Objective. Medication errors are a common cause of iatrogenic morbidity and mortality. The incidence of medication errors in pediatric emergency departments (EDs) has not been described. The objective of this study was to describe the incidence and type of drug errors in a pediatric ED and determine factors associated with risk of errors.
Methods. A retrospective cohort study was conducted of the charts of 1532 children who were treated in the ED of a pediatric tertiary care hospital during 12 randomly selected days from the summer of 2000. Two pediatricians, blinded to other study variables, independently decided whether a medication error occurred and ranked it according to a severity score. Disagreement was resolved by consensus.
Results. Prescribing errors were identified in 10.1% of the charts. The following variables were associated in univariate analysis with an increased proportion of errors: patients seen between 4 am and 8 am (odds ratio [OR]: 2.45; 95% confidence interval [CI]: 1.10–5.50), patients with severe disease (OR: 2.53; 95% CI: 1.18–5.41), medication ordered by a trainee (OR: 1.48; 95% CI: 1.03–2.11), and patients seen during weekends (OR: 1.48; 95% CI: 1.04–2.11). Among trainees, there was a higher rate of errors at the beginning of the academic year (OR: 1.67; 95% CI: 1.06–2.64). Logistic regression revealed increased risk for errors when a medication was ordered by a trainee (OR: 1.64; 95% CI: 1.06–2.52) and in seriously ill patients (OR: 1.55; 95% CI: 1.06–2.26).
Conclusions. In the pediatric ED, trainees are more likely to commit prescribing errors, and the most seriously ill patients are more likely to be subjected to prescribing errors.
It has been estimated that between 44 000 and 98 000 people die each year in the United States as a result of medical errors.1 Medication errors are a common cause for iatrogenic adverse events2–4 and can lead to severe consequences, including prolonged hospitalization, unnecessary diagnostic tests, unnecessary treatments, and death.3,5–8
Preventable errors are significantly more common in emergency departments (EDs) than in other hospital departments,3,4 and prescribing errors are most common in pediatric and EDs.9,10 Prescribing and administering medications in the ED is particularly complex for several reasons. Medications frequently need to be given urgently; thus, many are kept in ward stock. Auditing of these drug orders by a pharmacist is not current practice at our institution. In addition, patients in the ED are seen only for a short encounter, and in many cases the physician does not have complete information regarding the patients’ medical and drug history.
The incidence of medication errors in the pediatric ED has not been fully described, and factors associated with such errors have not been elucidated. In 1 retrospective review of medication errors in a pediatric ED, incorrect recording of patients’ weights, leading to incorrect dosing, was the most frequently reported error.11 That study, based only on voluntary reporting, could not estimate the true incidence of medication errors. The objectives of our study were to estimate the incidence and type of drug errors in a pediatric ED and to identify factors that may be related to an increased risk of medication errors.
We conducted a retrospective chart review of patients who were treated at the ED of the Hospital for Sick Children in Toronto, Canada. All charts of patients who were treated in the ED during 12 randomly selected days in the summer of 2000 were reviewed for potential medication errors. The Research Ethics Board of the Hospital approved the study.
Two medical students served as research assistants. They attended 2 training sessions that included extraction of data from charts under the supervision of 1 of the investigators, before data extraction. They subsequently extracted the data into a database that included information about patients’ demographics, clinical condition, diagnosis, acuity of condition (based on triage category), the prescribing physician, and all of the medications prescribed and given to the patient. Physicians were categorized according to their level of training as staff or trainees (interns, residents, and fellows). When the drug regimen was different (dose differed from the recommended dose by 20% or more, deviation by 2 hours or more from the recommended interval between doses, wrong units or route of administration) from the recommended regimen, it was flagged as a potential error. Medications ordered and not given and medications given without a written order were also flagged as an error. Delays of 1 hour or less in drug administration, the absence of date, or unclear signature were not considered errors. To establish the accuracy of data collection, a senior investigator reviewed a random sample of 3% of the charts. The entire database was further reviewed by both senior investigators for potential drug interactions and drugs incorrectly prescribed for any specific condition.
Ranking the Errors
Two pediatric emergency physicians who were aware only of age, weight, diagnosis, and drug details but blinded to the other variables (eg, trainee versus staff physician, day of the week) independently ranked the errors. Drug errors were ranked as insignificant/minimal risk, significant, or severe:
Insignificant/minimal risk: a drug error with minimal likelihood of harm (eg, a child receiving 5 mg of dexamethasone instead of 3 mg).
Significant: a drug error that could cause non-life-threatening consequences, or an error that would result in a less effective treatment for the child’s condition (eg, a 10-fold lower dose of amoxicillin for otitis media).
Severe: any drug error that could cause death or decrease the chance of successful treatment of a life-threatening condition (eg, a 10-fold error in insulin dosage, significant underdosing of antibiotics for a patient with meningitis).
Assessment of the potential effect of each error was based on available pharmacological data and clinical judgment of the physicians. The reference for drug dosages was the hospital’s formulary.12 However, drug dosages that were different from the hospital’s formulary were not considered an error if recommended by the Compendium of Pharmaceuticals and Specialties,13 the manufacturer, or other medical literature. Knowing that the date and time of orders is frequently omitted in our department, especially for children with single, rapidly treated encounters, the absence of these data were not analyzed as an error. It was believed that inclusion of these data would obscure the more important lessons that could be learned from this study.
The independent rankings were compared in 2 forms: 1) whether a medication error occurred and 2) the severity ranking of the error. In cases in which the raters did not agree, the case was discussed in an attempt to come to a consensus. If agreement was not reached, then a third investigator reassessed the rating of that error and the median rank was used.
The errors were also classified by type according to the system described by Lesar et al.14 Briefly, errors were classified as wrong dose, wrong drug, wrong frequency, wrong route, errors related to patient information (medication given to a patient with known allergy to the drug), and others.
The interrater reliability was measured using a κ statistic. Descriptive statistics related to the incidence of medication errors were calculated. Logistic regression was used to assess the likelihood of medication error among all patients for whom a medication was prescribed. The independent relationship between each variable and the outcome variable (error vs no error) was examined. Variables found to be significantly associated with drug errors or those that might confound the relationship between other variables and drug errors were included in a multivariate analysis. Multivariable logistic regression was used to identify factors such as the child’s age, gender, level of training of the physician, shifts (day, evening, or night), level of activity in the ED (estimated using patient waiting time), and triage category (a proxy measure for severity of illness) that independently influenced the rates of drug errors. In addition, interaction terms were included to test the interaction between level of training and time of day, and level of training and triage category. Variables were retained in the multivariate analysis when they met the significance level of .05 or when they changed the point estimate of another variable by >10%.
There were 1549 visits to the ED during the study days. A total of 1532 charts (98.9%) were available for review. Twenty-five patients had more than 1 visit to the ED during the study period. The characteristics of the study population are presented in Table 1. The total number of medications ordered was 1678 with a median of 1 and a range of 0 to 12 medications ordered per patient.
Potential errors in medications prescribed and given in the ED or for home use were identified in 403 charts. A senior investigator reviewed a random sample of 50 charts and identified 1 additional error (not found by the research assistants). In 330 cases (81.8%), there was initial agreement between the reviewers regarding whether an error occurred and on the ranking of the error. In 49 of the remaining 73 cases, agreement was achieved between reviewers after the case was discussed. Twenty-four cases were reviewed by a third researcher.
The κ for initial agreement between reviewers on whether an error occurred was 0.79, and the κ for ranking of the errors was 0.70, indicating a high level of agreement. Figure 1 presents the cohort flowchart and the methodology applied to identify the errors.
Incidence and Type of Errors
Prescribing errors were identified in 154 charts (10.1%), and drug administration errors were identified in 59 (3.9%). The most common types of prescribing errors were dosing errors, followed by drugs given with incorrect frequency (Table 2).
Characteristics of Medications Involved
The drugs most commonly involved in errors (Fig 2) were acetaminophen followed by antibiotics, asthma medications, and antihistamines. These drugs were also those most commonly prescribed in the ED (data not shown).
Severity of Errors
Two errors were ranked as severe, 47.5% were ranked as significant, and 51.7% were ranked as insignificant/minimal risk. Some examples of medication errors are presented in Table 3.
Factors Associated With Increased Risk of Prescribing Error
For exploring factors associated with increased risk of prescribing errors, the following variables were included in univariate analysis of the 766 patients for whom a medication was prescribed: age group, triage category, waiting time before being seen by a physician, time of the day, weekday versus weekend, drug prescribed before or after July 1, and level of training of the prescribing physician. The odds ratios (ORs) associated with each of these variables are presented in Table 4. The rate of medication errors peaked between 4 am and 8 am (OR: 2.45; 95% [CI]: 1.10–5.50). The proportion of errors was higher in patients with more severe disease as defined by triage category (OR: 2.53; 95% CI: 1.18–5.41 for severe compared with mild disease).
A total of 420 patients were treated by staff physicians, and 346 were treated by trainees. There was a higher rate of errors among trainees when compared with staff physicians (OR: 1.48; 95% CI: 1.03–2.11). A total of 157 patients were treated by trainees at the beginning of the academic year, and 189 were treated at the end of the academic year. Among trainees, there was a higher rate of errors at the beginning compared with the end of the academic year (OR: 1.67; 95% CI: 1.06–2.64). Patients who were seen during the weekend were more likely to have an error compared with patients who were seen during weekdays (OR: 1.48; 95% CI: 1.04–2.11). There was no increased risk of errors in any of the age groups when compared with children in the oldest age group, 12 to 18 years (Table 4).
All variables that were found to have a significant relationship with medication prescribing errors in univariate analysis were included in a stepwise logistic regression. Factors that were found to be associated with an increased risk of medication errors were medication ordered by a trainee as opposed to staff person (OR: 1.64; 95% CI: 1.06–2.52) and serious compared with least ill patients (OR: 1.55; 95% CI: 1.06–2.26).
In this retrospective chart review, we found that approximately 10% of patients who are seen in a pediatric ED are subjected to medication errors. We also found that compared with staff physicians, trainees were more likely to commit errors and that medication errors were more common among seriously ill patients. Approximately 50 000 children are treated each year at the Hospital for Sick Children ED. Extrapolating the data from this study to a 1-year period suggests that an estimated 5000 children each year may be subjected to a medication error and that 2500 may be subjected to significant errors. A previous study,15 reporting a much lower error rate, was based on incident reporting, which has been shown to identify only a small percentage of adverse drug events.16 Indeed, none of the events that we identified was reported through the hospital’s incident report system. Another study17 found errors in 95% of prescriptions issued in the ED. However, because of concerns related to the study methodology (eg, the definition of error included omission of instructions to shake the container or to keep the drug refrigerated), the high incidence reported should be interpreted with caution.
Previous studies have found that medication errors occur frequently in the intensive care setting,18 perhaps as a result of the complexity and intensity of treatment and the use of unfamiliar medications.19,20 Although many patients who are seen at the ED do not have life-threatening conditions, some of the circumstances found in intensive care units also exist in the pediatric ED, which treats patients with a high acuity of illness and has a rapid turnover of patients. The incidence of medication errors that we found is comparable with the 6.6% to 11.7% incidence reported from pediatric intensive care units.15,21 Our results suggest that seriously ill patients are susceptible to higher rates of medication errors. This may be related in part to increased use of medication in such patients.
Even in a favorable setting, a substantial proportion of pediatric residents make mistakes while calculating drug doses.22,23 It is therefore not surprising that dosing errors were the most common errors made by physicians. These findings are consistent with previous studies identifying dosage calculations as the most frequent reason for medication errors in pediatric patients.14,24
We found a significantly higher rate of errors in the ED between 4 am and 8 am compared with 8 am to 12 pm. Although one may surmise that this is the result of physicians’ tiredness, after adjusting for other variables, this association was no longer significant, suggesting that the association cannot simply be attributed to physicians’ tiredness. The higher rate of errors at night might be the result of a lower proportion of staff physicians during night shifts or of less supervision of trainees’ work at night or because patients who are seen late at night were more seriously ill. Our sample size was not sufficient to test all of these hypotheses; however, they should be considered in future research.
We used the waiting time before being seen by a physician as an indirect measure of the workload in the ED, with waiting time of >2 hours (which occurred in 20% of the patients) as an indicator for increased workload. This factor was not associated with an increased risk of errors. However, an association may not be ruled out on the basis of these data for 2 reasons. First, the study was conducted in the summer, when the workload in pediatric EDs is traditionally lighter. Second, in our study, the most seriously ill patients were seen promptly; thus, the effect of workload, as estimated by waiting time, could not be assessed in these patients.
Our findings suggest that the most important risk factor for errors in the pediatric ED is the physician’s level of training. Two previous studies demonstrated that the rate of medication errors in a pediatric service doubled when new physicians started their rotation18 and that first-year residents have higher rates of prescription errors.25 The finding that trainees made more errors at the beginning of the academic year strengthens the plausibility of the association between level of training and the rate of errors. Contrary to these studies and to our findings, another study did not find differences in rates of errors made by attending physicians and residents.26 One explanation for the effect of level of training on the proportion of errors in our study may be related to the setting in which this study was conducted. The Hospital for Sick Children is a tertiary care pediatric hospital, and its ED is the primary training site in pediatrics for University of Toronto medical residents. Approximately 360 trainees at different levels rotate through the ED each year. Only a small proportion of these trainees are pediatric residents. Most of them are in their first year of residency with very limited experience in pediatrics. In contrast, staff physicians are academic pediatricians with specialized training in emergency medicine.
The possibility of system errors as an underlying cause for medication mishaps has to be investigated. Although individuals make mistakes, characteristics of the system in which they are working can make errors more likely to happen and more difficult to detect.27 It has been postulated that the best strategy to reduce drug errors is to improve the system by which drugs are given, thereby making it safer.28 Examples of such approaches are the use of a unit-dose system, which has been shown to reduce significantly medication errors in committed hospital wards,29,30 and the use of a computerized system for prescribing medications.31,32 Including a pharmacist as part of the clinical team and having a satellite pharmacy in the ward were also suggested as measures to reduce medication errors.15,33 These strategies could be adapted to the environment of the ED and then evaluated in clinical trials.
This study has several limitations. The retrospective design limited available data to those documented on the ED chart. This design could not detect many errors in drug administration. Other factors that could influence the medication error rate, such as the interaction among parents, patients, nurses, aides, noise, and concurrent events in the ED, were not analyzed because they were not documented on the chart and could not be measured accurately. Therefore, our analysis focused on prescribing errors. In many cases, we could not determine whether an adverse drug event occurred as a result of the error. ED patients are either admitted to inpatient units or discharged from the hospital from the ED. It was not possible to assess dosing intervals in admitted patients because ward physicians rewrote the orders. For discharged patients, we could not determine whether medications were given as ordered. Therefore, we included all documented errors in our analysis. Similar to other groups,2 we believe that identifying these errors, even ones that were intercepted before harm was done, is important in developing strategies to reduce adverse drug events. An additional limitation of the study is that the only information collected on the prescribing physician was level of training. We therefore could not examine the effect by each physician. However, >80 different physicians worked in the ED on the study days and only a few of them were sampled >3 times. Therefore, it is unlikely that clustering of errors by 1 physician significantly affected the results. The relatively small sample size limited our ability to conduct stratified analyses. For example, we compared staff physicians with trainees but were unable to compare among trainees with different levels of training. It is also important to note that because the study was conducted in a tertiary care pediatric hospital, the results (especially the findings regarding trainees) may not be generalizable to other hospitals.
Approximately 10% of patients who are seen in a pediatric EDs may be exposed to medication errors. Our findings suggest that errors are more common among seriously ill patients and that trainees are more likely to make medical errors, particularly at the beginning of the academic year. Future research should focus on the development and evaluation of strategies to reduce medication errors in the pediatric ED.
Dr Kozer was a recipient of a fellowship grant from the Research Training Centre, the Hospital for Sick Children. Dr Koren is a senior scientist of the Canadian Institutes for Health Research.
- Received November 9, 2001.
- Accepted March 15, 2002.
- Reprint requests to (E.K.) Division of Clinical Pharmacology and Toxicology, the Hospital for Sick Children, 555 University Ave, Toronto, Ontario, Canada M5G 1X8. E-mail:
Presented, in part, at the Annual Meeting of the American Society for Clinical Pharmacology and Therapeutics; March 6–10, 2001; Orlando, FL; and the Pediatric Academic Societies Meeting, April 28–May 1, 2001; Baltimore, MD.
- ↵Kohn LT, Corrigen JM, Donaldson MS. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 1999
- Ferner RE. Is there a cure for drug errors? BMJ.1995;311 :463– 464
- ↵Koren G, Barzilay Z, Greenwald M. Tenfold errors in administration of drug doses: a neglected iatrogenic disease in pediatrics. Pediatrics.1986;77 :848– 849
- ↵The 2000–2001 Formulary of Drugs. 19th ed. Toronto, Ontario, Canada: The Hospital for Sick Children; 2000
- ↵Compendium of Pharmaceuticals and Specialties. 35th ed. Ottawa, Canada: Canadian Pharmacists Association; 2000
- ↵Tisdale JE. Justifying a pediatric critical-care satellite pharmacy by medication-error reporting. Am J Hosp Pharm.1986;43 :368– 371
- ↵Rowe C, Koren T, Koren G. Errors by paediatric residents in calculating drug doses. Arch Dis Child.1998;79 :56– 58
- ↵Bates DW, Teich JM, Lee J, et al. The impact of computerized physician order entry on medication error prevention. J Am Med Inform Assoc.1999;6 :313– 321
- ↵Nightingale PG, Adu D, Richards NT, Peters M. Implementation of rules based computerised bedside prescribing and administration: intervention study. BMJ.2000;320 :750– 753
- Copyright © 2002 by the American Academy of Pediatrics