 |
|
Reducing Infections Among Women Undergoing Cesarean Section in Colombia by Means of Continuous Quality Improvement Methods
Michelle Weinberg, MD, MPH;
Jose Maria Fuentes, MD;
Ariel I. Ruiz, MD;
Fred Wilson Lozano, MD;
Edith Angel, MD;
Hernando Gaitan, MD;
Brunhilde Goethe, RN;
Sonia Parra, RN;
Susan Hellerstein, MD, MPH;
Dennis Ross-Degnan, ScD;
Donald A. Goldmann, MD;
W. Charles Huskins, MD, MS
Arch Intern Med. 2001;161:2357-2365.
ABSTRACT
| |
Background Improving obstetric care in resource-limited countries is a major international
health priority.
Objective To reduce infection rates after cesarean section by optimizing systems
of obstetric care for low-income women in Colombia by means of quality improvement
methods.
Methods Multidisciplinary teams in 2 hospitals used simple methods to improve
their systems for prescribing and administering perioperative antibiotic prophylaxis.
Process indicators were the percentage of women in whom prophylaxis was administered
and the percentage of these women in whom it was administered in a timely
fashion. The outcome indicator was the surgical site infection rate.
Results Before improvement, prophylaxis was administered to 71% of women in
hospital A; 24% received prophylaxis in a timely fashion. Corresponding figures
in hospital B were 36% and 50%. Systems improvements included implementing
protocols to administer prophylaxis to all women and increasing the availability
of the antibiotic in the operating room. These improvements were associated
with increases in overall and timely administration of prophylaxis (P<.001) in both hospitals by time series analysis, with
adjustment for volume and case mix. After improvement, overall and timely
administration of prophylaxis was 95% and 96% in hospital A and 89% and 96%
in hospital B. In hospital A, the surgical site infection rate decreased immediately
after the improvements (P<.001). In hospital B,
the infection rate began a downward trend before the improvements that continued
after their implementation (P = .04).
Conclusion Simple quality improvement methods can be used to optimize obstetric
services and improve outcomes of care in resource-limited settings.
INTRODUCTION
COMPLICATIONS of pregnancy and childbirth are the leading cause of death
and disability among women of childbearing age in low- and middle-income countries.1 The majority of complications are caused by infection,
preeclampsia, hemorrhage, and obstructed labor, all of which can be treated
in facilities providing emergency obstetric services.2
Improving the availability, utilization, and quality of these services is
a major international public health priority.3-5
A number of publications have provided practical tools for evaluating
obstetric services and identifying areas for improvement.6-8
However, these publications offer limited guidance in how to actually improve
care. Clinicians need guidance in the process of improvement because systems
of patient care are often complex and require multidisciplinary problem solving
to achieve meaningful improvement.9
There is growing experience in the use of continuous quality improvement
(CQI) methods to evaluate and improve systems of patient care in North America
and Europe, although conclusive evidence that use of CQI methods has led to
improved patient outcomes is limited.10-11
Nonetheless, the rationale for applying this approach in resource-limited
settings is compelling, since it offers the potential to improve systems of
patient care and outcomes in the context of existing resources.12
This program was conducted in 2 nonprofit hospitals providing obstetric
care for low-income pregnant women in Bogotá, Colombia. The objective
was to reduce the incidence of infection after cesarean section by optimizing
systems of care for women by means of CQI methods. The program focused on
this objective for several reasons: (1) previous experience in both hospitals
indicated that infections after cesarean section (including infections of
the surgical incision and endometritis) ac counted for the vast majority of
postpartum infections; (2) there is a large base of evidence regarding modifiable
risk factors and strategies for prevention of these infections13-14;
and (3) risk-reduction strategies can be implemented effectively by hospital-based
programs.
METHODS
PARTICIPATING HOSPITALS
The program was conducted in 2 nonprofit hospitals in Bogotá,
Colombia, during 1996 to 1998. These hospitals are among the 5 regional obstetric
referral hospitals for low-income women in Bogotá and were the first
and only hospitals approached for participation in the study for logistic
reasons. Both hospitals are university-affiliated, teaching hospitals. Hospital
B also serves as a tertiary-care obstetric referral center for women with
high-risk pregnancies for the entire metropolitan area and surrounding regions.
The study leader was the chief of obstetrics and gynecology in hospital A
and the assistant hospital director (who was also a senior obstetrician) in
hospital B. The hospital administrations, clinical research committees, and
obstetric departments of each hospital approved the program.
STUDY DESIGN
The study design was a segmented time series analysis of the effect
of improvements in systems of care on quantitative process and outcome indicators.
Process indicators were defined during the course of the program by means
of CQI methods (see "CQI Methods" section below). The outcome indicator was
the rate of surgical site infection after cesarean section.
SURVEILLANCE OF SURGICAL SITE INFECTIONS AFTER CESAREAN SECTION
Surgical site infection after cesarean section was defined according
to criteria published by the Centers for Disease Control and Prevention.15 This definition divides surgical site infections
into incisional infections, which are subclassified as superficial or deep
incisional infections, and organ/space infections. Endometritis and endomyometritis
after cesarean section are included in this definition as organ/space infections.15
Active, prospective surveillance of surgical site infections was performed
at both hospitals. At hospital A, the obstetrical head nurse and 2 obstetrical
residents identified infections during twice-a-day bedside rounds and chart
review. The program leader performed an independent validation survey during
a 1-month period in 1997 and did not identify any additional infections or
any falsely identified infections. At hospital B, 2 infection control nurses
identified infections by every-other-day bedside rounds and chart review.
This surveillance was performed under the supervision of the infection control
head nurse. An obstetrician who was a subspecialist in obstetric infections
confirmed the presence of infection in all cases identified by the surveillance
nurses and consulted on any questionable cases. Women who were readmitted
for complications during the postpartum period were surveyed for infection
in both hospitals. Postdischarge surveillance was not performed in either
hospital.
Rates of infection were expressed as the number of infections per 100
cesarean sections. Infection rates were presented to clinicians during departmental
meetings at least every quarter in both hospitals. At hospital B, surgeon-specific
infection rates were also reported confidentially to individual obstetricians
and residents.
OTHER DATA COLLECTION
The number of deliveries per month was obtained from administrative
sources. The number of cesarean sections per month was determined by active
surveillance. Other data recorded during active surveillance of women undergoing
cesarean section were as follows: elective (presence of neither labor nor
rupture of membranes) vs nonelective cesarean section (presence of either
labor or rupture of membranes), the date and time of delivery, and the use
and time of administration of perioperative antibiotic prophylaxis.
MULTIDISCIPLINARY TEAMS AND CQI TRAINING
The program was supported by the hospital leadership as a demonstration
project and was the first organized, interdepartmental CQI project in either
hospital. No one in either hospital had previous formal CQI training or experience.
Hospital personnel formed multidisciplinary CQI teams containing at least
1 of each of the following personnel: an attending obstetrician, a resident,
an obstetric and/or operating room nurse, a pharmacist, and an administrator.
The leader of each team was the senior obstetrician. Facilitators with previous
training and experience in CQI methods provided training in CQI principles
during a 2-day workshop conducted during the second month of the program and
"just-in-time" training in the use of CQI tools during the teams' work sessions.
REVIEW OF THE LITERATURE
Teams reviewed authoritative texts and published reports to identify
risk factors for surgical site infections and effective prevention strategies.13-14,16-27
Electronic literature searches were performed by means of the MEDLINE database
(MeSH terms: cesarean section, endometritis, surgical wound infection, cross
infection, and infection control). Several team members summarized the literature
during the 2-day workshop.
CQI METHODS
The teams followed the Model for Improvement developed by Nolan and
colleagues and applied extensively to health care quality improvement initiatives
in the United States.28-31
The model poses 3 questions. The first question (What are we trying to accomplish?)
emphasizes that the team must define a consensus goal for improvement. The
second question (How will we know that a change is an improvement?) requires
the team to measure outcome indicators (eg, surgical site infection rates)
and/or process indicators (eg, the use and timing of perioperative antibiotic
prophylaxis). These indicators enable teams to analyze the existing performance
of a system of care and to determine whether a change in the system results
in performance improvement. The third question (What changes can we make that
will result in improvement?) and the accompanying Plan-Do-Study-Act (PDSA)
cycle emphasize that the team must design and test changes in systems of care:
a change is planned (Plan), it is implemented on a small scale (Do), its effect
is evaluated (Study), and, depending on the result, the change is accepted,
modified and tested again, or abandoned (Act).
The teams used standard CQI tools to apply this model.30, 32
First, they identified causes of post–cesarean section infection during
a brainstorming session. Then they organized these causes into a cause-and-effect
diagram (ie, fishbone or Ishikawa diagram)32
with 5 major groupings: preexisting host and antenatal factors; complications
or adverse events during labor; inappropriate use of perioperative prophylaxis;
improper preparation of the skin at the incision site; and surgical risk factors.
Teams then used a priority matrix to focus their improvement efforts on causes
that were both important and amenable to improvement. Using a semiquantitative
scale (1, least; 4, most), they evaluated the importance of each cause and
their capacity to address this cause effectively. They also assessed the time
frame required for change (short, weeks to several months; medium, several
months to 1 or more years; long, several years). The cost of improvement was
not ranked separately, but team members considered available resources in
determining their rankings.
Using these tools, the teams decided that their consensus goals were
to evaluate and optimize the systems for preparing the skin at the incision
site and prescribing and administering perioperative antibiotic prophylaxis.
Their decisions were based on the importance of these systems in preventing
infection and their capacity to improve the systems within a relatively short
time with available resources.
In both hospitals, the procedures for preparing the skin were reviewed.
No changes were made in hospital A. Hospital B modified its procedure to shave
women only as necessary immediately before surgery.
Because their respective systems for prescribing and administering perioperative
antimicrobial prophylaxis were more complex, involving multiple steps and
different individuals, teams used flow diagrams to describe these systems
of care qualitatively (hospital A, Figure
1; hospital B, Figure 2).
They also defined process indicators to evaluate the performance of these
systems quantitatively. The process indicators were the percentage of women
who received prophylaxis (utilization indicator) and the percentage of women
receiving prophylaxis in whom the antibiotic was administered within 1 hour
after delivery (timing indicator). Women who received an antibiotic for a
suspected or confirmed infection before the cesarean section were excluded
from these indicators. The timing indicator was defined as such because most
experts and published guidelines recommend administration of the antibiotic
after delivery to avoid transfer of antibiotic to the infant via the placenta.26-27,33 One hour after delivery
was used as a practical approximation, although it was recognized that the
antibiotic should be administered immediately after umbilical cord clamping.
|
|
|
|
Figure 1. Flow diagram of the existing and
revised systems for prescribing and administering perioperative antibiotic
prophylaxis at hospital A. The parallelogram is the initiating step in the
system, and the rounded rectangle is the end of the system. Diamonds represent
decision points; rectangles, actions.
|
|
|
|
|
|
|
Figure 2. Flow diagram of the existing and
revised systems for prescribing and administering perioperative antibiotic
prophylaxis at hospital B. The parallelogram is the initiating step in the
system, and the rounded rectangle is the end of the system. Diamonds represent
decision points; rectangles, actions.
|
|
|
Teams used these flowcharts to design and test changes in their systems,
monitoring the effect of these changes on the process indicators.
DATA AND STATISTICAL ANALYSIS
Surveillance data were recorded and analyzed using EpiInfo (version
6.02; USD Inc, Stone Mountain, Ga). In hospital B, the surveillance system
was interrupted for 2 months (months 15 and 16) because of a maternity leave.
Data from these 2 months are missing from the analysis. Data collection procedures
before and after this period were maintained as described previously.
The program was initiated at the same time in both hospitals. The entire
program was divided into 3 periods: the preintervention period (period 1),
the first intervention period (period 2), and the second intervention period
(period 3). Period 1 was 3 months long in each hospital. The length of periods
2 and 3 were different in the 2 hospitals because the start of each period
was defined by the date when major improvements in the system for prescribing
and administering antibiotic prophylaxis were undertaken. Period 2 was 7 months
long in hospital A and 9 months long in hospital B. Period 3 was 5 months
long in hospital A and 12 months long in hospital B.
Each delivery was regarded as an independent event because only a small
number of women delivered twice during the program.
Time series analysis was performed to determine the effect of the program
on monthly process indicator data and surgical site infection rates, adjusting
for the effects of time trends in the preintervention period and several covariates.34-35 Segmented linear regression models
were constructed by means of the PROC AUTOREG procedure in the SAS System
for Windows (release 6.12; SAS Institute Inc, Cary, NC).
Models contained 3 segments, 1 for each of the program periods. Each
segment was described by 2 independent variables (total of 6 variables for
the 3 segments). One variable represented the level of the segment (ie, the
y-intercept for the first segment and the join points of successive segments)
and the second represented the trend (ie, slope) of the segment. A change
in the level variable represented an immediate change at the beginning of
the period; a change in the trend variable represented a change over time
during the period. For example, implementation of a new protocol at the beginning
of an intervention period might have an immediate effect on a process indicator
(level change), and reinforcement of the protocol through additional staff
education may have a progressive effect during the intervention period (trend
change).
First-, second-, and third-order autoregressive terms were tested and,
if significant, were included in the models to control for autocorrelation.
A volume covariate (the total number of deliveries) and two case-mix covariates
(the percentage of deliveries by cesarean section and the percentage of cesarean
sections that were nonelective) were included in each model. These covariates
were included for several reasons. First, the participating hospitals are
understaffed and have limited material resources; therefore, large changes
in the volume of deliveries could affect care despite improvements in hospital
systems. Second, both changes in the characteristics of women presenting for
delivery and changes in antepartum care practices that affect the frequency
of cesarean section would be reflected in the case-mix covariates. Third,
the percentage of cesarean sections that were nonelective is an important
covariate in the analysis of surgical site infections to adjust for the presence
of labor and rupture of membranes.14
Nonsignificant level and trend variables (P>.05)
were excluded from the models by stepwise backward elimination. Autocorrelation
was assessed by means of the Durbin-Watson statistic.36
The adequacy of each model was tested by standard methods of residual analysis.37 Models using proportions of dependent and independent
variables were compared with models using counts of these variables. Because
the 2 approaches yielded nearly identical results and because models using
proportions are more easily interpretable, only the latter models are presented.
RESULTS
VOLUME AND CASE MIX OF DELIVERIES
The volume and case mix of deliveries at hospitals A and B are summarized
in Table 1. In hospital A, the
number of deliveries per month decreased during the program. There was no
change in the percentage of deliveries by cesarean section, but the percentage
of cesarean sections that were nonelective increased. In hospital B, the number
of deliveries per month increased during period 2, then decreased during period
3. The percentage of deliveries by cesarean section increased and the percentage
of cesarean sections that were nonelective decreased. Hospital B had a higher
percentage of delivery by cesarean section than did hospital A, probably as
a result of its status as a tertiary care obstetrics hospital.
|
|
|
|
Table 1. Volume and Case Mix of Deliveries in Hospitals A and B
|
|
|
IMPROVING SYSTEMS FOR PRESCRIBING AND ADMINISTERING PERIOPERATIVE ANTIMICROBIAL
PROPHYLAXIS
Monthly data for the process indicators and surgical site infection
rates are displayed in Figure 3
(hospital A) and Figure 4 (hospital
B). Lines in these figures represent the predicted values from the time series
models (Table 2).
|
|
|
|
Figure 3. Utilization and timing of perioperative
antibiotic prophylaxis and surgical site infection rates after cesarean section
at hospital A. Open squares represent the utilization indicator (percentage
of women undergoing cesarean section who received perioperative antibiotic
prophylaxis) and are plotted on the left y-axis. Open circles represent the
timing indicator (percentage of women receiving prophylaxis in whom the antibiotic
was administered within 1 hour of delivery) and are plotted on the left y-axis.
Solid diamonds represent the surgical site infection rate after cesarean section
and are plotted on the right y-axis. The hatched, dashed, and solid lines
represent the predicted values for the utilization indicator, the timing indicator,
and the surgical site infection rate after cesarean section, respectively,
as determined by the segmented linear regression models (Table 2). Period 1 was a baseline period. Periods 2 and 3 were successive
intervention periods (see the "CQI Methods" subsection of the "Methods" section
for a description of the interventions).
|
|
|
|
|
|
|
Figure 4. Utilization and timing of perioperative
antibiotic prophylaxis and surgical site infection rates after cesarean section
at hospital B. Open squares represent the utilization indicator (percentage
of women undergoing cesarean section who received perioperative antibiotic
prophylaxis) and are plotted on the left y-axis. Open circles represent the
timing indicator (percentage of women receiving prophylaxis in whom the antibiotic
was administered within 1 hour of delivery) and are plotted on the left y-axis.
Solid diamonds represent the surgical site infection rate after cesarean section
and are plotted on the right y-axis. The hatched, dashed, and solid lines
represent the predicted values for the utilization indicator, the timing indicator,
and the surgical site infection rate after cesarean section, respectively,
as determined by the segmented linear regression models (Table 2). Period 1 was a baseline period. Periods 2 and 3 were successive
intervention periods (see the "CQI Methods" subsection of the "Methods" section
for a description of the interventions).
|
|
|
|
|
|
|
Table 2. Segmented Linear Regression Models of Process Indicators for
Perioperative Antibiotic Prophylaxis and Surgical Site Infection Rates After
Cesarean Section in Hospitals A and B
|
|
|
Hospital A
In the existing system (Figure 1),
each obstetrician made a decision whether to prescribe prophylaxis and which
antibiotic to use. Because of problems with the availability of the antibiotic
in the labor and delivery unit and the pharmacy, family members were often
required to purchase the antibiotic outside the hospital.
During period 1 (Figure 3),
the median utilization and timing indicators were 71% (range, 65%-73%) and
24% (range, 16%-52%), respectively. There was an upward trend in the timing
indicator during period 1, but this change was not significant (Table 2). Antibiotics prescribed for prophylaxis were cephalothin
sodium (67% of regimens), penicillin G potassium (23%), ampicillin sodium
(9%), and penicillin and gentamicin sulfate (1%).
At the beginning of period 2, the department implemented a protocol
to routinely administer prophylaxis to all women undergoing cesarean section
(Figure 1, revised system 1), on
the basis of evidence of the efficacy of perioperative antibiotic prophylaxis
presented during the workshop.26-27
A first-generation cephalosporin was the antibiotic of choice. To improve
the availability of the antibiotic, the pharmacy director ensured that an
adequate supply of the antibiotic would be maintained on the labor and delivery
unit. The head nurse began including a vial of the antibiotic in the package
of surgical supplies that was transported with the woman to the operating
room.
These changes coincided with immediate increases in the utilization
(+31.6; P<.001) and timing (+62.2; P<.001) indicators (Table 2, Figure 3). Because the timing of antibiotic
administration was still suboptimal during period 2, the team reviewed the
system again and determined that the responsibility for administering prophylaxis
had not been assigned specifically. At the beginning of period 3, the protocol
was revised to specify that the anesthesiologist was responsible for administering
the antibiotic immediately after umbilical cord clamping (Figure 3, revised system 2). The timing indicator increased during
period 3, but this change was not significant. The utilization indicator degraded
slightly (-4.9; P<.001; Table 2) at the beginning of period 3.
During period 3 (Figure 3),
the median utilization and timing indicators were 95% (range, 89%-98%) and
96% (range, 82%-98%), respectively. Antibiotics prescribed for prophylaxis
during periods 2 and 3 were cephalothin (98% of regimens), ampicillin (1%),
and penicillin (<1%). The median time of administration was 5 minutes after
delivery (interquartile range, 0-23 minutes).
Hospital B
As in hospital A, in the existing system each obstetrician made a decision
about whether to prescribe prophylaxis and which antibiotic to use (Figure 4). This decision was made in the
operating room immediately before, during, or after the procedure. Because
a written prescription was required, administration of the antibiotic was
delayed if the decision to prescribe prophylaxis was made during or after
the procedure. Further delays were incurred if transport of the prescription
to the pharmacy or transport of the antibiotic back to the operating room
was slow.
During period 1 (Figure 4),
the median utilization and timing indicators were 36% (range, 31%-37%) and
70% (range, 50%-79%), respectively. The timing indicator had a significant
upward trend during period 1 (Table 2).
Antibiotics prescribed for prophylaxis were cephalothin (71% of regimens),
ampicillin (27%), clindamycin plus gentamicin (1%), and penicillin plus metronidazole
(1%). More than 1 dose of antibiotic was prescribed in 41% of regimens.
At the beginning of period 2, the department developed a prophylaxis
protocol (Figure 2, revised system
1) requiring use of prophylaxis for high-risk women only (women undergoing
a nonelective cesarean section or women with diabetes, preeclampsia, or other
medical conditions). It encouraged use of prophylaxis for low-risk women,
but allowed obstetricians to make their own decisions in this regard. A first-generation
cephalothin or ampicillin was the antibiotic of choice. The protocol specified
that the prescription should be written at the time of the decision to perform
a cesarean section to allow sufficient time to obtain the antibiotic from
the pharmacy.
The utilization indicator for all women increased gradually during period
2 (+5.4 per month; P<.001; Table 2, Figure 4). This
rate of increase was nearly identical among high-risk and low-risk women (data
not shown), and further review of the data showed that one third of high-risk
women undergoing cesarean section had not received prophylaxis. The experiences
of teams at both hospitals were discussed at a joint meeting, where hospital
A presented its protocol for prophylaxis of all women and its preliminary
data suggesting that infection rates had decreased during period 2 (Figure 3). The upward trend in the timing
indicator that began during period 1 in hospital B was not sustained during
period 2 (Figure 4, Table 2).
At the beginning of period 3, hospital B revised its protocol to require
use of prophylaxis for all women (Figure 4, revised system 2). In addition, a small pharmacy was established
in the operating room to dispense medications for patients undergoing surgical
procedures, including cesarean section.
These changes coincided with immediate increases in the utilization
(+7.1; P = .047) and timing (+15.2; P<.001) indicators (Table 2, Figure 4). The upward trend in the utilization
indicator continued during period 3, although at a lesser slope.
During period 3 (Figure 4),
the median utilization and timing indicators were 89% (range, 79%-98%) and
96% (range, 89%-100%), respectively. Antibiotics prescribed for prophylaxis
were ampicillin (67% of regimens), cephalothin (31%), clindamycin (2%), and
others (<1%). One dose of antibiotic was prescribed in 93% of the regimens.
The median time of administration was 5 minutes after delivery (interquartile
range, 0-10 minutes).
REDUCTIONS IN SURGICAL SITE INFECTION RATES
In hospital A (Figure 3),
the median surgical site infection rate was 10.5 per 100 cesarean sections
(range, 9.9-10.7) during period 1, 1.4 (range, 0-7.6) during period 2, and
0 (range, 0-3.0) during period 3. There was a large immediate decrease at
the beginning of period 2 (-9.8 infections per 100 cesarean sections; P<.001; Table 2).
In hospital B (Figure 4), the median
surgical site infection rate was 6.1 per 100 cesarean sections (range, 5.7-9.9)
during period 1, 5.9 (range, 4.3-10.6) during period 2, and 4.4 (range, 3.1-7.9)
during period 3. A downward trend began during period 1 and continued through
periods 2 and 3 (-0.1 infection per 100 cesarean sections per month; P = .042; Table 2).
COMMENT
This study demonstrates that multidisciplinary hospital quality improvement
teams can improve patient care systems and clinical outcomes for women undergoing
childbirth in resource-limited settings.
The significance of this work extends beyond optimizing perioperative
antibiotic prophylaxis for women undergoing cesarean section. Guidelines for
the evaluation of obstetric services in low- and middle-income countries have
been distributed widely,6-8
yet examples of how these guidelines have been used to improve obstetric care
are limited.38-39 Moreover, there
is little information about the actual process of improvement that clinicians
should use in optimizing care in their own hospitals or clinics. This program
illustrates how a general method for multidisciplinary teamwork and systems
analysis can be used to translate evidence-based guidelines into improved
patient care.
The key to the teams' success was their ability to analyze and improve
systems of care. This focus on systems has been termed "the key organizing
concept for an effective approach to improvement."40
While problems with the existing systems in each hospital seemed idiosyncratic,
they are representative of systems problems in all health care settings. According
to the schema of the Institute of Medicine's National Roundtable on Health
Care Quality,41 these problems included underuse
(eg, prophylaxis was underutilized), misuse (eg, less effective antibiotics
were used and the administration of the antibiotic was delayed), and overuse
(eg, too many doses of antibiotic prophylaxis were used). These problems are
often seen in systems that are complex and that involve multiple individuals
and a series of interdependent steps occurring over time in different locations.
As was seen in this program, improving these systems requires qualitative
and quantitative analysis, multidisciplinary problem solving, and often several
rounds of changes before improvement goals are met. In this program, the active
leadership of physicians and the opportunity for the teams to share their
experiences also likely contributed to their success.42-43
It was not the purpose of the study to compare the effectiveness of
specific changes in systems of care, nor was this appropriate given the study
design. Indeed, Berwick44 has argued against
such an approach, instead advocating use of rapid-cycle tests of small-scale
changes (ie, the PDSA cycle) and evaluating these changes in terms of their
contribution toward reaching an overall goal. Nonetheless, a few interinstitutional
comparisons are relevant. Improvement in the utilization of prophylaxis was
rapid in hospital A, where a protocol requiring use of prophylaxis for all
women undergoing cesarean section was implemented immediately. In contrast,
improvement occurred more slowly in hospital B, where a protocol was implemented
initially that required each obstetrician to decide whether a woman was at
high risk or low risk for infection and mandated prophylaxis only for high-risk
women. Hospital B ultimately implemented a protocol for universal prophylaxis,
because antibiotic utilization remained suboptimal with the initial approach
and the knowledge that the universal prophylaxis approach in hospital A appeared
to be lowering its infection rate. Both approaches were defensible by means
of the literature available at the time,26-27,45
although a recent meta-analysis supports universal prophylaxis.46
Systems improvements designed to increase the efficiency of antibiotic administration
also differed in each hospital, although both had a positive effect. However,
one could argue that the establishment of a small pharmacy in the operating
suite of hospital B was a more permanent change that will also benefit patients
undergoing other surgical procedures. Although their specific paths to improvement
differed, the fact that both hospitals achieved their goals is a powerful
example of the benefits of collaborative organizational learning and improvement
advocated by quality improvement experts.40, 43, 47
To guide their work, the teams used a CQI model, commonly referred to
as the Model for Improvement developed by Nolan and colleagues. Several CQI
models have been used to improve clinical practice,28, 48
but the Model has gained widespread acceptance in the United States, in part
because of its use by the Institute for Healthcare Improvement in its Breakthrough
Series.31, 40, 49 The
Model is appealing because it helps teams focus on an important, achievable
goal; use valid data to gauge improvement; and design, implement, and evaluate
rapid-cycle tests of changes in systems, a process akin to the scientific
method of experimentation. The Breakthrough Series has used this model to
assist groups of hospitals to optimally apply evidence-based interventions
by enabling them to exchange information and learn from each other's successes
and failures.31, 40, 49
The teams applied the Model and a collaborative learning approach in hospitals
where personnel had no previous formal training in CQI principles or methods,
demonstrating that broad-based training programs in CQI "culture" are not
a prerequisite for meaningful improvement. Instead, just-in-time training,
use of simple CQI tools, and facilitation by experienced individuals enabled
the teams to complete their work. While this approach was sufficient for this
demonstration project, more complex CQI initiatives involving chronic multisystem
diseases, collaboration among many departments, or diverse care settings may
require a greater up-front investment in the infrastructure and support for
CQI.
Other international projects are using CQI methods to improve the quality
of health care in low-resource settings.50-54
To our knowledge, this is the first report that has linked the use of well-defined
CQI methods with improvements in care practices and patient outcomes in resource-limited
settings.
The resource requirements of the program were modest. Surveillance of
infections, collection of other data, and data entry and analysis were performed
by hospital personnel largely during the course of routine nosocomial infection
surveillance mandated by existing hospital policies and governmental regulations.
Multidisciplinary team meetings were usually held as a part of previously
scheduled departmental meetings. With support from a modest extramural grant,
individuals with CQI experience provided training and guidance in CQI methods
to team members. The cost-effectiveness of the perioperative antibiotic prophylaxis
for cesarean section has been documented extensively,45, 55
and the actual antibiotics used (cephalothin and ampicillin) are relatively
inexpensive (approximately $1.50 for each 1-g dose).
The study had several limitations. First, there was no control group
to evaluate whether changes that were independent of the teams' activities
affected care practices or infection rates. Given the program's focus on improving
hospital systems, it was impossible to identify a suitable control group within
the same hospital. For logistical and ethical reasons, it was difficult to
use control hospitals in which process indicators and surgical site infection
rates were measured but no systems improvements were implemented. Nonetheless,
our analysis controlled for important threats to internal validity, such as
the effects of time trends in the preintervention period and changes in important
covariates. No effort was made during the program to change the management
of labor or prolonged rupture of membranes, and surgical techniques and obstetrician
characteristics (eg, number and responsibilities of staff and resident obstetricians)
remained constant. Consequently, these factors are not likely to have affected
our results substantially.
Second, feedback of process and outcome data to physicians and nurses
in the obstetrics departments was an integral part of the program and may
have had an effect independent of the systems changes implemented by the teams
(ie, a Hawthorne effect). In addition, the 2-day workshop was held in the
latter part of the preintervention period (period 1) and may have affected
the practices of some participants, even though changes in systems of care
had not yet occurred.
Third, the relatively small number of observations during the preintervention
period may have overestimated or underestimated preexisting trends in the
process indicator and surgical site infection data. While it would have been
helpful to have a longer preintervention period, it was not possible to collect
data retrospectively, because relevant data were not recorded reliably in
patient medical charts. A longer preintervention period would have also increased
the likelihood of contamination of the preintervention period as described
above. Nonetheless, in this case, the short preintervention period appears
to have biased our estimates of effects of the systems improvements toward
the null.
Finally, postdischarge surveillance of infections was not performed,
although women readmitted to the hospital during the postpartum period were
surveyed for evidence of infection. The number of infections occurring after
discharge that did not require readmission was likely to be relatively constant,
because there was little variation in the length of stay after cesarean section
in either hospital during the course of the program.
Although a recent meta-analysis concluded that a first-generation cephalosporin
or ampicillin is an adequate prophylactic agent,56
the increasing prevalence of antimicrobial resistance among bacteria causing
surgical site infections may lead some to question whether these antibiotics
are still adequate for prophylaxis. In addition, wider use of prophylaxis
may conceivably increase the prevalence of antimicrobial resistance. Although
these are legitimate concerns, they are unsubstantiated at present.
Childbirth is a leading reason for hospitalization in most low- and
middle-income countries,57 so programs to improve
the quality of obstetric care have broad-based importance. Although this program
focused on the prevention of surgical site infection after cesarean section,
its methods could be used to improve other critical aspects of obstetric care,
such as the management of obstructed labor, identification and management
of preeclampsia, and prevention and management of postpartum hemorrhage. Data
from the Centers for Disease Control and Prevention indicate that 73% of women
undergoing cesarean section in the United States from 1994 to 1999 received
perioperative antibiotic prophylaxis (Teresa C. Horan, MPH, written communication,
July 24, 2001). Use of these methods could also increase the use of antibiotic
prophylaxis and potentially lower infection rates after cesarean section in
the United States.
AUTHOR INFORMATION
Accepted for publication March 29, 2001.
This study was supported by the International Society for Infectious
Diseases. Dr Weinberg received support from the Paul Schliesman Memorial Traveling
Fellowship and the Von L. Meyer Award, both administered by Children's Hospital,
Boston, Mass.
Corresponding author: W. Charles Huskins, MD, MS, Mayo Clinic, 200
First St SW, Rochester, MN 55905 (e-mail: huskins.charles{at}mayo.edu).
From the Division of Infectious Diseases, Children's Hospital, Boston,
Mass (Drs Weinberg, Goldmann, and Huskins); Department of Medicine, Hospital
Simón Bolívar (Drs Fuentes and Lozano and Ms Goethe), Escuela
Colombiana de Medicina (Drs Fuentes and Lozano), Instituto Materno Infantil
(Drs Ruiz, Angel, and Gaitan and Ms Parra), and Departamento de Ginecologia
y Obstetricia, Universidad Nacional de Colombia (Drs Ruiz, Angel, and Gaitan),
Santafé de Bogotá, Colombia; Department of Obstetrics and Gynecology,
Beth Israel Deaconess Medical Center and Department of Obstetrics, Gynecology,
and Reproductive Biology, Harvard Medical School, Boston (Dr Hellerstein);
Department of Ambulatory Care and Prevention Drug Policy Research Group, Harvard
Medical School and Harvard Pilgrim Health Care, Boston (Dr Ross-Degnan); and
Department of Pediatrics, Harvard Medical School, Boston (Drs Goldmann and
Huskins). Dr Weinberg is now with the Division of Quarantine, National Center
for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta,
Ga. Dr Huskins is now with the Department of Pediatrics and Adolescent Medicine,
Mayo Clinic, Rochester, Minn.
REFERENCES
| |
1. Murray CJ, Lopez AD. Global and regional cause-of-death patterns in 1990. Bull World Health Organ. 1994;72:447-480.
ISI
| PUBMED
2. Walsh JA, Feifer CM, Measham AR, Gertler PJ. Maternal and perinatal health. In: Jamison DT, Mosley WH, Measham AR, Bobadilla JL, eds. Disease Control Priorities in Developing Countries. New York, NY: Oxford
University Press; 1993:363-390.
3. World Bank. World Development Report 1993: Investing in Health. New York, NY: Oxford University Press; 1993.
4. Nowak R. New push to reduce maternal mortality in poor countries. Science. 1995;269:780-782.
FREE FULL TEXT
5. AbouZahr CL. Lessons on safe motherhood. World Health Forum. 1998;19:253-260.
ISI
| PUBMED
6. World Health Organization. Essential Elements of Obstetric Care at First Referral
Level. Geneva, Switzerland: World Health Organization; 1991.
7. Maine D, Akalin MZ, Ward VM, Kamara A. The Design and Evaluation of Maternal Mortality Programs. New York, NY: Center for Population and Family Health, Columbia University;
1997.
8. UNICEF. Guidelines for Monitoring the Availability and Use
of Obstetric Services. Geneva, Switzerland: UNICEF, WHO, and UNFPA; 1997.
9. Nolan TW. Understanding medical systems. Ann Intern Med. 1998;128:293-298.
FREE FULL TEXT
10. Shortell SM, Bennett CL, Byck GR. Assessing the impact of continuous quality improvement on clinical
practice: what it will take to accelerate progress. Milbank Q. 1998;76:593-624.
FULL TEXT
|
ISI
| PUBMED
11. Blumenthal D, Kilo CM. A report card on continuous quality improvement. Milbank Q. 1998;76:625-649.
FULL TEXT
|
ISI
| PUBMED
12. Huskins WC, Soule BM, O'Boyle C, Gulacsi L, O'Rourke EJ, Goldmann DA. Hospital infection prevention and control: a model for improving the
quality of hospital care in low and middle income countries. Infect Control Hosp Epidemiol. 1998;19:125-135.
ISI
| PUBMED
13. Chalmers I, Enkin M, Keirse MJNC. Effective Care in Pregnancy and Childbirth. Oxford, England: Oxford University Press; 1989.
14. Mead PB. Prevention and control of nosocomial infections in obsetrics and gynecology. In: Wenzel RP, ed. Prevention and Control of Nosocomial
Infections. 3rd ed. Baltimore, Md: Williams & Wilkins; 1997:995-1016.
15. Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification
of CDC definitions of surgical wound infections. Am J Infect Control. 1992;20:271-274.
FULL TEXT
|
ISI
| PUBMED
16. Casey BM, Cox SM. Chorioamnionitis and endometritis. Infect Dis Clin North Am. 1997;11:203-222.
FULL TEXT
|
ISI
| PUBMED
|
