Comparison of two drug-burden computational procedures applied to anticholinergic drugs

Two computational procedures for estimating continuous drug-burden scores are compared, using a simulation study based on a sample of patients exposed to anticholinergic medications. Both procedures seek to quantify the medication burden to overcome the limitations of a binary classification of exposure. An artificial neural network procedure renders better statistical properties than a linear weighted score, even when both scores are correlated and are viable alternatives to binary classification.