A value-based evaluation of minimally invasive hysterectomy approaches

Conventional laparoscopy and robotic-assisted hysterectomy are performed with a wide variation in techniques and instrumentation that have been adequately described in the literature. We provide here the most important technical principles of the LRH technique as described in the operative reports. LRH is a variation of a total laparoscopic hysterectomy, with the distinguishing aspect of a standard retroperitoneal dissection (RPD), lateralization of the ureters, and ligation of the uterine artery at its origin from the anterior division of the internal iliac artery [14‐16]. Abdominal entry is obtained by 5mm direct entry trocar via the umbilicus and a 5mm suprapubic port. A third port is placed in the right or left lower quadrant in more complex cases where additional assistance is required. With the ureter lateralized and in view, a defect is made in the posterior leaf of the broad ligament, which aids in isolating the infundibulopelvic ligament. The anterior leaf of the broad ligaments is then opened on each side to create a bladder flap that is carried through to the midline along the vesicouterine peritoneum. The anterior vaginal fornix is delineated using a simple sponge stick, and a colpotomy is created using the Harmonic scalpel. A Zumi manipulator is used for the manipulation of the uterus during the procedure. The uterus is then extracted vaginally, and if needed in cases of large specimen, by using extra-peritoneal vaginal debulking techniques. With LRH, the vaginal cuff is closed transvaginally, which, in our experience has led to faster operative times. Though retroperitoneal dissection, lateralization of the ureters, and early ligation of uterine arteries at the origin are techniques that can be performed in various ways with other approaches, the LRH technique was performed consistently as described above in the group of patients included in our analysis.

Patient characteristics analyzed included age, race, weight, body mass index (BMI), surgical history, uterine weight, and uterine pathology. The Elixhauser Comorbidity Index was used to identify and record comorbid conditions that have been shown to potentially affect the operative outcomes [17], and those comorbidities were captured and recorded in each case. We also collected data on length of stay (LOS), which was counted as 0 if the patient was discharged before midnight on the same day of surgery. Surgical outcomes examined included the method of tissue extraction, estimated blood loss (EBL), skin-to-skin operative time (OT), intraoperative and postoperative complications, and rate of conversion from laparoscopic to abdominal hysterectomy. Minilaparotomy (< 5 cm) was reported but not included as a “conversion” in the adjusted analysis due to the associated decreased blood loss, pain, tissue trauma, and LOS when compared to standard laparotomy [18]. EBL was obtained from the anesthesiology reports. Complications were defined as any undesirable, unintended, and direct outcome of surgery which affects the patient, which would not have occurred, had the surgery gone as well as could reasonably be hoped [19]. Complications were categorized as intraoperative when they occurred at the time of the procedure. Complications that presented within 60 days of the hysterectomy procedure and were treated at the hospital or the outpatient clinic were captured and categorized as postoperative complications.

While most comparative studies on hysterectomy use reimbursement amounts or hospital charge data [11, 20], our cost analysis was conducted from a healthcare system perspective, incorporating direct hospitalization costs associated with each procedure. We sought to improve the accuracy of cost estimation by including data on unit costs and resources utilized at the patient level [21, 22]. Direct hospital costs included the following variables: room and board, physician reimbursement, operating room and anesthesia time, surgical equipment, and postoperative care (e.g., pharmacy, laboratory tests). To estimate physician reimbursement for each procedure, we used the 2013 OB/GYN Surgery Medicare reimbursement rates. The hospital’s billing department provided the cost of room and board per day, preoperative costs of setup or time spent in the preoperative holding area, and the average per-minute costs for the operating room and anesthesia times. The latter fee schedule is based on a combination of nursing labor and fixed equipment for the operating room and is adjusted for every calendar year. Detailed equipment costs were obtained from the hospital’s purchasing department. Costs of all disposable equipment (e.g., drapes, sealers, scalpel blades, trocars, forceps) required for each surgical procedure were included in the cost analysis (Appendix). Reusable instruments were assumed to have no cost as no additional investment was needed. Similarly, the acquisition and amortization of the da Vinci robot were not factored in, as the initial capital investment and the annual maintenance costs of the robotic system were previously funded by the hospital. Although we included a flat fee for preoperative times, it is important to note that this does not take into account personnel costs related to robot setup and disassembly and the impact on room turnover time [20]. The estimate of postoperative care, which includes pharmacy and laboratory tests, was also provided by the hospital’s billing department. A standard cost to charge ratio of .6 was applied to the postoperative care estimate, as only charge data was provided by the hospital. Average costs were computed at the procedure level as well as at the patient level in order to provide more accurate estimates of the average, median, and range of costs (operating room + LOS + equipment + surgeon fee + postoperative care).

The third arm of our value-based analysis drew on survey results of patient satisfaction within the context of an assessment of the patient decision-making process in choosing the route of hysterectomy. Patient satisfaction is multi-layered and contextual, and is conceptualized largely based on the congruence of expectations and subjective perception of the personalized service, care, and the final surgical outcome.

An IRB-approved survey was mailed to the same study population (N = 2689) in June 2015. A unique survey identifier allowed us to match surveys with patient medical charts in order to validate the procedure route. The questions were derived from the existing standardized instruments [23, 24]. We use the following questions as a proxy for “satisfaction” with the overall surgical experience:

For each question, we calculated the percentage of respondents that chose option “a” as our indicator of satisfaction. The cumulative average satisfaction score was used in the value equation described below.

The need to use quality and cost to assess value is well-established, but there is no clear consensus on how to operationalize these two fundamental parameters [7]. For the purpose of this study, we defined quality as short-term operative outcomes plus overall patient satisfaction with the hysterectomy procedure:

where

and

Based on Ryan and Tompkins’ “unconditional model,” quality was quantified by creating a summary composite score of the unadjusted average OT, EBL, complication and conversion rates, and patient satisfaction [4]. Scores were calculated using quartiles. Operative outcome scores were assigned a weight of 75%, and patient satisfaction score was assigned a weight of 25% of the quality score, which mirrors the Center for Medicare and Medicaid Services’ pay for performance model [3].

Patients’ demographic and clinical characteristics measured on a nominal or ordinal scale were summarized as counts and percentages, and compared across hysterectomy procedures by using Pearson’s chi-squared tests; variables measured in the interval scale were summarized as means and standard deviations and compared across hysterectomy procedures by using a one-factor analysis of variance (ANOVA). Operative outcomes were compared unadjusted and also adjusted for age, race, number of previous abdominal surgeries, uterine size, BMI, number of additional procedures, number of comorbidities, and weight. Surgeon’s experience was also controlled for by including the actual number of hysterectomies performed during the study period as a covariate in the models.

For unadjusted comparisons of continuous clinical outcome (EBL, LOS), one-factor ANOVA was used to test the overall mean differences, followed by all-pairs Tukey-Kramer post hoc comparisons to identify means that were significantly different from each other. Unadjusted comparisons of dichotomous clinical outcomes (intraoperative complications, postoperative complications) were performed using logistic regression without covariates. Similarly, conversion (none, minilaparotomy, laparotomy) was analyzed as a polygamous outcome with multinomial logistic regression. We used post hoc Bonferroni-adjusted comparisons of predicted probabilities to identify groups that differed significantly while maintaining the overall significance probability at 5%.

For adjusted analyses, linear regression was used to model EBL, logistic regression was used to model complications and conversions, and negative binomial regression was used to model LOS. Because of the concerns about non-normality of the dependent variable, we used median regression to model EBL and surgery time. We used adjusted (marginal) medians, proportions or counts, and corresponding delta method standard errors computed from the fitted models to perform Bonferroni-adjusted post hoc comparisons to identify groups that differed significantly while maintaining the overall significance probability at 5%.

There were significant cost differences across procedures, which reflected similar cost comparisons found in the literature [11, 25, 26]. Examining the average of the entire cohort, LRH was the least expensive with an average cost of $4061, ranging from $2782 to $10,687 (Fig. 1). This is approximately half the cost of LSH, RALH, TLH, and LAVH. The lower costs associated with LRH were attributed to the use of reusable instruments, shorter OT, and shorter LOS. TVH had the second lowest average cost at $6416. RALH had the highest cost with an average of $9354, ranging from $6019 to $19,815, which was driven primarily by OT, LOS, and disposable instruments (Table 6).

The response rate of the survey was 12% (N = 322), which falls within the expected average 10–15% rate of response for external mail surveys with no monetary incentive [27]. The average mean patient satisfaction scores ranged from 79.1 to 87.2. LRH had the highest combined patient satisfaction score, followed by RALH, while TLH, LAVH, TVH, and LSH averaged similar scores as shown in Table 7.

We applied our value formula (value = quality/costs) to each hysterectomy procedure approach, where a summary composite score of the unadjusted average OT, EBL, complication and conversion rates, and patient satisfaction were calculated via equally weighted quartiles [28].

LRH, TVH, and LAVH yielded the highest value scores, while LSH, TLH, and RALH yielded the lowest scores (Table 8). There was little meaningful difference between the lower scores.

The current study has several limitations. Its retrospective nature is limited by inherent selection bias. Although tests were conducted to ensure inter- and intra-rater reliability amongst data abstractors, the availability and accuracy of the medical records, as well as transcription errors, remain intrinsic limitations.

All hospital data on reoperation and readmittance within 60 days were collected; however, the total number of postoperative complications may be underreported, as patients with adverse events may have been seen in their physician’s office or at a different hospital. In the same vein, we were not able to capture the costs of readmissions or reoperations; however, these occurrences are reflected in the complication rates.

Another limitation is that the different hysterectomy approaches were not performed by the same surgeons. The three surgeons who performed all of the LRH procedures, and the three surgeons who performed 89% of RALH procedures, are skilled and experienced laparoscopists who perform a high volume of cases. The three high-volume robotics surgeons have each been performing robotics for over 10 years. Thus, there are limitations in making a direct comparison to other procedure types performed by different surgeons of varying experience, skill level, and practice patterns. As high-volume surgeons are associated with lower costs and better surgical outcomes [41], we attempted to make meaningful comparisons by controlling for surgeon volume in our regression models.

Further, length of stay was included in the cost analysis, but it should be noted that LOS is often a protocol-driven outcome; it can be as much influenced by surgeon preference and the time of day of the procedure, as by the condition of the patient. However, when we omitted LOS from the cost calculation, the result was largely the same. It did not change the ranking of the value scores (Table 8).

Although the US Panel on Cost-Effectiveness in Health and Medicine recommends micro-costing as the preferred approach to cost estimation [42], one of the limitations of micro-costing is that it lacks external validity; there is a limit to the extent that one can generalize to other settings. It should also be noted that costs were calculated from a healthcare system perspective, incorporating direct hospitalization costs associated with each procedure. A more rigorous cost analysis should also incorporate the indirect costs to society, such as lost wages from work absence and differential productivity costs. Determining the costs of surgical procedures is never a straightforward calculation; however, we sought not to compute the conclusive costs of hysterectomy procedures, but rather to compare the relative cost differences between procedural routes.

Additionally, our operational definitions may not accurately capture the intended construct of “value.” Past attempts to measure this concept required some type of weighting and threshold setting, which inevitably invite the stakeholder’s preference. This remains one of the inherent limitations in measuring value; how it is defined dictates how measures are selected and weighted. Our value ratio was based on a summary composite score of short-term operative outcomes and patient satisfaction using quartiles. Because this specific approach has not been used in previous studies, there is no assessment of the reliability and validity of measures combining quality and cost. We did not find an alternative approach for combining quality and cost measures into quantifiable measures of value to evaluate surgical modalities.

Further, the patient perspective was based on a mail survey with high non-response bias. Although a 12% response rate is a typical response rate for an external mail survey with no monetary incentive, the low response rate decreases the survey’s internal validity, thereby limiting its representativeness [28]. Recall bias may also be a limitation, as the survey was administered between 1.5 and 4.5 years postoperatively. Additionally, patient satisfaction score was constructed with just three standardized questions regarding satisfaction. In order to build a more rigorous value model that includes the patient perspective, a prospective study is needed to investigate time to return to normal daily activity for each surgical procedure, as well as changes in the patient’s quality of life using a validated quality of life instrument.

Rather than examining the individual metrics alone, we used a novel formula of operative outcomes plus patient satisfaction divided by direct hospital costs in an attempt to operationalize and compare the value of the most common minimally invasive hysterectomy techniques.

One of the main strengths of the current study is the large sample size, with a wide array of surgeon experience, surgical techniques, patient demographics, and variations in pathology. To our knowledge, it is also the only study in this field employing a sequential, mixed method design, which included a retrospective chart review and a mail survey of the same patient population. This allowed us to integrate the patient perspective into our value framework.

Another strength is the inclusion of direct hospitalization costs rather than the oft-reported insurance reimbursement amounts or charge data [11, 20]. Using a micro-costing approach from the healthcare system perspective, we included data on unit costs and resources utilized at the procedural and patient levels.