Potential drug interactions in a high risk ambulatory elderly population

Objective. This study was designed to determine the rates of occurrence of potential drug-drug interactions in ambulatory patients aged 60 years or older who were taking medications to treat chronic conditions. Variables which differentiated patients with potential drug-drug interactions from patients who had no interactions were examined. Design. Volunteer community pharmacists (N = 102) enrolled 762 patients who met criteria on use of medications for chronic conditions into the study. Detailed medication histories were obtained for these patients through telephone interviews. Analysis. All pairwise drug combinations taken by patients were compared with those cited in Hansten and Horn's Drug Interactions reference source as being of 'major clinical significance.' Results. Eighty-two patients (10.8%) were identified as having at least one potential drug interaction. Ninety-five percent of patients with at least one drug-drug interaction were taking four or more prescription drugs concurrently. The number of prescription drugs being taken by patients with at least one potential drug interaction (X̄ = 8.1, s.d. = 3.1) was significantly higher (t = 3.48, p < .001) than the number being taken by those with no interactions (X = 6.2, s.d. = 2.9). Results of the logistic regression indicated that, as the number of concurrent prescriptions per patient increased by one, the probability of having at least one drug interaction increased by 22.4% (controlling for age of the patient, having more than one prescribing physician and obtaining prescriptions from more than one pharmacy). None of the other variables in the model were significantly related to the probability of having an interaction. The drugs most commonly involved in the drug interactions are reviewed. Conclusions. Physicians and pharmacists who wish to target patients at greatest risk of problems resulting from problematic combinations should probably set the criteria based on number of drugs being used. Other variables, such as number of prescribers, may not add to the predictive ability of number of drugs in identifying patients who may be at risk.

Duke Scholars

Author Jane Frances Pendergast Biostatistics & Bioinformatics, Division of Biostatistics