Predicting urodynamic dysfunction from clinical features in incontinent elderly women

In a retrospective survey of 296 elderly women referred with urinary incontinence and in whom urodynamic studies were completed, multiple logistic regression was used to determine the optimal set of clinical features to predict the urodynamic dysfunction. Three formulae were obtained expressing the predicted probabilities (p) of the urodynamic dysfunctions as functions of the clinical variables. By choosing the cut-off values of p which gave equal weighting to sensitivity and specificity, we found that clinical features (symptoms, signs, residual volume) are helpful in diagnosing the presence of unstable detrusor and voiding dysfunction, but not incompetent urethra.