QUALITY IMPROVEMENT REPORT
Using real time process measurements to reduce catheter
related bloodstream infections in the intensive care unit
R J Wall, E W Ely, T A Elasy, R S Dittus, J Foss, K S Wilkerson, T Speroff
See end of article for
Dr R J Wall, Division of
Pulmonary and Critical
Care Medicine, University
of Washington, Box
359762, 325 Ninth
Avenue, Seattle, WA
Accepted for publication
27 May 2005
Qual Saf Health Care 2005;14:295–302. doi: 10.1136/qshc.2004.013516
Problem: Measuring a process of care in real time is essential for continuous quality improvement (CQI).
Our inability to measure the process of central venous catheter (CVC) care in real time prevented CQI
efforts aimed at reducing catheter related bloodstream infections (CR-BSIs) from these devices.
Design: A system was developed for measuring the process of CVC care in real time. We used these new
process measurements to continuously monitor the system, guide CQI activities, and deliver performance
feedback to providers.
Setting: Adult medical intensive care unit (MICU).
Key measures for improvement: Measured process of CVC care in real time; CR-BSI rate and time
between CR-BSI events; and performance feedback to staff.
Strategies for change: An interdisciplinary team developed a standardized, user friendly nursing checklist
for CVC insertion. Infection control practitioners scanned the completed checklists into a computerized
database, thereby generating real time measurements for the process of CVC insertion. Armed with these
new process measurements, the team optimized the impact of a multifaceted intervention aimed at
Effects of change: The new checklist immediately provided real time measurements for the process of CVC
insertion. These process measures allowed the team to directly monitor adherence to evidence-based
guidelines. Through continuous process measurement, the team successfully overcame barriers to change,
reduced the CR-BSI rate, and improved patient safety. Two years after the introduction of the checklist the
CR-BSI rate remained at a historic low.
Lessons learnt: Measuring the process of CVC care in real time is feasible in the ICU. When trying to
improve care, real time process measurements are an excellent tool for overcoming barriers to change and
enhancing the sustainability of efforts. To continually improve patient safety, healthcare organizations
should continually measure their key clinical processes in real time.
indwelling CVC at any given time.1CVCs expose patients to
increased risk of infection, and up to 80 000 catheter related
bloodstream infections (CR-BSIs) occur in ICUs in the US
each year.2At a cost of $34 500–56 000 per episode3 4and
with a crude mortality rate that may approach 35%,5
management of CR-BSIs remains an important issue for
patient safety and cost of care.
Experts advocate a broad preventive approach as the
key strategy for CR-BSI reduction,2and preventive inter-
ventions achieve impressive results.6–10However, sustaining
the effect of an intervention after cessation of a study is
difficult because multiple complex processes cause BSIs.
Organizations that monitor CR-BSI rates alone (the ‘‘term-
inal’’ outcome) find troubleshooting a failing system
after CR-BSI rates have already increased to be com-
plicated and time consuming. A better strategy involves
monitoring the processes that ultimately lead to the adverse
Continuous quality improvement (CQI) methodologies
provide an ideal framework for initiating and sustaining
improvements in complex systems.11By definition, CQI
engages front line staff in cycles of iterative problem solving,
with decision making based on real time process measure-
ments.12 13Through the systematic measurement of key
processes that lead to clinical outcomes, CQI allows an
organization to maximize improvement efforts and sustain
entral venous catheters (CVCs) are indispensable in the
management of critically ill patients, and half of all
intensive care unit (ICU) patients in the US have an
On 1 November 2002 we initiated a patient safety project
aimed at reducing CR-BSIs in the medical intensive care unit
(MICU) of a 640 bed tertiary university teaching hospital.
Approximately 600 patients are admitted to the 14 bed unit
each year with a variety of illnesses including acute
respiratory distress syndrome, asthma, chronic obstructive
pulmonary disease, respiratory failure, pneumonia, sepsis,
poisoning, drug overdose, and gastrointestinal bleeding. The
MICU also cares for critically ill solid organ and bone marrow
transplant patients. All patients undergoing CVC insertion
during their stay in the MICU were eligible for participation.
The hospital’s institutional review board approved the study
OUTLINE OF PROBLEM
For many years our busy MICU struggled with CR-BSIs. In
2002 the CR-BSI rate was 44% higher than the national
median of 5.2 per 1000 catheter-days. Like many hospitals,
we had historically focused on infection surveillance and staff
education. However, we recognized that more innovative
strategies were needed to tackle this ongoing complex
problem. The impressive utility of CQI in other healthcare
settings,14–18coupled with successful reports from other high
risk industries, encouraged us to consider a CQI approach.
Abbreviations: CR-BSI, catheter related bloodstream infection; CQI,
continuous quality improvement; CVC, central venous catheter; MICU,
medical intensive care unit
Although our infection control practitioners tracked the
number of CVCs (inputs) and the number of CR-BSI
(outputs), process measurements for the series of steps that
changed those inputs into outputs were not routinely
collected (fig 1). Before undertaking CQI we first needed a
systematic method for measuring the process of CVC care in
KEY MEASURES FOR IMPROVEMENT
Our primary goal was to show that real time measurement of
CVC care was feasible in the MICU. We anticipated these new
process measurements would guide CQI efforts and thereby
lead to a reduced CR-BSI rate. To increase staff appreciation
of the link between process measures and clinical outcomes,
we fed these bundled data back to providers.
STRATEGIES FOR CHANGE
We assembled a voluntary interdisciplinary team with at least
one MICU leader, infectious disease expert, front line staff
member, and quality improvement expert (fig 2). The team’s
goal was to develop a system for measuring the process of
CVC care in real time with the understanding that this
information would guide future improvement activities
aimed at reducing infections. The team compiled a list of
risk factors by reviewing the published literature on CR-BSIs,
and then classified these risk factors depending on whether
they occurred during insertion or daily maintenance (table 1).
When deciding where to focus their initial efforts, the team
selected CVC insertion as a high leverage starting point.20
Several issues guided this decision: (1) most CVCs in the
MICU were inserted by trainees and there was a high
likelihood of practice variability; (2) there was strong
evidence linking certain insertion behaviors with CR-BSIs;
and (3) CVC insertion was easily defined and amenable to
Identifying barriers to and facilitators of change
A trained research nurse randomly observed five CVC
insertions in the ICU over several days, collected data about
compliance with evidence-based insertion behaviors, and
identified potential barriers to and facilitators of change.21
After studying the local workflow, soliciting clinician input,
and mapping the process of CVC care,22the team established
priority areas for measurement. These areas included
provider education,23trainee supervision,24insertion site,25–27
hand hygiene,28skin antisepsis,29and use of maximal sterile
To capture the desired process measures, the team developed
a new standardized nursing checklist for CVC insertion
(available online31). While studying the local workflow,
the team noted that nurses handwrote a brief care note
after CVC insertions. By including the essential elements
from these nursing notes in the new checklist, the team
eliminated any need for the handwritten notes. The new
document employed check boxes and could be completed in
The new checklist recorded whether providers washed
their hands, used maximal sterile barriers, maintained a
sterile field, and were properly supervised. Although the
measurement of these items was essential for process
improvement, risk management did not believe these
sensitive items belonged in the permanent medical record.
To address this issue the team created a double paged
carbonless document. The sensitive items were intentionally
blacked out and unreadable on the second page.
Upon completing the checklist, the nurse detached the top
page (with all items readable) and dropped it in a secure
lockbox. The second page remained on the patient’s chart
with the sensitive items blacked out. The infection control
practitioners collected the checklists daily and scanned the
de-identified forms into a pre-established computerized
(Minneapolis, MN). Using a standard image scanner, this
application scanned pre-established fields on the checklist
and imported the information into a spreadsheet database.
These data were stored on a secure computer at the Center for
Clinical Improvement for future statistical analyses.
The checklist was initially pilot tested for 1 week with two
senior MICU nurses. They commented on readability of
items, missed opportunities for measurement, and overall
functionality of the checklist. We subsequently added,
modified, or deleted several items based upon their input.
After achieving a suitable final version, the checklist was
introduced to all nursing staff and implemented throughout
the entire MICU.
??????? ?? ??? ????
care. BSI, bloodstream infection.
Flow diagram for the process of central venous catheter (CVC)
Chief Hospital Epidemiologist
Infection Control Practitioners
Infectious Disease Experts
Center for Clinical Improvement*
National Quality Scholars†
a group of improvement experts at Vanderbilt University Medical Center
whose primary aim is to foster patient safety and improve the quality of
health care. The group engages in various activities including education
of caregivers and the advancement of improvement science. ?The
National Quality Scholars is a post-residency fellowship funded by the
Veterans Health Administration. Its primary aim is to equip physicians
with the skills necessary for a career focused on the scholarship,
research, and teaching of healthcare improvement.19
Interdisciplinary team. *The Center for Clinical Improvement is
catheter (CVC) related infections
Factors impacting on central venous
CVC insertionCVC maintenance
Maximal sterile barriers
Number of needle sticks
Insertion site care
Dressing and gauze changes
Specialized line care teams
296 Wall, Ely, Elasy, et al
OUTCOME MEASUREMENT: BLOODSTREAM
Our hospital Epidemiology and Infection Control section uses
the definition for primary CR-BSI outlined by the Centers for
Infections Surveillance System (NNIS).2Laboratory con-
firmed bacteremia (or fungemia) is attributed to a patient
with a CVC if they have been in the ICU for at least 48 hours,
provided the infection is not related to another distal source.
Patients with a CVC who develop a BSI within 48 hours of
ICU discharge also have CR-BSI.
Definitions for checklist items
Proper hand hygiene before insertion reduces CR-BSIs.28
Since alcohol based hand rubs are more efficacious than
soap and water,32we gave credit if either behavior was
Studies suggest that there is an increased risk for infection
when inexperienced providers insert CVCs.24Since house
staff insert most CVCs in our MICU, the checklist included
whether the proceduralist was properly supervised. Based on
guidelines established by the American Board of Internal
Medicine,33we considered trainees ‘‘supervised’’ if they had
previously performed five CVC insertions or if another
provider with experience of at least five CVC insertions
supervised them during the procedure. In most cases the
nurse directly inquired regarding the proceduralist’s experi-
ence. If the nurse was unable (or unwilling) to obtain this
information, the checklist included an option ‘‘didn’t ask.’’
Maximal sterile barriers
Studies show the importance of maximal sterile barriers
(gloves, gown, mask, patient drape) during CVC insertions.30
The checklist recorded which individual barriers were used
during the procedure. We defined ‘‘maximal sterile barriers’’
as the simultaneous use of all four component barriers.
Chlorhexidine skin preparation
Skin antisepsis with a 2% chlorhexidine preparation is
superior to povidone-iodine and alcohol for reducing CR-
BSIs.29We gave credit for chlorhexidine whenever it was
employed, even if other agents were also used.
Circumstances of CVC insertion
Providers may skip infection reducing steps during emer-
gency procedures. We considered a CVC insertion ‘‘emergent’’
if a life threatening situation mandated the procedure
immediately. Otherwise, the procedure was considered
‘‘elective’’. To simplify classification, the checklist intention-
ally avoided the ‘‘urgent’’ parlance.
Although the checklist was developed as a measurement tool,
its use during CVC insertions also made the checklist an
important intervention. In fact, clinical reminders at the
point of care are one of the most effective strategies for
affecting daily practice.34Since multifaceted interventions are
more likely to alter physician behavior than single interven-
tions,35the team simultaneously implemented other inter-
vention components aimed at reducing CR-BSIs. These
included provider education, continuous audit, and perfor-
Based on the successes of previous educational efforts,8–10the
team constructed a web based tutorial discussing catheter
related infections.31In addition to describing the new
checklist, the tutorial explained the scientific evidence for
recommended behaviors. The website recorded whether each
Leadership announced a new policy requiring annual CVC
education for all MICU providers. A letter was sent to house
staff asking them to independently complete the tutorial
before their MICU rotation. A similar letter was sent to
nursing staff explaining that the tutorial would be incorpo-
rated into their annual competencies curriculum. A repeat
letter was sent 3 months later to providers who had not
fulfilled the requirement. If individuals still did not complete
the tutorial, the MICU leadership approached them privately.
Audit and feedback
A premise of the Centers for Disease Control and Prevention
(CDC) National Nosocomial Infections Surveillance System
(NNIS) is that constantly feeding back surveillance data to
front line staff reduces hospital acquired infections.36As a
result, our infection control practitioners had traditionally
generated monthly NNIS reports for the MICU. Although
feedback alone is usually ineffective for altering provider
behaviors,37 38it is a useful complement to other strategies for
improving provider performance.39–42To increase our chances
for success and raise awareness for the link between
processes and outcomes, we bundled the new process
measurements with the monthly NNIS reports. When feeding
this information back to front line staff, the leadership
emphasized that process measurement was transparent and
blameless, data were de-identified, and the MICU was
committed to CQI.
Analysis of data
For many years our hospital infection control practitioners
tracked monthly CR-BSI rates using definitions established
by the NNIS system.2While providing an excellent frame-
work for detecting infections, the NNIS database presented
two statistical challenges. Firstly, CR-BSIs were statistically
rare events and the aggregate monthly reports limited our
ability to quickly detect changes. Secondly, NNIS counted all
CVCs in the MICU regardless of whether the CVC was
inserted by providers on another service. Since our efforts
only targeted MICU providers, we needed an alternative
approach that was capable of specifying the provider.
We overcame these barriers by using an alternative
Shewart-type statistical control methodology, called a g
chart.43By monitoring the number of days between infections,
the g chart has greater detection power for rare events
compared with conventional binomial based approaches.44 45
Whenever a CR-BSI occurred we added the data point to our
g chart and calculated the number of days from the previous
infection. Rather than waiting until the end of a pre-specified
time period, we took immediate advantage of each event and
improved our ability to rapidly detect changes (see examples
below). Furthermore, we were able to exclude CVCs inserted
by providers on another service.
The g chart is a type of statistical process control (SPC)
chart and therefore requires a basic understanding of the
principles inherent to SPC.46–49According to SPC theory, two
types of variation exist: special and common cause variation.
The managerial approach for reducing the variation depends
on which type is present. Special cause variation refers to
unnatural variation in a process, usually due to identifiable
events. Reducing special cause variation should focus on
investigating and eliminating the special events. Conversely,
common cause variation is natural and inherent to a process
on a regular basis. As an intrinsic part of the process,
common cause variation is best achieved through process
Reducing bloodstream infections in the ICU 297
At the start of the project we constructed a baseline (pre-
intervention) g chart by querying the NNIS database from 1
January 2000 to 31 October 2002 (fig 3, observations 1–39).
We only plotted CR-BSIs on the g chart for catheters inserted
by the MICU—for example, dialysis catheter CR-BSIs were
not included on the g chart. We measured the average time
between infections (27 days) and used Benneyan’s method
for calculating an upper control limit (UCL=109 days) at
three standard deviations above the pre-intervention mean.44
During the pre-intervention period the number of days
between infections was consistently below the UCL. The
absence of any points above the UCL suggested that the
variation in time between CR-BSIs was inherent to our
current process of care (that is, common cause),47and that
the proper way to reduce the CR-BSI rate was through
process redesign or CQI.50Our goal was to reduce the CR-BSI
rate and thereby increase the time between events. Based on
the g chart, we considered our intervention successful if data
points fell above the UCL since this would correspond to a
decreased CR-BSI rate.
EFFECTS OF CHANGE
In the 2 years between 1 November 2002 and 31 October
2004, a total of 630 CVCs were inserted using the new
checklist. Mortality rates, admission diagnoses, ventilator
days, and lengths of stay were similar for patients admitted
during the pre-intervention and post-intervention periods
(data not shown). Nursing staff immediately embraced the
new insertion checklist because it simplified their documen-
tation and replaced handwritten notes.
Following the introduction of the checklist, the MICU for
the first time collected frequency characteristics for CVC
insertions (table 2). Triple lumen catheters comprised over
half of all CVC insertions and the subclavian vein was the
most common site. As expected in a training hospital, interns
? ? ? ? ? ? ?? ? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ??
???? ??????? ??????????
???? ???????? ?????????
??? ? ??? ????
???? ? ?? ????
increase time between infections, hence higher points indicate better performance. As the CR-BSI rate decreases there are fewer data points per year—
this is reflected in shorter calendar year bars along the top. The baseline period is from January 2000 through October 2002 (inclusive of infections 1–
39). The intervention period is from November 2002 through October 2004 (initiated after infection 39 and inclusive of infections 40–45). During the
baseline period the number of days between infections was consistently below the upper control limit (UCL), suggesting the variation in time between
infections was random and inherent to the process of care. During the post-intervention period the number of days between infections was frequently
above the UCL, suggesting the intervention had introduced non-random improvements into the process of care.
Days between catheter related bloodstream infections (CR-BSIs) on a g chart. The g chart monitors time between events. The goal is to
November 2002 to October 2004
Characteristics of CVCs inserted from
Venous insertion site
PGY, postgraduate year.
Due to rounding, percentages may not total 100.
1-34-67-9 10-12 13-15 16-18 19-21 22-24
Percent of insertions
insertions. Process measurements of provider behaviors during CVC
insertions (post-intervention) are plotted over time. The data have been
combined into 3 month aggregates to simplify display. In practice, the
front line staff receive monthly updates with much greater detail.
Process measurements for central venous catheter (CVC)
298Wall, Ely, Elasy, et al
inserted most lines. The majority of procedures occurred
under non-emergent circumstances.
The checklist also generated real time measurements for
the CVC insertion process (fig 4). Using these new process
data, the team embarked on cycles of CQI and also began
exploring thereasons for
Examples of CQI (‘‘maximal sterile barriers’’) and an
exploratory analysis (‘‘femoral CVCs’’) follow.
Example 1: Maximal sterile barriers
Adherence to maximal sterile barriers declined unexpectedly
in early 2003 (fig 5). The team performed secondary analyses
of the process measures and noted that this trend was not
explained by emergent placement circumstances. A break-
down by individual barrier components revealed that use of
the patient drape was the main limitation to maximal sterile
barrier adherence. The team ensured adequate supplies were
stocked on the floor and confirmed providers had completed
the tutorial. Leadership asked front line staff to emphasize
appropriate sterile technique with trainees, particularly use of
the patient drape. In response, the MICU began purchasing
newer CVC kits prepackaged with drapes. In the following
months adherence to maximal sterile barrier gradually
Example 2: Femoral CVCs
Femoral CVCs have more infectious and thrombotic compli-
cations than other venous sites.25–27After noting a large
number of femoral CVCs early in the study, the ICU decided
to further explore this practice pattern (table 3). Femoral
insertions were more likely to have been preceded by
attempts at another site, suggesting that the femoral site
was often a secondary choice. Interestingly, femoral CVCs
exhibited lower rates of hand washing, chlorhexidine use,
and maximal sterile barriers. These trends were not explained
by insertions under emergent conditions (data not shown).
Contrary to popular belief, the femoral site was not more
popular than other venous sites during emergent situations.
As provider performance gradually improved for all inser-
tions, compliance with recommended behaviors likewise
improved at the femoral site. However, the percentage of
femoral CVCs has remained unchanged for 2 years.
Bloodstream infection rate
The last pre-intervention CR-BSI occurred 58 days before the
study (fig 3, infection number 39). When a CR-BSI occurred
55 days into the study, we totalled the number of days from
the last infection (58+55=113) and placed this data point on
the new g chart (number 40). After 233 days another
infection occurred (number 41). For the first time in 3 years,
data points fell above the UCL, suggesting that special cause
variation had affected the process of CVC care. We suspected
our improvement efforts were making a difference and
shared these preliminary results with the staff.
One year into the study the MICU experienced four CR-
BSIs in 6 months (fig 3, infections 41–44). The team began
investigating potential causes for this declining performance.
The process measures did not reveal any change in insertion
behaviors. The infection control practitioners could not
identify any pattern among organisms, catheter types,
providers, or insertion site. As a result, the team re-
emphasized the importance of daily CVC maintenance
??????? ???????? ???
??????? ???????? ???
? ? ???
? ? ??
measurements guided improvement efforts. (A) Declining use of maximal sterile barriers during January and February (arrow) prompted secondary
analyses. (B) The declining performance was not attributable to insertions during emergent circumstances. (C) Process measurements for the individual
barriers revealed that provider use of the patient drape was the main obstacle to maximal sterile barriers adherence. *Maximal sterile barriers denote
the simultaneous use of all four barrier components.
Adherence to use of maximal sterile barriers during insertion of central venous catheters (CVCs). Example showing how real time process
Reducing bloodstream infections in the ICU299