelling rather than on a consensus statement. This includes the effects of negative as well as positive associations. Further work is required on both the reproducibility of the clinical signs and the application of the criteria to other datasets.
BACKGROUND: The lack of universally agreed criteria has hampered population studies of the prevalence and causation of soft-tissue disorders of the upper limb. OBJECTIVES: To establish core variables for classification of the commonest disorders seen in population samples. METHODS: Consecutive new cases seen in clinical practice in five different centres were evaluated with respect to 30 variables shown to have discriminatory value in univariate analysis. Multivariate analysis using logistic regression modelling was carried out with these as the independent variables and with the clinical diagnosis as the dependent variable. RESULTS: A total of 1382 cases of soft-tissue disorder were recorded and only those diagnostic groups with 50 or more cases were included. In multivariate logistic regression, significant variables positively discriminating for each disorder were identified for carpal tunnel syndrome (n = 56), lateral epicondylitis (n = 87), tenosynovitis (n = 63), shoulder tendonitis (n = 157), non-specific upper limb disorder (n = 458), fibromyalgia (n = 124) and inflammatory arthritis (n = 100), which was used for comparison purposes. Significant discrimination for each model was demonstrated by the construction of receiver operating characteristic (ROC) curves and appropriate area under the curve statistics. CONCLUSIONS: This approach to classification criteria is based on multivariate modelling rather than on a consensus statement. This includes the effects of negative as well as positive associations. Further work is required on both the reproducibility of the clinical signs and the application of the criteria to other datasets.