• medication errors;
  • outpatients;
  • pediatrics;
  • chemotherapy;
  • leukemia;
  • children


  1. Top of page
  2. Abstract


Although medication errors are 1 of the most common types of medical errors, their frequency in pediatric patients receiving oral outpatient chemotherapeutic agents is unknown. The prescribing, dispensing, and parental administration of these medications to children receiving treatment for acute lymphoblastic leukemia (ALL) were systematically reviewed to determine the rate and types of medication errors occurring in these patients.


During a 2-month study period, parents of children with ALL were contacted and asked to participate in the study before a regularly scheduled clinic appointment. At the visit, the parent demonstrated how each medication was administered. A pediatric oncologist reviewed the medical record to determine the correct treatment regimen for study patients. After comparing the correct treatment regimen with what was administered, a classification of “no medication error,” “medication error,” or “cannot determine” due to insufficient information was made for each indicated drug. Identified medication errors were subclassified as prescribing, dispensing, or administration errors.


Data on 172 chemotherapeutic medications for 69 patients were analyzed. One or more errors occurred with 17 of the 172 (9.9%) medications; a classification of “cannot determine” was made for 12 (7.0%) medications. Among the 17 medication errors there were 12 (7.0%) administration errors and 5 (2.9%) prescribing errors. There were no pharmacy dispensing errors. All errors were due to incorrect dosing or failure to administer an indicated medication. At least 1 medication error occurred in 13 of the 69 (18.8%) study patients.


Prescribing and administration medication errors occurred with nearly 10% of chemotherapeutic drugs administered to outpatient children with ALL. Systematic changes, including computerized physician order entry and simplification of treatment protocols, should be considered. Cancer 2006. © 2006 American Cancer Society.

The Institute of Medicine's report, To Err Is Human: Building a Safer Health System,1 identified medical errors as a major public health problem in the U.S. Based on the results of 2 large studies,2, 3 it is estimated that as many as 44,000 to 98,000 patients die each year in the U.S. as a result of medical errors.1 If this estimate is correct, medical errors rank as the eighth leading cause of death in this country.1 Clearly, much research is needed to improve patient safety.

Medication errors are 1 of the most common,2, 3 and most studied, type of medical error. In order for a patient to receive the proper medication in an inpatient setting, the practitioner must prescribe the drug correctly, the pharmacist must understand and dispense the order accurately, and the nurse must administer the medication properly. Although the vast majority of research has been conducted on inpatients, the potential for prescribing errors and dispensing errors is similar in outpatient settings. Most important, proper administration of the medication is usually the responsibility of the patient or caregiver rather than a healthcare professional. Thus, the potential for medication errors among ambulatory patients is substantial.

The problem of medication errors among children receiving oral outpatient chemotherapeutic agents is particularly significant. Because these medications are toxic and have narrow therapeutic windows, the results of an error may be disastrous.4 Fortunately, most children receiving chemotherapy are on a specific treatment protocol, making identification of prescribing and dispensing errors easier than in other areas of pediatrics where medication decisions are more subjective. With accurate identification of these errors, changes in the systems of healthcare delivery can be made to improve medication safety. Finally, although parents of children receiving outpatient oral chemotherapy may be highly invested in properly administering the prescribed drugs, the number of required medications and complexity of dosing may be challenging for parents without medical training.

We conducted a study to assess the frequency of oral outpatient chemotherapeutic medication errors in children with acute lymphoblastic leukemia (ALL). Errors were classified as occurring during the prescribing, dispensing, and/or administration phase. The goal of the project was to identify errors so that specific changes in systems could be designed and implemented to improve pediatric patient safety.


  1. Top of page
  2. Abstract

A prospective case series study was conducted. Eligible patients were children age <18 years who were diagnosed with ALL without history of recurrence, were receiving oral outpatient chemotherapeutic agents, and were scheduled to be seen at the Hematology/Oncology Clinic at Children's Hospital and Regional Medical Center (CHRMC) during a 2-month time period from mid-April to mid-June 2005. All patients received chemotherapy as part of a Children's Oncology Group (COG) study or standard treatment based on the results of recent COG studies. Several methods were used to identify potentially eligible study participants. First, the list of patients scheduled to be seen in the Hematology/Oncology Clinic during the study period was reviewed and those children with ALL identified. In addition, healthcare providers in the clinic were asked to identify potential study children by reviewing lists of patients in their panels. Inpatients at CHRMC who were newly diagnosed with ALL during the study period and scheduled to be followed at the Hematology/Oncology Clinic were also recruited for the study. Finally, each day during the study period the study research assistant reviewed the list of patients to be seen in the Hematology/Oncology Clinic to determine whether any children with ALL who had not been previously recruited for the project were scheduled to be seen.

The parents or guardians of identified potential study patients were sent a letter explaining the project before their child's Hematology/Oncology Clinic appointment. Parents were asked to complete and return a postcard to indicate whether they were interested in participating in the project. Those parents who expressed interest in participating were contacted by the study research assistant, who asked that the person who usually administered outpatient chemotherapeutic medications to the child attend the upcoming clinic appointment and requested that they bring in all bottles of medications that had been given to the patient during the previous month. Telephone contact was attempted for parents who did not respond to the initial mailing. During this telephone call the research assistant explained the study and, for those parents interested in participating, reviewed the protocol in detail.

At the time of a Hematology/Oncology Clinic appointment by a study patient, the research assistant met with the parent or other caregiver and asked him/her to describe and demonstrate how each chemotherapeutic agent was given to the child. Specifically, the parent was asked how frequently the medication was given, the duration of therapy, and to show the number of tablets or volume of liquid administered, using the medication bottles to help her/him answer. Demonstration of the dose(s) administered was done using the child's medications. If the parent failed to bring the medications, or if the family had been recruited at the time of the visit, samples of the drugs commonly given to children with ALL were available for the demonstration. In addition, the research assistant reviewed the information on each bottle of medication brought in to the clinic by the parent and recorded the name of the pharmacy in which the drug was dispensed and the dosing instructions. Interpreters were provided for parents who were not conversant in English. In addition to the information regarding medications, data on the education level and native language of the parent or guardian who administered the chemotherapeutic agents to the child were collected at the visit.

After the visit, a pediatric oncologist (D.S.H.) reviewed the medical record of each enrolled patient, without any knowledge of the data collected by the research assistant. In the review, the oncologist noted the child's weight and/or body surface area at the previous visit and any pertinent laboratory data that might affect dosing of chemotherapeutic agents. Additional information abstracted from the medical record included length of time from diagnosis, phase of treatment, and initial risk classification. Each patient receiving care at the CHRMC Hematology/Oncology Clinic has a treatment roadmap included in her/his medical record. This roadmap provides specific information about which medications are to be prescribed at each stage of treatment, doses of these drugs, and modifications of these doses based on changes in the patient's condition. By carefully reviewing the roadmap, anthropometric and laboratory data, the oncologist determined the medications, doses, and frequency of dosing that each study patient should have received during the month before the study visit. In addition, the oncologist reviewed the medical record to determine if any telephone encounters with the patient's family had occurred since the previous visit that might have led to alterations in the administration of any chemotherapeutic medications.

The data on medication administration collected by the research assistant and the information on indicated medications, as determined by the pediatric oncologist, were independently screened by a pediatric oncology pharmacist (L.W.) and a pediatrician with experience in pediatric patient safety research (J.A.T.). For each oral chemotherapeutic medication indicated for a study patient, the reviewer determined whether the drug had been administered correctly, if there was a discrepancy between what was indicated and what was administered, or if it could not be determined if there was a discrepancy. A discrepancy was considered to have occurred if: an indicated medicated was not given, a nonindicated chemotherapeutic drug was administered, the duration of treatment was different, or if there was a >10% difference between indicated and administered dose of any medication. If both reviewers determined that there was no discrepancy, a classification of no error was made for an indicated chemotherapeutic medication. If either or both reviewers determined that a discrepancy existed, or if they could not determine if there was a discrepancy, the medical record was reviewed and a classification of “no error,” “error,” or “cannot determine” was made by consensus after discussion with the research team. If necessary, prescription information was collected from the pharmacy that dispensed the medication. If it was determined that a medication error occurred, classification as an administration error (by the parent or guardian), prescribing error, or dispensing error (by the pharmacy) was made after discussion among the 3 investigators. A final classification of “cannot determine” was made in situations in which the medication bottles were not brought to clinic, and the description of the administration by the parent or guardian was not specific enough to determine if an error had occurred.

Much of the analysis of study data was descriptive in nature. The rates and 95% confidence intervals (95% CI) of prescribing, dispensing, and administration errors were calculated, as was the overall rate of medication error. To determine whether errors were more likely to occur with some medications than others, logistic regression was used, after accounting for the clustering of multiple drug administrations in the same child. In addition, the rate of 1 or more medication errors in children of parents with higher levels of education (defined as education beyond high school) was compared with that of patients with parents with less education; differences were assessed with a chi-square test. This statistical test was also used to compare error rates in children whose parents spoke English or had another native language. Student t tests were used to assess the association between length of time from diagnosis, or age and medication errors, and chi-square tests were used to compare error rates during different phases of treatment or in patients in different patient risk classification groups.

The study was approved by the CHRMC Institutional Review Board. Written informed consent for participation was obtained from the parents of study children.


  1. Top of page
  2. Abstract

A total of 92 potentially eligible children with ALL were identified for possible participation in the study. One child had a recurrence during the study period and another developed a serious illness that precluded participation. Of the 90 remaining eligible patients, 70 (77.8%) were enrolled. The parent of 1 study child could not remember the administration of any of the child's medication, thus data were analyzed on 69 patients (76.7% of those eligible). The disposition of all potentially eligible patients is shown in Figure 1.

thumbnail image

Figure 1. Disposition of children identified as potentially eligible for the study.

Download figure to PowerPoint

Characteristics of study children and their parents are summarized in Table 1. At the study visit, parents of 49 (71%) children brought in all bottles of chemotherapeutic medications administered during the previous month, whereas 57 (80%) brought at least some of the medication bottles. Among the medication bottles available for review, 70.3% had been dispensed by the CHRMC pharmacy. A total of 172 chemotherapeutic medications were administered to the 69 study children with ALL. Five different drugs were administered: mercaptopurine, methotrexate, thioguanine, prednisone, and dexamethasone. Overall, 1 or more medication errors occurred with 17 of the 172 chemotherapeutic medications (9.9%, 95% CI, 5.9–15.4%); a classification of “cannot determine” was made for 12 drug administrations (7.0%; 95% CI, 3.7–11.9%). Among the 17 medication errors, 12 were categorized as administration errors (7.0% of all medication administrations; 95% CI, 3.7–11.9%), and 5 were prescribing errors (2.9%; 95% CI, 1.0–6.7%). No pharmacy dispensing errors were detected.

Table 1. Characteristics of Study Patients and Their Parents
  • *

    Parent or caregiver who usually administered chemotherapeutic agents.

Median patient age, y (range)6.43 (1.3–17.9)
Median time from diagnosis, wk (range)58.7 (0.7–166.6)
Risk classification
 Infant1 (1.5%)
 High risk19 (27.9%)
 Standard risk49 (72.1%)
Phase of treatment
 Induction4 (5.8%)
 Consolidation7 (10.1%)
 Delayed intensification10 (14.5%)
 Maintenance48 (69.5%)
Parent/caregiver education level*
 Less than high school graduate10 (14.5%)
 High school graduate13 (18.8%)
 Some college18 (26.1%)
 College graduate25 (36.2%)
 Not recorded3 (4.3%)
Parent/caregiver native language
 English54 (78.3%)
 Spanish11 (15.9%)
 Other4 (5.8%)

Details of administration errors are presented in Table 2. Nine of the 12 errors detected were related to corticosteroid medications. The medication bottle was unavailable for 6 of the 9 detected administration errors with corticosteroids. In these cases, parents were asked to recall the details of administration of a drug that had occurred up to 4 weeks before the visit. In situations in which a parent recalled a dosing regimen that was impossible (e.g., giving 1 tablet of dexamethasone twice daily when the prescribed dose was 2.5 mg twice daily), a classification of administration error was made. All three of the administration errors with medications other than corticosteroids were related to incorrect dosing, with an average discrepancy of 15.4% between the prescribed and administered dose.

Table 2. Description of Administration Errors
Indicated/prescribed medication*Administered medication*
  • BID indicates twice a day.

  • *

    Unless indicated there was no variance in dosing frequency and duration of treatment indicated/prescribed and duration of medication administration.

Mercaptopurine, 25 mg 6 d/wkMercaptopurine, 25 mg 7 d/wk
Methotrexate, 16.25 mg every wkMethotrexate, 13.75 mg every wk
Mercaptopurine, 50 mg dailyMercaptopurine, 50 mg daily Monday-Friday, 75 mg daily Saturday/Sunday
Prednisone, 50 mg a.m., 45 mg p.m.No prednisone administered
Dexamethasone, 2 mL BIDDexamethasone, 1.8 mL BID
Dexamethasone, 2 mg a.m., 1.5 mg p.m.Dexamethasone, 4 tablets BID
Dexamethasone, 2.5 mg BIDDexamethasone, 1 tablet BID
Dexamethasone, 2.5 mg a.m., 2 mg p.m.No dexamethasone administered
Prednisone, 20 mg a.m., 25 mg p.m.Prednisone, 5 mg BID
Prednisone, 40 mg a.m., 20 mg p.m.Prednisone, “1.5 or 2” tablets BID
Dexamethasone, 2 mg BIDDexamethasone, 1 mg BID
Dexamethasone, 2 mg BIDNo dexamethasone administered

Information on the 5 prescribing errors is shown in Table 3. All of the prescribing errors were miscalculated doses, with an average discrepancy of 22.0% between indicated and prescribed dose. The most significant prescribing error was a 57% overdose of the weekly indicated dose of mercaptopurine. This occurred in a child in whom a 50% reduction in the dose of mercaptopurine was indicated because of previous neutropenia. Although the dose was reduced, the prescribed dose was still higher than was indicated in the patient's roadmap. For the other administration errors, the discrepancy between indicated and prescribed dose ranged from 10.7% to 25.0% and were generally the result of “rounding errors” in which the provider incorrectly determined the best match between indicated dose based on weight or body surface area and available dosage forms of a specific medication.

Table 3. Description of Prescribing Errors
Indicated medication and dose*Prescribed and administered medication*
  • BID indicates twice a day.

  • *

    Unless indicated there was no variance in dosing frequency and duration of treatment indicated/prescribed and duration of medication administration.

Mercaptopurine, 112 mg dailyMercaptopurine, 100 mg daily
Mercaptopurine, 25 mg dailyMercaptopurine, 50 mg daily Monday-Thursday, 25 mg daily Friday-Sunday
Methotrexate, 10 mg weeklyMethotrexate, 12.5 mg weekly
Dexamethasone, 2.5 mg a.m., 2 mg p.m.Dexamethasone, 2 mg BID
Dexamethasone, 2.5 mg BIDDexamethasone, 2.5 mg every a.m., 2 mg every p.m.

Error rates for each of the specific chemotherapeutic medications are shown in Table 4. Compared with methotrexate, errors were significantly more common with prednisone (odds ratio [OR] 7.7; 95% CI, 1.1–53.4) and dexamethasone (OR 5.8; 95% CI, 1.6–20.8). There were no other significant differences in error rates between different medications. Among the 69 study patients at least 1 medication error was detected in 13 children (18.8%; 95% CI, 10.4–30.1%). The rate of medication errors was similar in children whose parents spoke English or had another native language (20.4% and 13.3%, respectively; P = .54). There was also no statistically significant difference in the rate of errors between children whose parent had a high school education or less (26.1%) and those whose parent had more education (16.3%, P = .34). All the medication errors occurred in patients who were in the maintenance phase of treatment. Compared with all other phases of treatment, the proportion of children in whom an error occurred was significantly higher among those in the maintenance phase (27.1% vs. 0%; P = .008). Neither length of time from diagnosis or patient risk classification were significantly associated with the risk of a medication error. The age of the child was not associated with whether an error occurred (P = .95). However, 3 patients (ages 13.0 years, 16.6 years, and 17.9 years, respectively) primarily administered their own medications. One or more medication administration errors were noted for 2 of these patients.

Table 4. Error Rates with Different Chemotherapeutic Medications
DrugNo. of prescriptionsNo. of errors (%)No. “Cannot determine” (%)*
  • *

    Information was insufficient to determine whether an error had occurred.

Methotrexate492 (4.1)1 (2.0)
Mercaptopurine564 (7.1)1 (1.8)
Thioguanine50 (0)0 (0)
Dexamethasone478 (17.0)7 (14.9)
Prednisone153 (20.0)3 (20.0)


  1. Top of page
  2. Abstract

The results of this study indicate that a medication error occurred in nearly 10% of chemotherapeutic agents prescribed for pediatric patients with ALL in an ambulatory setting. Perhaps more important, at least 1 error occurred in 18.8% of children. Thus, greater than one-sixth of the patients in this study were not actually receiving the chemotherapeutic regimen indicated. Although greater than two-thirds of the errors detected were related to administration, errors in prescribing also occurred. Most of the errors were probably of limited clinical significance and unlikely to alter the probability of survival. However, the 3 children who failed to receive a monthly pulse of corticosteroid may be at increased risk for a recurrence, and the patient who received a 57% overdose of mercaptopurine despite previous neutropenia might have been placed at a significantly increased risk of infection. Because of the size of our study and the lack of long-term outcome data, it is not possible to determine whether these medication errors could contribute to treatment failure.

Because of the paucity of data on medication errors in pediatric patients in general, and particularly among those seen in the ambulatory setting, it is difficult to place the results of this study in proper context. The 2.9% rate of prescribing errors that we found among pediatric oncology providers was not dissimilar from the 5.7% rate of prescribing errors reported in a large study of pediatric inpatients.5 The difference in rates is even less than seems because many of the errors detected in pediatric inpatients were trivial in nature, such as a missing date. Gandhi et al.6 recently reported a 3% error rate among 3244 patients receiving chemotherapy in 3 outpatient infusion centers; 77% of the medication errors occurred during the prescribing process. In that study, the rate of errors was higher in adult patients than pediatric patients (4% and 1%, respectively).6 This same group of investigators found a 7.6% rate of prescribing errors among adult patients seen in a primary care setting for a variety of conditions.7 The rate of outpatient prescribing errors may be higher in pediatric patients. McPhillips et al.8 recently reported that up to 15% of prescriptions written for children seen as outpatients were potentially errors. This analysis was based on comparing the prescribed dose/weight for selected medications to 110% of the upper and lower recommended doses on a weight basis in a given child, and projected the maximum possible daily dose of drugs prescribed on an “as needed” basis. The rate of true errors among the drugs and patients reviewed was undoubtedly lower.

There are even fewer comparative data for the administration errors. Although not directly comparable, Barker et al.9 observed nursing personnel administer medications to patients in 36 different healthcare facilities. An error was noted in 19% of 3213 medication administrations. However, 43% of these administration errors were related to timing of the dose, a variable unmeasured in our study. In the report by Barker et al.,9 47% of the administration errors were related to either omission of a dose of medication or administering an incorrect dose. The rate of these types of errors was 8.9%, slightly higher than the 5.9% rate of administration errors that we detected with parents giving the medication to their children. These data suggest that parents may be able to administer oral chemotherapeutic agents as well as medical professionals.

The rate of administration errors was likely influenced by multiple, competing factors. Parents of children with ALL are motivated to provide medications correctly because of the severity of the illness. In addition, most of the parents in this study had considerable experience in providing the chemotherapeutic agents. However, the dosing regimens were frequently complicated, with different doses prescribed for different days of the week, and, in the case of corticosteroids, a combination of 2 different size tablets to make up the prescribed dose. Because of these unique factors, the results of this study may not be generalizable to parents administering medications to children for other conditions.

We did not detect any pharmacy dispensing errors in this study. The upper limit of the 95% CI for a point estimate of 0% with a sample size of 172 prescriptions is 1.74%,10 which is at the lower end of the rate of dispensing errors found in other studies on unselected medications.11, 12 The majority of the chemotherapeutic agents were dispensed by the pharmacy at CHRMC staffed by pharmacists experienced with these drugs. This may account, in part, for the lack of dispensing errors detected.

Elimination of many of the errors we observed might best be accomplished by a human factors approach. This approach has proven to be effective in a number of high-risk industries, as well as in the reduction of medical errors.13 Most of the prescribing errors were the result of miscalculation of doses. One strategy to reduce the frequency of miscalculations would be to require 2 provider cosignatures for oral medications (as our institution already requires for clinic- or hospital-administered chemotherapy). Although not a panacea, the implementation of computerized physician order entry (CPOE) with a dose calculator could reduce this type of error.14 By linking CPOE to a standardized dosing table for chemotherapy agents by body surface area, dose miscalculations could be reduced. The majority of errors observed involved corticosteroid medications. If manufacturers used a specific shape, size, and/or color for each specific dose, there might be less confusion by parents. Perhaps more important, consideration should be given to modifying treatment protocols such that a given child receives the same morning and evening dose of corticosteroid medication and/or designing regimens so that patients are not prescribed 2 different strength tablets of dexamethasone or prednisone.

To further delineate the frequency and types of administration errors, future studies might include home visits by a research assistant. During the visit the research assistant could observe the actual administration of chemotherapeutic medications and conduct pill counts or measure volume of liquid medications to estimate compliance. In addition, future research efforts might include an assessment of interventions designed to reduce the frequency of administration errors such as the use of systematic instruction on proper administration, the use of calendars indicating the specific drug and dose to be given on each day of the month, or pill boxes containing the exact medications to be given during a specific time period.

Limitations in our study design necessitate cautious interpretation of our results. Our sample size was modest, preventing the detection of small differences in errors rates between different patient populations. The calculated rate of medication errors was likely biased by several factors. We observed parents preparing 1 hypothetical dose of each prescribed medication in a research setting; whether or not parents actually gave this dose to the child at home is unknown. Poor compliance could be considered an administration error. Using a variety of methodologies researchers have estimated that 10% to 40% of children with cancer do not receive all prescribed doses of oral chemotherapeutic agents due to noncompliance.15–18 Many of the administration errors were related to corticosteroids that had been given to the child several weeks before the study visit. It is possible that parents did not correctly recall how they administered the medication to their child. Our ability to determine whether an error occurred was limited when families did not bring the patient's medication bottles to the study visit. Thus, we were unable to determine if an error had occurred with 7% of the medications. Because modifications in dosing regimens may have been made after communication between providers and parents that was not documented in the medical record, we might have overestimated the rate of prescribing errors. Data were collected during a 2-month period; it is unclear whether the patients enrolled during this period were representative of all children seen in the Hematology/Oncology Clinic with ALL.

Despite these limitations, the results of this study have implications for the evaluation and design of treatment protocols for children with ALL. First, if our estimate that a medication error occurs in 18.8% of patients is correct and generalizable to other centers providing pediatric oncology care, it is possible that the efficacy of treatment regimens is reduced or toxicity increased because not all children are receiving the chemotherapeutic agents as indicated. In addition to investigating novel medication regimens, strategies to eliminate or substantially reduce oral chemotherapy errors could be incorporated into treatment protocols. Finally, in designing new protocols a balance needs to be struck between the precision of dosing regimens and simplification so that medication errors are minimized.


  1. Top of page
  2. Abstract