We retrospectively collected operating room data on 988 consecutive
primary total hip and knee arthroplasties at one tertiary care teaching
hospital. The protocol used in the Centers for Medicare and Medicaid Services
Premier Hospital Quality Incentive Demonstration Project
(,
)
was employed to select the case s from billing data on all hip and knee
replacements performed in one calendar year. Only unilateral primary total hip
and knee arthroplasties were selected. Patients with a history of trauma (for
example, a hip fracture) were excluded.
The operating room data were recorded prospectively as part of a continuous
quality-improvement project. The data include operating room times such as the
in-room time as well as the times when the anesthesia was started, when
induction was started, when induction was complete, when the incision was
made, when the procedure was completed, and when antibiotics were
administered. The type of antibiotic was recorded as well. The time at which
the infusion of antibiotics was started was recorded, by the anesthesiologist,
as the dosing time. Additional data include an identifier for the surgeon, an
identifier for the anesthesiologist, and the operating room number. We
collected International Classification of Diseases, Ninth Revision (ICD-9)
diagnostic
information6 from
the discharge summary data and calculated the Charlson comorbidity
index7.
We randomly selected 10% (ninety-nine) of the cases and confirmed the
operating room data and pharmacy data with the paper charts; no discrepancies
were identified. No patient had antibiotics withheld for a clinical reason,
such as suspicion of infection or a need for intraoperative cultures.
Preoperative administration of antibiotics is a part of all surgeons'
protocols at our institution. We also reviewed the charts of all patients who
had not received any antibiotics.
We then calculated the time in minutes between the antibiotic dosing and
the time of the surgical incision. Preoperative antibiotic dosing was
considered optimal when the antibiotics were administered within sixty minutes
before the surgical incision, as required by Medicare. We considered the
antibiotic dosing to be suboptimal if the drugs were administered too early
(more than sixty minutes before the incision) or too late (after the
incision). We did not analyze the appropriateness of the type of antibiotic
that was given because allergy information was not well documented for a large
percentage of the cases in the computer system, and this is not yet a
pay-for-performance measure. We did not search for an increased rate of
infection in patients who had not received antibiotics because the sample size
necessary (70,000 patients to have an 80% power to detect a 1% change) was not
feasible.
Statistical analysis was performed with use of SPSS software (version 10.0;
SPSS, Chicago, Illinois). For bivariate analysis, differences between groups
were analyzed with the chi-square test for categorical variables and with the
independent sample t test for continuous variables. Variables that were
significant at the p < 0.05 level on univariate analysis were included in a
multivariate analysis with use of logistic regression to calculate odds ratios
and 95% confidence intervals predicting optimal antibiotic use.
Baseline Measures
The characteristics of the study population are shown in
Table I. Thirteen percent (127)
of 988 patients undergoing elective hip or knee arthroplasty did not receive
optimal antibiotic treatment (within one hour before the surgical incision).
The antibiotics were given too early in 74% (ninety-four) of the 127 cases of
suboptimal antibiotic use, and they were given after the incision had been
made in 22% (twenty-eight) of the cases. If the optimal window was increased
to two hours before the incision, 3.4% (thirty-four) of the 988 patients would
have received suboptimal antibiotic treatment.
Five patients (0.5% of all patients and 4% of those who received suboptimal
antibiotic treatment) had no documented antibiotic prophylaxis. Chart review
confirmed that all of these patients had not received antibiotics (or it had
not been documented by nursing and anesthesia staff). None of the five
patients had a documented clinical reason for the antibiotics to have been
withheld. No infection occurred in these five patients. Review of the charts
revealed that all patients had received antibiotics for twenty-four hours
postoperatively, per our provider-order entry standard order sets.
Factors Associated with Suboptimal Antibiotic Dosing
Bivariate analysis revealed several factors to be associated with a greater
propensity to administer antibiotics outside of the one-hour window. Patients
undergoing total hip arthroplasty were more likely to receive suboptimal
antibiotic treatment than were those undergoing total knee arthroplasty (p
< 0.008). Patients with a long induction time were also significantly more
likely to receive antibiotic treatment outside of the one-hour window (p <
0.001). Patients treated by high-volume surgeons (those performing fifty total
hip or knee replacements or more per year) were more likely to receive optimal
antibiotic therapy (p < 0.044).
The type of antibiotic was not associated with the likelihood of suboptimal
antibiotic dosing. However, vancomycin, which is not stocked in the anesthesia
cabinet, was given significantly later than cefazolin, with the time of
administration averaging twenty-five minutes before the incision for
vancomycin compared with thirty-three minutes for cefazolin (p < 0.007),
but it was not given outside the one-hour window. Vancomycin also requires a
longer infusion time than cefazolin. We observed no effect of the case order
within the day or with regard to the day of the week or the month of the
year.
Multivariate Analysis
The independent risk factors for suboptimal antibiotic dosing are shown in
Table II. Both induction time
(odds ratio = 0.97, 95% confidence interval = 0.95 to 0.98) and total hip
arthroplasty (odds ratio = 0.67, 95% confidence interval = 0.45 to 0.99)
proved to be independently associated with antibiotic dosing outside of the
one-hour window.
The surgeon volume was no longer significant on multivariate analysis
because the high-volume-surgeon variable was strongly associated with lower
induction times (p < 0.001) and was weakly associated with a case profile
that included more knee arthroplasties (p = 0.073).
Personnel Factors
Analysis revealed a large range of suboptimal antibiotic dosing rates among
the surgeons. These rates ranged from 11% (thirty-four of 308) to 22.0% (nine
of forty-one). One surgeon had a significantly lower rate of suboptimal use of
antibiotics than did the general population of surgeons (p < 0.04,
chisquare test). That rate remained significantly lower when other positive
factors (induction time and procedure type) were controlled for in the
multivariate logistic regression models (odds ratio = 0.643, 95% confidence
interval = 0.42 to 0.98).
Analysis of the cases according to the anesthesiologist showed the rates of
suboptimal use of antibiotics to range from 8.1% (twenty of 246) to 23.4%
(twenty-six of 111). One anesthesiologist had a significantly lower rate of
suboptimal use of antibiotics (p < 0.002), and one anesthesiologist had a
significantly higher rate (p < 0.021). When surgeon, anesthesiologist,
induction time, and procedure volume were analyzed together in the same
multivariate model, the surgeon contribution was no longer significant (odds
ratio = 0.718, 95% confidence interval = 0.47 to 1.1). The anesthesiologist's
contribution was, however, the most significant factor (odds ratio = 0.53, 95%
confidence interval = 0.31 to 0.88).
Case Complexity and the Pay-for-Performance Measure
We observed that surgeons who received the highest marks (surgeons in the
highest decile of performance) for the antibiotic pay-for-performance measure
tended to perform surgery on patients with less complex conditions. Those
surgeons treated patients whose mean Charlson comorbidity indexes were lower
than the indexes of the patients treated by the other surgeons (p < 0.013).
The time required for anesthesia preparation, another measure of case
complexity, was also shorter for the patients of the surgeons in the highest
decile (p < 0.0001). High-volume surgeons (those performing fifty total
joint arthroplasties per year or more) also tended to treat less complex cases
as measured by lower Charlson comorbidity indexes (p < 0.001).
We found that 127 (13%) of 988 consecutive patients undergoing
elective primary knee or hip arthroplasty at our institution during a one-year
period did not receive preoperative antibiotics within one hour before the
incision. These results are in line with national benchmark data showing a
mean of 16% of patients not receiving antibiotics within one hour before the
incision
().
Patients undergoing hip arthroplasty and those with longer induction times
were more likely to receive antibiotics outside of the one-hour window. We
also noted that higher-volume surgeons had lower rates of suboptimal use of
antibiotics, although this effect was not independent. Certain
anesthesiologists had significantly higher rates of optimal antibiotic
dosing.
The trend toward the institution of pay-for-performance measures is strong
and unlikely to
dissipate5. The
development of evidence-based measures along with great interest on the part
of payers to reward better care is driving the United States health-care
system toward pay for
performance8.
Physicians must take the lead in designing systems that have clinical
merit—i.e., systems based on accurate measures that are likely to
improve patient care. Orthopaedic surgeons should play a role in the
development of pay-for-performance measures for hip and knee replacement
surgery. We analyzed one such measure, optimal timing of antibiotic dosing.
Antibiotic dosing within one hour before surgery has already been proposed by
the Centers for Medicare and Medicaid Services as a pay-for-performance
measure on the basis of national panel recommendations
().
Public reporting of these measures is likely to occur
soon9.
There is ample Level-I evidence supporting the use of prophylactic
antibiotics for patients undergoing total joint
arthroplasty10-12.
The evidence supporting administration of antibiotics within one hour before
the surgical incision is not as strong. The data primarily come from studies
of animals13 and
the general surgery literature. Platt et al. also demonstrated the importance
of giving antibiotics within two hours before the incision in a prospective
observational
study14. In a
randomized prospective study of patients undergoing breast and hernia surgery,
Classen et al. demonstrated that those who had received antibiotics thirty
minutes before the surgery had a decreased rate of wound infection (odds ratio
= 0.52, 95% confidence interval = 0.32 to
0.84)15. To our
knowledge, no study has shown post-incision administration of antibiotics to
be effective.
Given the evidence that prophylactic preoperative antibiotics are
effective, it is surprising that 13% of the patients at our institution did
not receive optimal antibiotic prophylaxis and 0.5% of the patients may have
received no antibiotics. While the rate is low, it begs the question as to
what techniques could be used to reduce the rate to 0%. Ultimately, the
administration of antibiotics before surgery depends on several individuals
remembering that antibiotics are indicated, carrying out the task in a timely
fashion, and documenting it accurately. Additional safeguards, such as adding
the question "Did the patient get antibiotics?" to the now
required pause prior to surgery, can be
instituted16 but
may be inadequate. Also, technologic solutions that provide automated
reminders to administer antibiotics and safeguards to prevent surgeons from
beginning operations without the patient having received antibiotics are
reasonable.
It is not surprising that patients undergoing total hip replacement were
more likely than patients undergoing total knee replacement to be given
antibiotics too early (p < 0.008). Total hip replacement requires more
preparation in terms of patient positioning and sometimes the placement of
invasive lines than does total knee replacement. For hip replacement at our
institution, the mean time (and standard deviation) required for induction,
positioning, and draping was 54 ± 12 minutes. Thus a substantial number
of patients who received antibiotics before entry into the operating room did
not meet the one-hour performance measure. With these data in mind, surgeons
and anesthesiologists may opt to administer antibiotics for hip replacement
after the patient is positioned but before the wound is sterilely prepared,
rather than when the patient is called for by the operating-room personnel or
at the start of induction.
It seems that increased medical complexity of the case is related to a
decreased rate of optimal antibiotic dosing. Patients with more complex
conditions tend to need more invasive lines, require slower induction, and
thus manifest longer induction times. Long anesthesia time is associated with
greater case
complexity17-20.
We found that longer anesthesia induction times were associated with higher
rates of suboptimal antibiotic dosing. It is likely that the additional
confounding issues related to case complexity lead to difficulty in
remembering to administer antibiotics at the correct time. For example, if the
patient is hypotensive on induction, it is likely that the anesthesiologist
will be otherwise occupied and may miss the antibiotic window. Again, systems
to prompt physicians to administer antibiotics would be of greater value for
complex cases. We specifically excluded revision hip and knee replacements
from this study because many surgeons elect to withhold antibiotics to obtain
intraoperative specimens for culture.
A side issue is that institutions that focus on less complex surgical
procedures and cases would be more likely to be rewarded under a
pay-for-performance scheme. In our study, the surgeons who had the highest
rates of optimal antibiotic dosing also had patients with lower-than-average
Charlson comorbidity indexes (p < 0.013) as well as shorter anesthesia
preparation times (p < 0.0001).
We also noted that surgical volume tended to contribute in a positive way
to the rate of optimal antibiotic dosing. High-volume surgeons were more
likely to administer antibiotics optimally. The role of surgical volume in
improved outcomes of joint replacement is controversial but has been
demonstrated in a number of
studies21-25.
Thus, pay-for-performance measures will likely tend to reward higher-volume
institutions.
Our study was limited by a small sample size and by the single-institution
nature of the study population. Furthermore, because the study included fewer
than thirty surgeons, it had limited power to truly address the surgeon-volume
issue. Also, we cannot address whether the patients actually did not receive
antibiotics or whether the documentation was incorrect. However, in
pay-for-performance systems, not documenting the administration of a
medication is the same as not giving it. Despite the small numbers, our data
provide guidance about specific interventions that may improve the quality and
documentation of the timing of antibiotic administration for patients
undergoing elective arthroplasty, but the findings need to be confirmed in
other settings.