Prospective, controlled study of an intervention to reduce ... - Nature

Journal of Perinatology (2015) 35, 631–635 © 2015 Nature America, Inc. All rights reserved 0743-8346/15 www.nature.com/jp

ORIGINAL ARTICLE

Prospective, controlled study of an intervention to reduce errors in neonatal antibiotic orders SS Garner1,2,3,4, TH Cox1,3, EG Hill5, MG Irving6, RL Bissinger7 and DJ Annibale4 OBJECTIVE: To evaluate the effectiveness of an interactive computerized order set with decision support (ICOS-DS) in preventing medication errors in neonatal late-onset sepsis (LOS). STUDY DESIGN: Prospective, controlled comparison of error rates in antibiotic orders for neonates admitted to the Medical University of South Carolina neonatal intensive care unit with suspected LOS (after postnatal day of life 3) prior to (n = 153) and after (n = 146) implementation of the ICOS-DS. Antibiotic orders were independently evaluated by two pharmacists for prescribing errors, potential errors and omissions. Prescribing errors included410% overdoses or underdoses, inappropriate route, schedule or antibiotic, drug–drug or drug–disease interactions, and incorrect patient demographics. Potential errors included misspelled drugs, leading decimals, trailing zeroes, impractical doses and error-prone abbreviations. Multiple errors and omissions in an order were counted individually. RESULTS: Overall error rate per order decreased from 1.7 to 0.8 (P o0.001) and potential error rate from 1.0 to 0.06 (P o 0.001). The reduction in omission error rate per order from 0.2 to 0.1 was not significant (P = 0.17). The prescribing error rate per order increased from 0.4 to 0.7 (P = 0.03) because of the use of incorrect patient weights (P o 0.001). Renal dysfunction was significantly associated with an increased risk of prescribing errors (odds ratio = 3.7, P = 0.01) which was not significantly different for handwritten versus ICOS-DS orders (P = 0.15). CONCLUSIONS: The ICOS-DS significantly improved the quality of neonatal LOS antibiotic orders although the use of incorrect patient weights was increased. In both groups, orders for patients with renal dysfunction were at risk for prescribing errors. Further evaluation of interventions to promote medication safety for this population is needed. Journal of Perinatology (2015) 35, 631–635; doi:10.1038/jp.2015.20; published online 2 April 2015

INTRODUCTION The National Coordinating Council for Medication Error Reporting and Prevention defines a medication error as ‘any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the health care professional, patient, or consumer.’1 Since the landmark study describing medication errors in pediatric inpatients,2 numerous studies and meta-analyses have confirmed that children remain at risk for drug misadventures and adverse events.3–6 Explanations include limitations in available literature, variance in dosing recommendations among published references, and weight-based dosing requiring frequent calculations. Neonates specifically are vulnerable to medication errors2,7 owing to the inability to buffer errors and heterogeneity among patients’ gestational age, postnatal age and weight. Antibiotics are commonly involved in pediatric medication errors.2,4,8,9 At the Medical University of South Carolina (MUSC), neonatal intensive care unit (NICU) pharmacists document interventions in internal databases as part of routine patient care. During database review by pharmacy and hospital administration, a disproportionate number of intercepted errors involving antibiotics led to a formalized review which detailed errors such as inappropriate empiric antibiotic selection, incorrect doses, dosing intervals and

dose calculations. Antibiotics were also inconsistently ordered stat causing delays in first doses.10 On the basis of these findings, an inter-professional NICU team developed an interactive computerized order set with decision support (ICOS-DS) to guide the antibiotic-ordering process. The objective of this study was to evaluate the effect of the ICOS-DS on errors in antibiotic orders for late-onset sepsis (LOS) in patients in the NICU. METHODS Study setting The MUSC NICU is a Level 3, 38-bed unit with 908 admissions annually. It is staffed by MUSC’s neonatologists, fellows, residents and neonatal nurse practitioners from the Division of Neonatology, Department of Pediatrics. Around-the-clock pharmacy services include an on-call service manned by residency-trained pediatric pharmacists. At baseline, diagnostic evaluations and empiric antibiotic selection for LOS were guided by evidence-based guidelines developed by an inter-professional subcommittee of the NICU’s Evidence and Value Committee. Antibiotic orders were handwritten and manually faxed to pharmacy by nurses.

Intervention An inter-professional group was convened to expand the existing evidence-based guidelines to address identified intercepted errors and

1 Department of Clinical Pharmacy and Outcome Sciences, South Carolina College of Pharmacy, Medical University of South Carolina, Charleston, SC, USA; 2Center for Medication Safety, South Carolina College of Pharmacy, Medical University of South Carolina, Charleston, SC, USA; 3Department of Pharmacy, Medical University of South Carolina, Charleston, SC, USA; 4Department of Pediatrics, Medical University of South Carolina, Charleston, SC, USA; 5Department of Public Health Sciences, Medical University of South Carolina, Charleston, SC, USA; 6Office of the Chief Medical Information Office, Medical University of South Carolina, Charleston, SC, USA and 7Department of Nursing, Medical University of South Carolina, Charleston, SC, USA. Correspondence: Dr SS Garner, Associate Professor, Department of Clinical Pharmacy and Outcome Sciences, South Carolina College of Pharmacy, Medical University of South Carolina, QF-411, Box 250132, 280 Calhoun Street, Charleston, SC 29425, USA. E-mail: [email protected] Received 28 October 2014; revised 12 February 2015; accepted 20 February 2015; published online 2 April 2015

Prevention of medication errors in neonatal LOS SS Garner et al

632 incorporate it as programmed decision support into an interactive computerized order set. The goals set forth for the ICOS-DS included improved quality with enhanced efficiency for both individual practitioners and system-wide processes, ease of use, and compatibility with an electronic medical record scheduled for implementation in the near future. The ICOS-DS is an interactive web-based Adobe PDF file that is electronically transmitted to pharmacy (fax to fax server). Interactive decision support logic was written in Adobe JavaScript and embedded in the PDF file. The document resides on the Clinician Order Forms website of the MUSC Medical Center, a familiar and easily accessible site to practitioners. The ICOS-DS includes fields for patient name, medical record number, weight, gestational age at birth, and date of birth, current date and time of day. Day of life is calculated and the field subsequently populated using the current date and date of birth. The ICOS-DS contains a series of check boxes for ordering blood and cerebrospinal fluid gram stains and cultures, urine cultures, serum chemistries, and cerebrospinal fluid protein, glucose and cell counts. There are required yes/no check boxes for the presence of a peripherally inserted central catheter 410 days, hemodynamic instability, serum creatinine 488.4 μmol l − 1 and suspicion of necrotizing enterocolitis. The antibiotic order section contains the empiric antibiotics included in the guidelines. To order an antibiotic, the prescriber selects a box beside the drug name. On the basis of the patient information provided (weight, day of life, presence of peripherally inserted central catheter 410 days, hemodynamic instability, serum creatinine 488.4 μmol l − 1 and suspicion of necrotizing enterocolitis), the box is checked whether the antibiotic matches the DS for appropriate empiric antibiotic selection. In addition, the antibiotic’s dosage regimen is automatically selected, dose and frequency calculated, and populated in the order form’s blanks. The dosage equation (for ex, mg kg − 1 per dose) is displayed next to the dosage regimen. For piperacillin/tazobactam, the dose is calculated based on the piperacillin component and converted to a piperacillin/tazobactam dose prior to populating the dose field. The practitioner has the option of modifying the dose and dosing interval. If the antibiotic selected does not match the DS, a warning message is displayed. For example, the guidelines recommend reserving empiric use of vancomycin and piperacillin/tazobactam for patients with a peripherally inserted central catheter 410 days, suspicion of necrotizing enterocolitis or hemodynamic instability. If a prescriber indicates none of these but selects vancomycin, a message is displayed stating that vancomycin and piperacillin/tazobactam are reserved for those clinical situations. The reverse would occur if either of those characteristics was indicated yet an antibiotic other than vancomycin or piperacillin/tazobactam was selected. In either situation, the prescriber has the option to override the suggestion if desired. If the prescriber indicates the patient has a serum creatinine 488.4 μmol l − 1, the unadjusted dosage regimen is selected and calculated as before, but in addition, a message is displayed warning that the patient has an elevated creatinine and suggesting the pediatric clinical pharmacist be contacted prior to the next dose to determine whether serum concentrations or dose modifications are needed. Each antibiotic is ordered stat and a note to administer all antibiotics stat upon delivery to the NICU is included. Once completed, the order form is electronically transmitted to pharmacy. Prior to implementation, the ICOS-DS was tested by the clinical informatics specialist (MI) and NICU pharmacists (SG, TC) and reviewed and approved by the NICU’s Evidence and Value Committee and hospital-wide forms committees.

Study design We performed a prospective, controlled, comparative study of errors in NICU antibiotic orders for suspected LOS to evaluate the impact of the ICOS-DS. Exemption status was granted by MUSC’s institutional review board. The study population included neonates admitted to the NICU for whom the clinical team ordered antibiotics for suspected LOS, defined as occurring after postnatal day of life 3. Baseline control data were collected prospectively for a 2-month period prior to implementation of the ICOSDS. There was a 2-month period of education and training of all neonatologists, fellows, neonatal nurse practitioners and nurses at mandatory staff meetings, in-services and through follow-up e-mails during which no data were collected. New residents were oriented to the ICOS-DS monthly. Post-implementation data were then prospectively collected over 4 months to serve as the intervention group. Journal of Perinatology (2015), 631 – 635

Order review Two NICU pharmacists (SG, TC) independently analyzed the pre- and postICOS-DS orders for prescribing errors, potential errors and omissions with consensus reached by discussion if needed. Prescribing errors included overdoses or underdoses, inappropriate route or schedule, inappropriate empiric or definitive antibiotic selection, incorrect patient information, drug–drug or drug–disease state interactions and contraindications. Dosing errors were defined as 410% variance and schedule errors defined as 42 h deviation from recommended.11,12 Inappropriate empiric antibiotic selection was defined as deviating from the guidelines without an explanation in the patient’s medical record. Inappropriate choice of definitive antibiotics was based on culture, susceptibility and the site of infection. Potential errors included misspelled drugs, illegible orders, leading decimals, trailing zeroes, impractical doses, first doses not ordered stat and error-prone abbreviations defined by MUSC’s Pharmacy and Therapeutics Committee or the Institute for Safe Medication Practices. Doses were considered impractical if ordered more precisely than tenths of a milligram for vancomycin and gentamicin or whole numbers for penicillins, cephalosporins and fluconazole. Omissions were missing information required by institutional policy or recommended by Institute for Safe Medication Practices. Multiple errors and omissions in a single order were counted individually. Our primary outcome was the error rate in NICU antibiotic orders for patients 43 days postnatal age.

Statistical analysis Data analysis was performed using SAS version 9.1.3. We estimated the error rate per order, both overall and by specific error type, and corresponding 95% confidence intervals (95% CIs) for control and intervention orders based on fitted generalized estimating equations with a log link function and exchangeable correlation structure, an approach akin to Poisson regression appropriate for clustered data applications (here, multiple orders from the same patient are clustered within patient).13,14 Specifically, we modeled the log error rate per order as a function of order type (control or intervention), estimated rates and 95% CIs based on fitted model parameters, and compared control and intervention per order error rates based on corresponding Wald tests for the order type model parameter. The proportion of orders with at least one error and by specific error types was also reported, with corresponding 95% CIs. Confidence intervals were constructed using an approach accounting for clustering, and comparisons between proportions were performed using the RaoScott chi-squared test.15,16 The same chi-squared test was used to assess the association between error probabilities and level of training or practitioner type. Associations between patient characteristics and error probabilities were summarized using odds ratios and corresponding 95% CIs based on fitted generalized estimating equation models with a logit link function and exchangeable correlation structure.

RESULTS There were 153 antibiotic orders among 34 patients analyzed in the pre-ICOS-DS control group and 146 orders among 45 patients in the post-ICOS-DS intervention group. Error rates and common examples of errors can be found in Figure 1 and Table 1, respectively. The ICOS-DS significantly reduced both the overall error rate per order and potential error rate per order (Figure 1). Although the omission error rate per order was reduced, the difference was not significant. Conversely, the prescribing error rate per order significantly increased because of an increase in incorrect patient weights. Incorrect patient weights were noted in 28.1% of intervention orders (95% CI = 17.0 to 39.2%) compared with 3.3% of control orders (95% CI = 0 to 6.7%; P o0.001). Of the 39 orders containing a patient weight varying from current, 7 were birth weights, 5 recent weights, 6 ‘dry’ weights for patients receiving extracorporeal membrane oxygenation, 7 rounded from current weights and 14 varied for unknown reasons. There was o10% discrepancy in patient weight in 24 of the orders, minimally affecting calculated dose. Most discrepancies 410% were due to using birth weights (n = 5) or ‘dry’ weights (n = 6). Similarly, the proportion of orders (Figure 2) with at least one error of any type was significantly reduced, as well as the proportion with at least one potential error. The reduction in the © 2015 Nature America, Inc.


633

Figure 1. Overall and specific error rates per order for control and intervention groups, *P o0.001.

Table 1. Common examples of individual error proportions for control and intervention orders Control orders (N = 153)

Intervention orders (N = 146)

Frequency Percent Frequency Percent P value

Error Dosage calculation Impractical dose Illegible order Scheduling error Not ordered stat

16 11 14 25 39

10 7 9 16 25

0 2 1 12 0

0 1 1 8 0

− 0.007 o0.001 0.02 −

Proportion (95% CI) of orders with errors

100 90

Control Intervention

80

*

70 60 50 40 30 20

*

10 0 Prescribing Errors

Potential Errors

Omissions

Any Error

Figure 2. Proportions of orders with at least one error of any type and specific types for control and intervention groups, *Po0.001. © 2015 Nature America, Inc.

Figure 3. Associations between patient characteristics and the likelihood of having at least one error of any type and specific types for control and intervention groups represented as odds ratios (95% CIs). The association between patient characteristics and error probabilities were analyzed as per week for gestational age, per 100 g for birth weight, and per day for day of life.

proportion with at least one omission was not statistically significant. The proportion with at least one prescribing error increased, but not significantly. There were no significant differences in the proportion of orders with at least one error of any type or of a specific type when analyzed by level of training or practitioner type. The associations between patient characteristics and having an order with at least one error of a given type are shown in Figure 3. Renal dysfunction was significantly associated with prescribing errors, with a 3.7-fold increase in odds of at least one prescribing error in orders for patients with renal dysfunction. Specifically, the percentage of orders with at least one prescribing error among orders for patients with renal dysfunction was 62% compared with 35% among orders for patients without renal dysfunction. Of the 46 prescribing errors in orders for patients with renal dysfunction, there were 23 overdoses, 2 underdoses, 11 incorrect schedules, 9 incorrect patient weights and 1 incorrect antibiotic choice. The increased risk of prescribing errors in orders for patients with renal dysfunction was not significantly different for control versus intervention orders (P-value for interaction = 0.15). There were also significant associations between renal dysfunction, gestational age and birth weight and having an order with at least one omission (Figure 3). A 1-week increase in gestational age resulted in an 11% increase in the odds of at least one omission (odds ratio = 1.11, 95% CI = 1.02 to 1.19; P = 0.01). Specifically, the average gestational age was 31 and 28 weeks, respectively, among orders with and without at least one omission. Likewise, a 100-g increase in birth weight resulted in a 5% increase in odds of Journal of Perinatology (2015), 631 – 635


634 an order with at least one omission (odds ratio = 1.05, 95% CI = 1.01 to 1.09; P = 0.02). Among orders with omissions, the average birth weight was 1691 g compared with an average birth weight of 1286 g for orders without omissions. There was a 90% reduction in the odds of an order with at least one omission for patients with renal dysfunction relative to those without (odds ratio = 0.10, 95% CI = 0.01 to 0.77; P = 0.03). Specifically, the proportion of orders with at least one omission for patients with renal dysfunction was 2% compared with 15% among patients without renal dysfunction. Associations between these patient characteristics and omissions were not significantly different for control versus intervention orders (interaction P-values 40.05). DISCUSSION We demonstrated a significant reduction in overall errors in antibiotic orders for neonatal LOS with an interactive computerized order set programmed with specific decision support. Similarly, others have documented reduced NICU medication errors following implementation of customized DS with commercial computerized physician order entry (CPOE) systems.17,18 One group studied the effects of CPOE with DS on gentamicin dosage errors for neonates either on admission or for LOS. This system’s DS functions like ours in that once the practitioner selects the medication, an order is generated which displays a dose per kilogram, dosing frequency and dose calculation. Although the sample size was small (28 pre-CPOE and 31 post-CPOE orders), dosing errors were eliminated from a rate of 6% pre-CPOE.17 Another NICU comparison of CPOE versus CPOE-DS documented no difference in non-intercepted error rate with CPOE but a 19% reduction (from 53 to 34%) with the addition of DS.18 In this system, the practitioner chose the medication and dosage regimen which was double checked by the DS. A warning message and request for correction was generated if the dose or frequency was outside the accepted range; however, the practitioner could choose to ignore it. The variation in outcomes between these two studies may be explained by their differences in DS design. The first provided a dose during the ordering process whereas the latter relied on warnings to correct errors which are often overridden.19 Our overall error rate was higher because these groups only tallied errors in drug dosing whereas our definition of an error was more comprehensive; however, our findings support the benefits of DS that prospectively provides and calculates antibiotic regimens. These pilot data provide insight into the benefits of dosage calculation packages for neonatal medication orders, an area previously identified as a knowledge gap in NICU patient safety.20 Although our intervention greatly reduced potential errors and omissions, these errors were not eliminated owing to practitioner behavior. For example, the ICOS-DS was designed for electronic transmission of the order to pharmacy. However, as a PDF file, at any step, a practitioner can print and manually fax it. Therefore, required fields could be left blank or handwritten doses could include leading decimals, trailing zeroes, etc. Even so, these behaviors were rare. One reason may be that ease of use and enhanced efficiency ultimately lead to a preference for the ICOSDS. Electronic transmission of inpatient orders to pharmacy was also a new process for NICU practitioners, possibly leading to uncertainty and subsequent manual printing and faxing. During this evaluation, order transmission could be verified; however, it was a cumbersome process. The process has subsequently been streamlined. It is reasonable to expect that integrating the ICOSDS with an electronic medical record and CPOE would eliminate these behaviors. The intervention orders contained a significantly higher number of prescribing errors per order because of incorrect patient weights. Potential explanations include practitioner misunderstanding of most appropriate weight for antibiotic dosing and our Journal of Perinatology (2015), 631 – 635

incomplete electronic medical record. Although birth and ‘dry’ weights are routinely used in other instances, current weight is typically preferred for antibiotic dosing based on the drug’s distribution pattern. During the study period, daily weights were recorded on handwritten nursing flow sheets while intervention orders were entered on computers in adjacent rounding rooms, preventing convenient double check during order entry. Similar alterations in workflow have been speculatively associated with an unanticipated increased mortality rate with CPOE implementation.21 It is noteworthy that none of the incorrect weights or gestational ages appeared to be keystroke errors. Most error types could be prevented by automated population from an electronic medical record. Whether orders were in the control or intervention group, there was a decreased rate of omissions for smaller or younger infants and those with renal dysfunction. Perhaps, this can be explained by additional attention paid to antibiotic orders for more fragile neonatal populations. We are not aware of similar reports specifically addressing these issues. In both control and intervention groups, we noted an association between renal dysfunction and prescribing errors. During design of the ICOS-DS, there was much discussion about displaying a warning message to practitioners for patients with renal dysfunction versus calculating a dose adjusted for renal dysfunction. The final decision was based on the concern that DS would oversimplify the judgment needed to assess neonatal renal function and drug clearance. For example, there was concern about basing doses on one serum creatinine value without considering trends over time given the lability commonly seen during a sepsis evaluation. In patients with renal dysfunction, our current practice for vancomycin or gentamicin is often to give one dose followed by monitoring serum concentrations to determine future doses. If similar logic was programmed, there was concern that ordering subsequent doses may be missed during practitioner handoffs. Previous NICU studies of CPOE-DS either did not describe incorporating an assessment of renal function17 or described the difficulties residents had understanding the DS’s assessment18 which supports our dilemma. Regardless, our preliminary results demonstrate the need for more specific DS to reduce errors for this patient population. As a result, we are currently evaluating a revised DS that adjusts doses and/or orders antibiotic serum concentrations for patients with renal dysfunction. This study has several limitations. First, we evaluated only errors in the ordering phase of the medication use cycle and not clinical outcomes such as adverse drug events, length of stay or mortality rates. We also did not include the proportion of errors corrected by pharmacist or nurse review as our goal was to prevent errors at the ordering phase. Our practitioners were not formally surveyed after NICU-wide implementation; however, their support for the ICOS-DS was demonstrated by nearly uniform use during this evaluation period. Given our study was limited to antibiotic orders for neonatal LOS, our results may not be generalizable to other drug classes and indications. Lastly, these findings may be less applicable to hospitals with fully implemented CPOE-DS. However, the most recent, albeit potentially outdated, survey of hospitals that provide care for children revealed only 6% had done so.22 A more recent 2012 survey of all US hospitals showed o 50% had adopted an electronic health record system.23 Through implementation of an order set with specific decision support guiding empiric antibiotic selection, dosing and dose calculation, we demonstrated significantly improved quality of antibiotic orders for LOS. Further refinements such as incorporating the order set into an electronic medical record and dosage adjustment for patients with renal dysfunction will need to be studied. These findings emphasize the potential for specific decision support and the importance of continued development and evaluation of interventions to promote medication safety for the NICU population. © 2015 Nature America, Inc.


CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGEMENTS We would like to acknowledge Lizbeth Hansen, Pharm.D., for her tireless efforts with data collection and entry.

REFERENCES 1 The National Coordinating Council for Medication Error Reporting and Prevention. Definition of medication errors http://www.nccmerp.org/ Accessed August 14 2013. 2 Kaushal R, Bates DW, Landrigan C, McKenna KJ, Clapp MD, Federico F et al. Medication errors and adverse drug events in pediatric patients. JAMA 2001; 285 (16): 2114–2120. 3 Gonzales K. Medication administration errors and the pediatric population: a systematic search of the literature. J Pediatr Nursing 2010; 25: 555–565. 4 Ghaleb MA, Barber N, Franklin BD, Yeung VW, Khaki ZF, Wong IC et al. Systematic review of medication errors in pediatric patients. Ann Pharmacother 2006; 40: 1766–1776. 5 Chedoe I, Molendijk HA, Dittrich ST, Jansman FG, Harting JW, Brouwers JR et al. Incidence and nature of medication errors in neonatal intensive care with strategies to improve safety: A review of the current literature. Drug Safety 2007; 30 (6): 503–513. 6 Snijders C, van Lingen RA, Fetter WPF. Incidents and errors in neonatal intensive care: review of the literature. Arch Dis Child Fetal Neonatal Ed 2007; 92: 391–398. 7 Ross LM, Wallace J, Paton JY. Medication errors in a paediatric teaching hospital in the UK: five years operational experience. Arch Dis Child 2000; 83: 492–497. 8 Holdsworth MT, Fichtl RE, Behta M, Raisch DW, Mendez-Rico E, Adams A et al. Incidence and impact of adverse drug events in pediatric inpatients. Arch Pediatr Adolesc Med 2003; 157(1): 60–65. 9 Di Pentima MC, Chan S, Eppes SC, Klein JD. Antimicrobial prescription errors in hospitalized children: Role of antimicrobial stewardship program in detection and intervention. Clin Pediatr 2009; 48(5): 505–512.

© 2015 Nature America, Inc.

635 10 Bissinger RL, Mueller M, Cox TH, Cahill J, Garner SS, Irving MG et al. Antibiotic timing in neonates with suspected hospital-acquired infections. Advances in Neonatal Care 2013; 13(1): 22–28. 11 Taketomo CK. Pediatric Dosage Handbook14th ed.Lexi-Comp, Inc.: Hudson, Ohio, 2007. 12 Lexi-Comp Online™ Pediatric Lexi-Drugs Online™ 2007, Lexi-Comp, Inc.: Hudson, Ohio. 13 Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73: 13–22. 14 Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating equation approach. Biometrics 1988; 44: 1049–1060. 15 Cochran WG. Sampling Techniques3rd EditionJohn Wiley & Sons: New York, 1977. 16 Rao JNK, Scott AJ. The analysis of categorical data from complex sample surveys: chi-squared tests for goodness of fit and independence in two-way tables. J Am Stat Assoc 1981; 76: 221–229. 17 Cordero L, Kuehn L, Kumar RR, Mekhjian HS. Impact of computerized physician order entry on clinical practice in a newborn intensive care unit. J Perinatol 2004; 24(2): 88–93. 18 Kazemi A, Ellenius J, Pourasghar F, Tofighi S, Salehi A, Amanati A et al. The effect of computerized physician order entry and decision support system on medication errors in the neonatal ward: Experiences from an Iranian teaching hospital. J Med Syst 2011; 35: 25–37. 19 Nightingale PG, Adu D, Richards NT, Peters M. Implementation of rules based computerized bedside prescribing and administration: intervention study. BMJ 2000; 320: 750–753. 20 Raju TNK, Suresh G, Higgins RD. Patient safety in the context of neonatal intensive care: Research and educational opportunities. Pediatr Res 2011; 70: 109–115. 21 Han YY, Carcillo JA, Venkataraman ST, Clark RS, Watson RS, Nguyen TC et al. Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system. Pediatrics 2005; 116: 1506–1512. 22 Teufel RJ, Kazley AS, Basco WT. Early adopters of computerized physician order entry in hospitals that care for children: A picture of U.S. health care shortly after the Institute of Medicine reports on quality. Clin Pediatr 2009; 48: 389–396. 23 DesRoches CM, Charles D, Furukawa MF, Joshi MS, Kralovec P, Mostashari F et al. Adoption of electronic health records grows rapidly, but fewer than half of U.S. hospitals had at least a basic system in 2012. Health Affairs 2013; 32: 1478–1485.

Journal of Perinatology (2015), 631 – 635