Temperature elevations during the postoperative period require evaluation for infection. Postoperative fever is common and does not always indicate infection. In 1970, Porges1 noted, “the patient's temperature may rise to 101 F by the second postoperative day without infection.” In many instances, temperature elevations are part of the normal physiologic response to surgical trauma and can be beneficial in the healing process.2,3
No standard definition exists for fever or febrile morbidity. In a review by Hemsell,4 32 separate definitions of fever and febrile morbidity were noted for gynecologic surgical patients. The definition used most commonly is two temperature elevations greater than or equal to 38.0C(100.4 F) 6 hours apart, excluding the first 24 hours after surgery.5,6 Using these criteria, Dicker et al7 noted febrile morbidity after 15.3% of vaginal hysterectomies and 32.3% of abdominal hysterectomies. Infection was found in 52% of those febrile women. The screening value of fever for detecting postoperative infection was not evaluated. The purpose of this study was to determine the screening value of fever for detecting infections after vaginal surgery.
Materials and Methods
Medical records of all women who had vaginal surgery at the M. S. Hershey Medical Center from September 1988 through June 1995 were abstracted at the time of discharge. Women were excluded from the study if they had defined sources of temperature elevation unrelated to procedures, such as preoperative infection or hyperthyroidism. General information was recorded, including age, operative time, estimated blood loss, patient weight, parity, uterine weight, prophylactic antibiotics, bladder drainage, estrogen exposure, history of gynecologic surgery, and types of procedures. Vaginal surgery included vaginal repair, vaginal hysterectomy with or without concomitant vaginal repair, or urethropexy. Premenopausal women (age less than 55 years without previous oophorectomy) and women receiving oral or transdermal estrogen replacement therapy at the time of admission were considered estrogen exposed.
After transfer of patients from the postanesthesia care unit, oral temperatures were taken every 4 hours during postoperative hospitalization, using an electronic thermometer with disposable sheaths. Febrile morbidity was defined as temperature greater than 38.0C (100.4F) on two separate occasions, excluding the first 24 hours after surgery.5,7 Postoperative evaluation of fever included physical examination, urinalysis, and complete blood count. Blood cultures, sputum cultures, and chest radiographs were taken at the discretion of the evaluating physicians.
The outcome variable, infectious morbidity, was defined as clinical infection in the immediate postoperative period requiring therapeutic antibiotics or abscess drainage. Urinary tract infection was defined as postoperative urgency, frequency, or dysuria with concomitant laboratory evidence of bacteriuria. Cuff infection was defined as a tender, hyperemic vaginal cuff with or without purulence. Suprapubic catheter site infection was defined as erythema, edema, induration, or purulence at the catheter exit site. Pelvic abscess was defined as fluid collection in the pelvis documented by examination, ultrasound, computed tomography, or drainage. Clostridium difficile colitis was diagnosed by the presence of gastric cramps, diarrhea, and Clostridium difficile toxin in stool samples. Sinusitis was diagnosed by rhinorrhea and tenderness on palpation of the maxillary sinuses. Subset analysis by bladder drainage was done to find a link between bladder drainage and febrile morbidity. Subset analysis by procedure was done to find whether fever or infection was related to extent of the procedure. Hysterectomy with vaginal repair was considered the most extensive surgery, followed by hysterectomy alone, and finally vaginal repair alone.
Shapiro-Wilk testing was used to evaluate normality of data distribution. Correlation coefficients were generated to assess the relation between variables. Mean and standard errors of continuous variables were evaluated using Wilcoxon rank-sum tests. Categoric data were evaluated using χ2 or Fisher exact test when appropriate. Sensitivity, specificity, positive predictive value, and negative predictive value were determined using standard calculations. Stepwise logistic regression was done to develop a final model predictive of febrile morbidity and infection. Initially, all continuous and categoric variables were included. Variables that failed to show statistical significance were removed sequentially from the model. Model checking for interaction and extreme values was done using standard methods including Hosmer-Lemeshow χ2 analysis. All statistics were calculated using STATA analytic software (STATA Corporation, College Station, TX).
Four hundred thirty-one vaginal surgery patients were identified. Hysterectomy alone was done in 185 (43%), hysterectomy with vaginal repair in 117 (27%), and vaginal repair without hysterectomy in 129 (30%). None was excluded for metabolic hyperthermia or preoperative infection. Fifty-four women (12.5%) had febrile morbidity. Thirty-five infections (8.1%) were identified, of which only 13 were accompanied by febrile morbidity. Forty-one women (9.5%) had unexplained fevers. Urinary tract infection was found in 20 women (4.6%), suprapubic catheter infection in five (1.2%), cuff cellulitis in three (0.7%), pelvic infection in three (0.7%), and Clostridium difficile colitis in two (0.46%). Two women (0.46%) had nonsurgical infections, including one sinusitis and one bronchitis. Specific infections and fevers are noted in Table 1.
The sensitivity of febrile morbidity for postoperative infection was 40%, specificity was 98%, positive predictive value was 26%, and negative predictive value was 94%. Univariate analysis is presented in Table 2. Univariate analysis found that women with febrile morbidity were older and had longer durations of procedures, greater blood loss, higher parity, lower uterine weights, and greater lengths of stay than women without febrile morbidity. Women with infections also had prolonged procedures, weighed less, had lower uterine weights, and had prolonged hospital stays compared with women without infections. Women who developed postoperative infections were more likely to have had previous gynecologic surgery. A positive correlation between blood loss and operative time (r = .561) was found. No additional relations between variables were found (r < .3).
Table 3 presents the distribution of patients by procedure and outcome variables of fever and infection. Women who had hysterectomies with vaginal repairs were more likely to develop infections than those who had hysterectomies or vaginal repairs alone. Bladder catheters were used in 407 of 431 women (94%), trans-urethral drainage was used in 248 (58%), and suprapubic drainage was used in 159 (37%). No significant difference was found between groups for febrile morbidity, urinary tract infection, or unexplained fever.
Logistic regression found blood loss (odds ratio 1.001/mL; confidence interval 1.0001–1.0035), uterine weight (0.987/g; 0.976–0.999), and parity (1.570; 1.146–2.050) as significant independent variables for developing fever. The final model for febrile morbidity was: Ln(odds ratio [OR] of developing febrile morbidity) = −2.791 + 0.00182 (estimated blood loss) + 0.01262 (uterine weight) + 0.4512 (parity). Patient weight (0.984/lb; 0.971–0.998) and type of procedure (2.16; 2.12–6.38) were confirmed as significant independent variables for postsurgical infections. The final model for infection was: Ln(OR of developing infection) = 2.8090 + −0.01581 (patient weight) + 1.3038 (procedure when 1 = hysterectomy; 2 = vaginal repair; 3 = hysterectomy with vaginal repair). Model checking failed to show interaction or extreme values.
In clinical practice, evaluation for infection is indicated when postoperative fever is noted. The threshold after gynecologic surgery is at least 38.0C (100.4F) on two separate occasions 6 hours apart, excluding the first 24 hours. Whereas temperature monitoring is useful as a screening device because of its ease, low cost, and acceptance by patients, our results show that it has low sensitivity and poor predictive value as a screening test for infection. This study also found that many infections were detected without fever. Sixty percent of infections in this study were not associated with febrile morbidity despite a 94% negative predictive value.
Our study is consistent with other work on postoperative febrile morbidity. Rates of infection in our cohort compare favorably with previous work on women who had vaginal surgery, with rates of febrile morbidity of 12.5%, infectious morbidity of 8.1%, and unexplained fever of 9.2%.7–11 Variables such as site of infection, age, duration of surgery, and estimated blood loss also correlated with previous studies.7,8
Multivariate analysis found a significant risk of febrile morbidity for women with smaller uteri, higher parity, and greater blood loss. A strong correlation between blood loss and operative time (r = .561) was also found. A significant risk of infectious morbidity was found in women with lower weights and those who had more extensive operations. Only five of 35 infections (14%) occurred in women who had vaginal repairs alone. The remaining infections were found in women who had hysterectomies alone (34%) or with repair (51%). In this study, postsurgical infection was linked to type of procedure, and not blood loss or duration of procedure, whereas blood loss and duration of procedure were risks for febrile morbidity. The notation that type of procedure is linked with infection might be another way of measuring blood loss, extent of dissection, and operating time in aggregate. This is consistent with the findings of Shapiro et al,10 who noted duration of procedure and blood loss as significant risk factors for postsurgical infections.
The value of the fever workup has been questioned in postsurgical patients.12 It is likely that the poor performance of fever evaluation was due in part to the failure of febrile morbidity to alter the prevalence of infection in women with fever. Therapy based on fever alone is often inappropriate, costs the health care system in terms of capital and manpower resources, and leads to the development of resistant bacterial flora.
Nonsurgical factors such as intraoperative hypothermia and the use of Foley catheters may influence fever and infection. Kurz et al13 found a reduced infection rate after surgery in women who were normothermic during their procedures, compared with women who developed hypothermia during surgery. Febrile morbidity was not addressed. Our study design precludes comment on this point. Summitt et al14 noted a 24% rate of febrile morbidity after vaginal hysterectomy with transurethral drainage, compared with an 8% rate in women without transurethral drainage. No difference in infectious morbidity was found between the groups. Bladder catheters were used in 94% of our subjects, transurethral drainage in 58%, and suprapubic drainage in 37%. In our study, bladder drainage did not significantly alter the distribution of febrile morbidity, unexplained fevers, or urinary tract infections.
Modification of the current fever evaluation to include ultrasound has been suggested. The incidence of sonographically detected fluid collections at the apex of the vaginal cuff in the immediate postoperative period ranges from 34% to 59%.15–17 The correlation between these findings and febrile and infectious morbidity is uncertain.16,17 Additional study of efficacy and cost will clarify this issue.
Fever remains the main indicator of postoperative infection. Increasing the threshold will improve specificity at the cost of sensitivity. A better approach might be to individualize fever evaluation. Febrile morbidity, as defined, is limited as a screening test for postoperative infection after vaginal surgery.
1. Porges RF. Vaginal hysterectomy at Bellevue hospital. An experience in teaching residents, 1963–67. Obstet Gynecol 1970;35:300–13.
2. Bernheim HA, Block LH, Atkins E. Fever: Pathogenesis, pathophysiology and purpose. Ann Intern Med 1979;91:26–70.
3. Dinarello CA, Wolff SM. Pathogenesis of fever in man. N Engl J Med 1978;298:607–12.
4. Hemsell DL. Prophylactic antibiotics in gynecologic and obstetric surgery. Rev Infect Dis 1991;13:S821–41.
5. Postoperative complications. In: Herbst AL, Mishell DR, Stenchever MA, Droegemueller W, eds. Comprehensive gynecology. 2nd ed. St. Louis: Mosby Year Book, 1991:749–800.
6. Martin DA, Gelder MS. Postoperative infections in gynecology and infectious complications in gynecologic oncology. In: Copeland LJ, Jarrell JF, McGregor JA, eds. Textbook of gynecology. Philadelphia: WB Saunders, 1993:531–59.
7. Dicker RC, Greenspan JR, Strauss LT, Cowart MR, Scally MJ, Peterson HB, et al. Complications of abdominal and vaginal hysterectomy among women of reproductive age in the United States. The Collaborative Review of Sterilization. Am J Obstet Gynecol 1982;144:841–8.
8. Harris WJ. Early complications of abdominal and vaginal hysterectomy. Obstet Gynecol Surv 1995;50:795–805.
9. Hardiman PJ, Drutz HP. Sacrospinous vault suspension and abdominal colposacropexy: Success rates and complications. Am J Obstet Gynecol 1996;175:612–6.
10. Shapiro M, Munoz A, Tager IB, Schoenbaum SC, Polk BF. Risk factors for infection at the operative site after abdominal or vaginal hysterectomy. N Engl J Med 1982;307:1661–6.
11. Hemsell DL, Cunningham FG, Kappus S, Nobles B. Cefoxitin for prophylaxis in premenopausal women undergoing vaginal hysterectomy. Obstet Gynecol 1980;56:629–34.
12. Freischlag J, Busuttil RW. The value of postoperative fever evaluation. Surgery 1983;94:358–63.
13. Kurz A, Sessler DI, Lenhardt R. Perioperative normothermia to reduce the incidence of surgical wound infection and shorten hospitalization. N Engl J Med 1996;334:1209–15.
14. Summitt RL Jr, Stovall TG, Bran DF. Prospective comparison of indwelling bladder catheter drainage versus no catheter after vaginal hysterectomy. Am J Obstet Gynecol 1994;170:1815–8.
15. Slavotinek J, Berman L, Burch D, Keefe B. The incidence and significance of acute post-hysterectomy pelvic collections. Clin Radiol 1995;50:322–6.
16. Toglia MR, Pearlman MD. Pelvic fluid collections following hysterectomy and their relation to febrile morbidity. Obstet Gynecol 1994;83:766–70.
17. Eason E, Aldis A, Seymour RJ. Pelvic fluid collections by sonography and febrile morbidity after abdominal hysterectomy. Obstet Gynecol 1997;90:58–62.