Cost drivers for benign hysterectomy within a health care system: Influence of patient, perioperative, and hospital factors.

Our sample includes 2,592 women who underwent benign hysterectomy. Without excluding patients without complete data and/or malignancy (N=1,459), the largest academic institution performed the most hysterectomies (N=1,444) which equates to 290 hysterectomies per year. The median age is 47 years (interquartile range [IQR], 43–52). Race breakdown is 50% White (n=1,287), 39% Black (n=1006), and 11% other (n=299). The distribution of BMI in our sample is 25% normal or underweight (n=656), 29% overweight (n=746), 23% class 1 obesity (605), and 23% class 2 or 3 obesity (n=585). Two-thirds of our sample have zero or one comorbidity (n=1,741) while 16% had three or more (n=403). The most common indication is fibroids or AUB (70%, n=1,815), Table 1 . Approach breakdown is 61% laparoscopic (n=1,585), 16% open (n=424), 14% robotic (n=355) and 9% vaginal (n=228). Most cases were performed by general gynecologists (48%, n=1,253), 20% each by gynecologic oncologists or MIS (n=582, 557), and 8% (n=200) by urogynecologists. Most hysterectomies (81%) were performed outpatient or ESR -- 21% were discharged on POD#0 (n=542), 58% on POD#1 (n=1,042), 13% on POD#2 (n=336), and 9% on POD#3 or later (n=243). Most cases were reimbursed with commercial third-party insurance (70%, n=1,816). Total OR time was 2 to 3 hours for 37% (n=953) of cases and 3 to 4 hours for 29% (n=754); the median was 185 minutes (IQR 146–238). Concomitant procedures are presented in Supplemental Table 1 by surgical approach, 43.5% had at least one. Unadjusted total charges for each approach (median, IQR, and the full range) are in Figure 1 . Before adjusting for potential confounders, laparoscopic is associated with the smallest median charges and open the highest. Using one-way ANOVA that is not adjusted for confounders, there is no statistical difference in mean total charges between laparoscopic and vaginal but both approaches had statistically lower mean charges than open or robotic (vaginal $13,202±6,174, laparoscopic $13,068±5,253, robotic $18,455±8,295, and open $20,591±12,246). After simple linear regression, the multiple linear regression model includes: approach, age, race, BMI, indication, comorbidities, year, case class, facility, payer type, LOS, OR time, specimen weight, EBL, concomitant procedures, need for transfusion, and resident participation ( Table 2 ). In this model, the following predictors no longer affect charges: age, race, indication, medical comorbidities, EBL, payer type, and resident participation. The only patient characteristic that influences charges is BMI. Most of the significant drivers are perioperative and hospital characteristics. After accounting for all covariates, the following charges can be attributed just to the surgical approach; compared to vaginal, open is associated with a mean increase of $692, laparoscopic with $712, and robotic with $1279. These findings could reflect charges associated with additional supplies, especially disposable charges associated with MIS. Other significant perioperative factors includes total OR time, uterine size, concomitant procedures, need for transfusion, and LOS. Each minute in the OR increases cost by $46 (95%CI $44–48). Uterine size >250g was associated with a mean increase of $388. The only concomitant procedures that are statistically significant were colpopexy, vaginal cuff suspension, ureterolysis/enterolysis, mid-urethral sling, and/or vascular or breast procedures. Transfusion increases total charges by a mean of $3000. Charges increase for each day beyond POD#1 in LOS. Hospital-level factors that influence charges are year, facility, and case class. Each subsequent year has increased charges. The costs at facilities C and D are statistically different than A, however, the difference was < $2000. Facility E is outside of the state of Maryland and thus not subject to the payer model and had significantly higher charges. Outpatient or ESR recovery is associated with a $3000 decrease compared to inpatient. Finally, we compare LOS, transfusion, and OR time, which are each significant drivers in the multiple linear regression model. Median LOS is longer for open (2 days) than other approaches (1 day). Transfusion occurred in 12% of open cases, but only ~1% in other approaches. Total OR time for vaginal and robotic was longer than laparoscopic and open, possibly due frequent concomitant urogynecologic procedures in vaginal hysterectomy ( Supplemental Table 1 ) and the additional robotic setup time ( Table 3 ).

This study was reviewed by the Johns Hopkins School of Medicine Institutional Review Board (IRB IRB00211405) and determined to be exempt. Neither Informed consent nor ethics approval was required. This is a retrospective cohort study of all hysterectomies between July 1, 2014 and February 28, 2019 at five academic and community hospitals which are all sites from a single, integrated, healthcare system in one geographic region. Facility A, D and E are community hospitals and facility B and C are academic institutions, with facility C being the largest academic teaching insitution. Inclusion criteria are patients 18 years or older undergoing laparoscopic, robotic, open, or vaginal hysterectomy (included supracervical). Surgeries were performed by general gynecologists, minimally invasive (MIS) surgeons, gynecologic oncologists, and urogynecologists. Exclusion criteria are those with surgeries done for malignancy or malignancy identified on final pathology and those without complete data (n=1,459). A database (Premier Healthcare Inc), was used to identify cases, assembled for quality and reporting purposes and includes clinical and financial data from hospital-based outpatient and inpatient encounters. Additional data were extracted via chart review. The primary outcome is total charges levied by the institution to the payer for an episode of care for hysterectomy. Cost incurred by the institution to provide the care nor the amount reimbursed by the payer or the patient is not assessed. As noted above, in the state of Maryland, charges closely approximate the total cost of care and the revenue collected. All charges were adjusted for inflation using the Consumer Price Index to February 2019. Within our system, the additional costs for a robotics (purchase price, leasing fee, maintenance, training and equipment costs) are allocated to individual patient cases indirectly. The total costs for all types of surgery are combined into one pool and levied as overhead charges that are amortized over all cases, regardless of which equipment is used per case. Thus, there is not a specific charge levied for use of the robotic console. Included variables are patient, perioperative, and hospital. Patient characteristics included age, race, body mass index (BMI), medical co-morbidities, payer type/insurance, and surgery indication. Race was categorized based on the electronic medical record (EMR). BMI was extracted from the preoperative visit. Medical comorbidities are assessed using the Elixhauser Comorbidity Index, which uses ICD-9 and ICD-10 diagnosis codes. 17 Payer type is condensed into Medicare/Medicaid, third-party commercial insurance, and other. Indications for surgery include fibroids, abnormal uterine bleeding (AUB), endometriosis, pelvic organ prolapse (POP), and other (ex. adnexal masses, cervical dysplasia, and gender affirmation). Perioperative characteristics include approach, total operating room (OR) time, estimated blood loss (EBL), specimen weight, concomitant procedures, resident participation, post-operative length of stay (LOS), and need for transfusion. Surgical approach includes open, vaginal, laparoscopic, or robotic. Laparoscopic-assisted vaginal hysterectomy (LAVH) was categorized as laparoscopic due to equipment used. The total OR time is measured from in-room to out-of-room; not just the time required for the hysterectomy. This is chosen due to data availability and because charges are levied on a per-minute basis of OR time. Hospital factors include year of surgery, facility, and case class. Case class is classified as inpatient, outpatient (same-day discharge), extended surgical recovery (ESR,23-hour) or surgical admission (2+ nights). Data analysis was performed using STATA 14.2 for Mac (StataCorp, College Station, TX). Standard descriptive statistics for patient, perioperative, and hospital characteristics were calculated. Measures of central tendency were selected based on each variable’s type and distribution. Total charges were initially compared between surgical approaches using one-way ANOVA. To assess whether total charges differed between approach, we performed multiple linear regressions to assess the impact of approach, adjusting for covariates. To determine which independent variables to include in the final regression model, we performed an initial simple univariate linear regression for every independent predictor variable. In the final regression model, we only included variables with a statistical significance of p<0.10 in initial simple linear regression. Vaginal hysterectomy was used as the comparison group. For other variables, the median or the mode was used as the comparison.

This study demonstrates that, before adjusting for other drivers, there was no significant difference in [unadjusted] charges between vaginal and laparoscopic approach but that both had statistically lower charges compared to an open or robotic approach. However, in the adjusted model accounting for other cost drivers, vaginal had the lowest mean charges, followed by open, laparoscopic, and then robotic. Most patient factors had no influence on charges, while perioperative and hospital factors had a significant impact. The payer type did not affect the total charges, consistent with the Maryland All-Payer Model . There was no influence of indication for surgery or medical comorbidities, as the impact of these indications were accounted for in the total OR time. The only patient factor influencing charges was BMI, likely reflecting additional supplies. The major drivers of charges were length of stay, year of surgery, facility, concomitant procedures and transfusion. This study contributes a large sample size with more granular patient and perioperative data due to inclusion of data from both a quality database and the medical record, facilitating an analysis of charge data to account for patient, perioperative, and hospital factors. Furthermore, our study included a racially diverse sample and included academic- and community-based hospitals as well as subspecialists, improving the generalizability of our findings. Importantly, the analysis was performed using charges levied by the hospital. In Maryland, this closely approximates the hospital’s costs incurred for providing care, and thus is a closer measure of the cost of care. In studies from other states, charges may differ significantly from the costs incurred by hospitals or the reimbursement provided by payers, thus comparing the charges in our study with other studies must be done cautiously. Limitations of our study include the retrospective nature of the study. Our study also contains a high proportion of laparoscopic and robotic cases, reflecting the high rate of MIS hysterectomy performed within our system, not mirroring national trends. Additionally, while OR time from door-to-door was a substantial driver of costs, it is important to analyze both other in-room factors that contribute to OR time and cost as well as out-of-room factors such as equipment setup or extra staff needed based on case complexity. At our institution, the majority of our hospitals have set turnover time targets in order to standardize this cost, although more complex MIS cases could influence this. Another major limitation for cost assessment is the unique billing within our Hopkins institution, as robotic costs are amortized across surgical cases which can significantly distort the cost of robotic cases, versus institutions that directly apply robotic program costs only to surgical cases using the system. It is important to be aware of the method of cost allocation in each institution when reviewing studies on cost of robotic use. We analyzed the impact of transfusion, but not other complications as complications are diverse ranging from simple (ex. UTI, superficial infections) to serious (ex. major injuries, ICU admission). Major complications are rare and from a statistical standpoint difficult to include in our model and could greatly skew cost directly related to the hysterectomy itself without being associated with the true cost of that surgical approach. Simple complications such as urinary tract infection or superficial infections occur following discharge and this analysis did not include post-discharge costs. Therefore, we chose to only analyze blood transfusion due to uniform cost allowing appropriate analysis. Regarding our time frame, the FDA issued its warning against power morcellators in April 2014 and updated in November 2014. 22 It is possible that some of the open cases performed between this time and before the introduction of contained specimen extraction could have been performed laparoscopically or robotically before 2014 with the use of power morcellation. Additionally, robotic programs incur several costs as listed above, and while some of these can be directly charged to the patient case, many are indirectly charged as overhead expenses The indirect application of costs for a robotic program that are amortized across surgical cases can distort the cost of robotic cases, versus institutions that directly apply robotic program costs only to surgical cases using the system. It is important to be aware of the method of cost allocation in each institution when reviewing studies on cost of robotic use. Patient factors such as BMI may not be modifiable at the time of surgery, but selection of the appropriate approach can help to mitigate some of the effects of unmodifiable factors. Multiple perioperative factors in this study reveal opportunities for intervention. For example, a case that could be performed vaginally, but was not, would be associated with higher charges. Similarly, an obese patient with complex fibroids would incur higher charges if the hysterectomy was performed through an open abdominal approach with a 2-day length of stay compared with a robotic approach with same-day discharge, as the higher charges for the increased LOS outweigh the robotic charges. Open surgical approaches also had a higher rate of blood transfusion, which may be related to case complexity necessitating an open approach. However, one study of benign hysterectomy using the NSQIP database suggests that both open and vaginal approaches are associated with higher odds of blood transfusion than laparoscopic (even after accounting for pre-operative hematocrit and patient complexity) – which may be due to the less frequent use of electrosurgical energy in these approaches leading to higher blood loss. It is important to consider which approach minimizes blood loss and subsequent probability of transfusion (with its associated medical risks and financial costs) when selecting a route of hysterectomy. Improving OR efficiency and being thoughtful about minimizing use of unnecessary disposable instruments can improve care and reduce charges, as each additional minute and instrument incurs significant costs. Efficiency strategies include use of checklists, parallel tasking, team and simulation training, and appropriate approach. Parallel tasking ensures OR team members perform their tasks simultaneously rather than in a consecutive series. 18 , 19 Extensive simulation training can improve performance, especially for trainees. 20 , 21 Finally, it is critical to train the OR team as a whole, including circulators and scrubs, to improve team performance and communication. Lastly, each additional day of LOS was associated with $3,000 or more in charges, and thus there should be a focus on enhancing protocols for same day discharge, when feasible. 10 , 22 – 24 Each of these improvement strategies may provide opportunities for reduction in the costs incurred by a hospital and limit the total charges.

In 2018, the U.S. National Health Expenditure was $3.6 trillion, representing 17.7% of the Gross Domestic Product. 1 There have been increasing efforts to control costs and improve healthcare value, and Maryland is an excellent environment to study healthcare costs as it is the only U.S. state with the unique reimbursement system adopting the “ All-Payer Model ” and “ Global Budget Revenue ” methodology, to reduce per-capita hospital expenditures and improve health outcomes. 2 The All-Payer Model has been in place > 40 years; all payers reimburse hospitals at a uniform rate for goods and services provided.. In the Global Budget Revenue , the Maryland Health Services Cost Review Commission signs agreements with individual hospitals regarding annual revenue before each fiscal year, determined by the population, payer mix, quality of care, and historic care volume. 3 – 6 During the fiscal year, if hospitals are trending high relative to budget, charges are lowered by a percentage. Hospitals are incentivized to lower internal costs as charges are lowered and are reimbursed for almost all charges; charges levied should closely represent both true cost as well as revenue collected. Hysterectomy has been studied as a driver of national healthcare expenditures and although vaginal hysterectomy is the gold standard, it is not always feasible based on pathology. Laparoscopy facilitates complex pathology, while allowing for decreased blood loss, recovery, pain, and complications. 7 , 8 Outpatient surgery with most laparoscopic hysterectomies also minimizes hospital stays. 9 – 11 Addition of the robotic platform provides improved visualization, dexterity, and range of motion, however, can be associated with higher costs. 12 – 14 The current American College of Obstetrics and Gynecology (ACOG) recommendation is “robot-assisted cases should be selected based on the likelihood of improved outcomes compared with other surgical approaches due to case complexity or patient factors.” 12 Studies have assessed the impact of approach hysterectomy cost, however, many do not investigate all surgical approaches and only include patient characteristics. One national study assessed hospital and patient characteristics, but did not assess perioperative characteristics. 15 Another study included operative time, length of hospitalization and complications, but did not assess laparoscopic approach. 13 One study assessed all the approaches and characteristics 16 , but in a state without an innovative payment system. In this study, we analyze the association between the total charges and surgical approach to benign hysterectomy to identify patient, perioperative, and hospital-level factors that drive expenses.