Study Highlights

• The 2 hospitals included (hospital A and hospital B) had 400 and 326 beds, respectively; these facilities were 40 miles apart.
• The investigators undertook direct observation of 98 nurses (63 at 4 wards in hospital A and 35 in 2 wards in hospital B) in the specialty areas of geriatrics, respiratory, renal medicine, orthopaedics, and neurology and included medical and surgical patients.
• Data were collected between 2006 and 2007 at hospital A (340 hours of direct observation) and between 2007 and 2008 at hospital B (164.75 hours).
• Medications for patients were stored in locked bedside cabinets, and drugs of addiction (eg, opioids) and injectable medications were stored in locked medication rooms.
• 82% of nurses recruited participated.
• Errors were identified in the administration and preparation of medications.
• Observers followed a "serious error" protocol in which they intervened if they observed a potentially dangerous error.
• This occurred once during training sessions and 9 times during actual observations.
• A structured observational tool was developed and incorporated into software on a handheld personal digital assistant.
• Observers entered data about nursing procedures and details of medication dose, timing, and route, as well as the nurses' status (full time, part time, or on a casual basis; classification; and years of experience).
• Interrater reliability was assessed at 16 observation sessions before data collection, and 528 drug administrations were observed by 2 researchers. The κ values, ranging from 0.94 to 0.96, indicated high reliability of agreement.
• Direct observation was chosen because it offered more accurate data than chart or incident report review.
• Procedural errors were classified as errors such as failure to read medication labels, failure to check patient identification, failure to record medications given, or failure to check glucose levels before administration of antidiabetic medications.
• Clinical errors were classified as errors with drug type, dose, route of administration, timing, or administration of medications not ordered.
• The errors were then classified by severity on a 5-point scale.
• A total of 4271 administrations for 720 patients were observed (2592 for 514 patients at hospital A and 1679 for 206 patients at hospital B).
• Patient mean age was 72.6 years for hospital A and 67.5 years for hospital B.
• Only 19.8% of administrations were free of procedural and clinical errors.
• At least 1 procedural error occurred in 74.4% of total drug administrations, and 25.0% had at least 1 clinical error.
• The most frequent procedural error was not checking the patient's identification, which was checked in only 41.3% of administrations (n = 1732).
• The most frequent clinical error was wrong timing (n = 688), but only 4.1% of clinical errors were of major severity.
• Wrong intravenous administration was the second most frequent clinical error, with 35.7% rated as major.
• Interruptions occurred in 53.1% of all administrations.
• Procedural failure rates increased with the number of interruptions (69.6% for no interruptions and 76.7%, 78.7%, 84.6%, and 92.2% for failures with 1, 2, 3, and ≥ 4 interruptions, respectively).
• There was a significant linear correlation with procedural failures.
• For clinical errors, there was no significant linear correlation with the number of interruptions, but the errors also increased with interruptions.
• The rates for clinical errors were 25.3%, 22.5%, 24.4%, 38.9%, and 30.4% for no interruptions and for 1, 2, 3, and 4 or more interruptions, respectively.
• The associations were not affected by patient age, sex, nursing classification, age, or years of employment.
• Overall, every interruption was associated with a 12.1% increase in procedural failures and a 12.7% increase in clinical errors.
• The risk of at least 1 clinical error occurring increased with every round of administration for each patient, from 39% with no interruptions to 61% with 5 interruptions.
• Mean severity rating of drug administration failures and errors was 1.13, and 79.3% were insignificant (severity level 1), 2.7% were severe (severity level 4), and none were rated at level 5.
• The authors suggested that strategies were needed to reduce interruptions to nursing administration of medications to reduce error rates.