Use of an administrative database to develop and test a model to predict the allocation of clinical pharmacy human resources.

Abstract

Pharmacy managers are handicapped by a lack of research in the area of resource allocation and utilization as it applies to clinical pharmacy staff. Using the decision support system at a tertiary care teaching facility in Ottawa, Ontario, two statistical models were developed and validated using over 7000 cases from each of two consecutive fiscal years. Multiple logistic regression analysis was used to identify the patient-level variables that were valid predictors of who would receive clinical pharmacy services: attending physician specialty, length of stay, complexity code, drug category. The model accurately classified 81.8% of patient encounters. A multivariate linear regression analysis was undertaken to identify the predictors of the cost of clinical pharmacy services: attending physician specialty, length of stay, drug category, drug cost, other hospital costs. These elements explained 43% of the variability in the outcome. The applicability of the models for daily operations and planning is described.