Prospective Implementation and Evaluation of a Decision-Tree Algorithm for Route of Hysterectomy.

The Mayo Clinic Institutional Review Board approved this study. All patients who had a hysterectomy performed at Mayo Clinic, Rochester, Minnesota, between November 24, 2015, and December 31, 2017, were eligible for inclusion. Patients who had their consultation and preoperative listing with a gynecologic surgeon before November 24, 2015, were excluded, even if their surgery occurred within the time frame of the prospective algorithm. Patients also were not included if they met any of the criteria listed in Box 1 , to exclude patients whose route of hysterectomy had a high probability of being planned on the basis of other surgical diagnoses. The current study added endometriosis and transgender patients undergoing gender-confirming surgery as exclusion criteria, which were not part of the retrospective algorithm ( 9 ). Endometriosis was added after reviewing the retrospective results and recognizing the therapeutic advantage of evaluating for, and removing, extrauterine endometriosis disease in affected patients. Transgender patients were excluded because of the complexity of the clinical scenario and several factors that would not be accounted for with our algorithm. The results of our retrospective study ( 9 ) showed substantial deviation from the algorithm when patients had a history of laparotomy. To improve the algorithm and allow its application in a prospective setting, a decision-tree branch was added to allow for a pelvic examination under anesthesia at the time of hysterectomy in patients with a history of laparotomy. An examination under anesthesia allowed the surgeon to evaluate the feasibility of a vaginal approach while the patient was fully relaxed (allowing observation of maximum uterine descent and vaginal caliber). The prospective algorithm was therefore designed to determine the optimal route of hysterectomy for benign indications according to a patient’s surgical history of laparotomy, uterine size, and vaginal access as decision-tree branches ( Figure 1 ). We developed the prospective algorithm and introduced it to our division of gynecologic surgery in August 2015 before it was implemented in November 2015. With the assistance of our institution’s radiology department, 3-dimensional pelvic models were created with variously sized interchangeable uteri. Modeled after computed tomographic images, the uteri represented 3 pathologic uterine weights from the retrospective cohort: 100 g, 280 g (12 weeks of gestation), and 500 g ( 10 ). These pelvic and uterine models were introduced at the division meetings and displayed in outpatient offices for reference during consultations to help accurately assess uterine weight on bimanual examination ( Figure 2 ) ( 11 ). Surgeons followed a strict physical examination template to standardize documentation of the pelvic examination in the office and under anesthesia to reliably assign an expected hysterectomy route. The examination findings were uterine size (estimation of gestational weeks comparison), uterine mobility (mobile or not mobile), uterine location (high in pelvis or not high), and vaginal caliber (narrow or normal). The prospective decision-tree algorithm and study exclusion criteria were posted in the outpatient offices and operating rooms for reference. An institutional surgical database was used to identify hysterectomies performed for benign indications, and data were collected on a rolling basis. Abstracted data included patient demographics; medical, obstetric, and surgical history; laboratory and pathologic results; clinically relevant intraoperative and postoperative events (including route conversion and Accordion grading for postoperative events within 6 weeks) ( 12 ); complications; and hospital readmissions. The expected surgical route for each patient was determined by using the prospective algorithm. If a patient underwent an examination under anesthesia, the expected surgical route was determined based on that path in the algorithm (pathway 2, Figure 1 ), in which the decision-tree branch of examination under anesthesia–TVH versus examination under anesthesia–robotic-assisted total laparoscopic hysterectomy is determined from the examination in the operating room. If clinical examination findings were not complete in the operative report dictation, no contraindications to TVH were presumed. If the actual hysterectomy route was less invasive than the route expected from the algorithm, the case was not considered a deviation. This was supported by the results of our retrospective cohort study, which showed similar or improved outcomes when a TVH was performed in patients expected to have robotic-assisted total laparoscopic hysterectomy ( 9 ), as well as our study that compared TVH and robotic-assisted total laparoscopic hysterectomy for perceived contraindications to TVH, which included uterine size and history of laparotomy ( 13 ). The results from the prospective use of the decision-tree algorithm are summarized descriptively: mean (SD) or median (interquartile range [IQR]) for continuous data and frequency (percentage) for categorical data. The previously published results from our group’s retrospective algorithm applied to 2009–2013 institutional data ( 9 ) were compared with the prospective use of the algorithm to evaluate its effect on the rate of deviation from the algorithm.

During the prospective study period, 365 hysterectomies performed at our institution met inclusion criteria. Of these, 202 patients (55.3%) met criteria for a TVH (pathway 1, Figure 1 ) and 57 (15.6%) were assigned to have an examination under anesthesia and then a TVH (examination under anesthesia–TVH) (pathways 2 and 4, Figure 1 ) according to the algorithm. Thus, combining the TVH and examination under anesthesia-TVH groups (pathways 1 and 4, Figure 1 ), 259 hysterectomies (71.0%) were expected to be performed through the vaginal route. Another 52 women (14.2%) were expected to undergo an examination under anesthesia followed by a robotic or laparoscopic approach (examination under anesthesia–robotic-assisted total laparoscopic hysterectomy; pathways 2 and 5, Figure 1 ), and 54 (14.8%) were expected to have an a priori abdominal, robotic, or laparoscopic hysterectomy (pathway 3, Figure 1 ). The procedures for 46 patients (12.6%) deviated from the algorithm to a more invasive route (44 were robotic when vaginal was expected and 2 were abdominal when vaginal was expected) ( Table 1 ). Seven patients (1.9%) underwent TVH when a laparoscopic, robotic, or abdominal hysterectomy was expected (less invasive than the algorithm expected), which was not considered a deviation from the algorithm. Among the 259 patients expected to have TVH or examination under anesthesia–TVH, the surgical indications included precancerous conditions or early (stage IA1) cervical cancer (30 patients, 11.6%), leiomyomas, abnormal uterine bleeding, or postmenopausal bleeding (216 patients, 83.4%), dysmenorrhea or painful menstrual cycles (68 patients, 26.3%), and other indications (28 patients, 10.8%); 78 patients (30.1%) had multiple indications for hysterectomy. Of the 202 hysterectomies that were expected to be performed as TVH per the algorithm, 170 (84.2%) were performed vaginally, in accordance with the algorithm ( Table 1 ). These 170 patients had a mean age of 44.5 years and mean body mass index of 28.3 kg/m 2 ( Table 2 ). Twenty-two patients (12.9%) were nulliparous, 32 of 168 (19.0%) had a vaginal parity of 0, and 22 of 168 (13.1%) had a history of cesarean delivery. The mean operative time was 95 minutes, which included an apical suspension. The median uterine pathologic weight was 116 g (IQR, 75–160 g), and manual uterine debulking techniques were used in 27 patients (15.9%). Three incidental cystotomies (1.8%) occurred, which were recognized intraoperatively and repaired. There were no intraoperative route conversions in the TVH group. Nearly all patients (161, 94.7%) were discharged within 24 hours of surgery. Within the 6-week postoperative period, 5 patients (2.9%) required a postoperative blood transfusion due to symptomatic anemia ( Table 2 ). Ten patients (5.9%) had a urinary tract infection, and 2 patients (1.2%) had an Accordion grade 3 or higher postoperative complication. Both of these patients required a return to the operating room. One patient had bleeding from an ovarian vessel, and a pelvic abscess subsequently developed that required drain placement by interventional radiology and readmission for intravenous antibiotics. The other patient had acute blood loss anemia, and diagnostic laparoscopy showed a contained retroperitoneal hematoma. She was also 1 of the 5 who required a postoperative blood transfusion. Of the 57 hysterectomies that were expected to be performed as examination under anesthesia–TVH per the algorithm, 41 patients (71.9%) had an examination under anesthesia followed directly by TVH ( Table 1 ). When these patients were combined with the expected TVH group, 211 of an expected 259 patients (81.5%) underwent vaginal hysterectomy. The 41 patients with examination under anesthesia–TVH had a mean age of 42.7 years and mean body mass index of 31.0 kg/m 2 ( Table 2 ). Two of the 41 patients (4.9%) were nulliparous, 16 of the other 39 (41.0%) had a vaginal parity of 0, and 19 of 40 patients (47.5%) had a history of cesarean delivery. The mean operative time was 119 minutes, which included an apical suspension. The median uterine pathologic weight was 185 g (IQR, 110–350 g). There were 2 intraoperative route conversions (4.9%) to complete the hysterectomy: the peritoneal cavity could not be entered due to adhesions in 1 patient, and bleeding was addressed in 1 patient, which resulted in an intraoperative blood transfusion. Almost all patients (95.1%) were discharged within 24 hours of surgery. One patient (2.4%) required a postoperative blood transfusion. There were no Accordion grade 3 or greater postoperative complications in the examination under anesthesia-TVH group within 6 weeks of hysterectomy. Among the combined 211 patients who underwent vaginal hysterectomy, the expected vaginal hysterectomies were completed successfully in 209 (99.1%). A total of 46 patients (12.6%) underwent a more invasive route than was predicted by the algorithm: 32 deviations among 202 hysterectomies expected to be TVH per the algorithm (15.8%; 95% CI, 10.8%−20.9%), and 14 deviations among 57 hysterectomies expected to be examination under anesthesia-TVH per the algorithm (24.6%; 95% CI, 13.4%−35.7%) ( Table 1 ). Among these 46 patients, 15 of 46 (33%) were nulliparous, 19 of 44 (43%) had a vaginal parity of 0, and 11 of 45 (24%) had a history of cesarean delivery ( Table 3 ). The median uterine pathologic weight was 172 g (IQR, 96–250 g). There were no intraoperative route conversions. One intraoperative incidental rectotomy occurred during robotic-assisted total laparoscopic hysterectomy, which was recognized and repaired at the time of surgery. Most patients (87%) were discharged within 24 hours. Of the 43 patients with follow-up, 1 patient (2%) was readmitted, with return to the operating room due to a ureteral injury, 1 (2%) had cellulitis at a port site requiring antibiotic treatment, and 2 (5%) had urinary tract infections. Among the 7 patients who underwent TVH when a more invasive route was expected by the algorithm ( Table 1 ), 1 (14%) was nulliparous, 4 of 6 (67%) had a vaginal parity of 0, and 5 of 6 (83%) had a history of cesarean delivery ( Table 1 ). The mean operative time in these cases was 137 minutes, which included an apical suspension. The median uterine pathologic weight was 405 g (IQR, 110–590 g). Manual uterine debulking techniques were used in 6 patients (86%). There was 1 intraoperative route conversion (14.3%). There were no intraoperative complications and no Accordion grade 3 complications. All 7 patients were discharged within 24 hours of surgery. In our previously published retrospective algorithm cohorts, 15.1% of hysterectomies deviated from the algorithm to a more invasive route before the initiation of robotic surgery at our institution, compared with 25.8% afterward ( 9 ). In the current study, when the algorithm was implemented prospectively, the deviation from the algorithm to a more invasive route decreased to 12.6%.

The prospective algorithm used in this study is a straightforward, clinically applicable decision-tree model that can assist in determining the optimal route of hysterectomy. The results show that the majority of patients are expected to have a TVH. This algorithm identified 2 groups of patients undergoing TVH: those deemed to be candidates a priori for TVH and those requiring initial examination under anesthesia, effectively distinguishing between more straightforward TVH and cases that may be more surgically challenging (examination under anesthesia–TVH). Prospective algorithm use predicts that 55.3% of all hysterectomies were expected to have an a priori TVH, which is higher than the actual TVH rate of 11.5% reported previously ( 2 ). The retrospective algorithm noted a deviation rate of 25.8% when a TVH was expected ( 9 ). Prospective use of the algorithm decreased the rate of deviation to 12.6%, which is lower than the baseline 15.1% before the initiation of robotic surgery ( 9 ). This further supports the use of the algorithm in combination with educating gynecologic surgeons about the feasibility of vaginal surgery. The median pathologic uterine weight was 116 g in the TVH group and 185 g in the examination under anesthesia–TVH group, which is much less than the 280 g (12 weeks of gestation) previously suggested as an upper limit for a priori TVH ( 7 , 14 ). Although large uteri are encountered by gynecologic surgeons, the median uterine weights in the current study are most likely representative of the vast majority of cases performed at the national level. Complications were low in our TVH cohort, with no ureteral injuries, which emphasizes the safety of this surgical approach and should encourage the increased use of vaginal surgery. Results of this study support that surgeons with advanced vaginal surgery skillsets may safely perform TVH in patients who would otherwise be expected to undergo a more invasive route per the algorithm. The subset of these patients in this study was small (n=7), but they encountered no intraoperative complications and had excellent outcomes. One patient had an intraoperative route conversion because the surgery could not be completed vaginally, but this was accomplished without complication. Skilled vaginal surgeons should continue to provide TVH as a safe and effective option, even in women for whom the algorithm would predict a more invasive approach. Substantial cost implications are associated with the route of hysterectomy. Woelk et al ( 5 ) reported an unadjusted cost that was $4,084 lower for vaginal than for robotic hysterectomies. In our cohort of 365 patients, the estimated cost savings would be $657,524 if the TVH rate (excluding examination under anesthesia-TVH) was 55% compared with 11%. As such, increasing the rate of vaginal hysterectomy could significantly decrease health care delivery costs at local and national levels. Our study has multiple strengths, including its prospective design and implementation. The education provided to surgeons before initiating the protocol included education on estimating uterine size using pelvic models for hands-on practice and review of the algorithm, both of which were helpful. Another strength was the alteration of the previous retrospective algorithm to include an examination under anesthesia option. This assisted the surgeon in determining vaginal accessibility to the uterus if it was unclear during the office examination. The exclusion criteria were used to create a cohort with clearly benign uterine disease, which improves generalizability to general gynecologic surgeons. Study limitations include an expertise bias at our institution, which has a long-standing history of vaginal surgery. However, even considering the TVH group alone, without the more complex cases (examination under anesthesia–TVH), the study supports the use of the vaginal route in routine gynecologic surgery and would allow more than 50% of these hysterectomies to be vaginal. The small number of patients with algorithm deviations or poor postoperative outcomes limited the power needed to perform valid statistical comparisons. Additionally, our study was limited by the lack of a control group; all patients undergoing hysterectomy at our institution were included. Further research comparing the use of this model with standard practice may highlight additional risks and benefits. Teaching residents-in-training the skillset and specific techniques required for vaginal hysterectomy is challenging, and the minimum number of cases required to graduate is only 15 ( 15 ). Additionally, the Minimally Invasive Gynecology fellowship does not focus on teaching fellows vaginal surgery, which thereby further decreases the fellowship-trained surgeons who can offer TVH. Unfortunately, this risks the future of vaginal hysterectomy. Surgeons who perform vaginal hysterectomies must mentor trainees so vaginal surgery can be offered as the optimal route of benign hysterectomy when feasible. In our study, a majority of patients qualified for vaginal hysterectomy (TVH and examination under anesthesia–TVH), and 99% were successfully completed. Vaginal surgery is feasible, carries a low complication rate with excellent outcomes, and should have a place in gynecologic surgery. Prospective use of this algorithm nationally may increase the rate of TVH and decrease health care delivery costs.

The rate of total vaginal hysterectomy (TVH) among women requiring hysterectomy decreased in the United States from 21.7% in 2007 to 19.8% in 2010. Furthermore, between 2010 and 2013, only 11.5% of commercially insured patients undergoing hysterectomy had TVH ( 1 , 2 ). Robotic-assisted hysterectomy increased from 0.5% to 9.5% from 2007 through 2010 ( 1 ). TVH is the most cost-effective route, with a low complication rate, and, therefore, should be performed when feasible ( 3 – 5 ). Algorithms using factors of vaginal access, uterine size, and concern for extrauterine disease have been suggested but not widely used, and the hysterectomy route is not standardized ( 6 , 7 ). TVH has been important historically. At our institution in 1958, over half of all hysterectomies were TVH ( 8 ). Recognizing this practice change, our group (J.J.S., J.A.O., J.N.B.-G., S.C.D., J.B.G.) created a clinical decision-tree algorithm for benign hysterectomies and previously applied it retrospectively to 2 cohorts: patients treated before (2004–2005) and after (2009–2013) the advent of robotic surgery (9). If the algorithm suggested TVH, patients who underwent TVH had better outcomes than those who had robotic hysterectomy. Initiation of robotic surgery increased algorithm deviations from 15.1% to 25.8% when TVH was assigned by the algorithm. Additionally, if the algorithm had been followed, the estimated 5-year savings would have been approximately $800,000 ( 9 ). Our primary objective was to evaluate the effect of the algorithm on the rate of TVH performed. The secondary aims were to evaluate short-term postoperative outcomes after prospectively implementing a clinical decision-tree algorithm to determine optimal hysterectomy route and to discuss estimated effects on health care costs.