The first two authors contributed equally to this article.
Medication safety in the ambulatory chemotherapy setting†
Article first published online: 24 OCT 2005
Copyright © 2005 American Cancer Society
Volume 104, Issue 11, pages 2477–2483, 1 December 2005
How to Cite
Gandhi, T. K., Bartel, S. B., Shulman, L. N., Verrier, D., Burdick, E., Cleary, A., Rothschild, J. M., Leape, L. L. and Bates, D. W. (2005), Medication safety in the ambulatory chemotherapy setting. Cancer, 104: 2477–2483. doi: 10.1002/cncr.21442
See related editorial on pages 2289–91, this issue.
The first two authors contributed equally to this article.
Fax: (617) 732-7072
- Issue published online: 18 NOV 2005
- Article first published online: 24 OCT 2005
- Manuscript Accepted: 24 JUN 2005
- Manuscript Revised: 23 JUN 2005
- Manuscript Received: 14 FEB 2005
- Harvard Risk Management Foundation
- medication error;
- ambulatory care;
- patient safety;
- computerized physician order entry;
Little is known concerning the safety of the outpatient chemotherapy process. In the current study, the authors sought to identify medication error and potential adverse drug event (ADE) rates in the outpatient chemotherapy setting.
A prospective cohort study of two adult and one pediatric outpatient chemotherapy infusion units at one cancer institute was performed, involving the review of orders for patients receiving medication and/or chemotherapy and chart reviews. The adult infusion units used a computerized order entry writing system, whereas the pediatric infusion unit used handwritten orders. Data were collected between March and December 2000.
The authors reviewed 10,112 medication orders (8008 adult unit orders and 2104 pediatric unit orders) from 1606 patients (1380 adults and 226 pediatric patients). The medication error rate was 3% (306 of 10,112 orders). Of these errors, 82% occurring in adults (203 of 249 orders) had the potential for harm and were potential ADEs, compared with 60% of orders occurring in pediatric patients (34 of 57 orders). Among these, approximately one-third were potentially serious. Pharmacists and nurses intercepted 45% of potential ADEs before they reached the patient. Several changes were implemented in the adult and pediatric settings as a result of these findings.
In the current study, the authors found an ambulatory medication error rate of 3%, including 2% of orders with the potential to cause harm. Although these rates are relatively low, there is clearly the potential for serious patient harm. The current study identified strategies for prevention. Cancer 2005. © 2005 American Cancer Society.
Medication safety is a pressing concern in health care, especially in oncology. Chemotherapeutic agents are well studied and highly beneficial medications, but they must be used carefully because of their high toxicity and narrow therapeutic window. Regimens are complex, usually involving multiple medications with different dosing schedules, and vary widely for different cancers. Clinical research protocols utilize investigational or commercial medications in various doses and schedules, further adding to the complexity. The need to calculate doses according to body size and laboratory results also complicates management for both children and adults.1 Furthermore, the process for administering chemotherapy involves individuals from multiple disciplines whose efforts must be coordinated to minimize risk.
Increasingly, chemotherapy administration occurs in the outpatient setting. Compared with the inpatient setting, outpatient chemotherapy is challenging because of high volume, time pressures, and lower levels of control. Decisions may hinge on the results of tests obtained at other facilities, provider continuity may be lacking, patients may need to self-administer essential medications, and side effects often occur at home.
Recent studies have shown that the overall inpatient medical-surgical medication error rate is approximately 5%2 and the primary care medication error rate is approximately 8%.3 In contrast, to our knowledge only a few studies to date have explored chemotherapy safety and chemotherapy errors,4, 5 except as case reports.6–8 Our organization had a highly publicized case of a chemotherapy overdose9 that was one of the main initiators of the patient safety movement, with changes in both the organizational culture and the chemotherapy ordering process, including the implementation of inpatient computerized chemotherapy order entry. To our knowledge, the safety of the outpatient chemotherapy process has not been assessed routinely across a broad range of patients and regimens. Therefore, we undertook a study of outpatient chemotherapy in adult and pediatric patients to: 1) evaluate the epidemiology of medication errors and potential adverse drug events (ADEs) at a comprehensive cancer care center and 2) identify potential system changes to improve the safety of outpatient chemotherapy processes.
MATERIALS AND METHODS
The current study was conducted in three outpatient chemotherapy infusion units at the Dana-Farber Cancer Institute (DFCI), the founding member of the National Cancer Institute-designated Dana-Farber Harvard Cancer Center in Boston. Two chemotherapy infusion units provide care for adult patients and one unit provides care for pediatric patients. One of the adult units cares for patients with various forms of cancer including sarcomas, brain tumors, skin cancers, and hematologic malignancies, as well as patients with benign hematologic conditions. The second adult unit cares for patients with a range of solid tumors, including gastrointestinal, head and neck, breast, gynecologic, thoracic, and genitourinary carcinomas. The pediatric unit cares for patients with hematologic and solid tumors.
Approximately 125 adults and 25 children receive outpatient chemotherapy treatments daily in these 3 units. The adult units utilize a computerized medication ordering system whereas the pediatric unit uses a handwritten ordering process. The computerized system is a provider order entry system, with specific screens for chemotherapy orders and ancillary and supportive care medications such as hydration and antiemetics.10 The computerized clinical decision support system included drug allergy and drug interaction checking and dose limits for chemotherapeutic agents, and was in place at the time of the current study. The attending physician enters the order into the computer (if a fellow enters the order, the system requires the cosignature of an attending physician). The nurse then reviews the patient demographic information, doses, clinical trial document if applicable, and place in therapy; performs an assessment; and provides a copy of the order to pharmacy. The pharmacist then reviews the patient's demographic information, recalculates the body surface area and doses, reviews laboratory parameters and any associated dose modifications, reviews the pharmacy profile to determine where the patient is in therapy, reviews pertinent components of the medical record, and dispenses the chemotherapy and any support medications. In the pediatric clinic, the same processes occurred with the exception that the orders were handwritten, with some completed on templated order forms.
Data were collected from March 2000 until December 2000 from 1 infusion unit at a time, each for 6–8 weeks. Data collection included order review and chart review. Multidisciplinary staff focus groups and in-service training were conducted at the beginning of the current study to identify any staff concerns about the chemotherapy process or the current study. The Institutional Review Board approved the study.
A random sample of approximately 50 adult or 25 pediatric patients was selected daily from patients receiving medications in the unit (at the time of the index visit). Patients were excluded if they were there only for blood tests or an intravenous line placement. A total of 1606 patients (Table 1) were enrolled, including 1380 adult and 226 pediatric patients.
|Characteristic||Adult population||Pediatric population||Total|
|No. of visits with orders reviewed||2122||575||2697|
|No. of medication orders reviewed||8008||2104||10,112|
|Total no. of patients||1380||226||1606|
|Females||815 (59%)||113 (50%)||928 (58%)|
|White||1237 (90%)||182 (81%)||1419 (89%)|
|On protocol||601 (44%)||126 (56%)||727 (45%)|
|Mean age (SE)||55 (0.4)||10 (0.4)||—|
|Chart reviews||1377 (99.8%)||225 (99.6%)||1602 (99.8%)|
Randomly selected patient orders were collected daily and reviewed by trained oncology nurse or pharmacist reviewers to identify medication errors. If nurses and pharmacists noted dose corrections on the orders, these interceptions were captured as possible events. Because patients often had multiple visits to the unit during the data collection periods, individual patient orders could be reviewed multiple times (until a maximum of three adult or four pediatric order reviews/patient were reached).
Nurses reviewed charts, starting from the index visit through the subsequent 3 months, for all patients whose orders were reviewed. The charts were examined for evidence of medication errors, ADEs, allergies, and comorbidities.
Clinicians were asked to report any near misses, medication errors, or ADEs to the study staff. To facilitate communication, study nurses and pharmacists were present in the units daily during data collection.
The events studied included medication errors and potential ADEs (Fig. 1). A medication error was defined as any error in the medication process, including ordering, dispensing, transcribing, administering, and monitoring, even if the error was intercepted and corrected prior to reaching the patient.2 Some medication errors had little potential for harm, whereas others had potential for patient harm and were classified as potential ADEs and, as noted below, were rated with regard to potential seriousness. Examples of medication errors with little potential for harm were the pharmacist dispensing oral prochlorperazine instead of the intravenous version or a pharmacist dispensing pamidronate (which the patient was known to receive monthly) without an order. Examples of potential ADEs (medication errors with the potential for harm) were missing parameters in a chemotherapy order, a required premedication written as an as-needed order, and a route omitted on a chemotherapy order.
Event Review and Classification
Trained oncology nurses and pharmacists reviewed medication orders and charts for possible medication errors and ADEs. All possible medication errors were reviewed independently by a panel of two physicians, including at least one oncologist. Medication errors were classified as potential ADEs if they had the potential for harm. The severity of potential ADEs was classified as fatal or life-threatening, serious, or significant.11 An example of a life-threatening potential ADE was an overdose of a high-risk chemotherapeutic agent such as doxorubicin. Examples of serious potential ADEs were an incorrect dose of vincristine, a missing order for leucovorin, and an incorrect drug ordered (vinblastine instead of vincristine). Examples of significant potential ADEs were a missing premedication order for an antiemetic or a duplicate order for pamidronate. Reviewers also specified error types and possible prevention strategies.
Differences in initial reviewer judgments regarding classification or severity were resolved by discussion. Common event classification was established for repetitive issues to ensure consistency. After the initial classification by the physician panels, selected potential ADEs and medication errors were reviewed by a different expert panel (oncologist, internist, and pharmacist) to ensure consistency.
Comparisons of continuous data were presented as the means ± the standard errors, and P values were calculated using the Student t test. Categoric data are presented as counts with percentages, and P values were calculated using chi-square tests. All reported P values were based on two-tailed tests of significance. All analyses were conducted using SAS software (SAS Institute Inc., Cary, NC) and Microsoft EXCEL (Microsoft Corporation, Redmond, WA).
During the study period, we reviewed 10,112 medication orders from 1606 patients (Table 1); within this group of patients, we also reviewed 1602 charts.
Medication Errors in Adults
In the 8008 medication orders reviewed in adult patients, we identified 249 (3%) medication errors. Of these, 203 (82%) were potential ADEs (Table 2), and 46 had no potential for harm. Ordering errors occurred most commonly (n = 191), followed by administering (n = 46) and dispensing (n = 41) errors. Of the 249 medication errors, 92 (37%) were related to chemotherapy, accounting for 4% of adult chemotherapy orders (Table 2).
|Adult patients||Pediatric patients||Total|
|Chemotherapy||2454 (31%)||790 (38%)||3244 (32%)|
|Medication errors||249 (3% of total orders)||57 (3% of total orders)||306 (3% of total orders)|
|Chemotherapy||92 (4% of chemotherapy orders)||9 (1% of chemotherapy orders)||101 (3% of chemotherapy orders)|
|Subset of potential ADEs||203 (2.5 % of total orders; 82% of medication errors)||34 (1.6% of total orders; 60% of medication errors)||237 (2.3% of total orders; 77% of medication errors)|
|Chemotherapy||80 (3% of chemotherapy orders)||8 (1% of chemotherapy orders)||88 (3% of chemotherapy orders)|
|Serious||53 (26%)||11 (32%)||64 (27%)|
|Significant||150 (74%)||23 (68%)||173 (73%)|
|Intercepted||88 (43%)||18 (53%)||106 (45%)|
Potential ADES in Adults
Potential ADEs were identified in 2.5% of the total adult orders reviewed (203 of 8008 orders). Of these, none were life-threatening, 53 (26%) were serious, and the remainder were significant. Examples of serious potential ADEs include missing premedication orders prior to chemotherapy, an order written for a dose of steroid that was well above the recommended guidelines for treatment, and missing treatment parameters for numerous chemotherapy orders. None resulted in patient harm, and 88 (43%) were intercepted before reaching the patient. Pharmacists intercepted 50 potential ADEs (57%), nurses intercepted 37 (42%), and a patient intercepted 1 (1%).
The most frequent error types involved in the 203 adult potential ADEs were missing dose (an order that should have been written was not) (n = 47 errors), treatment held and the computerized order not discontinued by the end of that day (n = 40 errors), missing chemotherapy treatment parameters (n = 26 errors), and orders written as free text within the order entry (n = 17 errors). The medications most commonly involved in adult potential ADEs were antineoplastics (n = 62 potential ADEs), followed by antiemetics (n = 38 potential ADEs), and antihistamines (n = 30 potential ADEs). The most commonly involved antineoplastics and nonchemotherapy therapeutic agents were pamidronate (n = 12 potential ADEs), 5-fluorouracil (n = 10 potential ADEs), and cyclophosphamide (n = 9 potential ADEs).
Of the 92 chemotherapy errors, 80 (87%) were potential ADEs, representing 39% of the total 203 potential ADEs and 3% of adult chemotherapy orders (Table 2). Chemotherapy-related potential ADEs were significantly more likely to be serious, compared with nonchemotherapy-related potential ADEs (48% vs. 12%; P < 0.0001).
Adult Potential ADE Prevention
Potential strategies for the prevention of adult potential ADEs were standardized chemotherapy treatment templates including standard chemotherapy orders as well as premedication and hydration orders (27%), guided dose algorithms that help the physician calculate dosing based on patient characteristics or laboratory information (14%), drug dose checks (12%), and eliminating the option to use free text in the order entry (10%) (Table 3).
|Adult patients||Pediatric patients||Total|
|Total no. of potential ADEs||203||34||237|
|Physician order entry||158 (78%)||25 (74%)||183 (77%)|
|Standardizing templates||54 (27%)||12 (35%)||66 (28%)|
|Drug dose check||25 (12%)||10 (29%)||35 (15%)|
|Guided dose algorithms||29 (14%)||7 (21%)||36 (15%)|
|Eliminating free text within the order entry||21 (10%)||—||21 (9%)|
|Drug–drug interaction check||11 (5%)||—||11 (5%)|
|Drug–patient characteristic check||8 (4%)||2 (6%)||10 (4%)|
|Other computer methods||37 (18%)||8 (24%)||45 (19%)|
|Noncomputer methods||41 (20%)||19 (56%)||60 (25%)|
Medication Errors in Pediatric Patients
Among 2104 pediatric medication orders reviewed, we found 57 medication errors (3%), including 34 potential ADEs (Table 2) and 23 with no potential for harm. Ordering errors occurred most often (n = 42), followed by dispensing errors (n = 13) and administering errors (n = 5). Of these medication errors, 9 (16%) were related to chemotherapy (Table 2), accounting for less than 1% of the total pediatric chemotherapy orders (Table 2).
Potential ADEs in Pediatric Patients
Potential ADEs accounted for 1.6% of the total pediatric orders reviewed (Table 2). Of the 34 potential ADEs, none were life-threatening, 11 (32%) were serious, and the remainder were significant. Examples of serious potential ADEs included an order written for too high a dose of chemotherapy, an omitted route for a chemotherapeutic agent, and missing hydration orders. None resulted in patient harm, and clinicians intercepted 18 errors (53%) before they reached the patient. Pharmacists intercepted the majority of potential ADEs (n = 14; 78%), followed by nurses (n = 2; 11%), and nurse practitioners (n=2; 11%).
The most frequent error types in pediatric potential ADEs were the order not being discontinued (nine potential ADEs), wrong dose (nine potential ADEs), and omitted frequency (seven potential ADEs). The medications most commonly involved were antineoplastics (eight potential ADEs), antibiotics (four potential ADEs), narcotics (four potential ADEs), and antiasthmatics (four potential ADEs). The chemotherapy agent most commonly involved was vincristine (three potential ADEs).
Of the 9 chemotherapy-related medication errors, 8 (89%) were potential ADEs, representing 24% of the total 34 potential ADEs and 1% of pediatric chemotherapy orders (Table 2). Similar to the findings in adults, chemotherapy-related potential ADEs in pediatric patients were more likely to be serious, compared with nonchemotherapy-related potential ADEs (63% vs. 23%; P = 0.04).
Pediatric Potential ADE Prevention
The most common strategies for the prevention of potential ADEs in pediatric patients were standardized templates (35%), drug dose checks (29%), and guided dose algorithms (21%) (Table 3).
Improvements Resulting from the Current Study
Based on the results of the current study, the DFCI implemented changes to their systems of care, including the development of a new and enhanced adult chemotherapy order entry system. All chemotherapy regimens are now standardized and templated within the computerized order entry system, thereby reducing the likelihood of drug entry errors and providing nurses and pharmacists with a more uniform set of therapies. Ancillary premedications, required hydration orders, and other important drugs and instructions are included in the order entry sets for each chemotherapy regimen. When ordering chemotherapy, the provider must now select specific treatment parameters that must be met prior to the initiation of chemotherapy. Hydration orders now must be included (or it must be stated specifically that hydration is not necessary) with each chemotherapy regimen to eliminate ambiguity. Any required premedications for chemotherapy are included as part of the chemotherapy order set and all investigational research protocol therapies are now templated into order sets. The pediatric infusion unit also has changed from handwritten orders to regimen-specific or protocol-specific computerized order templates.
The main objectives of the current study were to provide new information regarding medication errors in the ambulatory chemotherapy setting, and to characterize those errors so that error-reduction strategies could be developed. In assessing the safety of the outpatient chemotherapy process, we found a medication order error rate of 3%; approximately two-thirds of these errors (2% of orders) had the potential to cause harm. It is interesting to note that the majority of the errors were intercepted and none actually caused harm. Several strategies for improvement in the medication process were implemented as a result of the current study. We believe that the data presented herein regarding medication errors are novel and important, and can be used to drive improvement efforts at other institutions and clinics.
The medication error rate was similar to or lower than rates reported in other studies (3% vs. 0.4–5% in hospitals and 8% in primary care).2, 3, 12, 13 This finding may stem from the presence of multiple verification points in the chemotherapy ordering process, which have been instituted because of its intrinsic hazard. However, the error rate could still be lower. Conversely, we found a substantially higher percentage of these medication errors had the potential to harm patients (potential ADEs) (84% vs. 7%-33% in the inpatient setting and 43% in the primary care setting). This finding reflects the inherent risk of chemotherapy, and highlights why even relatively low error rates carry an unacceptable risk of serious patient harm, therefore highlighting the need for the continued development of error reduction strategies.
It is surprising that the medication error rate was the same in adult and pediatric units, despite the lack of computerized order entry in the pediatric unit, which has been demonstrated to reduce serious inpatient medication error rates.14, 15 A possible explanation is that a high proportion of pediatric patients receive investigational protocols with very specific dosing and dose modification parameters. This could account for the lower rate of chemotherapy-related medication errors noted in the pediatric clinic. Other possible reasons include the lower volume in the pediatric infusion clinic, fewer types of malignancies treated, and fewer regimens used. Moreover, the computerized order entry system in use in the adult units had relatively little decision support. Subsequent to the current study, a more sophisticated computerized order entry system was implemented which we believe will further reduce the medication error rate.
It is interesting to note that none of the medication errors or potential ADEs actually caused harm and the majority was intercepted by nurses or pharmacists. In both the adult and pediatric units, the chemotherapy process had many stages of order verification by a multidisciplinary clinical team. At all steps, the clinical team had access to the patient's medical records, including laboratory results, treatment history, and protocol information. As a result of the organization's commitment to error reduction and the culture of safety the executive leadership has promoted in recent years,16 each member of the clinical team feels a sense of personal responsibility to question an order if he or she is unclear about any aspect. The effectiveness of this culture and process demonstrates that the systems at the DFCI intercepted a high percentage of errors, and that none of the serious errors caused harm. Furthermore, after this study, the DFCI implemented several changes, such as enhancements to existing order entry systems and the development of specific computerized and noncomputerized pediatric order templates.
The current study has several limitations. First, it was performed at only one independent oncology center that has a strong culture of patient safety and a sophisticated chemotherapy ordering system. However, this center also has a high rate of use of investigational protocols and fewer standard regimens. Therefore, the results may not be generalizable to other cancer centers. However, we believe that these results should prompt readers to begin to evaluate their own processes of care because these data were found to have a significant impact on patient care at our own institution. Second, we did not examine whether treatments matched protocols, which may have led to an underestimation of the true error rate. Third, we have not yet measured the impact of our interventions on error rates, but hope to do so in the future.
Cancer patients are increasingly receiving highly toxic and complex chemotherapy in the outpatient setting. In addition, more new chemotherapeutic agents are being introduced each year that clinicians may be less familiar with. Delivering chemotherapy as safely as possible requires continually examining the systems by which chemotherapy is administered. We found a low serious error rate and, because of a rigorous chemotherapy order review process, none of the errors actually caused harm to patients. However, we still identified ways to improve medication systems. Clearly, outpatient chemotherapy infusion centers must have robust strategies in place to prevent errors from reaching patients. These data provide an important opportunity to think about systemic errors, prompt others to evaluate their own processes of care and chemotherapy administration, and further explore strategies for the reduction of medication errors.
The authors acknowledge the hard work and commitment of all the data collectors, reviewers, and steering committee members: Kate Abu; Sarah Alexander, M.D.; Amy Billet, M.D.; Josh Borus; Harold Burstein, M.D.; Demitrios Colevas, M.D.; Maureen Connor, R.N., M.P.H.; James Conway; Amy Dawson, R.Ph.; Dan DeAngelo, M.D.; Deb Duncombe; Craig Earle, M.D.; David Fisher, M.D.; David Frank, M.D.; Jeffrey Meyerhardt, M.D.; Deepa Narasimhan; Emily Olsen, R.N.; Pat Reid Ponte, R.N., D.N.Sc.; Charlie Roberts, M.D.; Ravi Salgia, M.D.; Stephen Sallan, M.D.; Robert Soiffer, M.D.; and Andrew Wagner, M.D. They also thank Erin Hartman, M.S., for her review of the article.
- 4A continuous-improvement approach for reducing the number of chemotherapy-related medication errors. J Health Syst Pharm. 2000; 4(Suppl 57): S4–9., , .
- 7Doctors' deadly mistakes. Medical errors kill up to 98,000 Americans yearly: a new report says that number could be cut drastically. Time. 1999; 154: 74–76..
- 9The Dana-Farber Cancer Institute. HBS Case #699-025. Boston: Harvard Business School Publishing, 1999., .
- 10An information system to improve the safety and efficiency of chemotherapy ordering. Proc AMIA Annual Fall Symposium 1996; 498–502., , , et al.