Among posthysterectomy morbid events, febrile morbidity is the most commonly reported adverse event. Postoperative febrile morbidity lasting more than 24 hours requires evaluation and is often followed by empiric treatment with antibiotics for suspected infection. Hence, reducing postoperative fever may significantly curtail expenditures by reducing length of stay and the need for evaluation and empiric treatment.
In the Collaborative Review of Sterilization study, of 1851 women undergoing elective hysterectomies for benign disease, 7.2% of patients experienced unexplained fever with vaginal hysterectomy, compared with 16.8% with abdominal hysterectomy.1 In that study, the difference in febrile morbidity was attributable to a greater use of prophylactic antibiotics in the former group. In a meta-analysis of antibiotic prophylaxis in abdominal hysterectomy, Mittendorf et al2 concluded that preoperative antibiotics were highly effective in preventing serious infections and concomitant morbidity associated with total abdominal hysterectomy.
Since this meta-analysis, many gynecologists and infectious diseases experts recommend prophylactic antibiotics for all hysterectomies. In his classic animal studies, Burke3 demonstrated that the effective period of prophylactic antibiotics extended from 1 hour before to up to 3 hours after the surgical incision. However, human data are limited regarding the most appropriate timing of prophylactic antibiotics. Most experts report that prophylactic antibiotic should be present in the wound, preferably before incision, and should reach a therapeutic concentration in the tissues of the operative site.4 Thus, administration within the 60-minute window before the surgical incision is recommended. Even with the more widespread use of prophylactic antibiotics, febrile morbidity is still a relatively common adverse outcome. Since the more widespread use of prophylactic antibiotic therapy for hysterectomy, risk factors for febrile morbidity have not been objectively assessed.
The specific aim of our study was to evaluate risk factors for febrile morbidity after hysterectomy. Our hypothesis was that the abdominal approach to hysterectomy results in higher risk for febrile morbidity than either the vaginal or laparoscopic approach, independent of other risk factors for febrile morbidity. We chose to study febrile morbidity instead of infectious morbidity because fever alone can increase hospital stay and costs. Our secondary aim was to assess the use of prophylactic antibiotics for hysterectomy procedures in our population. The secondary hypothesis was that prophylactic antibiotic therapy is not being used routinely, and if used, it is often inappropriately timed. Our ultimate goal was to increase our efforts to prevent postoperative morbidity by applying clinical algorithms and preventive measures, and thus improve the quality of care provided at our institution.
MATERIALS AND METHODS
We performed a retrospective cohort study to assess risk factors for febrile morbidity as part of a quality improvement project at the George Anderson Outcomes Measurement Unit at Women & Infants Hospital. Before initiating the study, we obtained institutional review board approval for chart abstraction. Our sample included all women who had undergone a hysterectomy for a benign condition at Women and Infants Hospital between January 1997 and September 1997. We excluded women who had either a pre- or postoperative diagnosis of malignancy of the genital tract. Women were categorized as having one of the following surgical approaches to hysterectomy by the corresponding International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM), procedure codes: abdominal (68.3), laparoscopic-assisted vaginal (68.4), or vaginal (68.5). In our sample, 408 women had abdominal, 90 laparoscopic-assisted vaginal hysterectomy, and 188 vaginal hysterectomy. Data were abstracted from the medical records by two trained data abstractors. Data abstractors were blinded to the research hypothesis. All the extracted information was then entered into a computerized database and verified for accuracy by the data management team.
We collected demographic and clinical information from the medical records, including women’s age, race/ethnicity, health insurance status, weight, height, and parity. Race/ethnicity was recorded as white, African American, Hispanic, and other (including unknown). Health insurance status was categorized as three major groups: private insurance or health maintenance organization, Medicaid/Medicare, and no insurance or self-pay. Body mass index (BMI) was calculated as weight (kg)/height (m2). Information regarding estimated blood loss during surgery, total operative time, use and timing of prophylactic antibiotic administration, and evidence of postoperative febrile morbidity was also obtained from patients’ medical records. Cefazolin is the most commonly used prophylactic antibiotic used at our institution and is usually provided in a single dose. Estimated blood loss was categorized as loss of 750 mL or more versus less than 750 mL. Although there is no standard definition in the medical literature for postoperative febrile morbidity in gynecologic surgery, we used the most commonly quoted definition of a temperature of more than 101F on a single occasion or 100.4F or more on two occasions more than 24 hours after hysterectomy.5
For our sample size calculations, we combined the vaginal and laparoscopic-assisted vaginal hysterectomy surgical approaches into a single category because the rates of medical complications, including febrile morbidity, after these two surgical approaches are quite similar.1,6 The sample size calculations were based on detecting a two-fold increase, from 7.5% to 15%, in the rate of febrile morbidity between vaginal (including laparoscopic-assisted vaginal hysterectomy) and abdominal hysterectomy approaches. Assuming a two-tailed α of .05, 170 women per group will provide 80% power and 225 women per group would provide 90% power to detect these differences in the rate of febrile morbidity. We performed bivariate and multivariable analyses to examine the relationships between potential risk factors and febrile morbidity. Bivariate comparisons were made by means of two sample t tests for continuous variables and by χ2 tests for comparing the differences in proportions in the categoric variables.
By means of logistic regression, we developed multivariable models in which febrile morbidity (present/absent) was analyzed as the dependent variable with independent variables age, health insurance type, race, BMI, abdominal versus laparoscopic-assisted vaginal hysterectomy plus vaginal hysterectomy, estimated blood loss (categorized as 750 mL or more versus less than 750 mL), and use of prophylactic antibiotic administration (before or after incision versus none). Finally, we compared the median length of hospital stay for those women who contracted a fever after hysterectomy compared with those who did not, both overall and by surgical approach. By means of the Kruskal-Wallis rank-sum test, we tested whether there were differences in hospital length of stay on the basis of the presence or absence of posthysterectomy fever for the full sample as well as within each surgical approach.
Among the 686 women who had a hysterectomy during the study period, 408 women (60%) had a hysterectomy by the abdominal approach, and 90 (13%) and 188 (27%) had a hysterectomy by laparoscopic-assisted vaginal hysterectomy and vaginal surgical approach, respectively. The age range of the study population was 25 to 83 years. Estimated blood loss ranged from minimal to 2500 mL. Table 1 contains the demographic and other characteristics of our study population, stratified by surgical approach to hysterectomy. Compared with women who underwent vaginal hysterectomies, women who had abdominal hysterectomies were more likely to be younger (17% versus 39% more than 50 years old), have private health insurance (87% versus 80%), be nulliparous (23% versus 7%), and have higher BMI (31% versus 16% having BMI more than 30). Women who underwent laparoscopic-assisted vaginal hysterectomy had a similar age distribution as women who underwent abdominal hysterectomies, as seen in Table 1. Differences between the other demographic and clinical factors were approximately the same as in the abdominal versus vaginal approach comparison. Women undergoing abdominal hysterectomies or laparoscopic-assisted vaginal hysterectomy had higher rates of estimated blood loss as compared with women who had vaginal hysterectomies (16% and 12% versus 8% with estimated blood loss 750 mL or more). On average, although abdominal and vaginal hysterectomies had similar operative times (33% versus 34% with operative time more than 120 minutes), laparoscopic-assisted vaginal hysterectomy required more surgical time (54% with operative time more than 120 minutes).
In Table 2, we present the risk of febrile morbidity, the proportion of antibiotics use and timing, and hospital length of stay by surgical approach. Of the 686 women, 96 (14%) had postoperative febrile morbidity. The risk of febrile morbidity was significantly higher in the abdominal compared with laparoscopic-assisted vaginal hysterectomy or vaginal hysterectomy groups (18% versus 9% versus 8%).
Overall, 342 women (50%) received prophylactic antibiotics before the surgical incision. The proportions varied by surgical approach, with women who had an abdominal hysterectomy being least likely to receive antibiotics before surgical incision (45%) followed by women who had laparoscopic-assisted vaginal hysterectomy (53%), compared with women who had vaginal hysterectomies (59%). Among the cases where prophylactic antibiotics were not appropriately administered, 198 women (29%) received antibiotics after incision, but 127 of all women in the sample (19%) did not receive antibiotic medication at all. The risk of not receiving any antibiotics differed by surgical approach: Nearly 27% of patients who underwent abdominal hysterectomies received no antibiotic therapy, compared with only 4% of women who underwent vaginal procedures and 11% of women who underwent laparoscopic-assisted vaginal hysterectomy. Although most of the women who received prophylactic antibiotics were administered the medication within 60 minutes before incision, 10% received the antibiotics more than 60 minutes before incision. Among women who received prophylactic antibiotics, cefazolin was the most common antibiotic used (68%).
In Table 3, we present the proportion of cases of febrile morbidity and the crude and adjusted odds ratios (ORs) for each risk factor under consideration. Except for surgical approach and estimated blood loss, none of the other patient characteristics were predictive of febrile morbidity. Women who had laparoscopic-assisted vaginal hysterectomy were 1.2 times as likely to experience fever after hysterectomy compared with women who had vaginal hysterectomies. The odds for women who underwent abdominal compared with vaginal hysterectomy was even higher, being nearly three times as likely to experience fever after surgery. These ORs were approximately the same even after adjustment for the other suspected risk factors. Women with higher-than-expected blood loss (estimated blood loss 750 mL or more versus estimated blood loss less than 750 mL) were at increased risk for febrile morbidity, with an OR of 2.4. However, in the adjusted model, the OR increased to 3.7. Therefore, women who had excess blood loss of 750 mL or more were 3.7 times as likely to experience a fever after hysterectomy, after adjusting for age, insurance, parity, BMI, operative time, prophylactic antibiotic, and surgical approach.
In Figure 1, we present the median length of stay for those women who had a fever after hysterectomy compared with those who had a fever after surgery. Consistently, across all surgical approaches, women who experienced fever had longer lengths of stay. For all comparisons of length of stay by febrile morbidity status, overall and within groups on the basis of surgical approach, P values were less than .01.
Febrile morbidity after surgery may be due to an operative site infection; it may be the result of an infection remote from the operative site; or it may be unexplained.1 Rates varying from 10% to 30% are found in the literature for posthysterectomy febrile morbidity.1,5,6 In our study, the incidence of febrile morbidity in the abdominal, laparoscopic-assisted vaginal hysterectomy, and vaginal hysterectomy groups were 18%, 9%, and 8%, respectively, which is a risk of more than 2:1 with abdominal versus laparoscopic-assisted vaginal hysterectomy/vaginal hysterectomy. These results are consistent with the existing literature.
It is widely known that prophylactic antibiotics are very effective in providing protection against postsurgical febrile morbidity and infection.2,4,7 Higher risk of febrile morbidity attributed to the abdominal approach may be due in part to the relative lack of appropriately timed administration of prophylactic antibiotics. In 1999, Shackelford et al8 proposed a model for predicting febrile morbidity, which consisted of the factors of blood loss, uterine weight, and parity. Unfortunately, because this study was limited to vaginal hysterectomies, it could not address the relationship of febrile morbidity to the specific surgical approach.
We found disparity in the use and timing of prophylactic antibiotic administration by surgical approach to hysterectomy. Additionally, the rate of administration of prophylactic antibiotic in the 60-minute time frame before surgery was lower in the abdominal hysterectomy group compared with the laparoscopic-assisted vaginal hysterectomy/vaginal approach (Table 2). We recognized that an opportunity for improvement existed if we could establish a clinical protocol that would require antibiotic prophylaxis before hysterectomy unless the clinician specifically canceled the order.
In general, surgical approach, blood loss at surgery, and use and timing of the administration of prophylactic antibiotics represent relatively modifiable elements of health care delivery in hysterectomy. In our multivariable analysis, we found that the abdominal surgical approach and excess blood loss were strong risk factors for febrile morbidity. Our study also indicates that the effect of abdominal approach and higher blood loss on febrile morbidity remained strong risk factors, even after adjusting for age, race/ethnic background, parity, BMI, health insurance type, operative time, and antibiotic use and timing.
Our study has several limitations. Because this was a retrospective study, we were limited to the variables collected in the patient chart. There may be other important but unmeasured factors that could be independent risk factors for incidence of febrile morbidity, but we were unable to assess their relationships. Additionally, these factors might also be related to the type of surgical approach the women received and therefore could potentially explain the relationship between abdominal hysterectomy and increased risk of febrile morbidity. Information regarding race and ethnicity was limited in our retrospective chart review with a relatively large percentage of patients listed under the category of other/unknown.
Another limitation was the lack of power to adequately measure the impact of appropriate timing of prophylactic antibiotic administration on febrile morbidity. It is also possible, given the observational design, that physicians were more likely to use prophylactic antibiotics in patients at highest risk for infection (eg, higher BMI, greater blood loss), thus limiting our ability to evaluate the effectiveness of prophylactic antibiotics. Future large-scale prospective studies attempting to assess the rates of febrile morbidity by surgical approach should evaluate both the timing of antibiotics and the specific types of antibiotics used for prophylaxis. A final limitation of our study was that we did not examine the impact of supplemental procedures (eg, colporrhaphy, urogynecologic procedure) or additional factors such as suprapubic drainage or use of indwelling catheters on febrile morbidity after various different surgical approaches.
Febrile morbidity, although limited as a screening tool for postoperative infection,8,9 is associated with longer lengths of stay and increased costs, especially, in patients who undergo abdominal hysterectomy (Figure 1). In our sample, women who experienced fever after hysterectomy had higher median lengths of stay after surgery compared with women who did not experience fever (median length of stay was 4 days when fever was present versus 3 days when fever was absent). This difference remained consistent when we stratified by surgical approach. Those women who had abdominal surgery and had a postoperative fever had the longest median length of stay.
Our results suggest that a reduction in the rate of febrile morbidity can be achieved by means of the following strategies: 1) using the vaginal approach whenever clinically appropriate; 2) paying meticulous attention to blood loss at surgery; and 3) liberally using and appropriately timing prophylactic antibiotics. Focusing on the modifiable risk factors can help reduce the costs associated with undue length of stay attributable to febrile morbidity.