The Effect of Computerized Physician Order Entry on Medication Prescription

Errors and Clinical Outcome in Pediatric and Intensive Care: A Systematic

Review
Floor van Rosse, Barbara Maat, Carin M. A. Rademaker, Adrianus J. van Vught,
Antoine C. G. Egberts and Casper W. Bollen
Pediatrics 2009;123;1184-1190
DOI: 10.1542/peds.2008-1494

The online version of this article, along with updated information and services, is
located on the World Wide Web at:
http://www.pediatrics.org/cgi/content/full/123/4/1184

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly

publication, it has been published continuously since 1948. PEDIATRICS is owned, published,
and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk
Grove Village, Illinois, 60007. Copyright © 2009 by the American Academy of Pediatrics. All
rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from www.pediatrics.org by on July 24, 2010

REVIEW ARTICLE

The Effect of Computerized Physician Order Entry on

Medication Prescription Errors and Clinical Outcome
in Pediatric and Intensive Care: A Systematic Review
Floor van Rosse, MSca, Barbara Maat, PharmDb, Carin M. A. Rademaker, PharmD, PhDb, Adrianus J. van Vught, MD, PhDa,
Antoine C. G. Egberts, PharmD, PhDc, Casper W. Bollen, MD, PhDa

aPediatric Intensive Care Unit and bDepartment of Clinical Pharmacy, University Medical Center Utrecht, Utrecht, Netherlands; cDepartment of Pharmaco-epidemiology

and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Faculty of Science, Utrecht University, Utrecht, Netherlands

The authors have indicated they have no financial relationships relevant to this article to disclose.

ABSTRACT
CONTEXT. Pediatric and intensive care patients are particularly at risk for medication
errors. Computerized physician order entry systems could be effective in reducing
medication errors and improving outcome. Effectiveness of computerized physician www.pediatrics.org/cgi/doi/10.1542/
peds.2008-1494
order entry systems has been shown in adult medical care. However, in critically ill
patients and/or children, medication prescribing is a more complex process, and doi:10.1542/peds.2008-1494
usefulness of computerized physician order entry systems has yet to be established. Key Words
CPOE, hospital information systems,
OBJECTIVE. To evaluate the effects of computerized physician order entry systems on medical record systems, ICU, children,
patient safety, medication errors, mortality
medication prescription errors, adverse drug events, and mortality in inpatient
pediatric care and neonatal, pediatric or adult intensive care settings. Abbreviations
MPE—medication prescription error
METHODS. PubMed, the Cochrane library, and Embase up to November 2007 were used CPOE— computerized physician order
entry
as our data sources. Inclusion criteria were studies of (1) children 0 to 18 years old ADE—adverse drug event
and/or ICU patients (including adults), (2) computerized physician order entry MeSH—Medical Subject Headings
versus no computerized physician order entry as intervention, and (3) randomized RR—relative risk
CI— confidence interval
trial or observational study design. All studies were validated, and data were ana-
Accepted for publication Aug 6, 2008
lyzed.
Address correspondence to Casper W. Bollen,
RESULTS. Twelve studies, all observational, met our inclusion criteria. Eight studies took MD, PhD, Wilhelmina Children’s Hospital,
University Medical Center Utrecht, Pediatric
place at an ICU: 4 were adult ICUs, and 4 were PICUs and/or NICUs. Four studies Intensive Care Unit, Room KG 1.319.0, PO Box
were pediatric inpatient studies. Meta-analysis showed a significant decreased risk of 85090, 3508 AB Utrecht, Netherlands. E-mail:
c.w.bollen@umcutrecht.nl.
medication prescription errors with use of computerized physician order entry.
PEDIATRICS (ISSN Numbers: Print, 0031-4005;
However, there was no significant reduction in adverse drug events or mortality Online, 1098-4275). Copyright © 2009 by the
rates. A qualitative assessment of studies revealed the implementation process of American Academy of Pediatrics
computerized physician order entry software as a critical factor for outcome.

CONCLUSIONS. Introduction of computerized physician order entry systems clearly reduces medication prescription
errors; however, clinical benefit of computerized physician order entry systems in pediatric or ICU settings has not
yet been demonstrated. The quality of the implementation process could be a decisive factor determining overall
success or failure. Pediatrics 2009;123:1184–1190

A CCORDING TO THE Institute of Medicine, medical errors lead to 44.000 to 98.000 deaths in the United States
annually.1 Currently, prevention of medical errors receives a large amount of attention and presents a major
challenge to health care. In particular, critically ill patients are vulnerable and at risk for medication prescription
errors (MPEs). Within this population, neonatal and pediatric patients present an even more vulnerable group. A
study by Kaushal et al2,3 underlined this by showing that potentially harmful errors occurred 3 times more frequently
in pediatric than in adult patients. Moreover, an increasing number of drugs, regimen complexity, and the
continuously growing knowledge base of drug indications and adverse effects create the need for automated systems
to deliver clinical support.4 Use of computerized physician order entry (CPOE) systems could possibly address these
problems. For example, it has been shown that computer support in drug dosing has resulted in more patients with
drug concentrations in the therapeutic range, reduced time to achieve therapeutic benefits, and resulted in fewer
adverse effects of treatment in adults.5 Computer systems, therefore, may support doctors in tailoring drug doses
more closely to the needs of individual patients.
CPOE can also improve patient safety in several ways. First, CPOEs are obviously more legible than handwritten
ones. Furthermore, CPOE can force physicians to include dose, route of administration, and frequency in the order
before authorizing the prescription, thus resulting in better structured and more complete medication prescriptions.

1184 van ROSSE et al

Downloaded from www.pediatrics.org by on July 24, 2010
CPOE systems can be linked to databases with back- cal decision-support system consists of at least basic dos-
ground information and deliver decision support by ing guidance for medication, formulary decision support,
warning for drug-dosage errors, interactions, or contra- and drug allergy, duplicate therapy, and drug-drug in-
indications.6 However, although it is generally assumed teraction checking.11 Clinical decision-support systems
that CPOE systems decrease medication error rates and are built into most CPOE systems.3 An MPE was defined
improve clinical outcome, unfavorable findings associ- as any error in prescription of medication irrespective of
ated with CPOE have been reported as well.7 In a study outcome. Potential ADEs were defined as medication
by Han et al,8 the mortality rate in a pediatric population errors with significant potential to harm a patient with-
increased after CPOE implementation. Therefore, spe- out reaching a patient, and ADEs were defined as actual
cific settings such as pediatric or neonatal care or com- harm that resulted from a medication error.2
plex environments such as ICUs could determine the
eventual clinical effect of CPOE systems. Data Extraction
We performed a systematic review of the use of CPOE The following data were extracted: year of study, study
systems in the most demanding and complex situations, design, study period, whether the study was performed
that is, adult ICUs, PICUs, and NICUs, and in general in an academic hospital, patient population (adult ICU,
pediatric and neonatal care. Meta-analysis was per- PICU, NICU, or pediatric ward), software manufacturer,
formed to estimate effects on MPEs, adverse drug events presence of decision regarding support. With respect to
(ADEs), and mortality rate. Factors associated with suc- the implementation process, use of classroom training
cess or failure of CPOE systems were identified. and individual training and on-site support present after
CPOE implementation was assessed.
METHODS
This systematic review was conducted according to the Validity Assessment
criteria as defined in the Quality of Reporting of Meta- Observational studies were evaluated by applying crite-
analyses (QUORUM) and MOOSE (Meta-analysis of Ob- ria from the STROBE (Strengthening the Reporting of
servational Studies in Epidemiology) statements.9,10 Observational Studies in Epidemiology) statement.12 We
determined validity by assessing whether control and
Literature Search intervention groups were defined, whether possible
Studies were identified by searching PubMed, the Co- sources of confounding, selection bias, or misclassifica-
chrane library, and Embase up to November 2007. The tion were identified and/or adjusted for, whether out-
literature search strategy was performed by using the come measures were clearly defined, whether the exact
following search terms: (child*[tiab] or paediatr*[tiab] or study period was mentioned, whether the implementa-
pediatr*[tiab] or infant*[tiab] or toddler*[tiab] or “pre tion process was described, and whether original out-
school”[tiab] or preschool[tiab] or adolescent*[tiab] come data were available in the publication. Validity of
or pediatrics[Medical Subject Headings (MeSH)] or randomized trials was assessed by using the criteria pub-
child[MeSH] or infant[MeSH] or adolescent[MeSH] or lished by Jadad et al.13
intensive care units[MeSH] or intensive care units, neo-
natal[MeSH] or intensive care, neonatal[MeSH] or in- Data Analysis
tensive care[tiab]) and (CPOE[tiab] or “computerized All data were analyzed on an intention-to-treat basis.
physician order entry”[tiab] or “computerized provider Risk rates for MPEs were calculated by dividing the
order entry”[tiab] or “computerized prescribing”[tiab] or number of errors by the total number of prescriptions in
“electronic prescribing systems”[tiab] or “computerized the intervention and control groups, respectively. Risk
order entry”[tiab] or “computer order entry”[tiab] or rates for ADEs and mortality were calculated by the
“medical order entry systems”[MeSH]). number of incidents divided by the population at risk in
the 2 groups: CPOE and no CPOE. Using the risk rates in
Study Selection both groups, relative risk (RR) estimates were calculated
After title screening, we examined abstracts and selected along with 95% confidence intervals (CI). Pooled RR
articles that met all of the following inclusion criteria: (1) estimates were calculated by using a random-effects
hospitalized children 0 to 18 years old and/or ICU pa- model. Heterogeneity was assessed by the I[r]2 statistic.14
tients (including adults); (2) intervention CPOE com- I2 describes the percentage of total variation across stud-
pared with no CPOE; and (3) randomized trial or obser- ies resulting from heterogeneity rather than chance. I2
vational cohort study design. Exclusion criteria were ranges from 0% to 100%; a value of 0% indicates no
descriptive studies (ie, case reports, narrative reviews, heterogeneity, and larger values indicate increasing het-
comments, etc) and CPOE research in populations tar- erogeneity. All analyses were conducted by using Excel
geted at specific diseases. Literature lists of included ar- 2007 (Microsoft, Redmond, WA).
ticles were searched for possible additional studies.
RESULTS
Definitions Search
A CPOE system was defined as a computer-based system Our literature search yielded 122 citations that were
that automates the medication-ordering process to en- screened for relevance, which left 12 articles that were
sure standardized, legible, and complete orders. A clini- included in the systematic review (Fig 1). We also cross-

PEDIATRICS Volume 123, Number 4, April 2009 1185

Downloaded from www.pediatrics.org by on July 24, 2010
 Titles screened for relevance: whether decision support was available.20,23 A quantita-
122 publications 43 exclusions:
- 3 language other than English, German, French, or Dutch tive data analysis on decision support also was not pos-
- 10 oncology/chemotherapy and CPOE
- 24 other specific medical topic and CPOE
sible, either because the studies poorly described the
- 4 reports/reviews decision-support systems or because of the different lev-
- 2 no CPOE research
els of decision support among studies.
There was considerable variation in timing and length
Abstracts screened for relevance: of the periods in which outcome was measured without
79 publications 51 exclusions:
- 9 specific medical topics and CPOE or before CPOE and with CPOE among studies (Fig 2).
- 22 no CPOE research
- 9 comments Five of the studies started their intervention period right
- 6 reports/reviews
- 4 no children/no ICU after CPOE implementation.8,18–20,23 Therefore, a so-
- 1 guideline called learning-curve in these studies was included in
Full texts screened for relevance
:
the measurements. The other 7 studies did not include
28 publications 16 exclusions: the period right after CPOE implementation in the mea-
- 7 no outcome data available
- 2 no CPOE research surement period.
- 3 no children/no ICU
- 3 no comparison
- 1 compares 2 CPOE systems
Adult ICU Studies
All 4 adult ICU studies described an intervention and a
control group, assessed potential confounding, and men-
12 publications
tioned quantitative outcome data on number of MPEs,
FIGURE 1 ADEs, and/or mortalities. Study periods varied among
Study selection. the ICU studies (Fig 2). For 2 of the studies, the imple-
mentation process was not described,7,17 for 1 study it
was mentioned only briefly,23 and for only 1 study was it
described extensively.22 An increase in MPEs was ob-
referenced the results of our literature search with lists served by Weant et al23 during the initial period after
of studies published in another systematic review.15 This CPOE implementation. Three studies showed a clinical
did not yield any additional studies that were not already beneficial effect.7,17,22 In the study by Colpaert et al,17
found in our search. Although the studies of Han et al8 CPOE only had a beneficial effect when potential ADEs
and Upperman et al16 took place in the same hospital, the were taken into account.
outcomes were different and both, therefore, were in-
cluded.
Pediatric, PICU, and NICU Studies
Included Studies In all 8 studies the intervention and/or control group
Among the 12 included studies, which are summarized were clearly defined. All studies reported patient and
in Table 1, there were no randomized trials. There was 1 clinical characteristics that implied comparability be-
controlled cross-sectional trial.17 Eight studies were ret- tween the intervention and control groups. The original
rospective,8,16,18–23 and 3 studies were prospective cohort outcome data could be extracted from all studies except
studies.7,24,25 Of the included studies, 4 were performed that of Upperman et al.16 In this study, only aggregate
with adult ICU patients,7,17,22,23 and 8 were performed outcome estimates were reported. Again, study periods
with pediatric patients.8,16,18–21,24,25 Of those 8 pediatric varied considerably (Fig 2). King et al21 did not describe
studies, 4 were performed on a PICU and/or NICU,18–20,25 their implementation process, Potts et al25 and Hold-
1 on a ward with a PICU,24 and 3 on a pediatric sworth et al24 mentioned it briefly, and the other 5
ward.8,16,21 authors8,16,18–20 described their implementation process
Three of the 12 studies reported mortality as out- more extensively.
come,8,19,20 1 analyzed workflow,22 7 of them studied Of 5 studies with MPEs and/or ADEs as outcome
MPEs and/or ADEs,7,16,17,21,23–25 and 1 study18 reported 3 measures, CPOE conferred a significant beneficial effect
outcomes: medication turnaround times, radiology pro- in 3 studies,18,24,25 and in 1 study a nonsignificant bene-
cedure completion time, and MPEs (only gentamicin ficial effect was reported.16 In the study by King et al,21
dosages). The definitions of MPEs and ADEs varied con- the overall result was beneficial: MPEs decreased, as did
siderably among studies (see Table 3, which is published ADEs, but potential ADEs increased. In the 3 studies
as supporting information at www.pediatrics.org/content/ with mortality rate as main outcome,8,19,20 results varied;
full/123/4/1184). in the study by Han et al8 the mortality rate increased,
Different kinds of CPOE software systems were used: whereas Del Beccaro et al19 reported a nonsignificant
Siemens (Munich, Germany), Eclipsys (Atlanta, GA), decrease in mortality rate, and Keene et al20 reported a
Cerner (Kansas City, KS), PHAMIS (Seattle, WA), Wiz- significant decrease in mortality rate.
Order (Nashville, TN), and homegrown systems. Be-
cause of a lack of consistency among studies, quantita- Implementation Process
tive data analysis across vendors was not possible. Four studies described classroom training before imple-
In 7 studies, implementation of decision support was mentation, extensive individualized instruction, and on-
explicitly mentioned,8,16–19,24,25 in 3 studies there was no site support during and after CPOE implementation.18–20,22
decision support,7,21,22 and 2 studies did not describe Two of those studies showed a significant beneficial ef-

1186 van ROSSE et al

Downloaded from www.pediatrics.org by on July 24, 2010
 TABLE 1 Included Studies
Study (Year) Study Design Academic Patient Set Setting Software Decision Comparison Outcome Implementation Conclusion
Support Process
Thompson et Retrospective cohort Yes Adult ICU Eclipsys No No CPOE Ordering times Classroom training, Improved timeliness
al22 (2004) personal of urgent tests
training, on-site
support
Shulman et al7 Prospective cohort Yes Adult ICU Cis No Handwritten MPEs and/or Not described Small decrease in MEs
(2005) orders ADEs and timeliness of
service
Colpaert et Controlled cross- Yes Adult ICU Not described Yes, moderate Paper-based MPEs Not described Decrease in MPEs
al17 (2006) sectional trial level unit
Weant et al23 Retrospective cohort No Adult ICU Not described Not described No CPOE MPEs and/or Training (kind of Increase in medication
(2007) ADEs training not errors during initial
mentioned) period after CPOE
implementation
Cordero et al18 Retrospective cohort Yes Neonatal NICU Siemens Yes No CPOE MPEs, medication Classroom training, Significant reduction
(2004) turnaround personal in medication
times training, on-site turnaround times
support and MPEs
Keene et al20 Retrospective cohort No Neonatal/pediatric NICU/PICU PHAMIS Not described No CPOE Mortality rate Classroom training, Mortality rate did not
(2007) personal increase during
training, on-site CPOE
support implementation
King et al21 Retrospective cohort Yes Pediatric Ward Eclipsys No No CPOE MPEs and/or Not described Decrease in MPEs, not
(2003) ADEs in ADEs
Potts et al25 Prospective cohort Yes Pediatric PICU WizOrder Yes No CPOE MPEs and/or Training (kind of Decrease in MPEs, and
(2004) ADEs training not potential ADEs
mentioned)
Upperman et Retrospective cohort No Pediatric Ward Cerner Yes No CPOE Number-needed- Classroom training, CPOE would prevent 1

Downloaded from www.pediatrics.org by on July 24, 2010

al16 (2005) to-treat analog extra training on ADE every 64
request patient-days
Han et al8 Retrospective cohort No Pediatric Ward Cerner Yes No CPOE Mortality rate Classroom training Increase in mortality
(2005) rate
Del Beccaro et Retrospective cohort Yes Pediatric PICU Cerner Yes No CPOE Mortality rate Classroom training, No increase in
al19 (2006) personal mortality rate, even
training, on-site shortly after CPOE
support implementation
Holdsworth et Prospective cohort Yes Pediatric PICU/ward Eclipsys Yes, substantial No CPOE ADEs Not described CPOE associated with
al24 (2007) reduction in ADEs
and potential ADEs

PEDIATRICS Volume 123, Number 4, April 2009

1187
 CPOE implementation
No CPOE
Colpaert et al 17 (2 x 5 wk)
CPOE

Cordero et al 18 (2 x 6 mo)

Del Beccaro et al 19 (2 x 13 mo)

Han et al 8 (13 vs 5 mo)

Holdsworth et al 24 (2 x 7 mo)

King et al 21 (2 x 35 mo)

Keene et al 20 (3 x 6 mo)

Potts et al 25 (2 x 2 mo)

Thompson et al 22 (2 x 1 mo)

Shulman et al 7 (a few days every 2 mo)

Upperman et al 16 (intervention period unknown)

Weant et al 23 (23 vs 3 mo)

-25 -23 -21 -19 -17 -15 -13 -11 -9 -7 -5 -3 -1 1 3 5 7 9 11 13 15 17 19 21 23 25

Months before/after CPOE implementation
FIGURE 2
Distribution of the study periods.

fect of CPOE.18,22 In the other 2 studies, mortality rates rates were pooled for pediatric and neonatal studies
did not increase after CPOE implementation.19,20 Han et only. There was a significant reduction in MPEs (RR:
al8 and Upperman et al reported 3 hours of classroom 0.08 [95% CI: 0.01– 0.77]), uniformly observed in all
computer practice 3 months before CPOE implementa- studies. The number of potential and actual ADEs
tion. In the Upperman et al16 study, CPOE had a positive showed a nonsignificant decrease with the use of CPOE
effect on ADEs, but in the Han et al8 study, introduction (RR: 0.65 [95% CI 0.40 –1.08]). However, there was
of a CPOE system increased mortality rates. significant heterogeneity (I2 ⫽ 65%) among the studies.
Quantitative analysis to explore the causes for this het-
Meta-analysis erogeneity was not possible because of the limited number
A meta-analysis was conducted to pool the outcome of studies available. Mortality rates were not signifi-
measures: MPEs, ADEs (potential and actual ADEs taken cantly influenced by CPOE (RR: 1.02 [95% CI: 0.52–
together), and mortality rate (Table 2). MPEs were 1.94]). This was observed in all studies except for the
pooled, taking all studies together. ADEs and mortality study by Han et al.8 In that study, an RR of 2.35 (95% CI:

TABLE 2 Meta-analyses
Study (Year) With CPOE No CPOE
n Errors/ADEs/ % n Errors/ADEs/ % RR 95% CI
Mortalities, n Mortalities, n
Errors
Potts et al (2004) 7025 12a 0 6803 2049a 30 0.01 0.00–0.01
Shulman et al (2005) 2429 117a 5 1036 71a 7 0.70 0.53–0.94
Colpaert et al (2006 1286 44a 3 1224 330a 27 0.13 0.09–0.17
Pooled (I2 ⫽34%) 0.08 0.01–0.76
ADEs
King et al (2003) 5786 7b 0.12 11 699 5b 0.04 2.83 0.89–10.33
Potts et al (2004) 246 88b 36 268 147b 55 0.65 0.58–0.73
Holdsworth et al (2007) 1210 72b 6 1197 170b 14 0.42 0.33–0.55
Pooled (I2 ⫽65%) 0.65 0.40–1.08
Mortalities
Cordero et al (2004) 100 9c 9 111 16c 14 0.62 0.29–1.35
Han et al (2005) 548 36c 7 1394 39c 3 2.35 1.51–3.65
Del Beccaro et al (2006) 1301 45c 3 1232 52c 4 0.82 0.55–1.21
Keene et al (2007) 374 9c 2 917 29c 3 0.76 0.36–1.59
Pooled (I2 ⫽0%) 1.02 0.52–1.94
a Errors.
b ADEs.
c Mortalities.

1188 van ROSSE et al

Downloaded from www.pediatrics.org by on July 24, 2010
1.51–3.65) was observed. Even after adjustment for pos- tended their postimplementation study period to 13
sible confounders, the mortality risk associated with months. The longer study period of Del Beccaro et al
CPOE remained elevated (odds ratio: 3.28 [95% CI: may have averaged out a potentially higher error rate in
1.94 –5.55]). the first few months after CPOE implementation (learn-
ing curve). Besides Del Beccaro et al and Han et al,
DISCUSSION Keene et al20 also studied the effect of CPOE introduction
In this systematic review we affirmed the important in a critically ill pediatric population with comparable
potential of CPOE systems to reduce MPEs. However, to results to those of Del Beccaro et al. Potential causes of
what extent the application of CPOE systems actually the increase in mortality rate in the study by Han et al
results in clinical benefit remains to be established. Our have been hypothesized as slowing down of adequate
meta-analysis showed a nonsignificant and heteroge- patient treatment resulting from (1) the inability to reg-
neously distributed reduction in ADEs. Overall, mortal- ister patients during transport to the hospital (medica-
ity did not seem to be affected by the use of CPOE. The tion could only be ordered when the patient had arrived
implementation process without individual practice and in the hospital), (2) an increase in time needed to enter
in-house support after CPOE-implementation could be orders, (3) a reduction in verbal communication, (4)
related with unfavorable clinical outcome. drug relocation from ward to central pharmacy, and (5)
This is the first systematic review that concentrates on technical problems with network connections.8,19,20,26
the effects of CPOE on pediatric care and critical care in Most of these causes cannot be attributed to the CPOE
general. It is necessary to specifically focus on these system itself but resulted from the implementation process.
groups because of their high vulnerability and the com- As can be concluded from the previous paragraph, the
plexity of their treatments. We pooled results on MPEs, implementation of a CPOE system could be critical. We
taking pediatric non-ICU, PICU, NICU, and adult ICU argue that 3 hours of training 3 months before the
studies together, because of the involved complexity of implementation day (Del Beccaro et al19 and Han et al8)
the prescription process mentioned above. We assumed is far from sufficient. House staff cannot learn enough in
that the effect of CPOE systems on MPEs would be just 3 hours, and 3 months later they probably will have
mainly influenced by the level of demand posed by the forgotten most of what they did learn.
setting in which the CPOE system was used and the Seven systematic reviews about CPOE have been
complexity of the patients. These patients probably de- published as yet,3,11,15,27–30 but none of them concentrated
mand a nonordinary CPOE system to improve MPEs and on CPOE in a pediatric and ICU population, which rep-
patient outcome, including ADEs. Obviously though, resent the most demanding and complex situations. For
pediatric non-ICU, PICU, NICU, and adult ICU patients 1 study the effect of CPOE on medication safety in
are quite different, and it would be interesting to distin- general was described,3 for 1 clinical decision support
guish between these groups and study them in more and clinicians’ behavior were described,29 for 1 the effect
detail with regard to clinical outcome. Unfortunately, on time records in clinical staff was studied,30 1 focused
only a limited set of clinical outcome data restricted to on the effect on pathology services,28 1 studied costs,
pediatric and neonatal patients was available. adherence, and safety in a noncritical adult population,27
It is evident that CPOE gives rise to better structured and 1 focused on decision support and examined cost-
and more clearly legible prescriptions. The dramatic de- effectiveness.11 Only 1 of these 7 reviews examined the
crease in MPEs experienced after CPOE implementation use of CPOE in a pediatric and/or critically ill popula-
in different studies clearly illustrates this aspect. More- tion.15 However, this review did not assess the exclusive
over, improvement in communication between physi- effects of CPOE systems on enhancing medication safety
cians, nurses, and pharmacists has been shown as well.22 but, rather, investigated other interventions as well. In
Ordering and prescribing by CPOE have been found to addition, this review applied other inclusion criteria and
be more efficient than handwritten prescribing. Al- so included studies that differed from ours.
though it might be expected that CPOE systems can Ideally, a large randomized trial would provide valid
introduce new errors, in the present study this was not evidence for the effect of CPOE systems on patient safety
demonstrated. However, reductions in MPEs did not and clinical outcome. However, because of the nature of
directly result in reduction in clinically relevant ADEs or the intervention, a randomized trial would be practically
improvement of clinical outcome. nearly impossible to conduct. Therefore, most studies
The increase in mortality rates associated with the were based on a before/after design; however, this de-
introduction of a CPOE system as reported by Han et al,8 sign permits limited conclusions about the causative na-
has been discussed extensively in the literature.19,20,26 Del ture of observed associations between CPOE introduc-
Beccaro et al19 studied the exact same CPOE system as tion and change in outcome. More valid effect estimates
Han et al but did not find a significant change in mor- could be obtained by using a “controlled before/after”
tality rates. Ammenwerth et al26 compared these 2 stud- design in a multicenter setting. An intervention setting
ies and stated that there were important differences in with and a control setting without the intervention are
design and implementation of these studies. Han et al both followed in time. Observed differences before and
studied CPOE use in a more critically ill and much after the intervention, thus, can be adjusted for general
younger patient population compared with Del Beccaro changes in time in the control setting. Furthermore, in
et al. Furthermore, Han et al only studied 5 months after future studies, strict criteria should be used to define
CPOE implementation, whereas Del Beccaro et al ex- MPEs and ADEs, and methods of detecting and evaluat-

PEDIATRICS Volume 123, Number 4, April 2009 1189

Downloaded from www.pediatrics.org by on July 24, 2010
ing should be clearly described. We found definitions of reports of randomized clinical trials: is blinding necessary?
detection and evaluation of MPEs and ADEs to vary Control Clin Trials. 1996;17(1):1–12
widely among studies, which possibly led to variable 14. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring
results and making comparison between studies difficult inconsistency in meta-analyses. BMJ. 2003;327(7414):
557–560
(Table 3). Finally, intervention data should preferably be
15. Conroy S, Sweis D, Planner C, et al. Interventions to reduce
collected directly after CPOE implementation to make
dosing errors in children: a systematic review of the literature.
assessment of a potential learning curve possible. Drug Saf. 2007;30(12):1111–1125
16. Upperman JS, Staley P, Friend K, et al. The impact of hospi-
CONCLUSIONS talwide computerized physician order entry on medical errors
CPOE systems indisputably reduce MPEs effectively. in a pediatric hospital. J Pediatr Surg. 2005;40(1):57–59
However, as to what extent this results in improved 17. Colpaert K, Claus B, Somers A, Vandewoude K, Robays H,
patient safety and better clinical outcome remains to be Decruyenaere J. Impact of computerized physician order entry
established. The implementation process of CPOE sys- on medication prescription errors in the intensive care unit: a
tems requires specific attention, because this may be controlled cross-sectional trial. Crit Care. 2006;10(1):R21
associated with adverse outcome. Multicenter studies, 18. Cordero L, Kuehn L, Kumar RR, Mekhjian HS. Impact of
preferably designed as controlled before/after studies, computerized physician order entry on clinical practice in a
newborn intensive care unit. J Perinatol. 2004;24(2):88 –93
are needed to ascertain the role and requirements of
19. Del Beccaro MA, Jeffries HE, Eisenberg MA, Harry ED. Com-
CPOE systems in improving hospital care for pediatric
puterized provider order entry implementation: no association
and critically ill patients. with increased mortality rates in an intensive care unit. Pedi-
atrics. 2006;118(1):290 –295
REFERENCES 20. Keene A, Ashton L, Shure D, Napoleone D, Katyal C, Bellin E.
1. Kohn LT. To Err Is Human: Building a Safer Health System. Wash- Mortality before and after initiation of a computerized physi-
ington, DC: National Academy Press; 1999 cian order entry system in a critically ill pediatric population.
2. Kaushal R, Bates DW, Landrigan C, et al. Medication errors and Pediatr Crit Care Med. 2007;8(3):268 –271
adverse drug events in pediatric inpatients. JAMA. 2001; 21. King WJ, Paice N, Rangrej J, Forestell GJ, Swartz R. The effect
285(16):2114 –2120 of computerized physician order entry on medication errors
3. Kaushal R, Shojania KG, Bates DW. Effects of computerized and adverse drug events in pediatric inpatients. Pediatrics. 2003;
physician order entry and clinical decision support systems on 112(3 pt 1):506 –509
medication safety: a systematic review. Arch Intern Med. 2003; 22. Thompson W, Dodek PM, Norena M, Dodek J. Computerized
163(12):1409 –1416 physician order entry of diagnostic tests in an intensive care
4. Schiff GD, Rucker TD. Computerized prescribing: building the unit is associated with improved timeliness of service. Crit Care
electronic infrastructure for better medication usage. JAMA.
Med. 2004;32(6):1306 –1309
1998;279(13):1024 –1029
23. Weant KA, Cook AM, Armitstead JA. Medication-error report-
5. Walton R, Dovey S, Harvey E, Freemantle N. Computer sup-
ing and pharmacy resident experience during implementation
port for determining drug dose: systematic review and meta-
of computerized prescriber order entry. Am J Health Syst Pharm.
analysis. BMJ. 1999;318(7189):984 –990
2007;64(5):526 –530
6. Bates DW. Using information technology to reduce rates of
24. Holdsworth MT, Fichtl RE, Raisch DW, et al. Impact of com-
medication errors in hospitals. BMJ. 2000;320(7237):788 –791
puterized prescriber order entry on the incidence of adverse
7. Shulman R, Singer M, Goldstone J, Bellingan G. Medication
drug events in pediatric inpatients. Pediatrics. 2007;120(5):
errors: a prospective cohort study of hand-written and com-
1058 –1066
puterised physician order entry in the intensive care unit. Crit
25. Potts AL, Barr FE, Gregory DF, Wright L, Patel NR. Comput-
Care. 2005;9(5):R516 –R521
8. Han YY, Carcillo JA, Venkataraman ST, et al. Unexpected erized physician order entry and medication errors in a pedi-
increased mortality after implementation of a commercially atric critical care unit. Pediatrics. 2004;113(1 pt 1):59 – 63
sold computerized physician order entry system [published 26. Ammenwerth E, Talmon J, Ash JS, et al. Impact of CPOE on
correction appears in Pediatrics. 2006;117(2):594]. Pediatrics. mortality rates: contradictory findings, important messages.
2005;116(6):1506 –1512 Methods Inf Med. 2006;45(6):586 –593
9. Clarke M. The QUORUM statement. Lancet. 2000;355(9205): 27. Eslami S, bu-Hanna A, de Keizer NF. Evaluation of outpatient
756 –757 computerized physician medication order entry systems: a sys-
10. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of ob- tematic review. J Am Med Inform Assoc. 2007;14(4):400 – 406
servational studies in epidemiology: a proposal for reporting. 28. Georgiou A, Williamson M, Westbrook JI, Ray S. The impact of
Meta-analysis of Observational Studies in Epidemiology computerised physician order entry systems on pathology
(MOOSE) group. JAMA. 2000;283(15):2008 –2012 services: a systematic review. Int J Med Inform. 2007;76(7):
11. Kuperman GJ, Bobb A, Payne TH, et al. Medication-related 514 –529
clinical decision support in computerized provider order entry 29. Kawamoto K, Lobach DF. Clinical decision support provided
systems: a review. J Am Med Inform Assoc. 2007;14(1):29 – 40 within physician order entry systems: a systematic review of
12. von Elm E, Altman DG, Egger M, et al. The Strengthening the features effective for changing clinician behavior. AMIA Annu
Reporting of Observational Studies in Epidemiology (STROBE) Symp Proc. 2003:361–365
statement: guidelines for reporting observational studies [pub- 30. Poissant L, Pereira J, Tamblyn R, Kawasumi Y. The impact of
lished correction appears in Ann Intern Med. 2008;148(2):168]. electronic health records on time efficiency of physicians and
Ann Intern Med. 2007;147(8):573–577 nurses: a systematic review. J Am Med Inform Assoc. 2005;12(5):
13. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of 505–516

1190 van ROSSE et al

Downloaded from www.pediatrics.org by on July 24, 2010
The Effect of Computerized Physician Order Entry on Medication Prescription
Errors and Clinical Outcome in Pediatric and Intensive Care: A Systematic
Review
Floor van Rosse, Barbara Maat, Carin M. A. Rademaker, Adrianus J. van Vught,
Antoine C. G. Egberts and Casper W. Bollen
Pediatrics 2009;123;1184-1190
DOI: 10.1542/peds.2008-1494
Updated Information including high-resolution figures, can be found at:
& Services http://www.pediatrics.org/cgi/content/full/123/4/1184
Supplementary Material Supplementary material can be found at:
http://www.pediatrics.org/cgi/content/full/123/4/1184/DC1
References This article cites 28 articles, 14 of which you can access for free
at:
http://www.pediatrics.org/cgi/content/full/123/4/1184#BIBL
Citations This article has been cited by 3 HighWire-hosted articles:
http://www.pediatrics.org/cgi/content/full/123/4/1184#otherartic
les
Subspecialty Collections This article, along with others on similar topics, appears in the
following collection(s):
Office Practice
http://www.pediatrics.org/cgi/collection/office_practice
Permissions & Licensing Information about reproducing this article in parts (figures,
tables) or in its entirety can be found online at:
http://www.pediatrics.org/misc/Permissions.shtml
Reprints Information about ordering reprints can be found online:
http://www.pediatrics.org/misc/reprints.shtml

Downloaded from www.pediatrics.org by on July 24, 2010