Fidelity of Electronic Documentation for Reactions Prompting Premedication to Iodinated Contrast Media

Purpose: The aims of this study were to assess the fidelity of electronic health record documentation prompting premedication to iodinated contrast media and to determine the appropriateness of administered premedication on the basis of that documentation. Methods: In this retrospective quality assurance cohort study, medication adverse events recorded in electronic health records between January 1, 2018, and August 31, 2019, to "iodine," "iodine-containing products," and "iodinated contrast media" were identified (N = 4,309); entries missing documentation (n = 1,651) and breakthrough reactions (n = 22) were excluded. Reaction description, severity, and free-text comments were used to categorize each entry as concordant (documentation matches recorded severity per the ACR Manual on Contrast Media version 10.3), discordant (description-severity mismatch, agent unrelated to iodinated contrast media, not a hypersensitivity reaction), or unclear. A subset of patients undergoing premedication was identified, and premedication was categorized as appropriate, inappropriate, or unsure on the basis of the index reaction using the aforementioned framework. Descriptive statistics were calculated. Results: There were 2,636 adverse event entries in 2,441 patients: 59.9% (1,578 of 2,636) were classified as concordant, 30.2% (797 of 2,636) as discordant (n = 377 not a hypersensitivity reaction, n = 317 description-severity mismatch, and n = 103 unrelated agent), and 9.9% (n = 261) as unclear documentation. For the premedicated subset, concordance classification was feasible for 202 unique patients premedicated 335 times. Premedication was appropriate in 72% (240 of 335) and inappropriate in 22% (73 of 335); 17% of premedication events (56 of 335) were inappropriately administered for a prior physiologic reaction. Conclusions: Premedication prompts in the electronic health record are often erroneous because of inaccurate coding, incomplete data, and reaction misclassification. These errors result in inappropriate premedication for a substantial minority of patients.