- Email: [email protected]
Effect of Electronic Health Record Implementation in Critical Care on Survival and Medication Errors
CLINICAL INVESTIGATION
Effect of Electronic Health Record Implementation in Critical Care on Survival and Medication Errors Jenny E. Han, MD, MSc, Marina Rabinovich, PharmD, BCPS, Prasad Abraham, PharmD, BCPS, Prerna Satyanarayana, MD, T. Vivan Liao, PharmD, BCPS, Timothy N. Udoji, MD, George A. Cotsonis, MA, Eric G. Honig, MD and Greg S. Martin, MD, MSc ABSTRACT Background: Electronic health records (EHR) with computerized physician order entry have become exceedingly common and government incentives have urged implementation. The purpose of this study was to ascertain the effect of EHR implementation on medical intensive care unit (MICU) mortality, length of stay (LOS), hospital LOS and medication errors. Materials and Methods: Prospective, observational study from July 2010-June 2011 in MICU at an urban teaching hospital in Atlanta, Georgia of 797 patients admitted to the MICU; 281 patients before the EHR implementation and 516 patients post-EHR implementation. Results: Compared with the preimplementation period (N ¼ 43 per 281), the mortality risk at 4 months post-EHR implementation (N ¼ 41 per 247) and at 8 months post-EHR implementation (N ¼ 26 per 269) significantly decreased (P o 0.001). In addition, the mean MICU LOS statistically decreased from 4.03 ⫾ 1.06 days pre-EHR to 3.26 ⫾ 1.06 days 4 months post-EHR and to 3.12 ⫾ 1.05 days 8 months post-EHR (P ¼ 0.002). However, the mean hospital LOS was not statistically decreased. Although medication errors increased after implementation (P ¼ 0.002), this was attributable to less severe errors and there was actually a decrease in the number of severe medication errors (both P o 0.001). Conclusions: We report a survival benefit following the implementation of EHR with computerized physician order entry in a critical care setting and a concomitant decrease in the number of severe medication errors. Although overall hospital LOS was not shortened, this study proposes that EHR implementation in a busy urban hospital was associated with improved ICU outcomes. Key Indexing Terms: Electronic health records; Computerized physician order entry; Quality improvement; Medication error. [Am J Med Sci 2016;351(6):576–581.]
INTRODUCTION
E
lectronic health record (EHR) has been endorsed by national healthcare organizations such as the Institute of Medicine and the Leapfrog Group to improve patient care, especially computerized physician order entry (CPOE).1 The American Recovery and Reinvestment Act of 2009 allowed for $2 billion in discretionary health information technology funding and $18 billion in investments and incentives through Medicare and Medicaid to adopt and to use EHR technology.2 The Centers for Medicare and Medicaid Services has initiated financial incentives with EHR implementation with the goal of improving healthcare efficiency and quality.1 However, the implementation of this technology is extremely costly for hospital institutions. Furthermore, the data are unclear if this new technology has any direct patient care benefit. A recent systematic review specifically looking at intensive care unit (ICU) mortality and length of stay (LOS) with EHR did not show any substantial effects because of the small number of studies and heterogeneity of the patient population.3 Another systematic
576
review article looking at 67 studies involving CPOE found that there were positive effects in adherence to guidelines, satisfaction and usability, but the various studies did not find a difference in mortality.4-6 The EHR studies often look at the effect of CPOE on patient safety outcomes. Schenarts et al7 reported that in patients with traumatic injury, EHR reduced the hospital LOS and ICU LOS, but did not change mortality. The focus on patient safety has expanded the role of technology to minimize the risk of medical errors and adverse events. It has been reported that 1.7 medical errors occur a day in the ICU and that medication errors account for 78% of these errors.8 A systematic review of CPOE implementation in various healthcare settings found a relative risk reduction ranging from 13-99% for medication errors and from 30-84% for adverse drug events.9 However, Weant et al10 reported that medication errors increased to more than 4 times higher in the first month after CPOE implementation in neurosurgical ICU, but that the number of errors causing patient harm decreased. The use of CPOE in a United Kingdom ICU demonstrated significant reduction in overall number of
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
errors attributed to prescribing errors.11 Although research shows that administration errors are the second most common cause of medication errors, to date, administration and dispensing errors with CPOE implementation have not been evaluated. Various studies have shown variable results regarding EHR implementation and EHR with CPOE. Different mortality results were published by 2 commonly cited studies done in the pediatric ICU. The University of Pittsburgh study showed an increase in mortality with EHR implementation, and a similar study done at Montefiore Medical Center found no mortality difference with EHR implementation in the pediatric ICU.12,13 These studies are more than 8 years old and original research in adult ICU has not been performed recently. Therefore, this study evaluates the effect of EHR with CPOE implementation on mortality, medical ICU (MICU) LOS, hospital LOS and medication errors in critically ill adults.
MATERIALS AND METHODS Design This prospective, observational study evaluated the effect of EHR with CPOE on patient care outcomes from July 2010-June 2011. The study was reviewed by Emory University Institutional Review Board as well as the hospital Research Oversight Committee, and was exempted as a quality improvement study. Setting The study was conducted in the county MICU in Atlanta, GA, (Grady Memorial Hospital). Before implementation, medical records were used for each patient where notes and orders were manually prescribed and then a clerk would enter the orders into the computer. The implementation software was EPIC Systems Corporation with computer order entry, which included all patient medical information and which consisted of all healthcare provider notes including medications, flow sheets, orders, diagnostic and laboratory results. Implementation Hospital-wide training sessions were required for every healthcare provider before the initiation of the EHR with CPOE system. Inpatient physicians were required to complete 4 hours of training, and nurses 3 days of training. For 3 weeks during the implementation process, “super-users” were available in all clinical areas at all times. Super-users were healthcare providers who underwent additional training to assist colleagues when needed to deliver effective patient care.
outcome data would not be included more than once. The EHR with CPOE implementation commenced on October 31, 2010. Design Patient outcomes were compared from 4 months before the EHR implementation (July, August, September and October), 4 months after implementation (November, December, January and February) and 8 months after implementation (March, April, May and June). The 2 time frames were selected to investigate if there was a possibility of increased mortality during the implementation process, whereas the new system was being used that one would expect to improve overtime after familiarity with the system. Additionally, medication errors were evaluated in this patient population by a clinical pharmacist (M.R. and V.L.) for 1 month before EHR with CPOE implementation (August 2010) and for 1 month after implementation (January 2011). Demographic data as well as APACHE II scores were obtained from all admitted patients. Outcomes were mortality, ICU LOS and hospital LOS. Secondary outcomes were frequency of medication errors before and after CPOE implementation. Medication errors were defined as the number of medication errors per 1,000 patient days, and as percentage change in the type of medication errors and percentage change in the severity of medication errors classified as A (capacity for error) through I (error reached patient and contributed to death), based on categories defined by the National Coordinating Council for Medication Error Reporting and Prevention Taxonomy.14 Medication errors were classified in categories of prescribing, dispensing and administration errors. Statistical Analysis Chi-square analysis, Fisher’s exact test, T tests, logistic regression, one-way analysis of variance and one-way analysis of covariance were performed. Analysis of covariance was used to compare the 3 periods after adjusting for the APACHE II score. Tukey’s post hoc pairwise comparisons were performed using these adjusted means. Before the analysis, ICU LOS and hospital LOS were log-transformed, due to skewed distribution of LOS in days. The results from the analysis of variance were back-transformed to the original units. Two-tailed P o 0.05 was considered statistically significant. All analyses were done with SAS/STAT (9.3) software (SAS Institute, Cary, NC).
RESULTS Conduct All patients admitted to the MICU team between July 1, 2010 and June 30, 2011 were included in the study. Patients were excluded if they were readmitted to the MICU service after being transferred out, so their
A total of 797 critically ill medical patients admitted from July 2010-June 2011 were included in the study. There were no significant differences across all groups in race, sex, age and disease category (Table 1). As a marker of illness severity, the APACHE II score increased
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
577
Han et al
TABLE 1. Baseline patient demographics.
Race (N, %) Black Hispanic White Other Sex (female, N (%)) Age (years) Disease (N, %) CNS CV Pulmonary GI Renal Sepsis Other APACHE II (mean, SD) ICU mortality
Pre-EHR
Post-EHR 4 months (November-February)
Post-EHR 8 months (March-June)
N ¼ 281 (%)
N ¼ 247 (%)
N ¼ 269 (%)
P Value
232 (83) 8 (3) 36 (13) 2 (1) 119 (42) 55 ⫾ 16
210 (86) 12 (5) 16 (7) 7 (3) 103 (42) 53 ⫾ 16
216 (81) 9 (3) 33 (12) 9 (3) 131 (49) 54 ⫾ 16
0.05
30 (11) 64 (24) 48 (18) 40 (15) 4 (2) 44 (17) 32 (12) 18.6 (⫾8.51) N ¼ 238 43 (15)
16 (7) 49 (20) 66 (28) 24 (10) 4 (2) 49 (20) 31 (13) 23.29 (⫾10.54) N ¼ 231 41 (16)
28 (11) 55 (21) 61 (23) 24 (9) 2 (1) 45 (17) 45 (17) 21.13 (⫾9.74) N ¼ 248 26 (9)
0.20 0.50 0.10
o0.001* 0.04
CNS, central nervous system; CV, cardiovascular; HER, electronic health record; GI, gastrointestinal; ICU, intensive care unit; SD, standard deviation. * APACHE II post 4 months EHR 4post 8 months EHR 4pre-EHR, P o 0.05 Tukey.
from the preimplementation (mean ¼ 18.6 ⫾ 8.51) to the immediate postimplementation period 4 months (23.29 ⫾ 10.54), and slightly decreased thereafter at 8 months (21.13 ⫾ 9.74), (P o 0.0001, Table 1). The median APACHE II score was 21 (Figure).
Compared with preimplementation levels, the mortality risk at 4 months post-EHR implementation 8 months post-EHR implementation period markedly decreased, (95% CI: 1.63-5.67) and (95% CI: 1.163.82) respectively, P o 0.001 (Table 2). The average
Distribution of APACHE II 60
Apache II Score
50
40
30
20
10
0 Pre
4 Months
8 Months
EHR Implementation FIGURE. Apache II score for medical ICU patients pre-EHR, 4 months post-EHR and 8 months post-EHR implementation.
578
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
TABLE 2. Mortality comparing study periods adjusted for APACHE II. Mortality Pre vs. 8 months Pre vs. 4 months 4 vs. 8 months * *
Odds ratio
95% CI
P Value
3.04 2.10 1.45
(1.63-5.67) (1.16-3.82) (0.78-2.67)
o0.001* o0.001* 0.48
Adjusted for APACHE II. α o 0.05, significant P values are marked with asterisks (*).
MICU LOS adjusted by continuous APACHE II score, statistically decreased from pre-EHR compared to 4 months (P ¼ 0.02) and 8 months post-EHR, (P ¼ 0.002). Although there was a numerical decrease in the hospital LOS at 4 and 8 months post-EHR; this was not statistically significant, (P ¼ 0.08) (Table 3). Even though overall the rate of medication errors increased after EHR implementation, 2,595 errors per 1,000 patient days compared with 1,972 errors per 1,000 patient days, (P ¼ 0.002) (Table 4), the severity of errors was reduced. There was a reduction in the incidence of more severe medication errors (category D-I), whereas there was an increase in the incidence of less severe (category C) errors.15 In terms of the severity of the errors, as defined by the NCC MERP taxonomy, the increase in administration errors and category C errors was primarily driven by an increase in the number of drugs that were administered in a delayed fashion (41 hour from scheduled time). Regarding the origin of the errors, there were significant reductions in dispensing related errors (pharmacy-related errors), 26.2% compared with 0.8%, P o 0.001, but an increase in administration-related errors (nursing administration), 29.8% compared with 60.4%, P o 0.001, which again is mostly related to delayed drug administration (Table 5).
DISCUSSION This study reports an association between severityadjusted reductions in mortality with the implementation of an EHR with CPOE in a critical care setting. In support of these benefits, the reductions in mortality were
accompanied by reductions in ICU LOS. In addition, this study documents the systematic changes in medication errors with EHR implementation, noting an overall increase in medication errors but, importantly, significant reductions in more severe errors. It is noteworthy that the survival benefit was apparent despite an increasing severity of illness during the observed postimplementation period, suggesting an even more positive effect of EHR implementation. There are a variety of mechanisms by which EHR implementation may have salutary effects on critically ill patients. Among the most commonly considered is the accessibility of medical information for patient care with the ability to track, monitor and manage patients in a multiprofessional manner regardless of healthcare provider location. There is a reduction in time to executing physician orders because there is not a time delay for a unit clerk to transcribe the order and input into the computer. Patient care can be initiated quickly because nursing orders or respiratory therapy orders can be seen in real time. In addition, the collaborative care of acutely ill patients by multiple providers is more efficient when each provider can simultaneously access the entire medical record from any location. Another important aspect to consider when implementing EHR is focusing on ergonomics to improve workflow in the ICU. Previous studies have analyzed the effect of ergonomics and having enough computers for healthcare providers and having mobile computers to increase the ease of bedside patient care delivery.16 In this hospital, computers were installed in every patient room with increased access to mobile computer stations assigned to various healthcare providers. Therefore, increased accessibility to patient medical records by several users at a time may have helped to increase the workflow and efficiency of care. Although a contradictory finding, the increase in medication errors found early in the rollout period could be related to challenges related to a lack of familiarity with a new system and also the increased detectability because of electronic medical record. Nonetheless, the reduction in severe medication errors may have been a contributing factor for improvement in patient outcomes.
TABLE 3. MICU and hospital length of stay in days.
Mean MICU LOS (days ⫾ SE) Mean hospital LOS (days ⫾ SE)
Pre
4 Months
8 Months
Pre vs. 4 months (NovemberFebruary) P Value
4.03 ⫾ 1.06 9.29 ⫾ 1.06
3.26 ⫾ 1.06 8.00 ⫾ 1.06
3.12 ⫾ 1.05 7.77 ⫾ 1.06
0.02* 0.19*
Pre vs. 8 months (MarchJune) P Value
4 vs. 8 Months P Value
0.002* 0.08*
0.83* 0.93*
Adjusted for continuous APACHE II, α o 0.05, significant P values are bold. LOS, length of stay; MICU, medical intensive care unit; SE, standard error. * Tukey P value
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
579
Han et al
TABLE 4. Medication error rate per 1,000 days. Time August 2010 January 2011
Total ICU days
Total error count
Error count/1,000 patient days
Errors per day
P Value
887 1,103
1,749 2,862
1,972 2,595
1.97 2.60
0.002*
ICU, intensive care unit. * α o 0.05, significant P values are marked with asterisks (*).
During the time frame of this study, the use of an EHR was still fairly new but was being widely implemented. The findings of this study are likely because of the improvement in the implementation process. This study was conducted after several previous studies revealed issues with EHR implementation. Our favorable results are likely because of the awareness of implementation process challenges that were addressed proactively during the system change. As EHR implementation has become more ubiquitous, the process of application has likely improved; for instance, establishing various training sessions for each type of provider. All providers at the institution underwent multiple hours of classroom and online training, and training was tailored to each provider type, which may have increased the effectiveness of the training sessions. The major difference from this implementation from earlier studies was the technical support by specific healthcare providers as “super-users.” There were specialty trained nurses, pharmacists and physicians, who were readily available to assist each specific user in their needs in real time, instead of an information technician who may not be familiar with the workflow needs. The availability of super-users who could help as needed during active clinical care may have helped to decrease provider errors and to mitigate any delays in patient care associated with a new care system, while also increasing acceptance of the new EHR system. As various types of EHR are being implemented, this is a key feature that is most generalizable that will likely provide the most benefit. There are several limitations to this study. Despite being a prospective, single-institutional pre- and postimplementation study design, comparisons were made across time and may have included temporal associations that were not anticipated or that could not be
accounted for. Our results could have been influenced by instrumentation bias because pre-EHR data collection was manual, whereas postimplementation data collection was electronic. This may have influenced the accounting of medication errors, both in number and in severity. Likewise, medication turnaround time is much quicker with electronic order entry compared with traditional paper order entry where the clerk needs to input the orders after the physician has written them.17 However, EHR implementation ensured a consistent and accurate recording and tracking of all medication orders and administration episodes. An additional limitation to the study that could have affected the outcome in the preimplementation group is the house staff, who started in July which coincided with the start of the EHR system. The residents and fellows may be a confounding variable influencing the average MICU LOS and mortality due to lack of experience or efficiency. However, our residency program assigns their strongest residents during the early months to supervise new interns to mitigate the effect of inefficiency or inexperience. The overall care is always supervised by an attending physician; therefore, the house staff effect on both medication errors and outcomes is unclear. Nonetheless, we were not able to measure or control this effect on patient care outcomes.
CONCLUSIONS Our study indicates that there is an association between improved ICU survival and implementation of an EHR, which has not been previously reported. As EHR implementation continues to expand in multiple healthcare settings, this study documents the clinically relevant changes in ICU care following EHR implementation and supports the hypothesis that EHR may improve outcomes in critically ill patients.
TABLE 5. Classification of medication error.
Origin of error Prescribing errors Dispensing errors Administration errors * * *
580
August 2010 N ¼ 1,749 (%)
January 2011 N ¼ 2,862 (%)
P Value
769 (44) 459 (26.2) 521 (29.8)
1,119 (39.1) 24 (0.8) 1,719 (60.4)
o0.001* o0.001* o0.001*
Definition of errors attached as Appendix 1. α o 0.05, significant P values are marked with asterisks (*). See Appendix 1 for the list of severity of medication errors.
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
ACKNOWLEDGMENTS The authors would like to thank the National Institutes of Health and the Atlanta Clinical and Translational Science Institute for their support in this study.
Appendix 1. Severity of Medication Errors Severity of error
Category Category Category Category Category Category Category Category Category
August 2010, N ¼ 1,749 (%)
A B C D E F G H I
4 529 1,015 185 16 0 0 0 0
(0.2) (30.2) (58) (10.6) (0.9) (0) (0) (0) (0)
January 2011, N ¼ 2,862 (%)
29 348 2,459 23 3 0 0 0 0
(1) (12.2) (85.9) (0.8) (0.1) (0) (0) (0) (0)
P Value o0.001 0.002 o0.001 o0.001 o0.001 o0.001 1.00 1.00 1.00 1.00
*α o 0.05, significant P values are marked with asterisks (*)
REFERENCES 1. The Leapfrog Group Fact Sheet. http://www.leapfroggroup.org/about_ leapfrog/leapfrog-factsheet; Accessed April 23, 2014. 2. Ledue C. Healthcare Finance News Physicians to receive incentives for EHR use. http://www.healthcarefinancenews.com/news/physicians-receiveincentives-ehr-use; Accessed April 23, 2014. 3. Thompson G, O’Horo JC, Pickering BW, et al. Impact of the Electronic Medical Record on mortality, length of stay, and cost in the hospital and ICU: a systematic review and metaanalysis. Crit Care Med 2015;43(6): 1276–82. 4. Eslami S, de Keizer NF, Abu-Hanna A. The impact of computerized physician medication order entry in hospitalized patients—a systematic review. Int J Med Inform 2008;77(6):365–76. 5. Ghazisaeedi M, Mohammadzadeh N, Safdari R. Electronic Health Record (EHR) as a vehicle for successful health care best practice. Med Arch 2014;68(6):419–21. 6. Clauser SB, Wagner EH, Aiello Bowles EJ, et al. Improving modern cancer care through information technology. Am J Prev Med 2011;40(5 suppl 2):S198–207. 7. Schenarts PJ, Goettler CE, White MA, et al. An objective study of the impact of the electronic medical record on outcomes in trauma patients. Am Surg 2012;78(11):1249–54. 8. Camire E, Moyen E, Stelfox HT. Medication errors in critical care: risk factors, prevention and disclosure. CMAJ 2009;180(9):936–43.
9. Ammenwerth E, Schnell-Inderst P, Machan C, et al. The effect of electronic prescribing on medication errors and adverse drug events: a systematic review. J Am Med Inform Assoc 2008;15(5):585–600. 10. Weant KA, Cook AM, Armitstead JA. Medication-error reporting and pharmacy resident experience during implementation of computerized prescriber order entry. Am J Health Syst Pharm 2007;64(5): 526–30. 11. Shulman R, Singer M, Goldstone J, et al. Medication errors: a prospective cohort study of hand-written and computerised physician order entry in the intensive care unit. Crit Care 2005;9(5): R516–21. 12. Han YY, Carcillo JA, Venkataraman ST, et al. Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system. Pediatrics 2005;116(6):1506–12. 13. Keene A, Ashton L, Shure D, et al. Mortality before and after initiation of a computerized physician order entry system in a critically ill pediatric population. Pediatr Crit Care Med 2007;8(3):268–71. 14. Baeke F, Takiishi T, Korf H, et al. Vitamin D: modulator of the immune system. Curr Opin Pharmacol 2010;10(4):482–96. 15. Holick MF. Vitamin D deficiency. N Engl J Med 2007;357(3):266–81. 16. Maslove DM, Rizk N, Lowe HJ. Computerized physician order entry in the critical care environment: a review of current literature. J Intensive Care Med 2011;26(3):165–71. 17. Weir CR, Staggers N, Phansalkar S. The state of the evidence for computerized provider order entry: a systematic review and analysis of the quality of the literature. Int J Med Inform 2009;78(6): 365–74.
From the Division of Pulmonary, Allergy and Critical Care Medicine (JEH, EGH, GSM), Department of Medicine, Emory University School of Medicine, Atlanta, Georgia; Department of Pharmacy and Drug Information (MR, PA), Grady Memorial Hospital, Atlanta, Georgia; Department of Medicine (PS), Tallahassee Memorial Healthcare, Florida State University, Tallahassee, Florida; Mercer Health Science Center (TVL), Mercer University College of Pharmacy, Atlanta, Georgia; Division of Pulmonary (TNU), WellStar Health System, Atlanta, Georgia; Rollins School of Public Health (GAC), Emory University, Atlanta, Georgia. Submitted October 6, 2015; accepted January 22, 2016. This work is supported in part by National Institutes of Health, United States, Grant Numbers T32 AA013528 and UL1 TR000454 to JEH and UL1 TR000455 to GSM. The authors have no conflicts of interest to disclose. Prior presentations: American Journal of Respiratory and Critical Care Medicine, 2012;185:A4009; Published abstracts: Critical Care Medicine 2011;39(S12). Correspondence: Jenny E. Han, MD, MSc, Division of Pulmonary, Allergy and Critical Care Medicine, Department of Medicine, Emory University School of Medicine, 49 Jesse Hill Jr. Drive SE, Atlanta, GA 30303 (E-mail: [email protected]).
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
581
Effect of Electronic Health Record Implementation in Critical Care on Survival and Medication Errors Jenny E. Han, MD, MSc, Marina Rabinovich, PharmD, BCPS, Prasad Abraham, PharmD, BCPS, Prerna Satyanarayana, MD, T. Vivan Liao, PharmD, BCPS, Timothy N. Udoji, MD, George A. Cotsonis, MA, Eric G. Honig, MD and Greg S. Martin, MD, MSc ABSTRACT Background: Electronic health records (EHR) with computerized physician order entry have become exceedingly common and government incentives have urged implementation. The purpose of this study was to ascertain the effect of EHR implementation on medical intensive care unit (MICU) mortality, length of stay (LOS), hospital LOS and medication errors. Materials and Methods: Prospective, observational study from July 2010-June 2011 in MICU at an urban teaching hospital in Atlanta, Georgia of 797 patients admitted to the MICU; 281 patients before the EHR implementation and 516 patients post-EHR implementation. Results: Compared with the preimplementation period (N ¼ 43 per 281), the mortality risk at 4 months post-EHR implementation (N ¼ 41 per 247) and at 8 months post-EHR implementation (N ¼ 26 per 269) significantly decreased (P o 0.001). In addition, the mean MICU LOS statistically decreased from 4.03 ⫾ 1.06 days pre-EHR to 3.26 ⫾ 1.06 days 4 months post-EHR and to 3.12 ⫾ 1.05 days 8 months post-EHR (P ¼ 0.002). However, the mean hospital LOS was not statistically decreased. Although medication errors increased after implementation (P ¼ 0.002), this was attributable to less severe errors and there was actually a decrease in the number of severe medication errors (both P o 0.001). Conclusions: We report a survival benefit following the implementation of EHR with computerized physician order entry in a critical care setting and a concomitant decrease in the number of severe medication errors. Although overall hospital LOS was not shortened, this study proposes that EHR implementation in a busy urban hospital was associated with improved ICU outcomes. Key Indexing Terms: Electronic health records; Computerized physician order entry; Quality improvement; Medication error. [Am J Med Sci 2016;351(6):576–581.]
INTRODUCTION
E
lectronic health record (EHR) has been endorsed by national healthcare organizations such as the Institute of Medicine and the Leapfrog Group to improve patient care, especially computerized physician order entry (CPOE).1 The American Recovery and Reinvestment Act of 2009 allowed for $2 billion in discretionary health information technology funding and $18 billion in investments and incentives through Medicare and Medicaid to adopt and to use EHR technology.2 The Centers for Medicare and Medicaid Services has initiated financial incentives with EHR implementation with the goal of improving healthcare efficiency and quality.1 However, the implementation of this technology is extremely costly for hospital institutions. Furthermore, the data are unclear if this new technology has any direct patient care benefit. A recent systematic review specifically looking at intensive care unit (ICU) mortality and length of stay (LOS) with EHR did not show any substantial effects because of the small number of studies and heterogeneity of the patient population.3 Another systematic
576
review article looking at 67 studies involving CPOE found that there were positive effects in adherence to guidelines, satisfaction and usability, but the various studies did not find a difference in mortality.4-6 The EHR studies often look at the effect of CPOE on patient safety outcomes. Schenarts et al7 reported that in patients with traumatic injury, EHR reduced the hospital LOS and ICU LOS, but did not change mortality. The focus on patient safety has expanded the role of technology to minimize the risk of medical errors and adverse events. It has been reported that 1.7 medical errors occur a day in the ICU and that medication errors account for 78% of these errors.8 A systematic review of CPOE implementation in various healthcare settings found a relative risk reduction ranging from 13-99% for medication errors and from 30-84% for adverse drug events.9 However, Weant et al10 reported that medication errors increased to more than 4 times higher in the first month after CPOE implementation in neurosurgical ICU, but that the number of errors causing patient harm decreased. The use of CPOE in a United Kingdom ICU demonstrated significant reduction in overall number of
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
errors attributed to prescribing errors.11 Although research shows that administration errors are the second most common cause of medication errors, to date, administration and dispensing errors with CPOE implementation have not been evaluated. Various studies have shown variable results regarding EHR implementation and EHR with CPOE. Different mortality results were published by 2 commonly cited studies done in the pediatric ICU. The University of Pittsburgh study showed an increase in mortality with EHR implementation, and a similar study done at Montefiore Medical Center found no mortality difference with EHR implementation in the pediatric ICU.12,13 These studies are more than 8 years old and original research in adult ICU has not been performed recently. Therefore, this study evaluates the effect of EHR with CPOE implementation on mortality, medical ICU (MICU) LOS, hospital LOS and medication errors in critically ill adults.
MATERIALS AND METHODS Design This prospective, observational study evaluated the effect of EHR with CPOE on patient care outcomes from July 2010-June 2011. The study was reviewed by Emory University Institutional Review Board as well as the hospital Research Oversight Committee, and was exempted as a quality improvement study. Setting The study was conducted in the county MICU in Atlanta, GA, (Grady Memorial Hospital). Before implementation, medical records were used for each patient where notes and orders were manually prescribed and then a clerk would enter the orders into the computer. The implementation software was EPIC Systems Corporation with computer order entry, which included all patient medical information and which consisted of all healthcare provider notes including medications, flow sheets, orders, diagnostic and laboratory results. Implementation Hospital-wide training sessions were required for every healthcare provider before the initiation of the EHR with CPOE system. Inpatient physicians were required to complete 4 hours of training, and nurses 3 days of training. For 3 weeks during the implementation process, “super-users” were available in all clinical areas at all times. Super-users were healthcare providers who underwent additional training to assist colleagues when needed to deliver effective patient care.
outcome data would not be included more than once. The EHR with CPOE implementation commenced on October 31, 2010. Design Patient outcomes were compared from 4 months before the EHR implementation (July, August, September and October), 4 months after implementation (November, December, January and February) and 8 months after implementation (March, April, May and June). The 2 time frames were selected to investigate if there was a possibility of increased mortality during the implementation process, whereas the new system was being used that one would expect to improve overtime after familiarity with the system. Additionally, medication errors were evaluated in this patient population by a clinical pharmacist (M.R. and V.L.) for 1 month before EHR with CPOE implementation (August 2010) and for 1 month after implementation (January 2011). Demographic data as well as APACHE II scores were obtained from all admitted patients. Outcomes were mortality, ICU LOS and hospital LOS. Secondary outcomes were frequency of medication errors before and after CPOE implementation. Medication errors were defined as the number of medication errors per 1,000 patient days, and as percentage change in the type of medication errors and percentage change in the severity of medication errors classified as A (capacity for error) through I (error reached patient and contributed to death), based on categories defined by the National Coordinating Council for Medication Error Reporting and Prevention Taxonomy.14 Medication errors were classified in categories of prescribing, dispensing and administration errors. Statistical Analysis Chi-square analysis, Fisher’s exact test, T tests, logistic regression, one-way analysis of variance and one-way analysis of covariance were performed. Analysis of covariance was used to compare the 3 periods after adjusting for the APACHE II score. Tukey’s post hoc pairwise comparisons were performed using these adjusted means. Before the analysis, ICU LOS and hospital LOS were log-transformed, due to skewed distribution of LOS in days. The results from the analysis of variance were back-transformed to the original units. Two-tailed P o 0.05 was considered statistically significant. All analyses were done with SAS/STAT (9.3) software (SAS Institute, Cary, NC).
RESULTS Conduct All patients admitted to the MICU team between July 1, 2010 and June 30, 2011 were included in the study. Patients were excluded if they were readmitted to the MICU service after being transferred out, so their
A total of 797 critically ill medical patients admitted from July 2010-June 2011 were included in the study. There were no significant differences across all groups in race, sex, age and disease category (Table 1). As a marker of illness severity, the APACHE II score increased
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
577
Han et al
TABLE 1. Baseline patient demographics.
Race (N, %) Black Hispanic White Other Sex (female, N (%)) Age (years) Disease (N, %) CNS CV Pulmonary GI Renal Sepsis Other APACHE II (mean, SD) ICU mortality
Pre-EHR
Post-EHR 4 months (November-February)
Post-EHR 8 months (March-June)
N ¼ 281 (%)
N ¼ 247 (%)
N ¼ 269 (%)
P Value
232 (83) 8 (3) 36 (13) 2 (1) 119 (42) 55 ⫾ 16
210 (86) 12 (5) 16 (7) 7 (3) 103 (42) 53 ⫾ 16
216 (81) 9 (3) 33 (12) 9 (3) 131 (49) 54 ⫾ 16
0.05
30 (11) 64 (24) 48 (18) 40 (15) 4 (2) 44 (17) 32 (12) 18.6 (⫾8.51) N ¼ 238 43 (15)
16 (7) 49 (20) 66 (28) 24 (10) 4 (2) 49 (20) 31 (13) 23.29 (⫾10.54) N ¼ 231 41 (16)
28 (11) 55 (21) 61 (23) 24 (9) 2 (1) 45 (17) 45 (17) 21.13 (⫾9.74) N ¼ 248 26 (9)
0.20 0.50 0.10
o0.001* 0.04
CNS, central nervous system; CV, cardiovascular; HER, electronic health record; GI, gastrointestinal; ICU, intensive care unit; SD, standard deviation. * APACHE II post 4 months EHR 4post 8 months EHR 4pre-EHR, P o 0.05 Tukey.
from the preimplementation (mean ¼ 18.6 ⫾ 8.51) to the immediate postimplementation period 4 months (23.29 ⫾ 10.54), and slightly decreased thereafter at 8 months (21.13 ⫾ 9.74), (P o 0.0001, Table 1). The median APACHE II score was 21 (Figure).
Compared with preimplementation levels, the mortality risk at 4 months post-EHR implementation 8 months post-EHR implementation period markedly decreased, (95% CI: 1.63-5.67) and (95% CI: 1.163.82) respectively, P o 0.001 (Table 2). The average
Distribution of APACHE II 60
Apache II Score
50
40
30
20
10
0 Pre
4 Months
8 Months
EHR Implementation FIGURE. Apache II score for medical ICU patients pre-EHR, 4 months post-EHR and 8 months post-EHR implementation.
578
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
TABLE 2. Mortality comparing study periods adjusted for APACHE II. Mortality Pre vs. 8 months Pre vs. 4 months 4 vs. 8 months * *
Odds ratio
95% CI
P Value
3.04 2.10 1.45
(1.63-5.67) (1.16-3.82) (0.78-2.67)
o0.001* o0.001* 0.48
Adjusted for APACHE II. α o 0.05, significant P values are marked with asterisks (*).
MICU LOS adjusted by continuous APACHE II score, statistically decreased from pre-EHR compared to 4 months (P ¼ 0.02) and 8 months post-EHR, (P ¼ 0.002). Although there was a numerical decrease in the hospital LOS at 4 and 8 months post-EHR; this was not statistically significant, (P ¼ 0.08) (Table 3). Even though overall the rate of medication errors increased after EHR implementation, 2,595 errors per 1,000 patient days compared with 1,972 errors per 1,000 patient days, (P ¼ 0.002) (Table 4), the severity of errors was reduced. There was a reduction in the incidence of more severe medication errors (category D-I), whereas there was an increase in the incidence of less severe (category C) errors.15 In terms of the severity of the errors, as defined by the NCC MERP taxonomy, the increase in administration errors and category C errors was primarily driven by an increase in the number of drugs that were administered in a delayed fashion (41 hour from scheduled time). Regarding the origin of the errors, there were significant reductions in dispensing related errors (pharmacy-related errors), 26.2% compared with 0.8%, P o 0.001, but an increase in administration-related errors (nursing administration), 29.8% compared with 60.4%, P o 0.001, which again is mostly related to delayed drug administration (Table 5).
DISCUSSION This study reports an association between severityadjusted reductions in mortality with the implementation of an EHR with CPOE in a critical care setting. In support of these benefits, the reductions in mortality were
accompanied by reductions in ICU LOS. In addition, this study documents the systematic changes in medication errors with EHR implementation, noting an overall increase in medication errors but, importantly, significant reductions in more severe errors. It is noteworthy that the survival benefit was apparent despite an increasing severity of illness during the observed postimplementation period, suggesting an even more positive effect of EHR implementation. There are a variety of mechanisms by which EHR implementation may have salutary effects on critically ill patients. Among the most commonly considered is the accessibility of medical information for patient care with the ability to track, monitor and manage patients in a multiprofessional manner regardless of healthcare provider location. There is a reduction in time to executing physician orders because there is not a time delay for a unit clerk to transcribe the order and input into the computer. Patient care can be initiated quickly because nursing orders or respiratory therapy orders can be seen in real time. In addition, the collaborative care of acutely ill patients by multiple providers is more efficient when each provider can simultaneously access the entire medical record from any location. Another important aspect to consider when implementing EHR is focusing on ergonomics to improve workflow in the ICU. Previous studies have analyzed the effect of ergonomics and having enough computers for healthcare providers and having mobile computers to increase the ease of bedside patient care delivery.16 In this hospital, computers were installed in every patient room with increased access to mobile computer stations assigned to various healthcare providers. Therefore, increased accessibility to patient medical records by several users at a time may have helped to increase the workflow and efficiency of care. Although a contradictory finding, the increase in medication errors found early in the rollout period could be related to challenges related to a lack of familiarity with a new system and also the increased detectability because of electronic medical record. Nonetheless, the reduction in severe medication errors may have been a contributing factor for improvement in patient outcomes.
TABLE 3. MICU and hospital length of stay in days.
Mean MICU LOS (days ⫾ SE) Mean hospital LOS (days ⫾ SE)
Pre
4 Months
8 Months
Pre vs. 4 months (NovemberFebruary) P Value
4.03 ⫾ 1.06 9.29 ⫾ 1.06
3.26 ⫾ 1.06 8.00 ⫾ 1.06
3.12 ⫾ 1.05 7.77 ⫾ 1.06
0.02* 0.19*
Pre vs. 8 months (MarchJune) P Value
4 vs. 8 Months P Value
0.002* 0.08*
0.83* 0.93*
Adjusted for continuous APACHE II, α o 0.05, significant P values are bold. LOS, length of stay; MICU, medical intensive care unit; SE, standard error. * Tukey P value
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
579
Han et al
TABLE 4. Medication error rate per 1,000 days. Time August 2010 January 2011
Total ICU days
Total error count
Error count/1,000 patient days
Errors per day
P Value
887 1,103
1,749 2,862
1,972 2,595
1.97 2.60
0.002*
ICU, intensive care unit. * α o 0.05, significant P values are marked with asterisks (*).
During the time frame of this study, the use of an EHR was still fairly new but was being widely implemented. The findings of this study are likely because of the improvement in the implementation process. This study was conducted after several previous studies revealed issues with EHR implementation. Our favorable results are likely because of the awareness of implementation process challenges that were addressed proactively during the system change. As EHR implementation has become more ubiquitous, the process of application has likely improved; for instance, establishing various training sessions for each type of provider. All providers at the institution underwent multiple hours of classroom and online training, and training was tailored to each provider type, which may have increased the effectiveness of the training sessions. The major difference from this implementation from earlier studies was the technical support by specific healthcare providers as “super-users.” There were specialty trained nurses, pharmacists and physicians, who were readily available to assist each specific user in their needs in real time, instead of an information technician who may not be familiar with the workflow needs. The availability of super-users who could help as needed during active clinical care may have helped to decrease provider errors and to mitigate any delays in patient care associated with a new care system, while also increasing acceptance of the new EHR system. As various types of EHR are being implemented, this is a key feature that is most generalizable that will likely provide the most benefit. There are several limitations to this study. Despite being a prospective, single-institutional pre- and postimplementation study design, comparisons were made across time and may have included temporal associations that were not anticipated or that could not be
accounted for. Our results could have been influenced by instrumentation bias because pre-EHR data collection was manual, whereas postimplementation data collection was electronic. This may have influenced the accounting of medication errors, both in number and in severity. Likewise, medication turnaround time is much quicker with electronic order entry compared with traditional paper order entry where the clerk needs to input the orders after the physician has written them.17 However, EHR implementation ensured a consistent and accurate recording and tracking of all medication orders and administration episodes. An additional limitation to the study that could have affected the outcome in the preimplementation group is the house staff, who started in July which coincided with the start of the EHR system. The residents and fellows may be a confounding variable influencing the average MICU LOS and mortality due to lack of experience or efficiency. However, our residency program assigns their strongest residents during the early months to supervise new interns to mitigate the effect of inefficiency or inexperience. The overall care is always supervised by an attending physician; therefore, the house staff effect on both medication errors and outcomes is unclear. Nonetheless, we were not able to measure or control this effect on patient care outcomes.
CONCLUSIONS Our study indicates that there is an association between improved ICU survival and implementation of an EHR, which has not been previously reported. As EHR implementation continues to expand in multiple healthcare settings, this study documents the clinically relevant changes in ICU care following EHR implementation and supports the hypothesis that EHR may improve outcomes in critically ill patients.
TABLE 5. Classification of medication error.
Origin of error Prescribing errors Dispensing errors Administration errors * * *
580
August 2010 N ¼ 1,749 (%)
January 2011 N ¼ 2,862 (%)
P Value
769 (44) 459 (26.2) 521 (29.8)
1,119 (39.1) 24 (0.8) 1,719 (60.4)
o0.001* o0.001* o0.001*
Definition of errors attached as Appendix 1. α o 0.05, significant P values are marked with asterisks (*). See Appendix 1 for the list of severity of medication errors.
THE AMERICAN JOURNAL
OF THE
VOLUME 351
MEDICAL SCIENCES
NUMBER 6
June 2016
EHR Implementation in Medical ICU
ACKNOWLEDGMENTS The authors would like to thank the National Institutes of Health and the Atlanta Clinical and Translational Science Institute for their support in this study.
Appendix 1. Severity of Medication Errors Severity of error
Category Category Category Category Category Category Category Category Category
August 2010, N ¼ 1,749 (%)
A B C D E F G H I
4 529 1,015 185 16 0 0 0 0
(0.2) (30.2) (58) (10.6) (0.9) (0) (0) (0) (0)
January 2011, N ¼ 2,862 (%)
29 348 2,459 23 3 0 0 0 0
(1) (12.2) (85.9) (0.8) (0.1) (0) (0) (0) (0)
P Value o0.001 0.002 o0.001 o0.001 o0.001 o0.001 1.00 1.00 1.00 1.00
*α o 0.05, significant P values are marked with asterisks (*)
REFERENCES 1. The Leapfrog Group Fact Sheet. http://www.leapfroggroup.org/about_ leapfrog/leapfrog-factsheet; Accessed April 23, 2014. 2. Ledue C. Healthcare Finance News Physicians to receive incentives for EHR use. http://www.healthcarefinancenews.com/news/physicians-receiveincentives-ehr-use; Accessed April 23, 2014. 3. Thompson G, O’Horo JC, Pickering BW, et al. Impact of the Electronic Medical Record on mortality, length of stay, and cost in the hospital and ICU: a systematic review and metaanalysis. Crit Care Med 2015;43(6): 1276–82. 4. Eslami S, de Keizer NF, Abu-Hanna A. The impact of computerized physician medication order entry in hospitalized patients—a systematic review. Int J Med Inform 2008;77(6):365–76. 5. Ghazisaeedi M, Mohammadzadeh N, Safdari R. Electronic Health Record (EHR) as a vehicle for successful health care best practice. Med Arch 2014;68(6):419–21. 6. Clauser SB, Wagner EH, Aiello Bowles EJ, et al. Improving modern cancer care through information technology. Am J Prev Med 2011;40(5 suppl 2):S198–207. 7. Schenarts PJ, Goettler CE, White MA, et al. An objective study of the impact of the electronic medical record on outcomes in trauma patients. Am Surg 2012;78(11):1249–54. 8. Camire E, Moyen E, Stelfox HT. Medication errors in critical care: risk factors, prevention and disclosure. CMAJ 2009;180(9):936–43.
9. Ammenwerth E, Schnell-Inderst P, Machan C, et al. The effect of electronic prescribing on medication errors and adverse drug events: a systematic review. J Am Med Inform Assoc 2008;15(5):585–600. 10. Weant KA, Cook AM, Armitstead JA. Medication-error reporting and pharmacy resident experience during implementation of computerized prescriber order entry. Am J Health Syst Pharm 2007;64(5): 526–30. 11. Shulman R, Singer M, Goldstone J, et al. Medication errors: a prospective cohort study of hand-written and computerised physician order entry in the intensive care unit. Crit Care 2005;9(5): R516–21. 12. Han YY, Carcillo JA, Venkataraman ST, et al. Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system. Pediatrics 2005;116(6):1506–12. 13. Keene A, Ashton L, Shure D, et al. Mortality before and after initiation of a computerized physician order entry system in a critically ill pediatric population. Pediatr Crit Care Med 2007;8(3):268–71. 14. Baeke F, Takiishi T, Korf H, et al. Vitamin D: modulator of the immune system. Curr Opin Pharmacol 2010;10(4):482–96. 15. Holick MF. Vitamin D deficiency. N Engl J Med 2007;357(3):266–81. 16. Maslove DM, Rizk N, Lowe HJ. Computerized physician order entry in the critical care environment: a review of current literature. J Intensive Care Med 2011;26(3):165–71. 17. Weir CR, Staggers N, Phansalkar S. The state of the evidence for computerized provider order entry: a systematic review and analysis of the quality of the literature. Int J Med Inform 2009;78(6): 365–74.
From the Division of Pulmonary, Allergy and Critical Care Medicine (JEH, EGH, GSM), Department of Medicine, Emory University School of Medicine, Atlanta, Georgia; Department of Pharmacy and Drug Information (MR, PA), Grady Memorial Hospital, Atlanta, Georgia; Department of Medicine (PS), Tallahassee Memorial Healthcare, Florida State University, Tallahassee, Florida; Mercer Health Science Center (TVL), Mercer University College of Pharmacy, Atlanta, Georgia; Division of Pulmonary (TNU), WellStar Health System, Atlanta, Georgia; Rollins School of Public Health (GAC), Emory University, Atlanta, Georgia. Submitted October 6, 2015; accepted January 22, 2016. This work is supported in part by National Institutes of Health, United States, Grant Numbers T32 AA013528 and UL1 TR000454 to JEH and UL1 TR000455 to GSM. The authors have no conflicts of interest to disclose. Prior presentations: American Journal of Respiratory and Critical Care Medicine, 2012;185:A4009; Published abstracts: Critical Care Medicine 2011;39(S12). Correspondence: Jenny E. Han, MD, MSc, Division of Pulmonary, Allergy and Critical Care Medicine, Department of Medicine, Emory University School of Medicine, 49 Jesse Hill Jr. Drive SE, Atlanta, GA 30303 (E-mail: [email protected]).
Copyright © 2016 Southern Society for Clinical Investigation. Published by Elsevier Inc. All rights reserved. www.amjmedsci.com www.ssciweb.org
581