case study  Intravenous infusion pumps

case study

Quality-improvement analytics for intravenous

infusion pumps
Susan J. Skledar, Cynthia S. Niccolai, Dennis Schilling, Susan Costello,
Nicolette Mininni, Kelly Ervin, and Alana Urban
An audio interview that supplements the
information in this article is available on Purpose. The implementation of a smart- on the collected CQI data and site-specific
AJHP’s website at www.ajhp.org/site/misc/ pump continuous quality-improvement requests, four systemwide updates of the
podcasts.xhtml. (CQI) program across a large health system smart-pump drug library were performed
is described, with an emphasis on key met- during the first 18 months of the program,
rics for outcomes analyses and program reducing “nuisance alerts” by about 10%

M
ultiple reports in the literature
have identified the continued
high frequency and degree of
patient harm associated with i.v. ad-
ministration of medications relative
to other administration routes. The
lack of i.v. medication administra-
tion standards within an institution
not only contributes to the propaga-
tion of errors but can hinder the i.v.
administration process. The Institute
of Medicine has reported that many
medications associated with a high
risk for error are delivered via i.v.
infusion.1
The administration step of the i.v.
medication process is prone to errors
and adverse events, with i.v. medica-
tion error rates exceeding 30%.2,3 The
United States Pharmacopeial Con-
refinement.
Summary. Three years ago, the University
of Pittsburgh Medical Center health system
launched a CQI initiative to help ensure
the safe use of 6000 smart pumps in its
14 inpatient facilities. A centralized team
led by pharmacists is responsible for the
retrieval and interpretation of smart-pump
data, which is continuously transmitted to
a main server. CQI findings are regularly
posted on the health system’s interdisci-
plinary intranet. Monitored metrics include
rates of compliance with preprogrammed
infusion limits, the top 20 drugs involved in
alerts, drugs associated with alert-override
rates of ≥90%, numbers of alerts by infu-
sion type, nurse responses to alerts, and
alert rate per drug library update. Based
vention’s Center for the Advance-
ment of Patient Safety has reported
per update cycle and enabling targeted
interventions to reduce rapid-infusion
errors, other adverse drug events (ADEs),
and pump-programming workarounds.
Over one 12-month period, bedside alerts
prompted nurses to reprogram or cancel
continuous infusions an average of 400
times per month, potentially averting i.v.
medication ADEs.
Conclusion. A smart-pump CQI program is
an effective tool for enhancing the safety of
i.v. medication administration. The ongo-
ing refinement of the drug library through
the development and implementation of
key interventions promotes the growth
and sustainability of the smart-pump initia-
tive systemwide.
Am J Health-Syst Pharm. 2013; 70:680-6
that errors originating during the ad-
ministration of i.v. medications lead

Susan J. Skledar, B.S., M.P.H., FASHP, is Director,
Pharmacy Drug Use and Disease State Management
Program, Department of Pharmacy, University of Pittsburgh Medi-
cal Center (UPMC), and Associate Professor, School of Pharmacy,
University of Pittsburgh, Pittsburgh, PA. Cynthia S. Niccolai, B.S.,
Pharm.D., is Clinical Pharmacist, Pharmacy Drug Use and Disease
State Management Program, Department of Pharmacy, UPMC.
Dennis Schilling, Pharm.D., is Pharmacy Manager, Department
of Pharmacy, UPMC Shadyside, Pittsburgh, PA. Susan Costello,
M.S.N., RN, is Advanced Practice Nurse, Nursing Education and
Research, UPMC Presbyterian, Pittsburgh, PA. Nicolette Mininni,
M.Ed., RN, CCRN, is Advanced Practice Nurse, Critical Care, Clini-
cal Administration, UPMC Shadyside. Kelly Ervin, B.S., CPhT, is
Data Analyst and Pharmacy Technician, Pharmacy Drug Use and
Disease State Management Program, Department of Pharmacy,
UPMC. Alana Urban, M.S.N., RN, CCRN, is Advanced Practice
Nurse and Infusion Safety Practitioner, UPMC.
Address correspondence to Dr. Niccolai at the Department of Phar-
macy, University of Pittsburgh Medical Center, UPMC Presbyterian
PFG 01-01-01, 200 Lothrop Street, Pittsburgh, PA 15213 (niccolaics@
upmc.edu).
At the time of writing, Ms. Urban was serving as an intermediary
between the UPMC Donald D. Wolff, Jr. Center for Quality, Safety,
and Innovation and CareFusion Corporation, San Diego, CA, for
educational initiatives on safe infusion practices. The other authors
have declared no potential conflicts of interest.
Copyright © 2013, American Society of Health-System Pharma-
cists, Inc. All rights reserved. 1079-2082/13/0402-0680$06.00.
DOI 10.2146/ajhp120104

680 Am J Health-Syst Pharm—Vol 70 Apr 15, 2013


to the most serious adverse patient
outcomes.4 Data compiled from 850
hospitals nationally and presented at
a 2008 i.v. medication safety summit
convened by the American Society
of Health-System Pharmacists and
other organizations indicated that
parenteral medication errors are
three times as likely as other reported
errors to cause harm or death, with
58% of such errors occurring during
medication administration and 79%
of harmful errors involving the i.v.
route.5
Background
Over the past decade, an increas-
ing number of institutions have
adopted smart-pump technology to
reduce errors and adverse events as-
sociated with i.v. infusions. Integrat-
ing decision-support software and a
customized drug library, this tech-
nology enables the nurse to receive
and respond to smart-pump alerts
generated at the patient’s bedside,
potentially intercepting harmful i.v.
infusion errors. Infusion program-
ming data, including alerts generated
and nurse responses, are stored on
the respective pump and (if wireless
technology is available) on a central
server. The regularly scheduled re-
trieval and analysis of these data and
distribution of the findings are key to
driving the growth and success of a
smart-pump program. The absence
of consistent feedback and subse-
quent interventions and education
can dilute the safety-enhancing ben-
efits of the smart-pump device due to
diminishing program commitment
and the eventual introduction of un-
checked, potentially harmful pump
programming “workarounds.”
Problem
Although many articles have been
published in the nursing and phar-
macy literature detailing the plan-
ning and implementation of smart-
pump devices, few have included
details about the experience of
developing and maintaining a formal
case study  Intravenous infusion pumps
smart-pump continuous quality-
improvement (CQI) program. 6-13
Even less information is available de-
scribing the unique challenges health
systems encounter when attempt-
ing to establish a smart-pump CQI
initiative, such as the inclusion of
numerous and diverse practice sites
(e.g., both urban and community
acute care hospitals); the location of
system facilities across an expanded
geographic area; variations in facility
leadership, policies, and educational
programs; and the use of multiple
smart-pump drug libraries.
In 2007, smart-pump technology
(Alaris System, CareFusion Corpora-
tion, San Diego, CA) was implement-
ed at the University of Pittsburgh
Medical Center (UPMC) to assist our
efforts in reducing i.v. medication in-
fusion errors and associated adverse
events. The UPMC health system
encompasses 20 hospitals, including
international sites, over 400 doctor’s
offices and clinics, health insurance
services (UPMC Health Plan), and a
commercial division and maintains a
strong affiliation with the University
of Pittsburgh Schools of the Health
Sciences.
The smart-pump technology is
supported by safety software (Guard-
rails Suite MX, CareFusion Corpo-
ration), enabling the construction
and maintenance of a customized
i.v. drug library and parallel CQI
program. The smart pump gener-
ates an alert if an i.v. infusion is
programmed outside of the health
system’s approved administration
parameters, or “Guardrail (GDR)
limits.” The drug library is populated
with drug-specific soft limits and,
when deemed necessary, hard limits
for continuous i.v. infusion dosages
and rates, intermittent infusion total
doses and durations, solution and
drug concentrations, and fluid infu-
sion rates. The bedside nurse has the
option to override soft GDR alerts
but must reprogram or cancel the
infusion when a hard GDR alert is
encountered. Hard GDR limits are
Am J Health-Syst Pharm—Vol 70 Apr 15, 2013
added for high-alert drug library
entries including insulin, opioids,
micronutrients, and anticoagulants.
Drug-specific advisories are created,
when needed, to provide the bedside
nurse with additional administra-
tion safety information, including
filtering requirements, infusion rate
warnings, potential adverse effects,
and incompatibilities.
Within Guardrails Suite MX,
“therapy tabs” are used to distinguish
multiple indications and different
dosing regimens for a single drug
entry. For example, the library entry
for bivalirudin differentiates distinct
dosing regimens with therapy tabs
denoting common indications (e.g.,
heparin-induced thrombocytopenia,
interventional cardiology). To sim-
plify drug searches within the smart-
pump monitor’s 1500-line scrolling
library screen, customized “care area
profiles” subcategorize the numerous
drug entries by the type and level of
care provided (e.g., “Adult Critical
Care,” “Adult-MOST” [medical-
oncology-surgery-telemetry], “Adult-
OB” [obstetrics]). The UPMC drug
library, initially developed collab-
oratively by pharmacists, nurses, and
physicians, is currently maintained
by pharmacists.
Analysis and resolution
Smart-pump technology was
initiated throughout UPMC’s inpa-
tient system facilities by December
2008. Calendar year 2009 allowed for
integration of a patient-controlled
analgesia device, expansion to our
outpatient cancer centers, and tar-
geted nurse education. To evaluate
the impact of this technology on
i.v. infusion safety, a comprehen-
sive systemwide CQI program was
implemented in January 2010 to
assist UPMC facilities in the growth
and success of their respective smart-
pump initiatives.
The smart-pump CQI software
facilitates the ongoing analysis of i.v.
infusion programming practices at
both the facility level and the health-
681
 case study  Intravenous infusion pumps
system level. Infusion programming
data, as well as alerts and nurse
responses generated from approxi-
mately 255,000 infusion starts per
month systemwide, are wirelessly
transferred from over 6,000 pumps
to a main server. These data are
instrumental in identifying GDR
compliance rates, alert trends, nurse
response patterns, drug library revi-
sions, discrepancies between clinical
practice and GDR limits, and pump-
programming workarounds. When
retrieved and reviewed at regular in-
tervals (e.g., monthly, quarterly), the
CQI data greatly assist in optimiz-
ing the safety benefits of the smart
pumps and maintaining the clinical
applicability of the drug library.
CQI program development. Iden-
tifying the need for data analytics
and the estimated time and person-
nel required to provide this ongo-
ing service is essential in the early
planning of a smart-pump initiative.
The assignment of this responsibil-
ity to a group or individual well in
advance of the go-live date provides
the opportunity for valuable CQI
software training. The CQI soft-
ware package provides multiple
user-friendly templates to describe
the smart-pump data; however, the
ability to create customized reports
and graphics is necessary to uncover
specific infusion programming issues
and trends. Due to the size of our
smart-pump initiative, the estimated
time required to abstract and review
the pertinent CQI data, report the
findings, and propose interventions
was substantial. By systemwide con-
sensus, CQI data analytics and main-
tenance of the drug library became
the responsibilities of a centralized
team of two clinical pharmacists and
a pharmacy technician with expertise
in quality-data analysis. The ability to
discern and abstract the most useful
information from the database, for
a single facility or the entire system,
gradually evolved as the team became
more familiar with the types of infu-
sion data collected, the GDR soft-
682 Am J Health-Syst Pharm—Vol 70 Apr 15, 2013
ware, the process of programming an
infusion and subsequent GDR alerts
generated, and facility-specific nurs-
ing infusion practices. The appendix
provides key questions to guide CQI
data analysis.
In addition to the scheduled
analysis of the CQI data, members of
the smart-pump CQI team often re-
ceive requests to abstract and review
smart-pump CQI data to recreate the
alert sequence for a reported infusion
adverse drug event (ADE). The team
also reviews event logs downloaded
from smart pumps sequestered due
to a reported malfunction or in-
volvement in an ADE. These services
have assisted with the root-cause
analysis of infusion-related ADEs
and prompted critical GDR library
changes, potentially resulting in the
interception of future events.
After the assignment of responsi-
bility for smart-pump data retrieval
and analysis, the next step is to de-
sign the CQI program components
and process. At UPMC, systemwide
participation and consensus were ob-
tained for the selection and report-
ing of meaningful quality metrics
thought to most accurately reflect
the use of GDR software and the sub-
sequent impact on patient safety. In
addition, the content and format of
the CQI report received system-level
approval, as did the team’s selection
of an appropriate medium to facili-
tate systemwide report distribution,
subsequent discussions, proposed
interventions, and feedback.
Initial CQI metrics included
facility-specific and systemwide
GDR compliance rates, defined as
the percentage of all infusion starts
programmed using either the “GDR
Drugs” or the “GDR Fluids” library
option; this metric is critical to pro-
gram success since the bedside nurse
can select an alternative program-
ming option, “Basic Infusion,” by-
passing the drug library GDR limits.
Additional metrics are GDR compli-
ance by clinical care area (useful in
pinpointing educational needs), the
top drugs and infusion types (con-
tinuous, bolus, intermittent, fluid)
generating alerts, actions taken by
the bedside nurse (override alert or
cancel and reprogram infusion), and
drugs associated with an alert over-
ride rate of ≥90% (useful in identify-
ing patterns of alert fatigue).
The actual CQI report is a compos-
ite of customized reports and selected
graphics templates describing the ap-
proved metrics for each facility and
the health-system aggregate. Informa-
tion is provided in both a PowerPoint
(Microsoft Corporation, Redmond,
WA) format for presentation at the
facility level and as Office Excel
(Microsoft Corporation) dashboards
for ease of facility comparisons.
Fourteen individual CQI reports
(13 facility-specific reports and 1
aggregated health-system report)
are posted quarterly to the inter-
disciplinary intranet site, which
is maintained using SharePoint
(Microsoft Corporation), with
monthly dashboard reports posted
in the interim. SharePoint also facili-
tates systemwide discussion of the
CQI data and provides a mecha-
nism for personnel at each facility
to evaluate the information, vote
on proposed interventions, and
provide suggestions and comments.
To complete the communication
loop, the CQI reports are distributed,
as appropriate, throughout each
facility and presented, for local feed-
back, at staff meetings and meetings
of medication safety committees,
pharmacy and therapeutics commit-
tees, and quality and patient safety
councils.
Implementation of approved
interventions. Translating findings
into actual practice is the final step
of the smart-pump CQI process and
is achieved through the approval
and release of the revised drug li-
brary and, when needed, targeted
educational initiatives. Systemwide
consensus is facilitated through the
posting of all proposed library revi-
sions to our smart-pump SharePoint

site. Using this site, designated phar-
macy and nursing clinicians from
each of the health system’s 13 facili-
ties have the opportunity to vote and
offer comments on library revisions.
Also, clinical pharmacists have an ad-
ditional opportunity for further dis-
cussion of smart-pump CQI issues
during our monthly Health System
Guideline Implementation Work-
group conference call (this group
also guides health-system formulary,
medication safety, and electronic
health record initiatives).
During the multifacility imple-
mentation phase, due to numerous
CQI findings and requests, wire-
less drug library updates occurred
quarterly; however, with all facilities
actively engaged in the smart-pump
CQI initiative since 2010, semiannual
updates have been deemed adequate
to maintain the drug library and are
preferred by nursing staff. A period
of up to 10–21 days after the release
of a revised drug library may be re-
quired for the wireless transfer of the
updates to all of the system’s 6000
smart pumps; therefore, the review
of CQI data reflecting the drug li-
brary revisions does not occur until
45 days from the release date.
Program experience. During the
past four years, the use of smart-
pump technology has continued
to grow within our health sys-
tem to include additional inpatient
sites, outpatient oncology clinics,
neonatal intensive care units, and
general pediatrics care areas. This
expansion has greatly increased
the volume of i.v. infusions pro-
grammed and associated alerts. To
describe our experience with the
CQI program, data representing the
13 inpatient facilities participating
in the initial implementation phase
(excluding data from a children’s
specialty hospital) were retrieved
for the following time frames: July–
December 2009, January–June 2010,
and January–June 2011.
Smart-pump drug library updates.
As previously described, drug library
case study  Intravenous infusion pumps
updates occur twice a year or when
an issue warranting an immediate
revision occurs. The smart-pump
CQI data are not the sole driver of the
drug library revision process but are
used in conjunction with reported i.v.
administration ADEs to identify ad-
ditions or modifications that can en-
hance safety and potentially intercept
future events. Other key impetuses for
library updates include recent litera-
ture reports, evolving practices, and
facility-specific requests.
In the first 18 months of the
smart-pump CQI initiative, we
released four systemwide library
updates, including 607 revisions
(Table 1). A systemwide initiative to
standardize continuous infusions
for concentration, volume, dose, and
infusion rate was implemented with
the May 2011 drug library update.
The success of this initiative was es-
sential due to the previous decision
to build one systemwide smart-pump
drug library. The approval of stan-
dardized concentrations enabled the
removal of several drug library “wild-
card” entries (e.g., ___mg/___mL) for
Table 1.
Summary of University of Pittsburgh Medical Center Health
System Smart-Pump Drug Library Revisions, 2009–11a
Description
Dosing limit
Duration limit
Concentration limit
Bolus limit
Addition of standard admixture
Removal of nonstandard admixture
Addition of drug library wildcardc
Removal of drug library wildcardc
Therapy addition
Advisory addition
Care area profileb (added or removed)
Drug library entry (added or removed)
Drug entry format
Pump configuration setting
Total
a
Cumulative data reflecting drug library updates in January 2009, January 2010, November 2010, and May
2011.
b
Denotes type and level of care provided (e.g., adult critical care, adult general medicine).
c
Free text admixture (drug amount/solution).
Am J Health-Syst Pharm—Vol 70 Apr 15, 2013
high-risk drugs, further decreasing
the risk of programming errors; such
wildcard entries are used when a
systemwide consensus on a standard
concentration cannot be reached or
when the dosing and concentration
of a drug are highly variable (i.e., in-
termittent total dose infusions).
With each library update, we
measure the success and patient
safety impact of library revisions
by a subsequent increase in GDR
compliance, a decrease in clinically
insignificant GDR alerts, a greater
frequency of “good catches” (de-
scribed below), and a reduction in
reported i.v. infusion administra-
tion errors and ADEs.
GDR compliance. Reported GDR
compliance metrics (monthly and
quarterly) include site-specific and
aggregated system percentage values
for all infusion starts programmed
using a drug library option (GDR
Drugs or GDR Fluids). Health-
system GDR compliance has re-
mained fairly constant at about 78%
(per a comparison of CQI data for
January–June of 2010 and the same
No. (%) Revisions
169 (27.8)
59 (9.7)
38 (6.3)
10 (1.6)
52 (8.6)
5 (0.8)
22 (3.6)
9 (1.5)
16 (2.6)
45 (7.4)
107 (17.6)
66 (10.9)
7 (1.2)
2 (0.3)
607 (100)
683
 case study  Intravenous infusion pumps
period in 2011), which is below our
target of 100%. The highest quarterly
systemwide GDR compliance rate
documented to date was 81% in the
fourth quarter of 2011.
To improve GDR compliance, a
systemwide direct-observation GDR
compliance audit was conducted by
a team of nurses led by an i.v. infu-
sion advanced practice nurse. Over a
selected one-month period (January
2010), the team documented system-
wide GDR compliance rates of 92%
for continuous infusions and 91% for
intermittent infusions; however, the
compliance rate for fluid infusions
was only 48%. Through direct obser-
vations, it was determined that the
Basic Infusion option was frequently
selected to program large-volume in-
fusions of maintenance fluids, effec-
tively bypassing pump safety features
and thereby increasing the risk of
errors and hindering initiatives to im-
prove GDR compliance. UPMC nurse
educators have launched awareness
and educational initiatives to pro-
mote GDR compliance, including
the use of workstation screensavers, a
web-based learning module, and one-
on-one consultations.
GDR alerts. The review of GDR
alerts generated over a designated
time period can help characterize
drug-specific smart-pump program-
ming issues and trend broad pro-
gramming problems involving care
area profiles and infusion types (con-
tinuous, intermittent, and fluid infu-
sions). From January through June
2011, 159,670 alerts were generated
across the health system, with 61%
occurring in the Adult Critical Care
profile, 34% in the Adult-MOST
profile (reflecting experience on gen-
eral medicine units), and 5% in the
Adult-OB profile. Of the generated
alerts, 34% involved the program-
ming of continuous infusions, 41%
related to intermittent infusions,
and 11% were triggered during fluid
infusions. Focusing on intermittent
infusions, 70% of alerts were attrib-
uted to the programming of dura-
684 Am J Health-Syst Pharm—Vol 70 Apr 15, 2013
tion times outside the minimum and
maximum GDR limits.
Good-catch analysis. A good-catch
analysis reflects the effectiveness
of the drug library in intercepting
potential i.v. infusion errors. This
reporting activity provides informa-
tion on infusions reprogrammed
within GDR limits or canceled in
response to an alert generated during
the initial programming. Although
the CQI software package provides
CQI report templates and graphs, we
review the raw data to identify the
specific drugs and verify the actual
number of good catches so as not
to overreport. Considering the large
number of infusions programmed
throughout the health system and
the size of our CQI database, we cur-
rently focus on continuous infusions
for the good-catch report. During the
period January–June 2010, continu-
ous infusion good catches averaged
462 per month; there were 450 per
month over the same time period
in 2011. A good catch often involves
the resolution of an incorrect “soft-
key” entry, including decimal-point
errors (e.g., the inadvertent entry of
22 units/hr rather than 2.2 units/hr),
the erroneous entry of “0” instead of
a decimal point due to soft-key mix-
ups (e.g., 503 mg/kg/hr rather than
5.3 mg/kg/hr), and “key-bounce”
errors (e.g., inadvertent entry of
11 mg/hr rather than 1 mg/hr). The
alerts generated from these program-
ming errors resulted in 190 good
catches during May 2011.
ADE analysis. During 2008 (con-
current with smart-pump imple-
mentation), 69 pump-related ADEs
were voluntarily reported across our
system’s facilities; that was a 27%
decrease compared with 2007 (before
smart-pump implementation) dur-
ing which 94 events were reported.
For 2008, 16% (n = 11) of reported
errors resulted in harm, with 29%
(n = 20) classified as “at the bor-
derline of harm.” An evaluation of
ADEs during the first 13 months of
smart-pump use (January 1, 2009–
January 31, 2010) identified 54 re-
ported events—a 43% reduction
from 2007; 11% of these ADEs re-
sulted in harm, and 50% were at the
borderline of harm. These outcomes
illustrate the successful reduction of
ADEs reported after smart-pump
implementation and a decrease of
harmful events. However, the find-
ings also highlight the need to focus
on borderline-harm events and pre-
ventable events.
Reduction of clinically insignificant
alerts. Infusions associated with a
high frequency of clinically insignifi-
cant alerts can induce overall nurse
alert fatigue. If this scenario contin-
ues, the use of the “override” option
can become routine during infusion
programming, increasing the prac-
tice of bypassing a truly important
alert. Causes of excessive alert rates
can include (1) overly conservative
GDR limit settings, (2) GDR limits
inconsistent with current clinical
evidence or practice, and (3) varia-
tions in nurses’ pump-programming
methods and workarounds.
The revised drug libraries released
in January 2010, November 2010,
and May 2011 each successfully
reduced the overall rate of unneces-
sary alerts by approximately 10%,
as determined by the first full CQI
quarterly alert analysis conducted
after each update.
Drug-specific CQI data for
January–June 2010 revealed that
norepinephrine infusion program-
ming was associated with the second-
highest alert frequency rate system-
wide. An average of 1700 GDR alerts
were generated each month due to
infusion rates exceeding the upper
GDR limit setting. The high alert
frequency and nurse alert override
rate of 93% prompted a literature re-
view, the results of which supported
a drug library revision increasing
the maximum GDR limit from 0.5
to 1 mg/kg/min. After the new GDR
limit was implemented with the
November 2010 library update, data
on norepinephrine use from Janu-
 case study  Intravenous infusion pumps

ary through June 2011 demonstrated

Table 2.
a 91% decrease in alerts compared
Comparative Rates of Clinically Insignificant Smart-Pump Alerts
with the previous year. In addition,
28 potentially excessive norepineph- Before and After Drug Library Revisions
rine dosing errors were intercepted. No. Alerts No. Alerts
From July through December During 6 During 6 Alert
2009, blood products and derivatives Mo Before Mo After Reduction,
Drug Library Entry Revisions Revisions %
(packed red blood cells [PRBCs],
fresh frozen plasma [FFP], platelets, Aminocaproic acid 103 6 94
albumin 25%) appeared consis- Amiodarone 2,011 157 92
Midazolam 6,703 668 90
tently in the list of the top 25 drugs
Albumin 25% 3,433 869 75
generating alerts; this was due to 16,325 4,858 70
Fentanyl
the frequent programming of infu- Blood product (red blood cells) 15,308 5,208 66
sion rates exceeding the drug library Pantoprazole 3,460 1,374 60
maximum GDR limits of 100 mL/hr Lorazepam 1,707 834 51
for albumin and 150 mL/hr for the Norepinephrine 10,743 982 91
other blood products. After consult- Tacrolimus 443 100 77
ing with our hematology groups Cyclosporine 777 189 76
and reviewing the pertinent litera- Magnesium sulfate 8,041 2,335 71
ture, the January 2010 drug library Paclitaxel 826 324 61
update included an increase in the Ampicillin–sulbactam 2,782 1,137 59
Oxacillin 1,864 791 58
maximum GDR limits to 400 mL/hr
Milrinone 439 187 57
for albumin and 350 mL/hr for the 834 400 52
Lorazepam
other blood products. 2011 CQI data Nicardipine 392 199 49
indicated that the GDR library revi- Alemtuzumab, natalizumab,
sions were successful in reducing the bevacizumab, cetuximab,
alert rates for PRBCs, FFP, platelets, infliximab, rituximab,
and albumin by 66%, 14%, 33%, rastuzumab 4,583 2,391 48
and 75%, respectively. Table 2 pro- Total 80,774 23,009 72
vides additional data on successful
reductions of clinically insignificant
alerts through approved drug library
adjustments. ception of similar rapid-infusion er- either the actual drug library entry
Addition of hard GDR limits. An rors. Subsequently, this library revi- or the pump-programming process
evaluation of four fentanyl infu- sion was deemed to have preempted during i.v. administration, both of
sion events that occurred within a 193 similar fentanyl infusion errors which can lead to an error.
two-month period identified overly during 2010; we highlighted this suc- For example, at one UPMC facil-
rapid infusion rates as the cause of cess to spearhead the UPMC “Hard ity, a standard cyclosporine infusion
each reported overdose. A review of Stop Halt” campaign launched in is only administered an estimated 15
the event log, manually downloaded June 2010. The ongoing educational times per month; therefore, the drug
from the smart pump, and the CQI program instructs nurses to halt does not appear in the facility’s list
data revealed that the rapid-infusion when a hard alert is encountered and of top 25 drugs generating alerts and
errors were due to the inadvertent obtain an independent nurse double- was not flagged for routine quarterly
programming of the “mg/hr” dose in check before proceeding. review. Following a reported cyclo-
the “mL/hr” field on the smart-pump Review of drugs with high alert- sporine rapid-infusion ADE at this
screen; in one case, the prescribed override rates. In addition to review- facility, a retrospective three-month
dose of 100 mg/hr was incorrectly ing the top drugs generating alerts, CQI data review was completed. The
programmed as 100 mL/hr, leading we also review all drugs with an alert- results indicated that cyclosporine
to the administration of the entire override rate of ≥90%. This compo- alerts were generated due to the pro-
100-mL infusion volume (2500 mg) nent of our program is important gramming of an infusion duration
over one hour. The drug library entry because these drugs do not always exceeding the GDR limit (24 hours)
for fentanyl was revised to include a appear in the list of the top 25 alert by only two minutes. The standard
hard maximum GDR limit of 1000 generators; however, the high over- programmed rate of 10.4 mL/hr con-
mg/hr (40 mL/hr) to enable the inter- ride rate indicates a problem with sistently generated this insignificant

Am J Health-Syst Pharm—Vol 70 Apr 15, 2013 685

 case study  Intravenous infusion pumps
alert and the subsequent override
response.
The event log downloaded from
the sequestered smart pump involved
in the event described above showed
that the rate was inadvertently pro-
gramming as 104 mL/hr instead of
10.4 mL/hr. Since cyclosporine alerts
were frequently generated during
infusion programming and over-
ridden due to alert fatigue, when
a substantially incorrect rate was
entered, the “meaningful” soft alert
generated was inadvertently overrid-
den, leading to the rapid-infusion-
rate ADE. Following this event, the
next drug library release included a
small increase of 1 hour (from 24 to
25 hours) in the cyclosporine upper
GDR duration limit. The subsequent
CQI data revealed a 76% decrease in
the frequency of cyclosporine dura-
tion alerts. This type of successful
nuisance alert reduction is intended
to elevate the awareness of and ap-
propriate response to the more
meaningful alerts generated.
Discussion
The use of programmable infu-
sion smart pumps is included in
the Joint Commission’s National
Patient Safety Goals,6 prompting
facilities across the country to adopt
smart-pump technology. Thought-
ful program design and systematic
implementation of the CQI program
have enabled the maintenance and
continued growth of a systemwide
smart-pump initiative. We sustain
an evidence-based and clinically ap-
plicable drug library through mean-
ingful interventions derived from
scheduled comprehensive analyses
of smart-pump CQI data, reported
ADEs, current scientific literature,
and, most important, the valued
participation of our clinicians. Our
models and extensive experience are
shared across the health system and
are applicable to facilities regionally
and nationally.
Targeted educational efforts and
facilitation of i.v. medication dosing
686 Am J Health-Syst Pharm—Vol 70 Apr 15, 2013
and concentrations have decreased
variances from recommended i.v.
administration practices and im-
proved GDR compliance systemwide.
The smart-pump technology is a
successful vehicle for standardizing
practices while allowing appropriate
flexibility for special patient popu-
lations. Small and large hospitals,
as well as outpatient clinics, benefit
from implementing evidence-based
standardized medication infusion
practices and guidelines. Our CQI
program, including the structured
quality metrics and report formats
described here, can be adapted by
any facility. The lessons learned in
reviewing i.v. medication ADEs and
interpreting GDR alerts with the use
of our algorithm can assist facilities
new to smart-pump integration.
Conclusion
A smart-pump CQI program is an
effective tool for enhancing the safety
of i.v. medication administration.
The ongoing refinement of the drug
library through the development and
implementation of key interventions
promotes the growth and sustain-
ability of the smart-pump initiative
systemwide.
References
1. Committee on Identifying and Prevent-
ing Medication Errors, Institute of
Medicine. Preventing medication er-
rors: quality chasm series. Washington,
DC: National Academies; 2006:100-
1,371.
2. Leape LL, Bates DW, Cullen DJ et al.,
for the ADE Prevention Study Group.
Systems analysis of adverse drug events.
JAMA. 1995; 274:35-43.
3. Taxis K, Barber N. Ethnographic study
of incidence and severity of intravenous
drug errors. BMJ. 2003; 326:684.
4. USP Center for the Advancement of
Patient Safety. Summary of information
submitted to MEDMARX in the year
2002. The quest for quality. Rockville,
MD: United States Pharmacopeial Con-
vention; 2003:10-7,35-6.
5. American Society of Health-System
Pharmacists. Proceedings from a summit
on preventing patient harm and death
from i.v. medication errors. Am J Health-
Syst Pharm. 2008; 65:2367-79.
6. Eskew JA, Jacobi J, Buss W et al. Using
innovative technologies to set new safety
standards for the infusion of intravenous
medications. Hosp Pharm. 2002; 37:1179-
89.
7. Hatcher I, Sullivan M, Hutchinson J et al.
An intravenous medication safety system:
preventing high-risk medication errors
at the point of care. J Nurs Admin. 2004;
34:437-9.
8. Wilson K, Sullivan M. Preventing medi-
cation errors with smart infusion tech-
nology. Am J Health-Syst Pharm. 2004;
61:177-83.
9. Fields M, Peterman J. IV medication
safety system averts high-risk medication
errors and provides actionable data. Nurs
Adm Q. 2005; 29:77-86.
10. Rothschild JM, Keohane CA, Cook EF
et al. A controlled trial of smart infusion
pumps to improve medication safety in
critically ill patients. Crit Care Med. 2005;
33:533.
11. Rohman C. Smart pump implementa-
tion: one hospital’s story. Nurs Manage.
2005; 36(6):49-51.
12. Williams C, Maddox RR. Implementation
of an i.v. medication safety system. Am J
Health-Syst Pharm. 2005; 62:530-6.
13. Bowcutt M, Rosenkoetter MM, Chernecky
CC et al. Implementation of an intrave-
nous medication infusion pump system:
implications for nursing. J Nurs Manage.
2008; 16:188-97.
Appendix—Questions for evaluating
smart-pump continuous
quality-improvement data to modify
drug library settings
What are the top 20 drugs causing alerts? For
each drug listed:
 1. What types of infusions have alerts been
generated (continuous, intermittent, fluid,
or bolus)?
 2. What programmed safety limit (dose,
rate, duration, or concentration) is being
encountered?
  3. What are the actions taken by nursing staff
in response to an alert (e.g., override, repro-
gram, cancel infusion)?
  4. Does overriding the alert have the potential
to cause harm (or are the alerts clinically
insignificant)?
  5. Can the addition of a hard safety limit pre-
vent potential programming mishaps?
  6. What steps are taken by the nurse to initially
program the infusion?
 7. Was a workaround created for infusion
programming?
 8. Can a minor, drug-specific revision of a
safety limit reduce an associated high rate
of clinically insignificant alerts?
 9. Is additional nursing staff education the
only resolution to reduce a high alert
frequency?
10. Are infusions programmed using the pre-
programmed drug library options?
If there is no improvement in alert compli-
ance after drug library or safety-limit updates,
assess nursing practices by direct observation to
identify issues.