Objective: To compare the distribution, causes, and consequences of medication errors in the ICU with those in non-ICU settings.
Design: A cross-sectional study of all hospital ICU and non-ICU medication errors reported to the MEDMARX system between 1999 and 2005. Adjusted odds ratios are presented.
Setting: Hospitals participating in the MEDMARX reporting system.
Measurements and Main Results: MEDMARX is an anonymous, self-reported, confidential, deidentified, internet-accessible medication error reporting program that allows hospitals to report, track, and share medication error data. There were 839,553 errors reported from 537 hospitals. ICUs accounted for 55,767 (6.6%) errors, of which 2,045 (3.7%) were considered harmful. Non-ICUs accounted for 783,800 (93.4%) errors, of which 14,471 (1.9%) were harmful. Errors most often originated in the administration phase (ICU 44% vs. non-ICU 33%; odds ratio 1.63 [1.43–1.86]). The most common error type was omission (ICU 26% vs. non-ICU 28%; odds ratio 1.00 [0.91–1.10]). Among harmful errors, dispensing devices (ICU 14% vs. non-ICU 7.1%; odds ratio 2.09 [1.69–2.59]) and calculation mistakes (ICU 9.8% vs. non-ICU 5.3%; odds ratio 1.82 [1.48–2.24]) were more commonly identified to be the cause in the ICU compared to the non-ICU setting. ICU errors were more likely to be associated with any harm (odds ratio 1.89 [1.62–2.17]), permanent harm (odds ratio 2.45 [1.17–5.13]), harm requiring life-sustaining intervention (odds ratio 2.91 [1.86–4.56]), or death (odds ratio 2.48 [1.18–5.19]). When an error did occur, patients and their caregivers were rarely informed (ICU 1.5% vs. non-ICU 2.1%; odds ratio 0.63 [0.48–0.84]) by the time of reporting.
Conclusions: More harmful errors are reported in ICU than non-ICU settings. Medication errors occur frequently in the administration phase in the ICU. When errors occur, patients and their caregivers are rarely informed. Consideration should be given to developing additional safeguards against ICU errors, particularly during drug administration, and eliminating barriers to error disclosures.