Objective
The purpose of this study is to compare costs of 3 strategies for women undergoing hysterectomy: (1) test all patients for bacterial vaginosis; treat if positive; (2) treat all patients for bacterial vaginosis; (3) neither test nor treat patients for bacterial vaginosis. For comparison purposes, a fourth strategy is examined: (4) no surgical site infection prophylaxis or bacterial vaginosis treatment.
Study Design
A cost minimization model was created using estimates obtained from the published literature, Medicare reimbursement data, and wholesale drug costs.
Results
In the base case, the optimal strategy was to treat all patients for bacterial vaginosis, with a cuff infection rate of 4.0% and mean cost of $593. The “test all patients for bacterial vaginosis; treat if positive” strategy was also inexpensive, with a mean cost of $623 and 4.2% cuff infection rate. “Neither test nor treat patients for bacterial vaginosis” and “no surgical site infection prophylaxis or bacterial vaginosis treatment” were more expensive and less effective than other strategies.
Conclusion
This model suggests that consideration should be given to adding metronidazole to standard surgical site infection prophylaxis before hysterectomy.
The American College of Obstetricians and Gynecologists (ACOG) recommends preoperative screening and treatment for bacterial vaginosis (BV) before elective hysterectomy to prevent vaginal cuff infection. Several studies have demonstrated an increased risk of cuff infection after hysterectomy if BV is present. In addition to screening for BV, ACOG recommends surgical site infection (SSI) prophylaxis with a broad-spectrum antibiotic, such as cefazolin.
BV is characterized by the disappearance of hydrogen peroxide–producing lactobacilli and by overgrowth of anaerobic bacterial species, including Gardenerella vaginalis , Mobiliuncus spp, and Prevotella spp, in the vagina. Patients with BV are asymptomatic in 50% of cases; tests to detect BV have a relatively low sensitivity and specificity. The cause of BV is not known; however, treatment includes standard antibiotics that have strong anaerobic coverage that do not effect lactobacillus species. The efficacy of oral metronidazole in returning vaginal flora to normal is approximately 80%.
Vaginal cuff infection is characterized by fever, foul-smelling vaginal discharge, and increasing pain in the abdomen or tenderness on bimanual examination at least 48 hours after surgery. Soper demonstrated that patients with cuff infection had a significantly longer duration of hospitalization compared with noninfected patients (8.1 vs 5.1 days). The increased length of stay and its associated costs may have a significant clinical and economic impact. Before universal application of SSI prophylaxis, several studies demonstrated that BV was strongly associated with both postoperative cuff infection and postoperative febrile morbidity. Soper found that women who had BV had a 3 times higher risk of developing cuff cellulitis, cuff abscess, or both after abdominal hysterectomy than women without BV (relative risk [RR], 3.2; 95% confidence interval, 1.5–6.7). Larsson et al examined premenopausal women undergoing total abdominal hysterectomy without antibiotic prophylaxis between 1987 and 1988. Of 105 study participants, 35% of patients with BV had postoperative cuff infection develop compared with 8% of patients without BV. Persson et al performed a prospective evaluation of postoperative infection in 1060 patients undergoing hysterectomy in Sweden in 1992. In this study, BV was associated with higher rates of wound, cuff, and deep infections (RR, 3.0; P = .01).
It is unknown whether preoperative treatment with a course of metronidazole decreases the risk of vaginal cuff infection in patients undergoing hysterectomy. Lin et al examined the risk of postoperative fever in patients undergoing hysterectomy. Subanalysis of patients who received SSI prophylaxis demonstrated a statistically increased risk of fever in the immediate postoperative period in patients who had BV compared with patients with normal (lactobacilli predominant) vaginal flora. Two studies investigating infection rates after elective abortion demonstrated a decrease in postoperative endometritis among groups treated preoperatively with metronidazole. These data suggest that BV may alter the efficacy of standard SSI prophylaxis to prevent infection and preoperative treatment with metronidazole could decrease the incidence of vaginal cuff infection in hysterectomy patients.
No studies published to date have examined the effect of BV on vaginal cuff infection rates in the era of universal SSI prophylaxis nor definitively addressed the costs and benefits of preoperative screening for BV before elective hysterectomy. A cost minimization model was created using best available clinical estimates to determine the impact of various factors on the best choice of strategy, as well as to define the direction of future research.
Materials and Methods
A decision model was constructed using commercially available software (TreeAge Pro; TreeAge Inc, Williamstown, MA) to compare the cost and effectiveness of 3 strategies for women who are to undergo hysterectomy, assuming standard SSI prophylaxis: (1) test all patients for BV with wet mount; treat if positive (TT strategy); (2) treat all patients preoperatively with oral metronidazole (TA strategy); (3) neither test nor treat patients for BV (TN strategy). For comparison purposes, a fourth strategy, not currently in standard clinical use, was also examined: (4) no SSI prophylaxis, no testing or treatment for BV (NP strategy). The software was used to create a simple decision model ( Figure 1 ) to estimate an average expected cost for each strategy by multiplying the costs associated with each scenario by the probability of occurrence of that scenario. Clinical probabilities were derived from the published literature; the study was declared exempt from review by the Duke Institutional Review Board. Extensive sensitivity analyses were performed to account for uncertainty in model parameters.
Key assumptions of the model were: (1) the risk of postoperative vaginal cuff infection is similar regardless of type of hysterectomy (laparoscopic, open, or vaginal); (2) the age and racial distribution of the US population presenting for hysterectomy is represented in the published studies used to create the clinical estimates; (3) all patients in TT, TA, and TN arms receive standard SSI prophylaxis per ACOG recommendations.
A PubMed search was performed for articles published between January 1980 and September 2010 using the search terms “Hysterectomy” or “Hysterectomy, Vaginal” and “Surgical wound infection” or “Postoperative Complications.” Parameter estimates were derived from the articles identified through this search and included the following: rate of cuff infection in hysterectomy patients who do not receive SSI prophylaxis, prevalence of BV, sensitivity and specificity of wet mount to detect BV, relative risk of cuff infection if a patient is infected with BV, cure rate of BV with a 7-day course of oral metronidazole within 1 month of surgery, and relative efficacy of cefazolin to prevent cuff infection in patients with compared with without BV. Baseline estimates and reasonable clinical ranges of each parameter were obtained for sensitivity analysis and are listed in Table 1 .
| Clinical parameter | Estimate | Estimate range (published literature) | Sensitivity analysis range |
|---|---|---|---|
| Prevalence of BV | 0.27 | 0.16–0.29 | 0.05–0.4 |
| Wet mount–sensitivity | 0.81 | 0.70–0.97 ,33-35 | 0.7–0.99 |
| Wet mount–specificity | 0.70 | 0.70–0.77 ,33-35 | 0.5–0.99 |
| Efficacy of treatment for BV | 0.80 | 0.80–0.91 | 0.01–1 |
| Overall rate of cuff infection (no SSI prophylaxis) | 0.1 | 0.08–0.19 | 0.01–0.3 |
| Efficacy of SSI prophylaxis | 0.49 | 0.34–0.86 ,36 | 0.3–0.9 |
| Difference in SSI prophylaxis efficacy if BV is present at time of surgery | 0.9 | N/A | 0.1–1 |
| Relative risk of vaginal cuff infection if BV is untreated | 2.3 | 2.3–4.4 | 1–4.4 |
Costs were derived in 2008 US dollars for wet mount, cuff infection, and antibiotics ( Table 2 ), using Medicare reimbursement was obtained using the Current Procedural Terminology ( http://www.cms.gov ). Treatment regimens and cost were compared between national statistics and estimates at the authors’ institution when possible. Costs of administration associated with treatment were not included in the model.
| Cost | Estimate | Estimate range (published literature) | Sensitivity analysis range |
|---|---|---|---|
| SSI prophylaxis | $75 | $50.27–100.28 | 2–500 |
| Wet mount | $5 | N/A | 1–10 |
| Treatment for BV | $9.80 | N/A | 2–30 |
| Treatment for postoperative cuff infection | $13,400 | $3382–13,486 ,37 | 0–15,000 |
Clinical estimates
To determine the baseline rate of cuff infection without surgical site prophylaxis, articles identified by the search were examined for rate of cuff infection. Cuff infection was defined as reported cases of either: “cuff cellulitis,” “cuff abscess,” “pelvic infection,” “pelvic cellulitis,” or “cuff infection,” divided by the study population. Articles were excluded if antibiotics were administered to some or all of patients, or documentation of whether perioperative antibiotic administration was performed was omitted. A range for the baseline rate of vaginal cuff infection without SSI prophylaxis was estimated from the findings of 6 studies: the baseline rate of cuff infection was found to be 0.08 to 0.19. A rate of 0.10 was used as our best estimate. Three of these studies specified rates of cuff infection in cohorts with or without bacterial vaginosis. These more specific data were used to estimate the relative risk of cuff infection in patients who have BV compared with those who do not have BV.
The prevalence of BV varies, depending on a population’s menopausal status and racial and ethnic backgrounds. Reported prevalence in US populations range from 6–28%. In a study published by Lin et al, the prevalence of the patient population in Chicago who presented for hysterectomy was 27%.
The Centers for Disease Control (CDC) recommends diagnosis of BV using clinical criteria as described by Workowski and Berman and Amsel et al. A positive result is defined by 3 or more of the following: (1) >20% squamous cells are clue cells; (2) thin white discharge coating wall of vagina; (3) vaginal pH >4.5; (4) positive whiff test. The sensitivity and specificity of clinical diagnosis in a research setting are 0.83 and 0.70, respectively.
Standard treatment for BV was metronidazole 500 mg orally twice daily for 7 days. The efficacy of this treatment (81%) was based on the results of extensive review of treatment literature performed by Joesoef et al.
The baseline efficacy of SSI prophylaxis to prevent vaginal cuff infection was estimated from randomized controlled trials. Three studies were identified and a baseline efficacy of 0.49 was calculated. A variable was created within the model (“delta efficacy”) to represent the ratio of the efficacy of cefazolin to prevent cuff infection in the presence of BV compared with its efficacy without BV. There were no clinical data available for this variable; the base-case estimate was 0.9 (efficacy of SSI prophylaxis with BV is 90% of its efficacy without BV).>
Cost estimates
According to the 2008 Medicare Laboratory Fee Schedule, the cost of wet mount using CPT code 87210 for North Carolina was $5 and was reasonable when compared with the authors’ institutional typical reimbursement.
The cost of medications was based on average wholesale cost of the drugs. The cost of metronidazole was $9.80, based on the average wholesale cost of 14 oral tablets of metronidazole. The majority of adverse effects associated with oral metronidazole are minor. The most common adverse effect is nausea (12%). Other minor adverse effects include dizziness (1-4%), headache (5-18%), and vaginal discharge/irritation (9-12%). Adverse events were not included in the cost estimate.
The cost of SSI prophylaxis was $25.20, estimated from the average wholesale cost of a 2 g intravenous bag of cefazolin. Adverse events associated with cefazolin, with the exception of hypersensitivity reactions (1-4%), are rare. The cost of hypersensitivity reaction, which has been previously estimated from national data in France in a publication by Flabbee et al, was converted from 2005 Euros to 2008 US Dollars (US Bureau of Labor Statistics, http://www.bls.gov ), resulting in an estimated cost of $2506 per case of anaphylaxis.
The cost of a vaginal cuff infection was estimated from the Agency for Healthcare Research and Quality Nationwide Inpatient Sample database for 2008. The ICD-9 code (614.3) for “Acute Parametritis” was used to retrieve data. The mean cost was stratified by age and excluded patients less than 45 years of age. This method was used to exclude the majority of confounding from pelvic inflammatory disease admissions, which could also fall under this diagnosis code.
Sensitivity analysis
Sensitivity analysis was performed on prevalence of BV, sensitivity and specificity of wet mount, BV cure rate with oral metronidazole, overall rate of cuff infection, efficacy of SSI prophylaxis to prevent cuff infection, difference in efficacy of SSI prophylaxis if BV is present, and relative risk of cuff infection if BV is present at time of surgery ( Figure 2 ) . Where possible, ranges were chosen that encompassed clinical estimates from the published literature. Sensitivity analysis was also performed on the cost estimates of SSI prophylaxis, wet mount, treatment for BV, and treatment for cuff infection. Ranges were estimated to encompass, at minimum, 50% to 200% of an estimate obtained from the literature ( Figure 3 ) .
