Characteristics of pediatric chemotherapy medication errors in a national error reporting database

Abstract

BACKGROUND.

Little is known regarding chemotherapy medication errors in pediatrics despite studies suggesting high rates of overall pediatric medication errors. In this study, the authors examined patterns in pediatric chemotherapy errors.

METHODS.

The authors queried the United States Pharmacopeia MEDMARX database, a national, voluntary, Internet-accessible error reporting system, for all error reports from 1999 through 2004 that involved chemotherapy medications and patients aged <18 years.

RESULTS.

Of the 310 pediatric chemotherapy error reports, 85% reached the patient, and 15.6% required additional patient monitoring or therapeutic intervention. Forty-eight percent of errors originated in the administering phase of medication delivery, and 30% originated in the drug-dispensing phase. Of the 387 medications cited, 39.5% were antimetabolites, 14.0% were alkylating agents, 9.3% were anthracyclines, and 9.3% were topoisomerase inhibitors. The most commonly involved chemotherapeutic agents were methotrexate (15.3%), cytarabine (12.1%), and etoposide (8.3%). The most common error types were improper dose/quantity (22.9% of 327 cited error types), wrong time (22.6%), omission error (14.1%), and wrong administration technique/wrong route (12.2%). The most common error causes were performance deficit (41.3% of 547 cited error causes), equipment and medication delivery devices (12.4%), communication (8.8%), knowledge deficit (6.8%), and written order errors (5.5%). Four of the 5 most serious errors occurred at community hospitals.

CONCLUSIONS.

Pediatric chemotherapy errors often reached the patient, potentially were harmful, and differed in quality between outpatient and inpatient areas. This study indicated which chemotherapeutic agents most often were involved in errors and that administering errors were common. Investigation is needed regarding targeted medication administration safeguards for these high-risk medications. Cancer 2007. © 2007 American Cancer Society.

Medication errors are common during pediatric hospitalizations, occurring in nearly 6% of all medication orders for pediatric inpatients.1, 2 The study of these errors remains in its infancy, because large holes exist in our understanding of which medications and children are at greatest risk.3, 4 To date, pediatric chemotherapy errors have been examined in only a few studies.5-8 This is humbling to consider in light of reports that 63% of all oncology nurses have witnessed a chemotherapy medication error.9

The pediatric chemotherapy error literature consists mostly of case reports.10-13 In 1 comprehensive pediatric oncology medication error article, it was reported that 13% of incidents reached patients and that 2% resulted in temporary patient harm requiring medical intervention. The authors of that article reported 73% ordering errors, 13% administration errors, 9% preparation errors, and 3% transcription errors. Vincristine, doxorubicin, and sodium mercaptoehanesulfonate were the most common medications involved in the reported errors. The most serious errors involved dexamethasone and methotrexate.5 A more recent study that examined the impact of computerized physician order entry on pediatric chemotherapy errors found that 2.3% of written pediatric chemotherapy orders in a tertiary care hospital contained an improper dose, and 5.8% contained an incorrect dosing calculation. With the institution of a computerized physician order entry system, these rates fell to 0.06% and 0.54%, respectively.6 The results from 1 outpatient study suggested that pediatric chemotherapy errors occurred in 18.8% of all patients and 9.9% of all medications prescribed.7 A prospective outpatient chart review identified errors in 3% of all pediatric outpatient chemotherapy orders, with 74% caused by ordering errors, 23% caused by dispensing errors, and 3% caused by administering errors.8

More information is needed regarding chemotherapy errors in pediatrics. These medications are among the most potent medications given to children and have narrow therapeutic windows and complex ordering requirements.10 In general, medication delivery for children is more complex than for adults because of necessities such as weight-based dosages and the availability of liquid oral medications formulated in a variety of concentrations.2, 14, 15 Chemotherapy medications often heighten this complexity by requiring dosages based on body surface area.

Working with United States Pharmacopeia (USP), we performed a quantitative and qualitative analysis of pediatric chemotherapy medication errors reported to MEDMARX between January 1, 1999 and December 31, 2004.16 Our objectives were to identify patterns in pediatric chemotherapy errors, including the types of medications involved, the types of errors created, causes of errors, level of harm of errors, location of errors, and characteristics of facilities associated with errors for children.

MATERIALS AND METHODS

The USP MEDMARX database is a national, voluntary, Internet-accessible error reporting system that consisted of 616 subscribing hospitals as of January, 2005. All participating hospitals are in the U.S., and all 50 states are represented. Each error report is standardized and provides both the reporting hospital and USP with information regarding the error's timing, unit location, inpatient versus outpatient location, phase of care in which the error occurred, harm caused by the error, cause of error, medication involved, type of error, and information on the facility in which the error occurred. A recent study of 101 MEDMARX users showed substantial inter-rater agreement (κ = .62) for medication error events using the National Coordinating Council for Medication Error Reporting and Prevention Index for Categorizing Medication Errors.17, 18

MEDMARX variables that are unique to error reporting and that we used in the current analysis included error category (reflecting the level of patient harm from error), error node (reflecting the macrolevel cause of error within the entire spectrum of medication administration, such as prescribing, transcribing, dispensing, or administration), error cause (reflecting the root cause of the error), and error type (reflecting the clinical manifestation of what happened regardless of cause). Comprehensive choice lists for error category, error node, error cause, and error type are available on request but can be seen in significant amounts within the presented tables. It is important to note that some data fields within the error reports are single pick lists (error category and error node), and some are multipick lists, allowing users to enter multiple answers (error cause, error type, and medications involved). An error that was classified as outpatient included any patient who was registered for services but was not admitted.

A key driver for the information reported in the MEDMARX error reporting system is the error category variable, which describes the degree of harm the patient experienced as perceived by the reporter. This categorization is not an ordinal scale. For example, category F errors are different from, but not necessarily “worse” than, category E errors. Category F errors specifically denote an extended hospitalization time, whereas category E errors denote an additional treatment change. These error category harm scores are the same as those reported by the National Coordinating Council for Medication Error Reporting and Prevention Severity Index.18 With increasing error category harm score, increasingly more data fields are required to be completed in MEDMARX, such as patient age. Patient sex is always an optional field. In addition, from 1999 to 2004, the MEDMARX database increased its participating institutions from 56 to 616 and went through a number of changes in required data fields as the concept of error reporting matured.

We queried the MEDMARX database for all error reports from 1999 to 2004 that involved chemotherapy medications, as defined by our institution, and that involved patients aged <18 years or patients in pediatric hospital locations. The list of chemotherapeutic agents was based on Section 10 of the American Hospital Formulary Service Drug Information Reference.19 We grouped ages into infant (age <1 year), preschool (ages 1–4 years), school aged (ages 5–9 years), preadolescent (ages 10–13 years), and adolescent (ages 14–18 years) based on Bright Futures Guidelines for Health Supervision and Neinstein's Adolescent Health Care.20, 21 For error type, which identifies the action that led to an error, we grouped response choices for the purposes of our analyses into categories that involved similar concepts (eg, deteriorated product and expired product). For error cause, which describes the root cause of the error, we grouped the MEDMARX pick list choices into a modified version of the National Coordinating Council for Medication Error Reporting and Prevention's error cause categories.18

For error reports that were logged during 2003 and 2004, we used the simultaneously supplied 2003 and 2004 facility profile data. For error reports that were logged in between 1999 and 2002, we extrapolated the 2003 facility data to these earlier years. This was done to enable analysis of the facility characteristics across all years despite differences in the data collected from 1999 to 2004 under the assumption that facility characteristics were unlikely to have dramatic shifts in this relatively short time span.

Analyses were done using Minitab software (release 14).22 The significance of trends across categorical variables was tested with chi-square tests and STATA 8.0 statistical software, and all tests of statistical significance were 2-sided.23

RESULTS

There were 829,492 errors submitted to USP MEDMARX as of January 1, 2005, and 29,802 of these errors involved patients aged <18 years. Our query returned 310 reported pediatric chemotherapy medication errors from January 1, 1999 through December 31, 2004. All age groups were represented well, and sex was distributed almost equally. In addition, 55.2% of errors occurred in inpatient units, 10% occurred in outpatient settings, and 34.8% had unknown locations. Twenty percent of the errors we identified occurred within pharmacies. From 1999 to 2004, the number of error reports increased in concert with increases in the number of participating institutions (Table 1). There were no significant trends observed in the Table 1 data based on reports in which age, sex, and inpatient/outpatient status were known or unknown.

Table 1. Descriptive Statistics on 310 Pediatric Chemotherapy Error Reports

Variable	Percentage of error reports
Age, y
<1	5.2
1–4	21.9
5–9	21
10–13	14.2
14–18	21.9
Unknown	15.8
Sex
Boy	18.4
Girl	20.6
Unknown	61
Location
Pediatrics, inpatient	37.4
Nursing unit	21.3
Pharmacy, inpatient	17.7
Oncology	5.8
Clinic, outpatient	5.8
Pharmacy, outpatient	3.9
Pediatric intensive care unit	3.5
Other (<2% each)	4.6
Setting
Inpatient	55.2
Outpatient	10
Unknown	34.8
Day of error
Weekday	82.9
Year of error
1999	2.3
2000	6.5
2001	11.9
2002	21
2003	33.5
2004	24.8
No. of participating institutions	No.
1999	56
2000	184
2001	368
2002	482
2003	570
2004	616

The 310 pediatric chemotherapy errors were reported by 69 separate facilities. Table 2 displays the characteristics of the facilities that participated in MEDMARX. Three facilities were no longer participating with USP MEDMARX by 2003; thus, the facility data for their 11 error reports were missing. University hospitals made up 26.1% of the involved institutions and reported 40% of the errors. Institutions with from 300 to 599 beds made up 31.9% of the facilities involved and reported 51% of the errors. Institutions that delivered ≥200,000 doses per year reported 66.1% of the errors, although they made up only 33.3% of the institutions involved. Chi-square analysis of these data revealed significant differences between the data representing a count of facilities participating in MEDMARX and reporting a chemotherapy error versus the count of error reports by each type of facility for the following variables: number of beds, number of doses, and inpatient pharmacist. The common theme was that institutions smaller in size (eg, lower bed count, lower number of doses dispensed per year, and no 24-hour on-site pharmacist support) had lower reported error prevalence than larger institutions with more beds, greater pharmacist availability, and higher volumes of drugs dispensed. For comparative purposes, Table 2 also summarizes the facility characteristics of all institutions reporting in MEDMARX. It is noteworthy that the institutions reporting pediatric chemotherapy errors are more likely to be university institutions or specialty children's hospitals, to have a greater number of beds, to dispense a greater number of medication doses each year, and to have greater availability of pharmacists onsite 24 hours a day, 7 days a week than the average profile of all facilities participating in MEDMARX.

Table 2. Facility Description

Variable	Profile of facilities participating in MEDMARX in 2004 that reported at least 1 pediatric chemotherapy error (N = 69), %	Error prevalence based on facility type (N = 310), %	P (Comparing 2 preceding columns)	Profile of all facilities participating in MEDMARX, 1999–2004 (N = 804), %
Facility type
General community hospital	53.6	46.5		69
University hospital	26.1	40		5.6
Ambulatory/outpatient clinic	8.7	2.6	NS	12.3
Specialty/children's hospital	5.8	7.1		1
Others (<5% each)	5.8	4.2		9.5
Facility owner
Nongovernment, nonprofit	56.5	63.9		56.1
Government, federal; military	23.2	7.1		19
Government, nonfederal	8.7	21.9	<.001	10.2
Government, federal; other	5.8	1.3		9.2
Others (<5% each)	5.8	5.8		1.4
No. of beds
<100	30.4	13.2		47.8
100–299	18.8	13.2		33
300–599	31.9	51	<.013	12.8
≥600	14.5	19		3.3
Missing	4.3	3.5		3.1
Total no. of doses
<9999	10.1	3.5		34
10,000–39,999	11.6	2.9		18.7
40,000–69,999	10.1	7.4		12.8
70,000–99,999	7.2	3.5	<.001	9.1
100,000–199,999	13	9		12.9
≥200,000	33.3	66.1		9.3
Missing	14.4	7.4		3.2
Inpatient pharmacist
Pharmacist is available onsite 24 h/d, 7 d/wk	60.9	82.9		32.7
When pharmacy closed, pharmacist on call (eg, after hours and emergencies)	24.6	9.7	<.005	44.8
No pharmacy services are available at this facility	10.1	3.9		18
No information	4.3	3.5		4.2
Outpatient pharmacist
Pharmacist is available onsite 24 h/d, 7 d/wk	20.3	18.1		10.1
When pharmacy closed, pharmacist on call (eg, after hours and emergencies)	27.5	26.8	NS	28.7
No pharmacist is available when pharmacy is closed	20.3	29.7		19.1
No pharmacy services are available at this facility	27.5	21.9		40.7
No information	4.3	3.5		1.4

• NS indicates not significant.

Table 3 describes the macrolevel characteristics of these 310 pediatric chemotherapy errors by looking at error category and error node. Overall, 264 of the reported chemotherapy errors (85%) reached the patient, and 49 errors (15.6%) created the need for additional patient monitoring or therapeutic intervention. The most common error node reported was administering (149 errors) followed by dispensing (94 errors). When examining the 171 inpatient errors versus the 31 outpatient errors, 42% of outpatient errors were administering errors, 32.3% were dispensing errors, and 22.6% were prescribing errors, compared with 50%, 23.4%, and 11.7%, respectively, for inpatient error reports (Table 3).

Table 3. Error Category and Error Node for Error Reports

Inpatient	Percentage of error reports
	Overall	Inpatient	Outpatient
Error category*
Unsafe conditions (category A)	3.5	6.4	0
Event did not reach individual because of chance alone or active recovery (category B)	11.5	15.8	22.6
Reached individual but did not cause harm (category C)	69.4	65.5	61.3
Reached individual and required additional monitoring (category D)	14.2	11.7	16.1
Individual experienced harm and required treatment (category E)	1	0	0
Individual experienced harm and required initial or prolonged hospitalization (category F)	0.6	0.6	0
Error node*
Administering	48.1	50.3	41.9
Dispensing	30.3	23.4	32.3
Prescribing	10.3	11.7	22.6
Transcribing/documenting	7.1	7.60	3.2
No data provided	3.5	6.40	0
Monitoring	0.6	0.6	0
Total no. of error reports	310	171†	31†

• * No significant differences comparing results coded as inpatient and outpatient.
• † Not all reports had the inpatient/outpatient variables coded.

Overall, there were 387 separate medications listed in the 310 error reports. The most common medication classes involved break down as follows: 153 medications (39.5%) were antimetabolites, 54 medications (14.0%) were alkylating agents, 36 medications (9.3%) were anthracyclines, and 36 medications (9.3%) were topoisomerase inhibitors. The most commonly involved chemotherapeutic agents were methotrexate (59 errors; 15.3% of 387 medications reported), cytarabine (47 errors; 12.1%), etoposide (32 errors; 8.3%), and doxorubicin (24 errors; 6.2%). The majority of error reports (81.3%) mentioned only 1 chemotherapy medication involved in the error (Table 4). There were 203 separate medications involved in the 171 inpatient error reports, and the most common medications were methotrexate (16.7% of 203 medications reported), cytarabine (15.3%), doxorubicin (6.9%), cyclophosphamide (6.4%), and etoposide (6.4%). There were 44 separate medications involved in 31 outpatient error reports. The most common outpatient medications involved were methotrexate (20.5% of 44 medications reported), tretinoin (15.9%), vincristine (11.4%), asparginase (6.8%), and cyclophosphamide (6.8%). For the tretinoin citations, 8 of the 11 total citations appeared to be for topical creams. Chi-square analysis of medications with >10 total citations revealed significant differences between lists of medications involved in inpatient versus outpatient settings (Table 4).

Table 4. Medications Involved for Error Reports

Variable	Total no. of all medications reported (%)	Percentage of inpatient medications	Percentage of outpatient medications
Medications involved	(N = 387)	(N = 203)*	(N = 44)*
Methotrexate	59 (15.3)	16.7	20.5
Cytarabine	47 (12.1)	15.3	4.5
Etoposide	32 (8.3)	6.4	4.5
Doxorubicin	24 (6.2)	6.9	2.3
Cyclophosphamide	23 (5.9)	6.4	6.8
Vincristine	22 (5.7)	2.5	11.4
Mercaptopurine	15 (3.9)	3.9	0
Thioguanine	13 (3.4)	3.4	0
Ifosfamide	12 (3.1)	3.9	0
Tretinoin†	11 (2.8)	0	15.9
Mesna	10 (2.6)	3.4	0
Daunorubicin	9 (2.3)	1.5	0
Asparaginase	8 (2.1)	1.0	6.8
Cisplatin	7 (1.8)	2.5	0
Hydroxyurea	7 (1.8)	3.4	0
5-fluorouracil	5 (1.3)	1.5	0
Carboplatin	4 (1.0)	0.5	2.3
Dexamethasone	4 (1.0)	0.5	2.3
Leucovorin	4 (1.0)	1	0
Megestrol	3 (0.8)	1.5	0
Floxuridine	2 (0.5)	0	4.5
Magnesium lactate	2 (0.5)	0	4.5
Others	64 (16.6)‡	17.8‡	13.7§
No. of medications listed
1	252 (81.3)	84.2	71.0
2	44 (14.2)	14.0	22.6
3	11 (3.5)	1.2	3.2
4	1 (0.3)	0	0
5	2 (0.5)	0.6	3.2

• Mesna indicates sodium mercaptoehanesulfonate.
• * Not all reports had the inpatient/outpatient variables coded. In a chi-square analysis of inpatient versus outpatient medications with >10 citations, P < .001.
• † Eight of 11 tretinoin citations were for topical creams and ointments based on error descriptions.
• ‡ All medications in the “other” group constituted <1% of the medications cited.
• § All medications in the “other” group constituted 2.3% of the medications cited.

Within the 310 error reports, there were 324 error types cited and 3 error reports with no error type identified. The most common error types were improper dose/quantity (75 citations; 22.9% of 327 possible error types), wrong time (74 citations; 22.6%), omission error (46 citations; 14.1%), and wrong administration technique/wrong route (40 citations; 12.2%). The majority of error reports (n = 292) cited only 1 error type, 13 reports cited 2 error types, and 2 reports cited 3 error types. Because wrong time error types were cited more often in this data set than in previously published MEDMARX datasets, we cross tabulated these error reports with error category.24, 25 Seventy-two wrong time errors (97%) reached the patient, and 7 wrong time errors (9.5%) created the need for additional patient monitoring or therapeutic intervention, suggesting that wrong time errors were less likely to harm a patient than other errors in our dataset. Outpatient error types were nearly twice as likely to be classified as prescribing errors (17.6% vs 9.4%), were more likely to be classified as improper dose/quantity (26.5% vs 18.9%), and were less likely to be classified as omission errors (11.8% vs 16.1%) (Table 5).

Table 5. Error Type and Error Cause for Error Reports

Variable	Percentage of errors
	Overall*	Inpatient	Outpatient
Error type‡	N = 327	N = 180†	N = 34†
Improper dose/quantity	22.9	18.9	26.5
Wrong time	22.6	18.9	11.8
Omission error	14.1	16.1	11.8
Wrong administration technique/wrong route	12.2	13.3	8.8
Prescribing error	8	9.4	17.6
Drug prepared incorrectly	6.1	6.7	2.9
Not determined, no data, expired/deteriorated product	4.9	5.6	8.8
Extra dose	3.4	5	0
Unauthorized/wrong drug and wrong dosage form	3.4	3.9	5.9
Wrong patient	2.4	2.2	5.9
Error cause‡	N = 547	N = 294†	N = 53†
Performance deficit	41.3	39.5	35.8
Equipment and medication delivery devices	12.4	10.5	17
Communication	8.8	9.5	7.5
Knowledge deficit	6.8	8.8	3.8
Written order errors	5.5	7.5	5.7
Transcription error	4.2	5.1	5.7
Miscalculation of dose	3.8	3.7	3.8
Human-related computer error	3.3	4.1	3.8
System safeguards	2.9	1.7	3.8
Missing	2.7	3.1	1.9
Drug preparation error	2.4	3.4	0
Labeling/packaging/reference materials	2.4	1.4	3.8
Brand name similarities	1.5	0	5.7
Stress/high workload	0.2	0	1.9
All others (<1% each)	1.8	1.7	0

• * Error reports could cite 1 or more error types and error causes. Error causes were grouped into a modified version of the National Coordinating Council for Medication Error Reporting and Prevention's error cause categories. (see National Coordinating Council for Medication Error Reporting and Prevention, 200618).
• † Not all reports had the inpatient/outpatient variable coded.
• ‡ There were no significant differences when comparing results that were coded as inpatient and outpatient.

There were 542 error causes cited and 5 error reports with no error cause cited. The most common error causes were performance deficit (226 citations; 41.3% of 547 possible error causes), equipment and medication delivery devices (68 citations; 12.4%), communication (48 citations; 8.8%), knowledge deficit (37 citations; 6.8%), and written order errors (30 citations; 5.5%). One hundred eighty-nine error reports cited 1 error cause, 58 reports cited 2 error causes, 27 reports cited 3 error causes, and 33 reports cited ≥4 error causes. Using the broadest categories from the National Coordinating Council for Medication Error Reporting and Prevention, the 2 most common error cause groups were human factors (368 errors; 67%) and miscommunication (78 errors; 14%).18 The causes of outpatient errors were more likely to be classified as equipment and medication delivery devices (17.0% vs 10.5%), were more likely to be classified as brand name similarities and stress/high workload (5.7% and 1.9%, respectively vs. 0%), were less likely to be classified as knowledge deficit (3.8% vs 8.8%), and were less likely to be classified as drug preparation errors (0% vs 3.4%) (Table 5).

There were 5 errors that resulted in individual harm and required treatment or prolonged hospitalization (error categories E-I) (Table 6). Of these errors, only 1, which involved bleomycin, occurred in a facility that did not have a pharmacist on site 24 hours a day. In general, only 2 of the 5 high harm errors involved a chemotherapy-specific issue, namely, improper calculation of an absolute neutrophil count and a delay caused by an insufficient quantity of methotrexate available. The other 3 high harm errors involved issues that are generic to all medication errors, such as reversing infusion rates between 2 infusions, a delay in medication delivery, and switching of patient labels for the same medication. Four of the 5 most serious errors occurred at community hospitals.

Table 6. Summary of the 5 Most Serious Errors

Error category	Error description	Error location; facility description	Error node	Error type	Error cause
F: Individual experienced harm and required initial or prolonged hospitalization	Not enough methotrexate to replace dose for leaking bag; replacement bag not ready until approximately 1400 h the next day, because the large quantity of medication had to be obtained from 2 other hospitals	Pharmacy, inpatient; general community hospital; 400–499 beds	Dispensing	Wrong time	Drug distribution system; packaging/container design
F: Individual experienced harm and required initial or prolonged hospitalization	Patient arrived to floor at 1300 h for IV chemo after IT methotrexate; checked with pharmacy at 1600 h regarding when chemo would arrive and was told about 1800 h; prechemo hydration initiated; at 2045 h, checked with pharmacy regarding when to expect chemo and was advised it had just been delivered from being prepared and would be right up; parents upset regarding delay, which resulted in extending stay by 1 day	Pharmacy, inpatient; general community hospital; 500–599 beds	Dispensing	Wrong time	Drug distribution system
E: Individual experienced harm and required treatment	Two medications (etoposide and ifosfamide) were hung at the same time for the same patient, and their infusion rates were reversed	Nursing (patient care) unit; university hospital; 100–199 beds	Administering	Improper dose/quantity	Monitoring inadequate/lacking; packaging/container design; performance (human) deficit; procedure/protocol not followed; system safeguard(s)
E: Individual experienced harm and required treatment	Order read, “treat if ANC >1000”; in calculating ANC, WBC, poly, and lymph were used to obtain result; error was to use the no. of lymphs in this calculation, which resulted in a higher ANC; patient now receiving filgrastim; drugs involved include bleomycin, dacarbazine, doxorubicin, and vincristine	Oncology department; general community hospital; 100–199 beds	Administering	Improper dose/quantity	Procedure/protocol not followed
E: Individual experienced harm and required treatment	Two syringes for vincristine, 0.4 mg and 1.8 mg, were prepared by pharmacist; labels were switched; the patient who was to receive 0.4 mg received the 1.8-mg dose; however, the physician noticed the excess volume and stopped the injection at about 0.7 mg	Pharmacy, inpatient; general community hospital; 400–499 beds	Dispensing	Improper dose/quantity	Performance (human) deficit

• IV indicates intravenous; chemo, chemotherapy; IT, intrathecal; ANC, absolute neutrophil count; WBC, white blood cells; Poly, polymorphonuclear leucocyte; lymph, lymphocyte.

DISCUSSION

To our knowledge, this study is among the first to describe the epidemiologic foundation of chemotherapy errors in children by analyzing voluntarily reported chemotherapy errors from a 6-year database that represents 69 facilities across the nation. It is noteworthy that 85% of the errors we identified did reach the patient. Despite the large focus nationally on computerized ordering systems to fix medication prescribing errors, 48% of the chemotherapy errors we studied were caused by administering mistakes. In terms of the chemotherapeutic agents, 39% involved antimetabolite drugs, and 15% involved methotrexate. The multicausal nature of medication errors is reflected in our data, because the number of reported error causes is large. Outpatient chemotherapy errors were more likely to involve dispensing mistakes, prescribing mistakes, and equipment failures. There was also a clear trend toward more error reports in larger institutions. We have no reason to believe these data are indicative of larger institutions simply making more errors, as opposed to the simple fact of greater volumes of oncology patients and more internal resources devoted to quality and error reporting.

Differences between the current study and prior pediatric chemotherapy error investigations may be because of the anonymous nature of the MEDMARX error reporting system; the increasing presence of computerized physician ordering systems, which decrease prescribing errors; and innate differences between tertiary care hospitals, in which all prior studies were conducted, and community hospitals, in which 46% of our error reports were generated.5-8 More research is needed to examine these differences.

These data suggest clear areas for intensive patient safety research and development of targeted interventions to improve the safe delivery of pediatric chemotherapy medications. First, a renewed focus on issues and barriers to correct medication administration should be explored. The national focus on computerized order entry does not ameliorate the likelihood of administering errors anywhere to the same degree that it does for prescribing errors. Related to this, many vendor-supplied, computerized, order-entry systems are weakest in their ability to handle the prescribing and electronic Medication Administration Record posting of chemotherapy agents as opposed to other medications; and their use in pediatrics has been brought into question by a recent, albeit controversial, study.26 Interventions such as Health Care Failure Mode and Effect Analysis have been proven useful in the pediatric oncology setting.6, 27 For maximal effectiveness, a collaborative Health Care Failure Mode and Effect Analysis between multiple institutions could create a forum for discussion of best practices among pediatric oncology centers. Such an effort likely would identify fruitful areas for targeted safety work in areas as relatively simple as standardizing the logic and order of protocols and nomenclature used (eg, is the first day of a chemotherapy protocol Day 0 or Day 1?). In addition, point-of-care bar coding interventions have shown promise at preventing administration type errors and could be applied in pediatric chemotherapy situations, although the recent Institute of Medicine report, Preventing Medication Errors, acknowledges that there is currently very limited evidence on the efficacy of bar coding for medication error prevention.28-30

Second, a discussion of the different chemotherapy error types and error causes that occur in outpatient versus inpatient settings needs to be translated into different safety systems tailored for each of these settings. Third, increased communication between adult and pediatric chemotherapy delivery systems could prevent similar errors from occurring. It is likely that the adult oncology community, with its greater patient volume, has determined some best practices that could be translated and disseminated in the pediatric oncology arena. Supporting this idea, prior studies utilizing adult MEDMARX data have produced results similar to those from pediatric MEDMARX studies.24, 25, 31, 32 Based on this, we speculate that, although significant differences exist between adult and pediatric chemotherapy delivery, common themes may be found to decrease errors across both patient populations.

The USP MEDMARX database does involve a number of limitations.24, 25, 31, 32 First, all data are reported voluntarily and, thus, do not convey a representative picture of all errors and are not suitable for determining national error rates because of the lack of a “denominator.” The anonymous nature of the data also precludes the identification of geographic trends. The distribution of facilities indicates that the MEDMARX database consists mainly of general community hospitals, although our error reports had a greater tendency to be generated at university hospitals. It is unclear how this distribution of hospitals affected the data collected and how representative these facilities are of all pediatric chemotherapy centers in the U.S. Furthermore, despite evidence of substantial interrater agreement, MEDMARX users may have varied impressions of errors and their categories.17 Nevertheless, error report data may be one of the most robust ways to obtain snapshots of errors that are useful for identifying trends and issues.33 Second, the vast majority of error reports are generated by nurses and pharmacists.34 It appears logical that nurses would be positioned the best to identify errors, because they are the last link in the chain of medication delivery. It is unclear, however, the extent to which their perspectives bias the data or provide the best vantage point from which to observe the entire spectrum of the medication use process. Collectively, an accurate picture of the scope of medication errors needs both chart reviews and error reporting to robustly capture medication errors from all possible phases.8 For the patient, errors at any step in the medication process are significant and need to be understood and addressed. Third, because of the changing requirements of the USP MEDMARX system, errors from different years have different data fields completed. In addition, this temporal issue in the data also most likely is effected by the 2000 Institute of Medicine's report To Err is Human.35 Finally, not all of the medications in our query are used solely to treat oncologic problems, which is clear by the multiple citations of tretinoin topical creams.

In conclusion, for this study, we examined 310 pediatric chemotherapy medication errors that were collected through a national, voluntary error reporting database. These errors often reached the patient, were potentially harmful to the pediatric patient population, and differed in quality between outpatient and inpatient oncology areas. Commonly used chemotherapeutics, such as methotrexate and cytarabine, frequently were involved, and medication administration was cited as the macrolevel cause in nearly half of the errors. Pediatric hospitals and future quality improvement research should target medication administration safeguards for these high-risk medications and should consider different and specific solutions for inpatient and outpatient pediatric oncology units, respectively.

Acknowledgements

We thank Alix Butler, PharmD, for her advice and assistance on this project.