Int J Clin Pharm
DOI 10.1007/s11096-014-9925-x
RESEARCH ARTICLE
Performance of a clinical decision support system and of clinical
pharmacists in preventing drug–drug interactions on a geriatric
ward
Pieter Cornu • Stephane Steurbaut •
Sabina Šoštarić • Aleš Mrhar • Alain G. Dupont
Received: 7 November 2013 / Accepted: 11 February 2014
 Koninklijke Nederlandse Maatschappij ter bevordering der Pharmacie 2014
Abstract Background Drug–drug interactions (DDIs) can
lead to adverse drug events and compromise patient safety.
Two common approaches to reduce these interactions in
hospital practice are the use of clinical decision support
systems and interventions by clinical pharmacists. Objec-
tive To compare the performance of both approaches with
the main objective of learning from one approach to
improve the other. Setting Acute geriatric ward in a uni-
versity hospital. Methods Prospective single-centre, cohort
study of patients admitted to the geriatric ward. An inde-
pendent pharmacist compared the clinical decision support
alerts with the DDIs identified by clinical pharmacists and
evaluated their interventions. Contextual factors used by
the clinical pharmacists for evaluation of the clinical rel-
evance were analysed. Adverse drug events related to DDIs
were investigated and the causality was evaluated by a
clinical pharmacologist based on validated criteria. Main
outcome measure Number of alerts, interventions and the
acceptance rates. Results Fifty patients followed by the
clinical pharmacists, were included. The clinical pharma-
cists identified 240 DDIs (median of 3.5 per patient) and
advised a therapy change for 16 of which 13 (81.2 %) were
accepted and three (18.8 %) were not. The decision support
system generated only six alerts of which none were
P. Cornu (&)
S. Steurbaut
A. G. Dupont
Faculty of Medicine and Pharmacy, Research Group Clinical
Pharmacology and Clinical Pharmacy (KFAR), Vrije
Universiteit Brussel, 1090 Brussels, Belgium
e-mail: Pieter.Cornu@vub.ac.be
S. Šoštarić
A. Mrhar
Faculty of Pharmacy, University of Ljubljana, 1000 Ljubljana,
Slovenia
accepted by the physicians. Thirty-seven adverse drug
events were identified for 29 patients that could be related
to 55 DDIs. For two interactions the causality was evalu-
ated as certain, for 31 as likely, for ten as possible and for
12 as unlikely. Mainly intermediate level interactions were
related to adverse drug events. Contextual factors taken
into account by the clinical pharmacists for evaluation of
the interactions were blood pressure, international norma-
lised ratio, heart rate, potassium level and glycemia.
Additionally, the clinical pharmacists looked at individual
administration intervals and drug sequence to determine
the clinical relevance of the interactions. Conclusion
Clinical pharmacists performed better than the decision
support system mainly because the system screened only
for high level DDIs and because of the low specificity of
the alerts. This specificity can be increased by including
contextual factors into the logic and by defining appropri-
ate screening intervals that take into account the sequence
in which the drugs are given.
Keywords Adverse drug events
Belgium
Clinical
decision support systems
Clinical pharmacy

Computerized physician order entry system
Drug–drug
interactions
Geriatrics
Medication safety
Impact of findings on practice
• Drug–drug interaction screening by clinical pharma-
cists may be useful even in the presence of a clinical
decision support system.
• Clinical decision support systems should not only
screen for high level drug–drug interactions.
• Contextual factors such as lab values should be
incorporated in the decision support rules.
123
 Int J Clin Pharm

• Appropriate screening intervals that also take into analysis of (1) ADEs related to DDIs and the ability of both
account the sequence in which the drugs are given approaches for detecting and resolving DDIs related to
should be defined for individual drug–drug interactions. ADEs; (2) factors relating to the occurrence of ADEs; and
(3) contextual factors used by the clinical pharmacists for
evaluation of the clinical relevance of DDIs.
Introduction
Ethical approval
Drug–drug interactions (DDIs) can lead to adverse drug
The study was approved by the UZ Brussel Medical Ethics
events (ADEs) and compromise patient safety [1, 2]. Dif-
Committee.
ferent approaches can be used to reduce unwanted DDIs in
hospital practice. One of these is the use of clinical deci-
sion support (CDS) systems [1, 3, 4]. At the moment of
Methods
prescribing, CDS embedded in a computerized physician
order entry (CPOE) system has the potential to reduce the
Setting
occurrence of DDIs [3, 5]. However, DDI checking with
CDS often produces many alerts with low clinical rele-
The UZ Brussel is a 721-bed university hospital in Brus-
vance resulting in low alert adherence caused by alert
sels, Belgium. The software system is self-developed and
fatigue [3, 4, 6–10]. Several studies report high alert
includes basic CDS such as DDI screening [28, 29]. CDS
overriding with up to 96 % of alerts being overridden [11,
for DDIs is provided as non-interruptive alerts during
12]. One way to reduce alert fatigue is by carefully
prescribing. The knowledge base used for DDI checking is
selecting which interaction alerts are presented to the
the commercially available DelphiCare database [30].
physicians and to decide for every DDI alert if it should be
The interactions are divided into 6 categories depending on
interruptive or not, based on severity level [12–14]. It is
their clinical relevance and seriousness. Currently, the
also important to increase the specificity of alerts by inte-
system screens only for level 1 and 2 interactions, the two
grating individual patient data to create patient-specific
most severe categories (contraindicated and precautionary
DDI alerts [15]. Furthermore, attention should be given to
contraindicated), in order to avoid alert fatigue. Because
human factors principles such as the sociotechnical aspect
the CDS system is self-developed, there are opportunities
of alerting and workflow issues [16, 17].
to improve DDI sensitivity and specificity.
Another approach is to let clinical pharmacists actively
On the geriatric ward, clinical pharmacists perform several
screen for DDIs and perform interventions for clinically
routine tasks such as medication reconciliation and screening
relevant DDIs [18–20]. Several studies report high accep-
for DRPs. We have a site license for Lexicomp and Stock-
tance rates for interventions by clinical pharmacists on
ley’s so clinical pharmacists can use these screening tools.
drug related problems (DRPs) in general and for inter-
They conduct an intervention for DDIs judged clinically
ventions on DDIs [18, 20–24]. However, the number of
relevant and work independently from the CDS system [31,
clinical pharmacists in hospitals is limited and active DDI
32]. During the study period, a senior year pharmacy student
screening is a very intensive and time-consuming activity.
(ŠS) documented which screening tool they used and which
Therefore, it is important to optimize CDS systems for DDI
patient characteristics they took into account to determine
screening saving time for the clinical pharmacists to spend
whether a DDI required an intervention or not.
on tasks like medication reconciliation and medication
review that can’t be performed completely with CDS [22,
Design and study population
25, 26]. Zaal et al. [27] recently observed that clinical
pharmacists identified many DRPs in addition to CDS and
This was a prospective single-centre, cohort study of patients
that the specificity of CDS is low.
admitted to the 29-bed acute geriatric department between
March 2nd 2011 and June 15th 2011. Patients were con-
secutively followed-up by the clinical pharmacists whenever
Aim of the study possible and excluded for the study if their follow-up was not
possible, if they died during hospitalisation, when they took
The primary aim of this study was to compare the routine no or only one drug and in case of rehospitalisation. An
performance of a CDS system and of clinical pharmacists independent research pharmacist (PC) compared the CDS
in reducing DDIs on a geriatric ward in order to learn from alerts with the DDIs identified by the clinical pharmacists
one approach to improve the other. Other objectives were a and evaluated their interventions. The patients’ electronic
comparison of the different interaction databases used and medical record (EMR) was screened by the pharmacy

123
Int J Clin Pharm
student under the supervision of the research pharmacist to
identify possible ADEs and hospital admissions potentially
related to DDIs. An ADE was defined according to the ICH
guidelines as ‘‘any untoward medical occurrence that may
present during treatment with a pharmaceutical product but
which does not necessarily have a causal relationship with
this treatment’’ [33]. The causality between DDIs and ADEs
was evaluated by a clinical pharmacologist based on the
causality assessment criteria of Edwards and Arondson [34].
He was given a short overview of the clinical cases and had
access to the patient’s EMR.
Additionally, the patients’ drug therapy was analysed
with the web-based DelphiCare application to identify
lower risk level DDIs that were not triggered by the CDS
but could have been triggered by it (level 3–5). The pur-
pose of this analysis was to investigate if screening on
lower level interactions could be useful and to identify
possible missing or additional DDIs when compared with
the results from the clinical pharmacists.
Statistical analysis
Descriptive and statistical analyses were carried out with
IBM SPSS Statistics version 22 (SPSS Inc., Chicago, IL).
In case of parametric data, the mean and standard deviation
were used. In case of non-parametric variables, the median
and range were used. Binary logistic regression analyses
were conducted to identify possible patient-related risk
factors associated with the presence of ADEs. The depen-
dent variable was the presence or absence of any ADE. The
independent variables analysed were age, gender, the res-
idential situation before admission (home or nursing
home), the length of hospital stay, the estimated glomerular
filtration rate category (eGFR, MDRD-IDMS), the number
of drugs, and the number of DDIs. Statistical analyses were
performed with a 0.05 significance level and 95 % confi-
dence intervals (CI) were calculated where relevant.
Results
Characteristics of the study population
Fifty patients, 17 males (34.0 %) and 33 (66.0 %) females,
were followed by the clinical pharmacists and included for
the comparison. Twenty patients were excluded because
follow-up was not possible (13 patients, mostly related to
lack of time or communication problems), follow-up started
at discharge (three patients), two were rehospitalisation’s,
one died and one took only one drug. The mean age was
84.6 ± 4.8 years, 24 patients (48.0 %) had an eGFR
C60 ml/min/1.73 m2 while 26 (52.0 %) had a lower eGFR.
The median length of hospital stay was 17 days (6–44) and
patients took a median of 8 drugs (3–15). The most frequent
reasons for admission were a decline in health status (12
patients), fall (ten patients) and dyspnea (four patients). The
majority of patients (36; 72.0 %) resided at home prior to
hospitalization and 14 patients (28.0 %) resided in a nursing
home. After discharge, 21 (42.0 %) patients returned home,
22 (44.0 %) were transferred to a nursing home and 7
(14.0 %) were transferred to a revalidation centre.
Identified DDIs
A total number of 240 DDIs with a median of 3.5 (range
0–18) per patient were identified by the clinical pharma-
cists, 239 using Lexicomp and 1 additional interaction
using Stockley’s. An overview of the DDI classification
levels of the different databases is provided in Table 1. The
clinical pharmacists advised a therapy change for 16 DDIs
of which 13 (81.2 %) were accepted and 3 (18.8 %) were
not. For 8 other DDIs the advice was to monitor the patient
closely. We were not able to determine if the physician
followed this advice. The CDS system generated only six
alerts of which none were accepted by the physicians. Two
of these alerts were for DDIs that were not identified by the
clinical pharmacists. The four other DDIs were also iden-
tified by the clinical pharmacists and they conducted two
interventions of which one was accepted. Analysis by the
research pharmacist of the patients’ drug therapy with the
web-based DelphiCare application resulted in the
Table 1 DDI levels of DelphiCare, Lexicomp, and Stockley’s
DelphiCare [30]
1 Contraindicated
2 Precautionary contraindicated
3 Monitoring or therapy adjustment needed
4 In some cases monitoring or therapy adjustment is
needed
5 Precautionary monitoring
6 No action needed
Lexicomp [31]
X Avoid combination
D Consider therapy modification
C Monitor therapy
B No action needed
A No known interaction
Stockley’s [32]
Avoid Contraindicated drug pair, avoid drug combination
Adjust It is recommended that either one of the drugs is
changed, or the dose is altered on initiating the
combination
Monitor Monitor
Informative Informative
No action No action
123
 Int J Clin Pharm

Table 2 Number of identified DDIs by the clinical pharmacists, CDS potassium level for 13 DDIs and glycemia for ten DDIs.
system and with the web-based DelphiCare application divided into For the four DDIs identified by both the CDS system and
action levels
the clinical pharmacists, only two interventions were con-
Screening by Database used Level Number DDIs ducted because for the other two DDIs an intervention was
judged unnecessary based on available lab values. Another
Clinical pharmacists Lexicomp X 5
factor was the duration of the screening interval and the
D 40
sequence in which the drugs were given. The CDS system
C 192
screened with a fixed interval going back 3 days in the
Ba 2
medication history and 2 days ahead in the medication
Stockley’s Informative 1
planning when a prescription was made. The clinical
CDS system Delphicare 1 3
pharmacists took into account a specific relevant screening
2 3
interval for each individual DDI and the sequence in which
Research pharmacist DelphiCare 1 1b
the drugs were given was also taken into account.
2 5c
3 53
4 4 Identified ADEs and causality assessment
5 47
a
A total of 37 ADEs were identified for 29 patients that could
The clinical pharmacists only reported type B DDIs in case an ADE
occurred that could be related to the DDI
be related to 55 DDIs. Often, several DDIs could be related
b
Two level 1 DDIs were initially not identified by the research
to the same ADE. Ten ADEs occurred already before hos-
pharmacist because his analysis was based on the drug therapy doc- pital admission and 27 during hospitalization. For five of the
umented by the clinical pharmacist and for these two DDIs one of the ten ADEs that occurred at home, the potentially responsible
interacting drugs was prescribed one-time only and the clinical DDI(s) was/were still present during hospitalization, for the
pharmacists did not capture these prescriptions
c
other 5 the drug therapy was adjusted at admission. For 10
Two level 2 DDIs were not triggered by the CDS because of the
(20.0 %) patients, the main reason for hospital admission
system settings
was an ADE that could be related to one or more DDIs. Of
these ten patients, seven resided at home and three in a
identification of 110 DDIs. Of these 110 DDIs, 105 were
nursing home prior to hospitalisation. An overview of the
also identified by the clinical pharmacists with Lexicomp
types of ADEs is provided in Table 3. Based on the clinical
and 5 were exclusively included in DelphiCare. Of the
effect of the eight DDIs related to falls, five could be related
240 DDIs identified by the clinical pharmacists, 135 were
to CNS depression, two to orthostatic hypotension and one
not included in DelphiCare. The numbers of identified
to bradycardia.
DDIs with the different approaches broken down according
Of the 55 DDIs that could be related to an ADE, 18 were
to action level are summarized in Table 2. In general, there
triggered by both Lexicomp and Delphicare, 36 were
was similarity between the severity classification of DDIs
triggered only by Lexicomp and 1 DDI was triggered only
identified with both Lexicomp and Delphicare although
by Stockley’s. Analysis of the action levels of these DDIs
there were some remarkable differences. The type X DDIs
indicates that the 18 mutual DDIs were level 3 (5), level 4
in Lexicomp were not classified as level one but as level
(1) and level 5 (12) in Delphicare and type D (3) and C
two in Delphi. DDIs classified as type D in Lexicomp
(15) in Lexicomp. The 36 DDIs triggered only by Lexi-
were classified as level 1, 2, 3 or 5 in Delphicare and one
comp consisted of type D (2) and type C (34) DDIs. The
type B Lexicomp interaction was classified as level 3 in
DDI triggered by Stockley’s was categorized as ‘‘infor-
Delphicare. The action levels of the DDIs that were only
mative’’. None of the DDIs related to an ADE were trig-
included in one of the databases were analysed. There were
gered by the CDS system.
3 type X, 14 type D and 117 type C DDIs that were only
A clinical pharmacologist evaluated the causality
included in Lexicomp. One level 2, 1 level 3 and 3 level 5
between the 37 ADEs and 55 DDIs. For 2 DDIs the cau-
DDIs were only included in Delphicare and 1 informative
sality was evaluated as certain, for 31 as likely, for 10 as
DDI was only included in Stockley’s. This indicates that
possible and for 12 as unlikely.
also high level DDIs like type X, D (Lexicomp) and level
2 (Delphicare) were not included in all databases.
Factors associated with the occurrence of ADEs
Contextual factors for evaluation of DDIs
The univariate binary logistic regression analyses identified
Blood pressure was taken into account for 52 DDIs, the the number of DDIs (OR 1.33; 95 % CI 1.07–1.66;
INR value for 27 DDIs, heart rate for 16 DDIs, the p = 0.01) as a significant risk factor for the presence of one

123
Int J Clin Pharm

Table 3 Overview of the types of identified ADEs Discussion

Type of ADE Number
This study showed that clinical pharmacists (16 interven-
Hypotension 8 tions advising therapy change, 13 accepted) performed
Fall 8 better than the CDS system (six alerts, none accepted) in
Bradycardia 3 reducing DDIs on a geriatric ward. Several reasons may
High INR 2 explain this finding.
Symptoms of serotonin syndrome 2 A first reason relates to the extensiveness of the data-
Rigidity, extrapyramidal symptoms 2 base’s content and the set alert trigger level. In general, we
Aggression and confusion 2 found similarity between the severity classifications of
Gastric irritation and/or cramps 1 DDIs included in both databases (Lexicomp and Del-
Hypercalcemia 1 phicare) but several DDIs, including high risk level DDIs,
Orthostasis 1 were not included in both databases. More importantly, of
Symptoms of CNS depression 1 the 55 DDIs that could be related to an ADE, 36 (65.5 %)
Hyperglycemia 1 were not included in the database on which the CDS is
Malaise due to hyponatremia 1 based. This indicates that additional DDIs should be
Hypoglycemia 1 included in the CDS knowledge base. As a start, critical
High blood pressure 1 DDIs should be included [1, 12]. It is however important
Bleeding 1 not to focus on high risk level DDIs only given the finding
Loss of consciousness 1 that mainly intermediate risk level DDIs could be related to
Total 37 ADEs. Horn et al. [5] stated that showing only the most
severe alerts is the most common knowledge base cus-
tomization but this approach can lead to patient harm.
Therefore, the CDS system should be set to screen also for
lower risk level DDIs. However, adding more DDIs will
not solve the problem of alert overriding. There were only
Table 4 Univariate binary logistic regression analysis of possible
patient-related risk factors for ADEs 6 CDS alerts and all were overridden; hence, this is prob-
ably not resulting from alert fatigue but rather a problem of
Characteristics Odds ratio (95 % CI) p
low alert specificity which is a second reason for the poor
Age 1.00 (0.89–1.12) 0.98 performance of the CDS. To determine whether a DDI
Length of hospital stay 1.07 (0.98–1.16) 0.11 required an intervention or not, the clinical pharmacists
eGFR category often took into account individual patient characteristics.
C60 ml/min/1.73 m2 a – – This is an important factor to increase alert specificity. As
<60 ml/min/1.73 m2 0.53 (0.17–1.65) 0.27 indicated by Duke et al. [9], we believe that just showing
Sex the relevant lab values or patient characteristics together
Malea – – with an alert will probably not reduce alert overriding as
Female 1.37 (0.42–4.46) 0.60 the prescribers can already easily lookup these parameters
Residential situation before admission in the EMR. Therefore, lab values and other characteristics
Homea – – should be included in the logic of the interaction rules so
Nursing home 1.44 (0.40–5.16) 0.58
that a patient specific advice can be generated. The severity
No. of drugs 1.12 (0.91–1.39) 0.28
of the alert and the required action can then depend on the
No. of DDIs 1.33 (1.07–1.66) 0.01
context in which the DDI occurs. Another factor to increase
alert specificity is to adjust the duration of the screening
a
Reference category interval in function of individual DDIs and to take into
eGFR estimated glomerular filtration rate account the sequence in which the drugs are given. We
only found one study that mentions this type of custom-
ization, although no results are provided regarding the
or more ADEs (Table 4). For every additional DDI, the effect on alert specificity and overriding [35].
odds of experiencing one or more ADEs increased by The scarce provision of solutions together with the DDI
33 %. Age, gender, the residential situation before admis- alerts probably also contributed to the low performance of
sion (home or nursing home), the length of hospital stay, the CDS system. The system provided suggestions for
the eGFR category and the number of drugs had no sig- solutions for some DDIs but the information was often not
nificant effect. specific. For example, advice for specific combinations of

123

drugs with risk for QT-prolongation was not provided. The
advice was of a more general nature reflecting towards the
whole group of QT-prolonging drugs. The clinical phar-
macists on the other hand, provided a specific solution for
the DDIs. Providing solutions is indicated as a success
factor in other studies [36]. A last reason for the low per-
formance of the CDS system may be the difference
between the human/computer and human/human interface.
The personal intervention of a clinical pharmacist has
probably more effect than an alert on a computer screen.
The clinical pharmacists’ approach also had important
limitations. The CDS system identified two additional
DDIs that were not identified by the clinical pharmacists
because one of the interacting drugs was prescribed one-
time only and the clinical pharmacists did not screen
continuously for DDIs. Another disadvantage of the clini-
cal pharmacists’ approach was that the DDIs were identi-
fied after they had already occurred in practice while the
CDS system had the potential to prevent DDIs from
occurring during hospitalisation.
The binary logistic regression analyses showed that the
number of drugs had no significant effect on the occurrence
of ADEs while the number of DDIs did. It may therefore be
advisable not to select geriatric patients at higher risk for
ADEs caused by DDIs merely based on the number of
drugs they are taking but rather on the number of DDIs, in
particular specific types of DDIs.
Our study has several limitations. It was performed at
one hospital ward with a limited number of patients in a
single hospital. The geriatric ward was chosen because of
the expected high prevalence of DDIs and because clinical
pharmacy services were already implemented for several
years [22, 25]. Physicians could have been biased for DDIs
first detected by the CDS for which the clinical pharmacists
conducted an intervention afterwards, resulting in a higher
acceptance rate of clinical pharmacists’ interventions.
However, the low number of CDS alerts for DDIs identi-
fied by both approaches makes this unlikely. The fact that
different DDI knowledge bases were used by the clinical
pharmacist and the CDS system can be seen as a bias for
the comparison but we wanted to compare the everyday
situation, and this allowed us to identify shortcomings in
the current CDS database for DDIs.
Conclusion
We conclude that clinical pharmacists performed better
than the CDS system in identifying and solving DDIs on a
geriatric ward mainly because the CDS system screened
only for high risk level DDIs and because of the low
specificity of the CDS alerts. A CDS system should also
trigger lower risk level DDIs because we found that mainly
123
Int J Clin Pharm
intermediate level DDIs could be related to ADEs.
Increasing alert specificity by including contextual factors
into the CDS logic, adjusting the screening interval in
function of individual DDIs and taking into account the
drug sequence is also of utmost importance. The database
of a CDS system should be compared with other available
databases and inclusion of additional DDIs may be
necessary.
Acknowledgements We thank Claudine Ligneel, Tinne Leysen,
Eva De Baere and Hilde De Ridder for their support as clinical
pharmacists.
Funding The study was performed with the support of the Agency
for Innovation by Science and Technology in Flanders, Belgium,
which provided a research grant for the first author. The funding
agreement ensured the authors’ independence in designing the study,
interpreting the data, writing, and publishing the report.
Conflicts of interest The authors have no conflicts of interest
regarding this study.
References
1. Classen DC, Phansalkar S, Bates DW. Critical drug–drug inter-
actions for use in electronic health records systems with com-
puterized physician order entry: review of leading approaches.
J Patient Saf. 2011;7(2):61–5.
2. Espinosa-Bosch M, Santos-Ramos B, Gil-Navarro MV, Santos-
Rubio MD, Marin-Gil R, Villacorta-Linaza P. Prevalence of drug
interactions in hospital healthcare. Int J Clin Pharm.
2012;34(6):807–17.
3. Kuperman GJ, Bobb A, Payne TH, Avery AJ, Gandhi TK, Burns
G, et al. Medication-related clinical decision support in com-
puterized provider order entry systems: a review. J Am Med
Inform Assoc. 2007;14(1):29–40.
4. Osheroff JA. Improving medication use and outcomes with
clinical decision support: a step-by-step guide. Chicago: The
Healthcare Information and Management Systems Society; 2009.
5. Horn JR, Gumpper KF, Hardy JC, McDonnell PJ, Phansalkar S,
Reilly C. Clinical decision support for drug–drug interactions:
improvement needed. Am J Health Syst Pharm.
2013;70(10):905–9.
6. van der Sijs H, Aarts J, Vulto A, Berg M. Overriding of drug
safety alerts in computerized physician order entry. J Am Med
Inform Assoc. 2006;13(2):138–47.
7. van der Sijs H, Mulder A, van Gelder T, Aarts J, Berg M, Vulto
A. Drug safety alert generation and overriding in a large Dutch
university medical centre. Pharmacoepidemiol Drug Saf.
2009;18(10):941–7.
8. van Zwart Rijkom JE, Uijtendaal UV, ten Berg MJ MJ, van
Solinge WW, Egberts AC. Frequency and nature of drug–drug
interactions in a Dutch university hospital. Br J Clin Pharmacol.
2009;68(2):187–93.
9. Duke JD, Li X, Dexter P. Adherence to drug–drug interaction
alerts in high-risk patients: a trial of context-enhanced alerting.
J Am Med Inform Assoc. 2013;20(3):494–8.
10. Payne TH, Nichol WP, Hoey P, Savarino J. Characteristics and
override rates of order checks in a practitioner order entry system.
Proc AMIA Symp. 2002; 602–6.
11. Shah NR, Seger AC, Seger DL, Fiskio JM, Kuperman GJ, Blu-
menfeld B, et al. Improving acceptance of computerized
Int J Clin Pharm
prescribing alerts in ambulatory care. J Am Med Inform Assoc.
2006;13(1):5–11.
12. Phansalkar S, Desai AA, Bell D, Yoshida E, Doole J, Czochanski
M, et al. High-priority drug–drug interactions for use in electronic
health records. J Am Med Inform Assoc. 2012;19(5):735–43.
13. Paterno MD, Maviglia SM, Gorman PN, Seger DL, Yoshida E,
Seger AC, et al. Tiering drug–drug interaction alerts by severity
increases compliance rates. J Am Med Inform Assoc.
2009;16(1):40–6.
14. Phansalkar S, van der Sijs H, Tucker AD, Desai AA, Bell DS,
Teich JM, et al. Drug–drug interactions that should be non-
interruptive in order to reduce alert fatigue in electronic health
records. J Am Med Inform Assoc. 2013;20(3):489–93.
15. Duke JD, Bolchini D. A successful model and visual design for
creating context-aware drug–drug interaction alerts. AMIA Annu
Symp Proc. 2011;2011:339–48.
16. Berner ES. Clinical decision support systems: state of the Art.
Rockville, Maryland: agency for healthcare research and quality,
U.S. Department of Health and Human Services; 2009.
17. Phansalkar S, Edworthy J, Hellier E, Seger DL, Schedlbauer A,
Avery AJ, et al. A review of human factors principles for the
design and implementation of medication safety alerts in clinical
information systems. J Am Med Inform Assoc. 2010;17(5):
493–501.
18. Kaboli PJ, Hoth AB, McClimon BJ, Schnipper JL. Clinical
pharmacists and inpatient medical care: a systematic review.
Arch Intern Med. 2006;166(9):955–64.
19. Bosma L, Jansman FG, Franken AM, Harting JW, Van den Bemt
PM. Evaluation of pharmacist clinical interventions in a Dutch
hospital setting. Pharm World Sci. 2008;30(1):31–8.
20. Lea M, Rognan SE, Koristovic R, Wyller TB, Molden E. Severity
and management of drug–drug interactions in acute geriatric
patients. Drug Aging. 2013;30(9):721–7.
21. Kucukarslan SN, Peters M, Mlynarek M, Nafziger DA. Phar-
macists on rounding teams reduce preventable adverse drug
events in hospital general medicine units. Arch Intern Med.
2003;163(17):2014–8.
22. Cornu P, Steurbaut S, Leysen T, De Baere E, Ligneel C, Mets T,
et al. Effect of medication reconciliation at hospital admission on
medication discrepancies during hospitalization and at discharge
for geriatric patients. Ann Pharmacother. 2012;46(4):484–94.
23. Khalili H, Karimzadeh I, Mirzabeigi P, Dashti-Khavidaki S.
Evaluation of clinical pharmacist’s interventions in an infectious
diseases ward and impact on patient’s direct medication cost. Eur
J Intern Med. 2013;24(3):227–33.
24. Hasan SS, Lim KN, Anwar M, Sathvik BS, Ahmadi K, Yuan
AW, et al. Impact of pharmacists’ intervention on identification
and management of drug–drug interactions in an intensive care
setting. Singap Med J. 2012;53(8):526–31.
25. Steurbaut S, Leemans L, Leysen T, De Baere E, Cornu P, Mets T,
et al. Medication history reconciliation by clinical pharmacists in
elderly inpatients admitted from home or a nursing home. Ann
Pharmacother. 2010;44(10):1596–603.
26. Cornu P, Steurbaut S, Leysen T, De Baere E, Ligneel C, Mets T,
et al. Discrepancies in medication information for the primary
care physician and the geriatric patient at discharge. Ann Phar-
macother. 2012;46(7–8):983–90.
27. Zaal RJ, Jansen MM, Duisenberg-van Essenberg M, Tijssen CC,
Roukema JA, van den Bemt PM. Identification of drug-related
problems by a clinical pharmacist in addition to computerized
alerts. Int J Clin Pharm. 2013.
28. Van de Velde R. Framework for a clinical information system. Int
J Med Inform. 2000;57(1):57–72.
29. Lanssiers R, Everaert E, De Win M, Van De Velde R, De Clercq
H. An integrated drug prescription and distribution system:
challenges and opportunities. Stud Health Technol Inform.
2002;93:69–74.
30. APB DelphiCare 2013; http://delphicare.apb.be/DutchHTML/
inbfLegende.html; 17 Oct 2013.
31. UpToDate, Lexi-Comp Online: Lexi-Interact, 2013; http://www.
uptodate.com/crlsql/interact/frameset.jsp; 17 Oct 2013.
32. Press TP, Stockley’s Interaction Alerts, 2013; http://www.medi
cinescomplete.com/mc/alerts/current/actions.htm; 17 Oct 2013.
33. Nebeker JR, Barach P, Samore MH. Clarifying adverse drug
events: a clinician’s guide to terminology, documentation, and
reporting. Ann Intern Med. 2004;140(10):795–801.
34. Edwards IR, Aronson JK. Adverse drug reactions: definitions,
diagnosis, and management. Lancet. 2000;356(9237):1255–9.
35. Resetar E, Reichley RM, Noirot LA, Dunagan WC, Bailey TC.
Customizing a commercial rule base for detecting drug–drug
interactions. AMIA Annu Symp Proc. 2005;1094.
36. Mollon B, Chong JJ, Holbrook AM, Sung M, Thabane L, Foster
G. Features predicting the success of computerized decision
support for prescribing: a systematic review of randomized
controlled trials. BMC Med Inform Decis Mak. 2009;9:11.
123