Impact of Computerized Provider Order Entry on Hospital Medication Errors Daniel L. Roberts, MD, Brie N. Noble, Mary J. Wright, RN, MN, Eric A. Nelson, MS, RPh, Judd D. Shaft, and Jorge Rakela, MD Abstract Objective: To assess the impact of a phased implementation of computerized provider order entry (CPOE) on the incidence of medication errors and preventable adverse drug events (ADEs). Design: Retrospective observational analysis. Setting: Mayo Clinic Hospital (Phoenix, AZ), a 232- bed teaching hospital. Participants: All patients admitted to Mayo Clinic Hospital from 6 November 2006 to 7 May 2008. Measurements: Medication errors and preventable ADEs per 1000 patient-days. Results: Relative to the period prior to CPOE implementation, the frequency of medication errors per 1000 patient-days was unchanged during implementation (13.7 vs. 14.1, P = 0.66) and significantly reduced after implementation was completed (10.4 vs. 14.1, P < 0.001). CPOE implementation was associated with a reduced rate of insulin orders containing errors (0.05 vs. 0.20, P = 0.007) but not heparin (0.62 vs. 0.52, P = 0.56). There was a significant decrease in preventable ADEs after implementation (6 vs. 0, P = 0.01). Conclusion: CPOE systems can improve medication safety. Implementing CPOE in a step-wise fashion, rather than all at once, does not appear to have been associated with an increase in medication errors. Computerized provider order entry (CPOE) has been suggested as a potential solution for the substantial problem of inpatient prescribing errors [1]. King et al demonstrated a reduction in medication errors (but not adverse drug events [ADEs]) with the implementation of CPOE [2], while Potts reported a reduction in both medication errors and ADEs [3]. Supporters of CPOE have pointed to its ability to reduce or even eliminate verbal orders [4] and errors related to drug-drug interactions and allergies [5]. Enthusiasm for CPOE as it relates to inpatient prescribing is by no means unanimous, however. A widely cited article identified several types of medication errors that can be expedited, rather than reduced, by CPOE [6]. Others have conceded benefits of CPOE but pointed to remaining substantial vulnerabilities [7]. Finally, an institution that had previously published the positive impact of CPOE on its ADE rate [8] later reported an unexpected increase in hospital mortality [9]. Mayo Clinic Hospital (Phoenix, AZ), a 232-bed teaching hospital, introduced CPOE in 2007 using IDX LastWord, version 4.3. The design team leveraged customization tools and rules-based functionality internal to IDX software. Medication safety was a principal focus during design and implementation. Specific features designed to reduce the rate of medication errors included automatic allergy and interaction checking, automatic duplicate medication checking, guidance for renal dosing and the provision of standardized dosing and frequency recommendations. The process of ordering heparin was substantially altered by the introduction of CPOE; the hospital s weight-based nomogram (which required the provider to calculate and hand-write bolus and drip doses, with a pharmacist checking the calculations) was automated so that all calculations were made by the CPOE system and confirmed by the provider. CPOE also introduced changes to the process of ordering insulin, albeit less dramatic changes than those regarding heparin. The hospital s pre-printed sliding scales were converted directly to the electronic environment and duplicatechecking was applied to the different formulations of insulin. Rather than implementing CPOE throughout the hospital at one time, CPOE was rolled out on a unitby-unit basis over a 6-month period. Investigators from From the Mayo Clinic, Phoenix, AZ. www.jcomjournal.com Vol. 20, No. 3 March 2013 JCOM 109
Hospital Medication Errors Table 1. Medication Errors and Adverse Drug Events Period A Period B Period C P (A vs. B) P (A vs. C) Medication errors (MEs), n 474 421 362 MEs/1000 patient-days 14.1 13.7 10.4 0.66 < 0.001 Adverse drug events (ADEs), n 6 4 0 ADEs/1000 patient-days 0.18 0.13 0.00 0.62 0.01 our institution have previously reported a nonsignificant trend towards fewer medication errors in the operating room during and after CPOE implementation [10]. We have included medication errors in all aspects of the inpatient practice in the current analysis. Methods Error Documentation All medication errors at Mayo Clinic Hospital are reported to a centralized medication events committee independent from the CPOE design and implementation process. Errors can be reported by anyone involved in the care of patients, including physicians, nurses, pharmacists and technicians. Each error is categorized by quality assurance personnel and, if patient harm occurred, categorized according to the National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP) index [11]. Errors are categorized as relating to administration, preparation/ dispensing, prescribing/ordering or transcription, then further categorized as follows: IV infiltrate, wrong IV concentration, wrong IV rate, documented contraindication/allergy, drug not given, extra dose given, pain management/sedation event, prescribing issue, unordered drug, wrong dose, wrong route, or wrong time. The committee can optionally code 1 or more causes (calculation issue, computer entry/transcription issue, documentation issue, illegible/unclear order, IV pump issue, IV rate/flow issue, labeling issue, look alike/ sound alike, medication availability issue, order not faxed, out of sequence, Pyxis fill issue, or override access prior to pharmacist review) and/or 1 or more factors (chart check issue, communication issue, distractions, medication five rights [right medication, right patient, right dose, right route, right time] not followed, patient ID issue, computer issue, environmental issue or other). This process was unchanged before, during, and after CPOE implementation. Data Assessment We examined medication error data for three 6-month time periods: Period A Preimplemention (Nov 2006 May 2007), Period B Implementation (May 2007 Nov 2007) and Period C Postimplementation (Nov 2007 May 2008). Errors that occurred outside of the inpatient practice (eg, emergency department, outpatient practices housed at the hospital) were excluded. Data about hospital census and length of stay were collected. The study was approved by the Mayo Clinic institutional review board. The primary outcome was the number of medication errors and preventable ADEs per 1000 patient-days. Medication error was defined as an error in the process of ordering or delivering a medication, regardless of whether an injury occurred or the potential for injury was present [12]. A preventable ADE was defined as an injury resulting from a medication error. Secondary outcomes were the rate of medication errors related to heparin and insulin per 1000 orders. Rates of medication errors across the 6-month periods (before, during and after CPOE) were compared using Pearson s chi-square test. Comparison of ADEs was conducted using Fisher s exact test because the expected cell count was less than 5. P values less than 0.05 were considered statistically significant. All computations were performed by using SAS software version 9 (SAS Institute, Cary, NC). Results Table 1 shows the rate of medication errors per 1000 patient-days. The rate was unchanged during implementation (13.7 vs. 14.1, P = 0.66) and decreased after full implementation (10.4 vs. 14.1, P < 0.001). Table 2 summarizes the error rate for orders involving heparin and insulin across time periods. With regard to heparin, no statistically significant differences in the rate of medication errors were noted during or after implementation. Regarding insulin, there was a signifi- 110 JCOM March 2013 Vol. 20, No. 3 www.jcomjournal.com
Table 2. Medication Errors (MEs) Involving Heparin and Insulin Period A Period B Period C P (A vs. B) P (A vs. C) MEs involving heparin 35 30 39 MEs involving heparin/ 5.2 4.9 6.2 0.81 0.47 1000 heparin orders MEs involving insulin 13 12 4 MEs involving insulin/ 1000 insulin orders 2.0 1.9 0.5 0.82 0.007 Table 3. Medication Errors by Category Category Period A Period B Period C P (A vs. B) P (A vs. C) s 8.09 (272) 8.30 (255) 7.22 (252) 0.77 0.19 Preparation/dispensing errors 1.01 (34) 0.94 (29) 1.37 (48) 0.78 0.17 Prescribing/ordering errors 1.13 (38) 2.28 (70) 1.32 (46) < 0.001 0.49 Transcription errors 3.87 (130) 2.18 (67) 0.46 (16) < 0.001 < 0.001 cant decrease in the rate of medication errors after full implementation had occurred (0.05 vs. 0.20, P = 0.007). Table 3 shows medication error rates by category of medication error. The rate of errors attributed to prescribing and ordering more than doubled during implementation (2.28 vs. 1.13 per 1000 patient-days, P < 0.001) but returned close to baseline (1.32 per 1000 patient-days, P = 0.49) after full implementation. On the other hand, transcription errors dropped sharply during implementation (2.18 vs. 3.87, P < 0.001) and after implementation (0.46, P < 0.001). There was a nonsignificant increase in errors related to preparation and dispensing after full implementation (1.37 vs. 1.01, P = 0.17). With regard to types of errors (Table 4), errors of omission dropped by more than half (2.00 vs. 4.23, P < 0.001) and there were significant drops in errors attributed to extra doses (0.86 vs. 1.49, P = 0.02) and wrong doses (1.95 vs. 2.80, P = 0.02). There was a significant increase in errors attributed to the catchall type prescribing issue (0.26 vs. 0.06, P = 0.04). A transient nonsignificant trend toward increased errors attributed to unordered drugs was noted during implementation (2.80 vs. 2.14, P = 0.09) but stabilized after full implementation (2.18, P = 0.92). Regarding factors and causes for medication errors (Table 5 and Table 6), there was a significant decrease after implementation in errors attributed to distractions and labeling. There was, on the other hand, a nonsignificant increase during the period of implementation in errors attributed to communication problems, and this improved but did not return to baseline after full implementation. An increase in errors attributed to nurses not following the 5 rights of medication administration was noted during implementation and continued (as a nonsignificant trend) after full implementation. Table 7 describes the ADEs that occurred. ADEs were rare before CPOE implementation, but there was a significant decrease after implementation compared to baseline (0 vs. 6, P = 0.01). Of the 4 ADEs that occurred during implementation, 2 occurred on floors not yet using CPOE and 2 occurred on CPOE floors. Discussion The implementation of CPOE at Mayo Clinic Hospital was associated with a significant decrease in the rates of medication errors and adverse medication events. Although supporters of CPOE have long emphasized its medication safety benefits, in actual practice, its impact on medication safety remains to be demonstrated. Possible explanations for the decrease in medication errors in this study include the marked reduction in verbal orders, the elimination of handwriting errors, computerized guidelines for renal dosing and the elimination of heparin dosing calculations by providers. www.jcomjournal.com Vol. 20, No. 3 March 2013 JCOM 111
Hospital Medication Errors Table 4. Medication Errors by Type Error Type Period A Period B Period C P (A vs. B) P (A vs. C) IV infiltrate 0.54 (18) 0.42 (13) 0.37 (13) 0.52 0.31 Wrong IV concentration 0.06 (2) 0.00 (0) 0.00 (0) 0.18 0.15 Wrong IV rate 0.33 (11) 0.16 (5) 0.20 (7) 0.19 0.31 Documented contraindication/allergy 0.45 (15) 0.29 (9) 0.32 (11) 0.31 0.38 Drug not given 4.23 (142) 3.06 (94) 2.00 (70) 0.01 < 0.001 Extra dose given 1.49 (50) 1.69 (52) 0.86 (30) 0.51 0.02 Pain management/sedation event 0.06 (2) 0.03 (1) 0.03 (1) 0.62 0.54 Prescribing issue 0.06 (2) 0.65 (20) 0.26 (9) < 0.001 0.04 Unordered drug 2.14 (72) 2.80 (86) 2.18 (76) 0.09 0.92 Wrong dose 2.80 (94) 2.21 (68) 1.95 (68) 0.14 0.02 Wrong route 0.45 (15) 0.36 (11) 0.32 (11) 0.58 0.38 Wrong time 1.52 (51) 2.02 (62) 1.89 (66) 0.13 0.24 Table 5. Medication Errors by Cause* Cause Period A Period B Period C P (A vs. B) P (A vs. C) Calculation issue 0.36 (12) 0.42 (13) 0.20 (7) 0.67 0.22 Computer entry/ 2.71 (91) 2.31 (71) 0.86 (30) 0.32 < 0.001 transcription issue Documentation issue 0.21 (7) 0.26 (8) 0.14 (5) 0.67 0.52 Illegible/unclear order 0.09 (3) 0.16 (5) 0.03 (1) 0.40 0.30 IV pump issue 0.03 (1) 0.00 (0) 0.11 (4) 0.34 0.19 IV rate/flow issue 0.39 (13) 0.23 (7) 0.17 (6) 0.25 0.09 Labeling issue 0.36 (12) 0.13 (4) 0.09 (3) 0.07 0.02 Look alike/sound alike 0.30 (10) 0.39 (12) 0.29 (10) 0.52 0.93 Medication availability 0.12 (4) 0.29 (9) 0.40 (14) 0.12 0.02 issue Order not faxed 0.15 (5) 0.10 (3) 0.14 (5) 0.56 0.95 Out of sequence 0.09 (3) 0.10 (3) 0.03 (1) 0.91 0.30 Pyxis fill issue 0.24 (8) 0.23 (7) 0.20 (7) 0.93 0.74 Override access prior 0.00 (0) 0.03 (1) 0.09 (3) 0.30 0.09 to pharmacy review Other 2.38 (80) 3.22 (99) 1.75 (61) 0.04 0.07 *Assignment of cause is optional. Our data demonstrate a statistically significant decrease in the rate of medication errors after CPOE implementation, whereas a previous report regarding the surgical practice at our hospital demonstrated a decrease without statistical significance. As the previous report comprises only 38 of the 1257 medication errors we report, we suggest that the broader scope of this study accounts for the difference in statistical significance. The decision to roll out CPOE gradually, rather than as a single big bang, afforded us the opportunity to study the implementation phase to determine if it was particularly error-prone. There was no statistically significant increase 112 JCOM March 2013 Vol. 20, No. 3 www.jcomjournal.com
Table 6. Medication Errors by Assigned Factor* Factor Period A Period B Period C P (A vs. B) P (A vs. C) Chart check issue 2.80 (94) 2.31 (71) 0.86 (30) 0.22 < 0.001 Communication issue 0.89 (30) 1.30 (40) 1.03 (36) 0.12 0.56 Computer issue 0.39 (13) 0.85 (26) 0.46 (16) 0.02 0.65 Distractions 2.89 (97) 2.57 (79) 2.06 (72) 0.44 0.03 Environmental issue 0.21 (7) 0.29 (9) 0.17 (6) 0.50 0.73 5 Rights not followed 3.51 (118) 5.08 (156) 4.35 (152) 0.002 0.08 Patient identification issue 0.51 (17) 0.78 (24) 0.72 (25) 0.17 0.27 Other 1.13 (38) 1.37 (42) 0.97 (34) 0.40 0.53 *Assignment of factor is optional. Table 7. Preventable Adverse Drug Events Time Period Type of Error Description paper floor CPOE floor paper floor CPOE floor Unordered drug Drug not given Pain management/ sedation event Pain management/ sedation event Transcription error Extra dose given Extra dose given Wrong dose Drug not given Transcription error Pain management/ sedation event Drug not given Patient on heparin drip received unordered subcutaneous heparin. Monitored in Intermediate Care unit. Ordered diltiazem drip was interrupted, leading to transfer to intermediate care. Same patient and date as above. Post-operative patient on morphine PCA became unresponsive, requiring code blue, naloxone drip and transfer to intermediate care. Patient on prn IV hydromorphone became unresponsive, requiring code blue and transfer to intermediate care. Extra dose of oral potassium given, leading to hyperkalemia that delayed discharge to home hospice by 1 day. Extra dose of digoxin administered, requiring monitoring of digoxin level. 12.5 mg of oral quietapine ordered twice daily for agitation. Patient received 212.5 mg twice before error was caught. Patient s discharge to skilled nursing facility delayed by 1 day. Physician s written order was clear, but was mistranscribed when entered into computer. Patient with Cushing s disease was post-op from adrenalectomy and on IV hydrocortisone taper. Three doses of hydrocortisone doses ordered but not given, leading to adrenal crisis marked by hypotension, tachycardia, tachypnea and agitation. Patient admitted for nausea and abdominal pain, became unresponsive after receiving IV morphine, hydromorphone and promethazine, leading to code blue and transfer to intermediate care. Two doses of IV dihydroergotamine ordered for intractable headaches, but only one given. in medication errors or adverse medication events during the implementation process, suggesting that hospitals implementing CPOE in the future can consider a phased rollout as a safe alternative to the big bang approach. Closer examination of the data, however, demonstrates that the implementation period was not without significant risks. It is not surprising that the expected decrease in transcription-related errors was coupled to www.jcomjournal.com Vol. 20, No. 3 March 2013 JCOM 113
Hospital Medication Errors an increase in prescribing errors as users learned the new system. It is reassuring that this rate decreased after full implementation, but it is notable that it did not return fully to baseline. Similarly, the increase in errors attributed to nurses not following the five rights of medication administration (and the transient increase in unordered drugs being administered) suggests that the new system took time for nurses to learn as well, and that this learning period was associated with dangerous distractions. This rate, too, improved after implementation, but did not return to baseline. The safety gains from CPOE, however, appear to have been substantial and sustained. It is not difficult to trace the decreases in errors due to drugs not given, extra doses and wrong doses to the automation that CPOE introduces. It is surprising that there was not a significant decrease in errors attributed to calculation, but there was a non-significant trend towards improvement, and this particular cause of error was rare during all time periods (12 errors before implementation, 13 errors during implementation, 7 errors after implementation). The significant decrease in the rate of insulin errors suggests that mechanizing the process of insulin ordering may reduce the error risk. Our insulin ordering was largely unchanged with the advent of CPOE; the sliding scale orders were converted from paper to software with no modifications. The addition of duplicate checking may have provided substantial benefit, since pharmacists will often not call about duplicate insulin orders (since many forms of insulin are designed to be used together), whereas an automated system alerts the provider to duplicate orders regardless of context. It should also be noted, however, that a resident education program about insulin safety was undertaken the year before CPOE implementation began [13]; if its benefits were not immediate, it may have impacted the rate of insulin errors during our study. Conversely, the lack of a significant reduction in the rate of heparin errors suggests that mechanization and the avoidance of provider calculation errors is not sufficient to improve safety with this drug. Several factors may account for our CPOE system s lack of demonstrable benefit. First, heparin errors were relatively rare (although not nearly as rare as insulin errors, for which a significant benefit was demonstrated). Second, many of the errors involving heparin occur at the administration phase and are unaffected by the process of calculating the dose and ordering the medication. Finally, the hospital s CPOE system involves a somewhat rigid protocol, and any deviation from the protocol (eg, to skip a dose adjustment when the partial thromboplastin time is close to a threshold, or to hold anticoagulation for a procedure) requires the provider to use a separate set of non-protocol orders. It is therefore possible that the CPOE heparin protocol has substituted protocol-change and timing errors for the previous handwriting, verbal order and calculation errors, a phenomenon Koppel et al [6] described in general terms. Our study has several limitations that should be considered. First, we relied on self-reporting for medication errors and AMEs. This method, while inexact, was unchanged before, during and after CPOE implementation. Second, our implementation period unavoidably involves errors that occurred on floors that had not yet adopted CPOE. In summary, the implementation of CPOE at Mayo Clinic Hospital was associated with a significant decrease in the rate of medication errors and adverse medication events. A phased approach to implementation appears to be safe. Acknowledgements: This publication was made possible by Grant No. 1 UL1 RR024150 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and the NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH. Corresponding author: Daniel L. Roberts, MD, 5777 East Mayo Blvd., Phoenix, AZ 85259, roberts.daniel@mayo.edu. References 1. Bobb A, Gleason K, Husch M, et al. The epidemiology of prescribing errors: the potential impact of computerized prescriber order entry. Arch Intern Med 2004;164: 785 92. 2. King WJ, Paice N, Rangrej J, et al. The effect of computerized physician order entry on medication errors and adverse drug events in pediatric inpatients. Pediatrics 2003;112:506 9. 3. Potts AL, Barr FE, Gregory DF, et al. Computerized physician order entry and medication errors in a pediatric critical care unit. Pediatrics 2004;113:59 63. 4. Kaplan JM, Ancheta R, Jacobs BR; Clinical Informatics Outcomes Research Group. Inpatient verbal orders and the impact of computerized provider order entry. J Pediatric 2006;149:461 7. 5. 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. 6. Koppel R, Metlay JP, Cohen A, et al. Role of computerized 114 JCOM March 2013 Vol. 20, No. 3 www.jcomjournal.com
physician order entry systems in facilitating medication errors. J Am Med Assoc 2005;293:1197 203. 7. Ferner RE. Commentary: computer aided prescribing leaves holes in the safety net. BMJ 2004;328:1172 3. 8. Upperman JS, Staley P, Friend K, et al. The impact of hospitalwide computerized physician order entry on medical errors in a pediatric hospital. J Pediatr Surg 2005;40:57 9. 9. Han YY, Carcillo JA, Venkataraman ST, et al. Unexpected increased mortality after implementation of a commercial sold computerized physician order entry system. Pediatrics 2005;116:1506 12. 10. Stone WB, Smith BE, Shaft JD, et al. Impact of a computerized physician order-entry system. J Am Coll Surg 2009;208:960 9. 11. National Coordinating Council for Medication Error Reporting and Prevention website. Accessed 3 Feb 2012 at www.nccmerp.org. 12. Bates DW, Boyle DL, Vander Vliet MB, et al. Relationship between medication errors and adverse drug events. J Gen Intern Med 1995;10:199 205. 13. Cook CB, McNaughton DA, Braddy CM, et al. Management of inpatient hyperglycemia: assessing perceptions and barriers to care among resident physicians. Endocr Pract 2007;13:117 24. Copyright 2013 by Turner White Communications Inc., Wayne, PA. All rights reserved. www.jcomjournal.com Vol. 20, No. 3 March 2013 JCOM 115