Of 55 992 new prescriptions processed and reviewed by the pharmacists in the study period, 532 were reported by pharmacists to contain inconsistent communication (0.95%). A computerized search of the 830 823 prescriptions written in our study period (which included both refills and new prescriptions) found 157 055 containing a comment in the free-text window (18.9%) (). From the latter group, we randomly selected 500 for manual review and excluded 9 written for glucose-monitoring supplies. From the remaining 491 prescriptions containing comments, 26 were found to contain inconsistent communication (5.3%); none reported by pharmacists. An absence of overlap between reported (n=532) and unreported (n=26) errors of inconsistent communication was not unexpected based on previous literature.17
In addition, pharmacists do not review refills routinely, and almost a third of the pharmacists reported no errors. Unreported errors discovered through the computerized search and review were used to estimate frequency of the problem. Based on a 5.3% error rate in prescriptions with comments, we estimated the 157 055 prescriptions with comments might contain 8324 prescriptions with inconsistent communication (5.3%). Because these estimated 8324 errors derive from a larger sample of prescriptions with both refills and new prescriptions, we estimate the overall rate of errors of inconsistent information to be 1% (8324 of 830 823).
lists the types of inconsistent communication errors in prescriptions reported by pharmacists, frequency of pharmacist contact with providers, and use of free-text information for final prescription fill. The most common inconsistent element across reported prescriptions was drug dosage (44.9%), followed by drug administration schedule (20.5%), duration of medication administration for inpatients (24.4%), medication administration time for inpatients (12.4%), route of administration (3.2%), and quantity of drug requested (1.7%). Pharmacists used the free-text comment in 88.9% of cases to fill the final prescription, the structured template in 9.2% of cases, and in combination in the remainder (1.9%). Pharmacists called providers in 149 cases (28.0%). In the 26 unreported errors, dosage inconsistency was also the most common (n=10; data not shown).
Types of Inconsistent Communication Errors in Prescriptions Reported by Pharmacistsa
lists the ordering provider characteristics, setting (inpatient or outpatient), medication classes, and other predictor variables of the 558 total study prescriptions containing inconsistent communication and the 465 controls. The following variables differed significantly between the 2 groups: ordering provider type (P=.04) and specialty (P<.001), inpatient vs outpatient setting (P<.001), and medication class (P<.001). Orders with inconsistent communication were most commonly entered by trainees (59.0%), followed by attending physicians (24.9%), physician assistants (10.2%), and nurse practitioners (5.4%). Most errors (68.3%) occurred in prescriptions for inpatients, and 19.0% (n=103) had 2 or more inconsistencies. The following medication classes were most commonly affected by errors: cardiovascular and antihypertensives (15.8%), analgesics (14.7%), antibiotics (10.8%), psychiatric medications (10.4%), and laxatives (7.5%). Only 2 reported errors were discovered after reaching the patient and did not result in harm. Medications needing a complex order (eg, a tapering dose of prednisone) were highly correlated with error. We also compared characteristics of reported and unreported errors in logistic regression analysis. Prescriptions with inconsistencies in schedule and administration route were less likely to be reported by pharmacists than prescriptions without schedule and route inconsistencies (OR, 0.18; 95% CI, 0.06–0.54 for schedule and OR, 0.14; 95% CI, 0.01–0.89 for route; data not shown in ).
Comparison of Prescriptions With and Without Errors of Inconsistent Communication in Computerized Provider Order Entrya
summarizes the results of the logistic regression model of predictors for inconsistent communication. Inpatient setting (OR, 3.30; 95% CI, 2.18–5.00) and non–primary care specialties such as surgery (OR, 2.45; 95% CI, 1.57–3.82) were more likely to have errors. Certain medication classes were also more likely to have errors, with the highest odds of error occurring for steroids (OR, 7.62; 95% CI, 3.11–18.63) compared with analgesics.
Logistic Regression Model of Predictors for Errors of Inconsistent Communication in Computerized Provider Order Entry
lists the potential harms associated with the 558 errors based on the Human Error Consequence Scale20 and the 3 most common drug classes for each category of harm. Even though 29.4% of errors were categorized as inconvenience, 112 errors could have resulted in moderate to severe harm (considerable, very serious, serious permanent damage, or death) (20.1%). Anticoagulants and cardiovascular drugs were the most frequent categories involving moderate to severe harm. summarizes the results of the logistic regression analysis of the 112 errors that could potentially have led to moderate to severe harm. Medications ordered by providers in the outpatient setting had higher risk (OR, 2.23; 95% CI, 1.15–4.35) for causing moderate to severe harm than those ordered for inpatients. Medications ordered by providers in specialties other than primary care and surgery had lower risk (OR, 0.30; 95% CI, 0.14–0.63) than those ordered by providers in primary care. Of all medication classes, anticoagulants were associated with the highest risk of causing potentially moderate to severe harm (OR, 61.89; 95% CI, 10.73–356.93). As expected, medications considered high alert (OR, 3.43; 95% CI, 1.62–7.27) had a high potential for causing moderate to severe harm. Of note, 23.1% of unreported errors could have resulted in moderate to severe harm, statistically no different from reported errors (P=.62).
Outcome Categories for Harm From Errors of Inconsistent Communication in Computerized Provider Order Entry
Logistic Regression Analysis of Predictors for Moderate to Severe Harma From Errors of Inconsistent Communication in CPOE