ORIGINAL RESEARCH

Computerized Physician Order Entry, a Factor in Medication Errors: Descriptive Analysis of Events in the Intensive Care Unit Safety Reporting System David A. Thompson, DNSc, MS, RN, Laura Duling, Christine G. Holzmueller, BLA, Todd Dorman, MD, Lisa H. Lubomski, PhD, Fern Dickman, MPH, Maureen Fahey, MLA, Laura L. Morlock, PhD, Albert W. Wu, MD, MPH, and Peter J. Pronovost, MD, PhD

Abstract • Objective: To describe the impact of computerized physician order entry (CPOE) on medication errors submitted to a voluntary Web-based reporting system. • Setting: 18 intensive care units (ICUs) in the United States. • Reporting system: The ICU Safety Reporting System (ICUSRS) is an anonymous Web-based incident reporting system developed by a team of medical and public health researchers at the Johns Hopkins University as part of a demonstration project funded by the Agency for Healthcare Research and Quality in September 2001. • Results: 55 incidents were related to CPOE. The majority (85%) of CPOE incidents resulted in a medication error, while 10 events (15%) did not. Of the CPOE incidents that resulted in a medication error or near miss (an event that did not result in patient harm), 37 (67%) were coded as user errors, 11 (20%) as software errors, and 7 (13%) as computer malfunction. • Conclusion: CPOE can lead to new types of errors. When implementing CPOE systems, hospital leaders should create independent checks to monitor for mistakes and ensure sufficient resources for training, be sensitive to users’ suggestions for improvement, and plan ahead regarding strategies for addressing the likely disruption in work flow and staffing.

E

ach year, a significant number of people are harmed rather than helped by medical care [1]. Adverse drug events (ADEs) are a significant cause of injury in hospitalized patients that can increase patient length of stay and total hospitalization costs [2–6]. ADEs are frequently preventable. In one study, 56% of preventable ADEs were the result of prescribing errors [2].

www.turner-white.com

Computerized physician order entry (CPOE) has been recommended as a way to reduce medical errors and thereby improve patient safety [7–9]. CPOE systems are designed to reduce medication errors by standardizing orders and making relevant patient information accessible [10–12]. CPOE systems provide legible documentation of orders and quicker processing and can be programmed to prompt users during medication order entry regarding dosage recommendations, drug interaction warnings, and patient allergies [12,13]. Ideally, CPOE systems will assist physicians and other prescribers in making better clinical decisions for their patients. For these reasons, health care systems are installing CPOE systems to help solve their medication error problems. A consensus statement developed by a panel of CPOE experts outlines considerations for successful CPOE implementation, including cost, motivation to implement, vision and leadership, and value to users [14]. Despite their potential benefits, computerized medication ordering systems were associated with 10% of the 278,607 medication errors reported to the MedMarx medication incident reporting system in 2001–02 [15,16]. In addition, MedMarx data indicated that use of a CPOE system was the fifth leading cause of all medication errors and the third leading cause of errors involving wrong patients. Errors occurred in all phases of the medication use process (ie, prescribing, transcribing, dispensing, administering, monitoring). Some examples include a prescribing physician selecting the wrong patient in the computer system, nursing staff entering orders for a patient under another patient’s profile, a dispensing pharmacist entering orders into the computer under the wrong patient room, and a nurse administering medications

From The Johns Hopkins University, the School of Medicine, Department of Anesthesiology and Critical Care Medicine Quality and Safety Research Group; the School of Nursing; and the Bloomberg School of Public Health, Department of Health Policy and Management, Baltimore, MD.

Vol. 12, No. 8 August 2005 JCOM 407

MEDICATION ERRORS to the wrong patient because she accessed a different patient’s computer screen [17]. Although CPOE may reduce medication errors, it introduces new risks and new types of errors, particularly as systems are initially implemented. Potential factors that could increase risk of error include diversion of resources from other critical areas and adjusting to a new medication delivery system that creates the need for additional training and new redundant systems. The Intensive Care Unit Safety Reporting System (ICUSRS) has been evaluating near misses and adverse events in 18 intensive care units (ICUs) throughout the United States for over 2 years [17,18]. During the first year of our study, 3 participating ICUs implemented commercially available CPOE systems. Here we describe incidents identified in the ICUSRS attributed to CPOE during the first year. Methods Participating Centers Of 18 ICUs submitting reports to the ICUSRS between 1 July 2002 and 30 June 2003, 3 ICUs (17%) implemented CPOE during this time period. The first ICU implemented CPOE within 5 months after the study period began, while the other ICUs implemented CPOE in the last quarter. The first ICU was part of an approximately 300-bed hospital in a network of hospitals within an academic medical center, the second ICU was in a private, not-for-profit community teaching hospital with approximately 300 beds, and the third ICU was in a private, not-for-profit community hospital with approximately 400 beds. The 3 CPOE systems were not used exclusively by physicians. Reporting System The ICUSRS is an anonymous Web-based incident reporting system developed by a team of medical and public health researchers at the Johns Hopkins University as part of a demonstration project funded by the Agency for Healthcare Research and Quality in September 2001. Through a partnership with the Society of Critical Care Medicine, 30 ICUs across the United States were identified for participation in the ICUSRS project. The intention was to recruit geographically diverse hospitals that were willing to participate in reporting and improvement efforts rather than a representative sample of hospitals or ICUs. Both adult and pediatric ICUs were identified that included surgical, medical, trauma, and cardiac services. A study team was assembled at each site, including the ICU director or an ICU attending (principal investigator), nurse manager (coprincipal investigator), nurse educator, and risk manager. The study team remains the clinical champion for the ICUSRS project throughout the demonstration project. After a site obtained internal Institutional Review Board approval and the study team discussed participation in the 408 JCOM August 2005 Vol. 12, No. 8

project with its risk management department, a 2-day site visit was scheduled with the site study team. All members of the ICU staff participated in both a didactic portion and a hands-on training session to learn to complete the ICUSRS form. In order to provide training to staff who were not present during the site visit, the nurse educator received “train the trainer” instruction so that, theoretically, all staff working within the ICU would have received the same education and training. All members of the ICU staff could complete a report, which was sent and stored in a central server at Johns Hopkins. All data collected were anonymous to patient and provider; reporters elected whether to identify their ICU. Data collected are considered a limited data set under the Health Insurance Portability and Accountability Act (HIPAA). Further details on the development and implementation of the ICUSRS have been previously described [17,18]. Reporting Form A copy of the reporting form can be found on the ICUSRS Web site (www.icusrs.org). A free-text box is provided for reporters to describe the incident in their own words, including details before, during, and after the event. Incidents were classified as follows: (1) orders not completed/delayed, (2) incorrect/incomplete care delivered, (3) patient identification problems, (4) assessment and review problems, (5) airway management, (6) lines/tubes/drains, (7) skin integrity, (8) patient testing, (9) patient fall, (10) medication and therapeutics, (11) equipment/medical device, (12) information technology, (13) patient restraint, and (14) hazardous situations. Event types were assigned by expert coders and were not mutually exclusive. Interrater reliability was 99.6% for the first year of event coding [17]. Check boxes were provided for reporters to indicate patient demographics, type of provider reporting and those involved in the incident, location of incident, degree of patient harm (none, mild, moderate, severe), and system factors that contributed to or could have prevented the incident. System factors in the first year of reporting included training/ education, team, patient, provider, task, ICU environment, and institutional environment [19]. The reporter can select more than 1 contributing factor for each incident and select subfactors within each category (Table 1). Reporters identified CPOE as a factor in the event as a free-text entry. A CPOE-related incident was defined as any event in which CPOE user error, software error, or computer malfunction occured that led to or could have led to patient harm. At the time of the study, CPOE was not a stand-alone check box, but it was added in July 2003. All reporters were asked “Did this incident involve a medication error?” If the reporter answered “yes” to this question, a screen would appear asking when in the medication process the error occurred. The reporter could select www.turner-white.com

ORIGINAL RESEARCH among check boxes and could select all that applied. A medication error was defined as any event involving a medication at any step in the medication process that led to or could have led to harm. A prescribing error was defined as any incident in which a medication, dosage, or frequency was ordered incorrectly. A dispensing error was defined as any time a medication or dosage was dispensed differently from a physician’s order. An administration error was defined as any time a patient received an incorrect medication, an incorrect dose of a prescribed medication, an incorrect frequency of a prescribed medication, or a drug infusion over a time period not ordered. A documentation error was defined as incorrect transcription of a medication order or a CPOEbased Medication Administration Record (e-MAR) printout of medications that differed from those ordered. A monitoring error was defined as any time a medication was not monitored as prescribed. Reporters also filled in the name of the medication and the route of administration. Results Between 1 July 2002 and 30 June 2003, 18 ICUs submitted 854 reports to the ICUSRS, averaging 4 incidents per month per ICU (range, 0–14). The 3 ICUs with CPOE systems submitted 55 incidents that were classified as related to CPOE. Forty-five (85%) of the CPOE incidents resulted in a medication error while 10 events (15%) did not. Thirty-seven (67%) were coded as user errors, 11 (20%) as software errors, and 7 (13%) as computer malfunctions (Table 2). Table 3 shows examples of CPOE errors. Table 1 shows where CPOE was coded in the contributing factor list before adding it as an individual factor category. The examples in Table 1 are taken directly from the text entered into the report and entered next to the related contributing factor that was selected by the staff member completing the report. Of the 55 CPOE incidents, training and education was reported as a contributing factor in 44 (43%), team factors in 21 (20%), and provider factors in 16 (16%) of incidents. The majority (88%–98%) of CPOE events reported did not result in patient harm (Table 4). Discussion Incident reporting provides a crucial mechanism to identify hazards and improve patient safety. Through our incident reporting system, we have learned valuable lessons regarding the implementation of CPOE. First, significant CPOE errors can occur when these systems are implemented and in the months thereafter. As such, hospital leaders should recognize that implementing a CPOE system creates new risks and types of errors. One physician stated that one had to be extremely careful when scrolling the medication list as it was easy to select the wrong dose or incorrect medication from the list. So although transcription errors are avoided, one www.turner-white.com

may have incorrect medications or doses because of selection errors. Steps should be taken to anticipate and protect against these new errors. For example, 2 ICUs protected against mistakes by maintaining a paper record for comparison with the electronic record as an independent check. Second, user error was the most frequent type of CPOE error, and insufficient training was the most commonly cited subfactor contributing to this mistake. Two of the 3 hospitals stated there was a formal educational requirement before implementing CPOE. This information is not known for the third institution, however. In narrative descriptions of the incident, several physicians expressed concern regarding insufficient training. This underscores 2 comments that we have heard on site visits at many participating ICUs: (1) physicians at all levels, from the ICU attending to house officers, do not receive the orientation or inservice education at the same frequency as received by nursing personnel; and (2) there is an undersupply of physician coverage that would allow physicians to attend training beneficial to their clinical practice. Hospital leaders need to ensure that physicians prescribing medications receive adequate CPOE training. This will require that hospitals establish a system that has sufficient resources to train and periodically evaluate the adequacy of training and that provides for cross-coverage of patient care to ensure that physicians can allow adequate time to complete the training process. Third, there were some problems with the software in the commercially available products. For example, medications entered into the CPOE system suddenly disappeared from the e-MAR, placing patients at risk of not receiving their medications. It may be necessary to have a CPOE vendor representative present when the system is implemented to fix software and hardware malfunctions and potentially retrofit a system to the customer’s needs, especially in the initial implementation phases. One study recommended providing technical and training support 24 hours a day, 7 days a week [14]. An important conclusion of the 2002 MedMarx report is that technology has the potential to reduce many types of errors while also increasing risk and creating new types of errors. Facilities are encouraged to develop interdisciplinary teams with expertise in content and process redesign to prioritize choices of new technology and to ensure implementation that does not create new patient safety risks [16]. There are limitations to this study. First, our results may not be generalizable to other hospitals, as participating hospitals may differ in enthusiasm and leadership from hospitals that did not volunteer. Second, due to the perceived requirement for anonymous reporting and resource constraints, we were not able to independently verify the accuracy of submitted incidents or reporters’ coding of events. Independent verification of incidents could undermine the Vol. 12, No. 8 August 2005 JCOM 409

MEDICATION ERRORS Table 1. Factors Contributing to CPOE Incidents (n = 55)

Contributing Factor

No. of Times Factor Selected (%)*

Training/education Knowledge, skills, and competence Failure to follow established protocol Supervision and seeking help

44 (43)

Team factors Verbal or written communication during hand off

21 (20)

Verbal or written communication during routine care Patient factors Condition (complexity, agitation) ICU environment Skills mix

Distribution of Subfactors within Category, %

57 39

5

43

57

Example

Physician and nurse with limited dosing knowledge increased phenylephrine to 10 times the recommended maximum dose. Standard policy requires that physicians flag all new written orders and oncoming nurse check order sheet to catch unflagged orders. Neither remembered, and the order to reduce hydrocortisone from 100 mg to 50 mg was missed. A patient was to be given fosphenytoin. The drug must be prescribed in phenytoin equivalents, which are abbreviated on CPOE as “pe.” The nurse giving the drug saw “PE” and administered it NG instead of IV. Local abbreviations include PE (for polyethylene tubing). Patient transferred units. At the time of admission, antibiotics ordered in CPOE system and automatic time defaulted to start time of 9 PM. Admitting nurse and physician assumed antibiotics started at outlying hospital or in transport. 3 days later patient positive N. meningitidis. Antibiotics ordered but never started. Order written to change heparin dose not communicated to nurse, causing delay in rate change.

1 (1) 100

Trauma patient who was unstable and deteriorating develops hypotension refractory to norepinephrine, dobutamine, and fluid resuscitation.

15 (15) 7

Workload

7

Availability/ maintenance of equipment

40

Administrative and managerial support Physical environment (eg, space, noise level)

33

13

New nurse who recently completed orientation gave wrong concentration of furosemide. Several nurses sent home sick, resulting in staffing shortage. Charge nurse caring for 2 patients with 31-bed unit full, all of which led to incorrect dose programmed into dobutamine pump. Dilantin dose ordered in CPOE as a future order with start 12 hours after loading dose in the “order comments” field. The team realized that drug would not be given unless order “activated,” and nurses cannot see the “order “comments” field. Equipment not functional for unit needs. Medication order faxed 2 hours late because unit was extremely busy and unit clerk off that day. Noise on the unit so loud nurse could not hear physician orders.

(continued on next page)

level of trust necessary for open and candid reporting. Because our system relied on voluntary reporting, we did not capture all incidents and do not know to what extent staff may have underreported events. Our rate of error reporting may appear low; higher error rates have been reported in studies that used direct observation or chart review to calculate error rates [2–4]. Third, as an observational study we were not able to obtain baseline data on severity, type, and 410 JCOM August 2005 Vol. 12, No. 8

frequency of events reported to their existing reporting systems. However, as our study describes medication errors that were a result of CPOE, this would provide baseline data for this type of event and related system factors. Fourth, we found harm to be underreported in the majority of cases based on our clinical experience. For example, in a case submitted to the ICUSRS that was not related to CPOE, the text description in one incident reported that the patient died, yet www.turner-white.com

ORIGINAL RESEARCH Table 1. (continued)

Contributing Factor

No. of Times Factor Selected (%)*

Provider factors Fatigue Motivation/ attitude Task factors Availability of protocols Institutional environment Financial resources

Distribution of Subfactors within Category, %

Example

16 (16) 63

15 hours into an overtime shift, nurse received a new admission; as a result of fatigue, antibiotic order was missed. Trimethoprim-sulfamethoxazole ordered at 26 kg/mg/day instead of the recommended 20 mg/kg/day.

37 3 (3) 100

Patient found with swollen hand and arm due to peripheral IV infiltrating with dopamine—no protocol available noting use of central venous catheters for vasoactive drugs.

50

Unit having problem with pump infusing at improper rate. Solution was to use smart pumps, which ensures that infusion is correct; however, resources not immediately available to buy smart pumps. Nurse reviewing orders in computer noticed that intravenous fluid (IVF) was changed at 3 PM but old IVF still infusing. She called pharmacy. Pharmacist stated that new order was not processed and would make the IVF.

2 (2)

Time pressures

50

*Reporters usually identified more than 1 category of contributing factor for each incident. Therefore, the total number of contributing factors is greater than the number of incidents.

Table 2. CPOE Error and Medication Error (n = 55) Medication Error

CPOE Error User error Software error Computer malfunction Total

Prescribing n (%)

Documentation* n (%)

Dispensing n (%)

Administration n (%)

Monitoring n (%)

10 (27)0 1 (9)0 1 (14) 12

12 (32) 02 (18) 02 (29) 16

5 (14) 1 (9)0 1 (14) 7

9 (24) 6 (55) 2 (29) 17

1 (3)0 1 (9)0 1 (14) 3

*Refers to the Medication Administration Record (MAR).

Table 3. Examples of CPOE-Related Errors User error leading to prescribing error Same order for lymphocyte immune globin entered into CPOE system twice by attending. He did not cancel first entry, assuming the first order would be canceled by the CPOE system.

Outcome: mild harm Patient missed a dose of medication before the error was picked up by staff.

Software error leading to prescribing error Critical medication (trimethoprim-sulfamethoxazole) dropped out and canceled by the CPOE system, unknowingly and without orders.

Outcome: near miss Error was picked up by the patient’s nurse; might have resulted in an opportunistic infection.

Computer malfunction leading to documention error In reviewing orders, the e-MAR from the CPOE system read for medication to be given every 8 hours. The correct frequency was every 6 hours. The system went down shortly thereafter until 6 PM.

Outcome: near miss Nurse caught the error and administered the medication on time.

e-MAR = CPOE-based Medication Administration Record.

www.turner-white.com

Vol. 12, No. 8 August 2005 JCOM 411

MEDICATION ERRORS Table 4. Degree of Harm in Reports Submitted to ICUSRS (n = 55) Harm

n*

None, %

Mild, %

Moderate, Severe, % %

Physiologic Physical injury Psychological distress Discomfort Patient/relative dissatisfaction Anticipated/ actual prolonged stay

52 55 55

88 96 98

10 4 2

2 0 0

0 0 0

55 54

96 90

2 6

2 4

0 0

55

96

4

0

0

*Number varies because reporters had the option of not selecting any harm category when reporting.

no harm was checked. Overall, caregivers underreported harm and contributing factors. Conclusion Generally, successful implementation of a CPOE system enhances multidisciplinary collaboration in the larger context of delivering safe patient care [20]. In the long run, CPOE is likely to prevent some types of medication errors. Nevertheless, when systems are first implemented, the risk for a medication error may increase. When implementing CPOE systems, hospital leaders should create independent checks to monitor for mistakes and ensure sufficient resources for training. In addition, leaders should be sensitive to users’ suggestions for improvement and plan ahead regarding strategies for addressing the likely disruption in work flow and staffing. Corresponding author: David A. Thompson, DNSc, MS, RN, Johns Hopkins University School of Medicine, 901 S. Bond St., Ste. 318, Baltimore, MD 21231, [email protected]. Funding/support: This project was funded by a grant from the Agency for Healthcare Research and Quality (AHRQ), #U18 HS11902. Financial disclosures: None. Author contributions: conception and design, DAT, LD, TD, LHL, AWW, PJP; analysis and interpretation of data, DAT, LHL, FD, MF, AWW; drafting of the article, DAT, LD, CGH, AWW; critical revision of the article, DAT, CGH, TD, LLM, AWW, PJP; statistical expertise, DAT, PJP; obtaining of funding, TD, LLM, AWW, PJP; administrative support TD, LHL; collection and assembly of data, LHL, FD, MF.

References 1. Kohn LT, Corrigan JM, Donaldson MS, editors. To err is human: building a safer health system. Washington (DC): Na-

tional Academy Press; 2000. 2. Bates DW, Cullen DJ, Laird N, et al. Incidence of adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Study Group. JAMA 1995;274:29–34. 3. Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med 1991;324:377–84. 4. Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients. Results of the Harvard Medical Practice Study I. N Engl J Med 1991;324:370–6. 5. Kaushal R, Bates DW, Landrigan C, et al. Medication errors and adverse drug events in pediatric inpatients. JAMA 2001; 285:2114–20. 6. Leape LL, Lawthers AG, Brennan TA, Johnson WG. Preventing medical injury. QRB Qual Rev Bull 1993;19:144–9. 7. The Leapfrog Group for patient safety: fact sheet. Computer physician order entry. Available at www.leapfroggroup.org/ media/file/Leapfrog-Computer_Physician_Order_Entry_ Fact_Sheet.pdf. Accessed 18 Jul 2005. 8. Doolan DF, Bates DW. Computerized physician order entry systems in hospitals: mandates and incentives. Health Aff (Millwood) 2002;21:180–8. 9. Eikel C, Delbanco S. John M. Eisenberg Patient Safety Awards. The Leapfrog Group for Patient Safety: rewarding higher standards. Jt Comm J Qual Saf 2003;29:634–9. 10. Bates DW, Leape LL, Cullen DJ, et al. Effect of computerized physician order entry and a team intervention on prevention of serious medication errors. JAMA 1998;280:1311–6. 11. 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;6:313–21. 12. Bates DW, Kuperman G, Teich JM. Computerized physician order entry and quality of care. Qual Manag Health Care 1994;2:18–27. 13. Johnston ME, Langton KB, Haynes RB, Mathieu A. Effects of computer-based clinical decision support systems on clinician performance and patient outcome. A critical appraisal of research. Ann Intern Med 1994;120:135–42. 14. Ash JS, Stavri PZ, Kuperman GJ. A consensus statement on considerations for a successful CPOE implementation. J Am Med Inform Assoc 2003;10:229–34. 15. Summary of information submitted to Medmarx in the year 2001: a human factors approach to medication errors. Rockville (MD): U.S. Pharmacopeia; 2002. 16. Summary of information submitted to Medmarx in the year 2002: a human factors approach to medication errors. Rockville (MD): U.S. Pharmacopeia; 2003. 17. Holzmueller CG, Pronovost PJ, Dickman F, et al. Creating the web-based intensive care unit safety reporting system. J Am Med Inform Assoc 2005;12:130–9. 18. Wu AW, Holzmueller CG, Lubomski LH, et al. Development of the ICU safety reporting system. J Patient Saf 2005;1:23–32. 19. Vincent C. Understanding and responding to adverse events. N Engl J Med 2003;348:1051–6. 21. Gorman PN, Lavelle MB, Ash JS. Order creation and communication in healthcare. Methods Inf Med 2003;42:376–84.

Copyright 2005 by Turner White Communications Inc., Wayne, PA. All rights reserved.

412 JCOM August 2005 Vol. 12, No. 8

www.turner-white.com