Assessment of Adverse Drug Events Among Patients in a Tertiary Care Medical Center

Philip E. Johnston; Daniel J. France; Daniel W. Byrne; Harvey J. Murff; Byron Lee; Renee A. Stiles; Theodore Speroff


Am J Health Syst Pharm. 2006;63(22):2218-2227. 

In This Article


In total, 84,021 admissions occurred during the study period. After removing 23,708 duplicate admissions from the control group, 59,531 were retained in that group for the analysis. A total of 782 AEs were reported and included a combination of ADRs (n = 83) and medication errors (n = 699). Medication errors that did not cause patient harm and process errors that did not involve patients were removed from the database (n = 107), leaving 675 patients with AEs to be included in the analysis. The number of control patients and those with AEs totaled 60,206. This population was then randomly split into two equal groups of patients: the training data set (n = 30,103) and the validation data set (n = 30,103) (Figure 1). Initial comparisons involved the training data set, consisting of 29,764 admissions with no AEs and 339 with an AE. While there was no significant difference in mean age of the subjects between the two groups (p = 0.591), there was a significant difference in the age groups that were more susceptible to an AE (p < 0.001) ( Table 1 ). AEs occurred in a higher percentage of patients who were <1 year of age, 1-15, 47-59, and ≥60 years than in other groups (Figure 2). A higher percentage of AEs were reported in men than women, but the groups were not significantly different when comparing those with an AE and those without an AE (p = 0.089). Among the various races of patients involved, the highest percentages of AEs were reported for whites and blacks. While the Asian Indian race showed a high rate of AEs, we believe the rate is a statistical artifact, reflecting its very small percentage in the study population. No significance among races was found between the AE and non-AE groups (p = 0.071).

Figure 1.

Flowchart of patient record selection and methodology. ADRs = Adverse Drug Reactions; AE = Adverse Event.

Figure 2.

Nonparametric regression estimates of the relationship between age and the proportion of patients with an adverse event.

Evaluation of admission sources showed that a doctor's office, clinic referrals, and local hospital transfers accounted for higher rates of AEs than other sources and that patients experiencing an AE were more likely admitted from these sites (p < 0.001).

Insurance class was evaluated ( Table 1 ) for significance. There were significant differences (p < 0.001) between the control and AE groups in the third party, Medicaid (out of state), and Medicare part A patients.

A comparison of diagnoses between the AE and control groups is provided in Table 2 . Nine of the diagnostic groupings evaluated were significantly different between the AE and control groups (p ≤0.001). These diagnoses included respiratory failure; multiple myeloma; blood disorders; disorders of fluid, electrolytes, or acids-bases; respiratory distress syndrome; myeloid leukemia; pneumonia; congestive heart failure; and aplastic anemia.

A comparison of the top medications associated with an AE compared with the use of these agents in patients without an AE is provided in Table 3 . Of the several hundred medications used by these patients, 10 agents and three drug groups were significantly more likely to be associated with an AE (p ≤0.001). The individual medications were vancomycin, lorazepam, dobutamine, insulin, phenobarbital, theophylline, heparin, cyclosporine, gentamicin, and tobramycin. Medication groupings associated with a large number of AEs included antineoplastics and blood formation and coagulation agents.

When the validity of the findings from the training set was compared with the validation set, age, high-risk admission sources (doctors' offices or the local hospital), and high-risk insurance class (third party, Medicaid [out of state], or Medicare part A) were found to be important risk factors for an AE, with odds ratios ranging from 0.9 to 1.5 for the various measures. To determine if each measure significantly differed between that set of admissions and the patients who had an AE, controlling for all other variables, p values were calculated. High-risk admission sources remained statistically significant in the validation data set ( Table 4 ). When performing the same statistical evaluation for diagnosis measures, all were higher than expected, and in every case the measure was statistically different, verifying these nine diagnoses as key indicators for AEs. Finally the results of a three-stage logistic regression model are shown in Table 5 . All medications included in Table 5 were more often used in patients with AEs than in patients without AEs, both in the training set and in the validation set, confirming the need to further investigate the risks of these medications. In the validation set, the occurrence of the same medications (except phenobarbital, lorazepam, warfarin, and blood formation and coagulation agents) was significantly different (p < 0.05), indicating that these may be the most commonly involved in AEs at our institution.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.