Society of Gynecologic Oncology Future of Physician Payment Reform Task Force: Lessons Learned in Developing and Implementing Surgical Alternative Payment Models

Introduction

The total cost of cancer care in the United States continues to rise and is estimated to range from 172 to 206 billion dollars per year by 2020 [1]. This has important implications for our healthcare system’s sustainability as well as for cancer patients who are at increased risk for experiencing significant financial strain due to out-of-pocket costs [1–3]. The National Academy of Medicine (formerly Institute of Medicine) endorses efforts to “improve the affordability of cancer care by leveraging existing efforts to reform payment and eliminate waste” [4]. Toward this end, there has been a recent emphasis on creating value in cancer care, which is a measure of patient health outcomes relative to the total cost [5]. This has resulted in the development of new alternative payment models (APMs) that aim to maximize outcomes and minimize costs [6]. APMs are defined as interventions that change the financing of healthcare with an expressed goal of incentivizing improved clinical outcomes while also reducing unnecessary utilization and the overall costs of care [6]. The experience with surgical APMs in oncology has been limited, and resources to guide clinicians who are entering into an APM contract are even scarcer.

The Society of Gynecologic Oncology (SGO) Future of Physician Payment Reform Task Force (PPRTF) has previously published on its early-stage efforts to develop an Endometrial Cancer Alternative Payment (ECAP) model [7]. The purpose of the current report is to present additional details regarding this ECAP model and share others’ experiences in developing APMs, with a focus on providing guidance and lessons learned for those seeking involvement in or confronted with future APMs. We will summarize the development of the ECAP, describe institutional efforts to develop other surgical APMs, present challenges and lessons learned from these APM experiences, and discuss future directions for APM development in the field.

Summary of ECAP development and the Centers for Medicare and Medicaid Services’ (CMS) feedback

Well before the Medicare Access and CHIP Reauthorization Act (MACRA) passed in 2015, SGO had published on the importance of developing novel care delivery and payment models to improve care [7]. In light of the MACRA, the SGO PPRTF developed an APM focused on endometrial cancer [8–10]. Endometrial cancer was selected because it is the most common gynecologic cancer diagnosed in the United States and is almost uniformly treated by surgical resection. The PPRTF reviewed and incorporated evidence-based best practices, clinical care pathways, data from national cancer registries and administrative claims databases, as well as input from multiple SGO committees, board members, and consultants.

The ECAP included three clinical phases of care: preoperative, surgical, and postoperative care through 60 days. The preoperative phase included all testing, consultation and office visits by surgeons, and ancillary medical services. The surgical management phase included all procedural physician and hospital costs associated with surgery. The postoperative phase included all routine postoperative care and management of complications. Practice patterns were captured through national cancer data registries including the National Cancer Database (NCDB) and the SGO Clinical Outcomes Registry (COR) [11–12]. Associated costs were gathered from national administrative claims data (Marketscan and Medicare) [13]. Altogether, these were used to develop models of overall care and cost. Critical elements driving this surgical APM included selection of the modality of surgery, hospital length of stay, and readmission rates.

Using methods described in the companion paper, the PPRTF compared the current (“status quo”) costs of care to the costs that might be incurred in various alternative (“optimal”) scenarios [14]. Three potential areas of cost savings were examined: increasing the rate of minimally invasive hysterectomy; reducing length of stay; and decreasing readmissions/emergency department visits [15–22]. The PPRTF modeled a scenario incorporating favorable changes in all three parameters. From the baseline scenario in which 35% of hysterectomies were abdominal, 18% laparoscopic, and 47% robotic-assisted, the PPRTF estimated the effect on cost of increasing the rate of minimally invasive surgery by 5%, reducing length of stay by 10%, and reducing readmissions and emergency department visits by 10%. Under all these assumptions, the potential cost savings to the system totaled $980 per patient or 3.2% of the cost of the episode of care, with more optimistic assumptions associated with even higher achievable savings. The baseline mean reimbursement by commercial payers was $30,839, while the mean Medicare reimbursement was $19,825. In addition to reducing cost, improving quality was an equally important goal of the ECAP. After review by the SGO Policy, Quality and Outcomes Taskforce, seven proposed quality measures for endometrial cancer were selected (Table 1) [23–35].

Table 1:

Proposed quality measures applicable to the Endometrial Cancer Alternative Payment (ECAP) model versus an all-hysterectomy bundled payment model

Quality measure	Phase of care	What is measured	Included in ECAP	Included in an all-hysterectomy bundle
Preoperative counseling [13–25, 40]	Preoperative care	Percentage of patients who have been given the purpose for the recommended procedure and goals of care discussion documented in the medical record	Yes	Yes
Women’s health surgical considerations counseling [49–50, 52–53]	Preoperative care	Percentage of patients ≤65 years of age who have been given the risks and benefits of ovarian preservation documented in the medical record Percentage of women ≤35 years of age who have been given the risks and benefits of fertility-sparing options documented in the medical record Percentage of patients not having planned oophorectomy who have been given the risks and benefits of opportunistic salpingectomy documented in the medical record	No	Yes
Use of minimally invasive surgery [28, 29, 51]	Perioperative/surgical care	Percentage of patients who underwent minimally invasive surgery (vaginal, laparoscopy, or robotic-assisted laparoscopy)	Yes	Yes
Timeout checklist [40]	Perioperative/surgical care	Percentage of patients who had timeout checklist performed prior to surgery documented in the medical record	No	Yes
Timing of antibiotic prophylaxis [26]	Perioperative/surgical care	Percentage of patients who have an order for prophylactic parenteral antibiotic to be given within 1 hour (if fluoroquinolone or vancomycin, 2 hours) prior to the surgical incision	Yes	No
Venous thromboembolism prophylaxis administered [27]	Perioperative/surgical care	Percentage of patients who received low molecular weight heparin (LMWH), low-dose unfractionated heparin (LDUH) adjusted-dose warfarin, fondaparinux or mechanical prophylaxis, given within 24 hours prior to incision time or within 24 hours after surgery end time	Yes	No
Surgical site infection [32]	Perioperative/surgical care	Percentage of patients who had a surgical site infection	Yes	No
Regional or national level risk-standardized composite complication rate [47]	Postoperative care	Percentage of patients who experienced at least one of the following complications: Surgical site infection Venous thromboembolism Ileus/obstruction Sepsis Bleeding requiring transfusion Pulmonary Cardiac Central nervous system/stroke Reoperation	No	Yes
Unplanned hospital readmission within 30 days of principal procedure [32]	Postoperative care	Percentage of patients who had an unplanned hospital readmission within 30 days of the principal procedure	Yes	Yes
Emergency department visit within 30 days of principal procedure [14]	Postoperative care	Percentage of patients who had an emergency department visit within 30 days of the principal procedure	No	Yes
Post-discharge review of patient goals of care [23–25, 33–35, 40]	Postoperative care	Percentage of patients who have updated goals of care (including pathology review and follow-up plan) documented in the medical record within 90 days after discharge from the principal procedure	Yes	Yes
Patient-reported outcome measure [54]	Postoperative care	Satisfaction score using Consumer Assessment of Healthcare Providers and Systems Surgical Care Patient Survey (S-CAHPS)	No	Yes

^*The Endometrial Cancer Alternative Payment (ECAP) model was designed to include hysterectomies for endometrial cancer only. A broader, all-hysterectomy model was also conceptualized to include hysterectomies performed for both benign and malignant indications.

The ECAP has since been reviewed by both public and commercial (private) payers. It has been used as a backbone for individual practices and institutions to initiate discussions and the development of APMs locally. The APM was also reviewed with the CMS. The CMS welcomed further development while emphasizing the following points: 1) the need for a larger number of beneficiaries covered by the model to substantiate the CMS’ investment in further development of the APM, 2) requirement that the data be captured predominantly through available claims-based or publicly reported information, 3) that the care pathway transformation not be already occurring naturally, i.e., the shift from abdominal to minimally invasive surgery, and 4) a clear delineation of attribution and risk adjustment strategies that would be tied to payment. The full list of criteria upon which APMs are reviewed by the CMS are available through a publicly available resource [36]. Subsequently, the PPRTF directed its efforts toward a more in-depth understanding of strategies to address attribution, risk adjustment, and quality outcome measurement, as these comprise the key features of any APM and are detailed in the following three sections.

Attribution

Attribution is a methodology that public and commercial payers use to assign patients or episodes of care to the providers accountable for the patient’s outcomes and costs. This is often difficult, as multiple clinicians and healthcare systems may be involved in providing different aspects of a patient’s care, particularly in the setting of multiple medical conditions or multimodal therapy. Two main approaches for attribution are possible: prospective or retrospective. Prospective attribution assigns patients to a payment model once the episode is triggered by a particular ICD-10 or CPT code. Providers are paid a single, risk-adjusted price at the time care is delivered. In retrospective attribution, patients are assigned based on services completed in the previously defined time period (i.e., performance year).

Given the feedback from the CMS, we broadened our scope to a theoretical bundled payment model for an episode of care that would include all hysterectomies for benign and malignant disease. The attribution methodology for the all-hysterectomy bundled payment model included four areas of focus: clinical triggers, episode length, fiscal allocation, and payment reconciliation. One core assumption of this model is that participation would be voluntary during pilot years, with the potential for required participation once the methodology had been proven to be beneficial for all parties. Initially, attribution would be retrospective, with the possibility of phasing in prospective triggers during follow-up years.

Clinical triggers.

An inpatient claim from an acute care hospital with a qualifying ICD-10 Procedure Coding System (PCS) code or a hospital outpatient claim with a qualifying Healthcare Common Procedure Coding System (HCPCS) code would be used to trigger an episode (Table 2). The clinical episode could be triggered by submission of claims by an acute care hospital or a Medicare-enrolled benign gynecologist, gynecologic oncologist, reproductive endocrinologist, or urogynecologist. The Medicare beneficiary would need to be entitled to Part A benefits and enrolled in Part B during the entirety of the clinical episode.

Table 2:

Proposed episode and conditional termination trigger codes for an all-hysterectomy bundled payment model

Episode triggers	ICD-10-PCS	0UT90ZZ, 0UT94ZZ, 0UT90ZZ±0UTC0ZZ; 0UT94ZZ±0UTC4ZZ, 0UT97ZZ±0UTC7ZZ, 0UT9FZZ±0UTC7ZZ
	CPT/HCPCS Code	58150, 58152, 58180, 58200, 58210, 58240, 58260, 58262, 58263, 58270, 58275, 58280, 58285, 58290, 58291, 58292, 58293, 58294, 58541, 58542, 58543, 58544, 58548, 58550, 58552, 58553, 58554, 58570, 58571, 58572, 58573
Conditional termination triggers for adjuvant therapy in cases of cancer diagnosis	CPT/HCPCS Code for radiation therapy planning/delivery	77261-77263, 77280-77290, 77295, 77300, 77331, 77301, 77316-77318, 57155, 57156, 55920, 58364, 58999
	CPT/HCPCS Code for chemotherapy	96400, 96408-96425, 96530

Episode length.

Bundled payment models are often delivered in either a 30-day, 60-day, or 90-day period. We proposed a 30-day episode length based on data indicating that after hysterectomies for both benign and cancer indications, approximately 75-82% of readmissions occur within 15 days of discharge [37–39]. Furthermore, a shorter episode length would reduce the likelihood of large variations in cost, particularly within the context of a diagnosis of malignancy. The attributed cost and outcomes of the proposed bundle would ideally focus on the surgical procedure. We proposed a conditional termination of the episode based on trigger codes that would indicate that adjuvant therapy had started in cases of cancer, so the surgical team would not be held accountable for issues related to radiation or chemotherapy (Table 2).

Fiscal attribution.

The model would hold participating practices financially accountable for the quality and cost of the episode of care. The participating practice would be required to distribute funds using a methodology that best suits its needs. One option for the distribution of risk and reward would be based on the institutional historic proportion of payments received by each clinical role involved in the episode. This would require hospitals to use internal data to compute their own payment and risk distributions. A second option would include distributing payments and savings based on the methodology proposed by the American College of Surgeons (ACS), with portions of the total payment allocated to the surgeon (50%), supporting operative staff (25%), supporting clinical staff (10%), and supporting diagnostics staff (15%) [40].

Payment reconciliation.

Target prices for each episode trigger code would be based on historical claims data less an agreed upon with Medicare discount and further adjusted based on risk stratification and performance on quality metrics, which are described in subsequent sections. We proposed that payment reconciliation occur biannually.

Risk adjustment

Risk adjustment is a method to estimate the resources required to administer care to patients taking into account different baseline health status and needs. Risk adjustment that generates an accurate assessment and quantification of perioperative risks is critical to the success of any model attempting to determine fair payments when moving away from a fee-for-service paradigm. When risk adjustment is inaccurate, physicians are at risk of losing significant reimbursement for patients whose complexity of care extends outside of the expected perioperative course or requires increased resources [41].

We considered two categories we believed were critical to assess risk: 1) surgical factors related to surgical approach and 2) patient factors, including comorbid conditions. With respect to surgical factors, while surgery for uterine cancer historically has been performed via laparotomy, a series of randomized trials have confirmed the safety of minimally invasive techniques in this disease. These trials not only established parity in oncologic outcomes, but they also demonstrated a clear advantage in reduced perioperative morbidity with a minimally invasive approach [28–30,41–43]. An accurate model should therefore reflect the expected decrease in complications and length of stay associated with minimally invasive techniques. The performance of additional surgical procedures, such as lymph node assessment, bowel resection, or hernia repair, can also contribute to increased perioperative risk. Meanwhile, the shift from full lymphadenectomy to sentinel lymph node assessment is associated with decreased risks related to operative time, blood loss, infection, and lymphedema [44]. With respect to patient factors, comorbidities such as obesity, previous surgery, previous hernia repair with mesh, cardiovascular disease, pulmonary disease, hypertension, and diabetes, can be expected to affect postoperative morbidity and mortality, leading to an increased cost of care [38]. Age and fragility can also impact perioperative complications and length of stay [45]. Therefore, an effective risk adjustment model needs to account for preoperative factors to assign a level of risk to each patient entering the care episode.

There are many validated preoperative risk assessment tools currently available, though many are limited to one specialty. One of the most extensive is the ACS National Surgical Quality Improvement Program (NSQIP) surgical risk calculator, which includes over 700 participating hospitals [46]. The surgical risk calculator is an online tool designed for physicians and their patients that leverages NSQIP data to predict the risks of a specific surgery. The calculator utilizes 20 characteristics (2 surgical and 18 patient factors) to estimate the risk of 13 postoperative complications, which range in severity from urinary tract infection to sepsis to death. Length of stay, readmission, and reoperation rates are estimated as well. The ACS calculator is based upon a large, validated, prospective data set that utilizes trained surgical abstractors as opposed to retrospective, claims-based data and includes a comprehensive set of surgical codes, including those used for gynecologic oncology procedures. By quantifying the likelihood of complicated care, the risk calculator can be used to stratify patients into risk categories.

We found that there were two primary risk adjustment strategies, which we characterized as either exclusive or inclusive. The exclusive approach to risk adjustment uses a risk calculator to define the attributes of low-risk patients and excludes patients at significantly higher risk from being included in the APM. This approach results in a homogeneous population and reduces variability in outcomes and costs. Conversely, an inclusive approach identifies all patients with a certain diagnosis or procedure and then uses a risk calculator to sort patients into risk categories for which there would be a designated price per episode at each risk level. This risk adjustment strategy is utilized in the CMS-approved Comprehensive Joint Replacement APM, in which target prices are risk-adjusted into four categories based upon presence of hip fracture and occurrence of major complication [47]. While this approach has the benefit of a larger participant pool, the resulting population is heterogeneous, with inherently greater variability in outcomes and costs; therefore, an inclusive approach can lead to greater financial risk to healthcare providers. Ultimately, no single risk adjustment tool was identified as being most or least effective for an all-hysterectomy bundle.

Quality outcome measurement

Quality outcome measurement is important for value-based payment models and relies on the incorporation of relevant and measurable quality measures that are used to adjust payments based upon performance. We reviewed existing surgical APM efforts, including those that are CMS-approved, such as the Comprehensive Joint Replacement model and the Bundled Payments for Care Improvement (which includes several cardiac and orthopedic procedures), and one submitted to the CMS but not approved, the ACS Physician-Focused Payment Model (ACS-Brandeis Advanced APM) [40, 47–48]. We also examined existing quality measures, such as those used by the Physician Quality Reporting System (PQRS) and now under the Merit-based Incentive Payment System (MIPS) [32].

We modified the quality measures proposed in ECAP (Table 1) as we considered potential expansion to an all-hysterectomy bundled payment model that was not specific to endometrial cancer [22–35]. We developed 9 quality measures, divided into 3 phases of care: preoperative, surgical, and postoperative (Table 1) [32, 40, 48–54]. Preoperative measures were centered on preoperative counseling documentation: 1) patient-centered discussion of goals and risks of surgery and 2) specific women’s health surgical considerations discussion, including ovarian preservation, fertility-sparing options, and opportunistic salpingectomy. Surgical measures included: 1) minimally invasive surgery offered and 2) a timeout safety checklist performed immediately before surgery. Postoperative measures included: 1) regional or national level risk-standardized composite complication rate, 2) 30-day unplanned readmission rate, 3) 30-day emergency department visit rate, 4) patient-reported outcome (PRO) or satisfaction measure, and 5) documentation of patient-centered postoperative discussion, including review of pathology and follow-up plan.

To measure performance and set payments based on the quality measures, we proposed a methodology similar to that of the ACS-Brandeis APM whereby participating practices would be required to meet a minimum reporting threshold for a percentage of patients eligible for the APM to be considered acceptable [40]. Most APMs build in graduated reporting threshold requirements during the implementation phase. Measurement of quality metrics would differ based upon the type of quality measure: positive indicators (higher compliance desired), negative indicators (lower percentage of the outcome desired), or PRO measure. Performance on a pre-specified number of the 9 measures would be used to categorize participating practices into performance-based quality tiers (i.e., acceptable, good, and excellent). Subsequently, payment could flow in either direction, depending on the tier: reconciliation (a share of savings is retained by the participating practices) versus repayment (a percentage of excess spending is reimbursed to the payer for losses). The risk limit would be capped (i.e., upside risk capped at 15-20% and downside risk capped at 8% of the benchmark cost for the episode), and these levels would be mutually decided upon between the provider and payer. Most APMs defer downside risk (repayment from participating practices back to payer) during the first few years of implementation. Ultimately, the complexity of quality outcome measurement and payment adjustment would necessitate validation or piloting.

Case studies

Development of the ECAP was an effort to design an APM that was initially intended for submission to the CMS; however, simultaneously, several SGO members and their institutions were actively negotiating with private payers to develop and implement surgical APMs. Below we present case studies describing the experiences of three of these institutions.

Case Study 1 - Women’s Cancer Care experience

Women’s Cancer Care Associates, LLC (WCCA) is a seven-provider single-specialty private practice group based in Albany, NY, that began to explore alternatives to fee-for-service reimbursement in 2009. Historical challenges for WCCA included low reimbursements in the geographic catchment area of Upstate NY (considered a Medicare rural district), conservative reimbursement multipliers from private payers, and dominance of two physician-owned insurance companies in the local market. The changing landscape that was moving away from fee-for-service models led WCCA to initiate dialogues with the two physician-owned companies and to enroll as the only gynecologic oncology practice in the CMS Oncology Care Model [55].

WCCA’s first step in negotiations began with a notice to both carriers (accounting for 40% of the practice) that WCCA would be terminating participation with all of their lines of service, with the caveat that the practice would like to explore an APM in place of the current fee-for-service arrangement. Nationally, this approach of giving termination notification is common for private practices and institutions seeking enhanced reimbursement from payers over historical amounts. Sometimes these negotiations can result in increased multipliers on the physician conversion factor, as had occurred with WCCA in the past.

WCCA received two disparate responses. The first payer declined negotiation, leading to contract termination and resulting in a negative impact on patients’ access to care in their geographic region and the practice’s volume. After 4 years, a new executive team representing the payer arrived, and a very limited APM was negotiated and implemented. This APM included a per-patient care management fee and an enhanced physician multiplier for fee for service. This agreement was not innovative from a clinical care or patient-centered perspective, but allowed patients to be cared for locally and improved the practice’s finances so that further innovation could be contemplated through other channels.

Negotiations with the second payer were much more successful and ultimately resulted in a modified capitation plan for reimbursement that is still ongoing. While capitation plans are not novel, the key innovation was collaboration and communication on a larger scale. The initiative included a dedicated team that met quarterly to share data and feedback from both sides, with the common goal to measurably improve the care of beneficiaries while lowering the overall cost of care. The parent agreement called for a significant increase in total payment over the previous year’s aggregate and was divided into 12 monthly payments. The annual amount was determined by the number of patients seen in the previous year, which allowed an increase or decrease annually based upon accounting of actual patients seen in the previous period. This agreement had two-sided risk, because if the number fell out of a 10% risk window, either the practice or payer owed a lump sum. This risk window was reduced to 5% in subsequent years. Importantly, chemotherapy drug reimbursement was excluded, as both parties considered the increasing sums that would be involved and the lack of predictability of use to be impediments to a capitated payment structure.

In addition to reimbursement changes, WCCA’s involvement in the APM has allowed for the successful completion of the following projects: a behavioral health management project that implemented depression/anxiety screening for all chemotherapy patients, with a defined path for immediate case management by the payer for at-risk patients; analysis of brand name versus generic chemotherapy and supportive drugs, which led to a significant cost reduction without any deleterious effect on patient care; use of a modified Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey to measure satisfaction with provider communication; and a collaboration to reduce the use of emergency department services, high-tech imaging (i.e., CT, MRI, PET), and inpatient short stays, which led to a significant reduction in all measured parameters [56]. Finally, a recently completed collaboration between the payer, WCCA, and SGO allowed for the testing of the SGO-proposed ECAP.

Case Study 2 - The Ohio State University experience

The James Cancer Hospital at Ohio State is a National Cancer Institute (NCI)-designated Comprehensive Cancer Center with prospective payment system (PPS) exempt status in Columbus, Ohio. PPS exemption means that the institution is paid by Medicare on a fee-for-service basis rather than using the Diagnostics Related Group (DRG) system like most institutions. Hospital leadership recognized the value of applying what was structurally developed through the SGO’s ECAP to gain a better understanding of the drivers of cost at the institution. Initial efforts were spent understanding institutional data, followed by initiating conversations with managed care organizations to ensure that negotiated contracts reimbursed on the value of the care delivered. The process undertaken was developed with the intent of working with currently existing technologies and utilized an iterative approach.

Clinical experts collaborated with cancer analytic experts early in the process to develop an endometrial cancer bundle. Having these co-champions allowed for a more complex understanding of outliers and how they impact value. First, important data elements were identified prior to developing a system that housed the administrative data along with the clinical data. Utilizing Strata Decision Technology (a cloud-based financial planning, analytics, and performance healthcare platform), clinical data were combined with detailed payer-based data as well as hospital and physician billing data to accurately allocate costs. Utilizing Strata also enabled the use of one system for both bundled payment design and tracking bundled payment metrics once implemented.

Basing the bundle on the SGO ECAP, a care pathway was replicated defining the flow of patients through their initial diagnosis, surgical staging, and postoperative care. Reviewing initial patient selection demonstrated that the cost per patient was clustered around a few values based on surgical approach and adjuvant therapy. There were a few outliers, which subsequently required modification of the bundle to eliminate large variability. For example, recognizing that inclusion of patients who had bowel surgery as part of their initial procedure led to increased complications, length of stay, and therefore cost, these patients were omitted in subsequent iterations of the bundle. With a few additional exclusions, the institution began to minimize this variability and gain a better understanding of the cost of a typical patient. This step required continued collaboration between the clinical and analytics teams. Additionally, this allowed for selecting and deselecting specific populations based on, for example, receipt of chemotherapy or robotic versus laparoscopic hysterectomy.

Gaining an understanding of the drivers of cost in the care of patients with endometrial cancer will allow for continued negotiations with managed care organizations. These discussions have been focused on structuring payments within a shared-savings model based on an established baseline and incentives for meeting quality metrics. Through this experience, a challenge that has become apparent is that both managed care organizations and medical centers are comfortable providing and receiving fee-for-service payments and less comfortable with the operational idea of providing and receiving a lump-sum bundled payment without intermittent billing.

Case Study 3 - Moffitt Cancer Center experience

Moffitt Cancer Center is an NCI-designated Comprehensive Cancer Center located in Tampa, FL, that also has PPS exempt status. Moffitt has pursued various value-driven strategies, such as participation in a Total Cost of Care Shared Savings Program, bundled payments, and a Chemotherapy Medical Home. In 2016, Moffitt entered into an agreement with a commercial payer for early-stage non-small cell lung cancer. The 1-year pilot included a single payment for facility and professional fees incurred at Moffitt only. Patient attribution or identification was performed prospectively by Moffitt. The bundle was triggered by either surgery or radiation treatment planning. There were various risk mitigation strategies, including bundle disqualifiers.

Framework for Model Development.

First, the Payer Strategies team identified key stakeholders. Roles included a multidisciplinary group of oncologists, an engaged payer, and members from Revenue Cycle, Financial Analytics, and Clinical Pathways. A detailed roadmap was developed with a disease site-specific market assessment to identify opportunities for bundle development. Due to the clinical data and internal resource requirements, Moffitt focused on disease stages with high volumes for pilot bundles. To increase the reliability of the model, 3 years of data were analyzed. Commercial and Medicare claims were used to maximize sample size. Claims data helped to stratify patients, understand cost and clinical variation, and inform pricing options that are appropriate for the population treated. An important aspect of the analysis was to identify sources of variation from both a clinical and cost perspective. Using financial and clinical data, the team developed an attribution method, target population, trigger event, and episode length. In partnership with physicians and the payer, quality metrics were evaluated and agreed upon. A payment model was developed after analysis of the representative population, with attention to variation in cost as it relates to clinical variables, risk stratification if feasible, and a means to account for the impact of new treatments, and exclusion criteria. The goal was to ultimately develop a pricing model that was specific to the organization.

Clinical Pathways.

Moffitt has developed an extensive set of evidence-based, disease-specific, multidisciplinary clinical pathways that integrate surgical, medical, and radiation treatment. These pathways serve as the basis for defining each bundle and expected treatments for a specific disease and stage. Moffitt routinely assesses clinical alignment with these pathways and has seen high levels of concordance, especially for early-stage cancers, which are the frequent target for bundled payments. These pathways support bundle design by prospectively defining appropriate treatment strategies as well as allowing the retrospective analysis of predictive factors associated with cost outliers and other considerations for bundle exclusions.

Moffitt is in the early stages of bundled payment development and deployment. The creation, implementation, and sustainability of a bundled payment require significant institutional focus, analysis, commitment from physicians and executive leadership, and alignment between the institution and payers. The organization must dedicate analytic resources with reporting capabilities, have access to the entirety of internal clinical and financial data, and have a dedicated care coordination team to support patient navigation. Without such capability and access, the potential impacts of such an arrangement could exceed the risk tolerance of the organization.

Common themes - Pitfalls and potential solutions

The PPRTF noted challenges and common themes across each of the categories of APM development (attribution, risk adjustment, and quality outcome measurement) within existing and proposed APMs, including the SGO ECAP model. Moreover, stakeholder engagement was and is a critical component of successful APM development. Challenges and potential strategies include the following:

Attribution: How might episode length be used to minimize provider risk associated with participation in an APM?

Limiting the occurrence of non-surgical healthcare events in the bundle contract is critical. When developing a bundled payment contract, there are ways in which attribution can be used to minimize risk. One strategy to minimize risk is to limit the length of the episode such that gynecologic oncologists are responsible for only the costs associated with care and complications related to the surgery performed. Using existing literature, we found that most surgical complications occur within the first 30 postoperative days. Additionally, for cancer patients who need adjuvant therapy, radiation and chemotherapy can typically be expected to start between 4 to 6 weeks after surgery. We used these data points to advocate for a 30-day episode length. If the payer is unwilling to accept a 30-day episode, a 60-day bundle could be considered where the occurrence of any triggers or billing codes associated with adjuvant radiation or chemotherapy would automatically cause a termination of the surgical episode. Through this mechanism, the provider would not be held responsible for healthcare utilization that was unrelated to surgical management and the payer may be more assured that all costs associated with surgical complications would be captured in the bundle. The concurrent development and implementation of pilot oncology APMs focusing upon administration of chemotherapy and radiation represent additional opportunities to transition between APMs when adjuvant therapy is indicated after surgery. Examples of these efforts include the CMS Oncology Care Model, the American Society of Clinical Oncology (ASCO) Patient-Centered Oncology Payment framework, the Making Accountable Sustainable Oncology Networks model, and the CMS proposed Radiation Oncology APM [55, 57–59].

Attribution: Should gynecologic oncologists sponsor an APM that incorporates hysterectomy for both benign and malignant indications?

During discussions related to the ECAP model with the CMS, a key recommendation was to increase the number of potential beneficiaries to be included in our bundle given the limited number of Medicare-eligible endometrial cancer patients. This led to discussions of whether to design a general all-hysterectomy bundled payment model that includes all hysterectomy procedures whether for benign, premalignant, or malignant conditions. Since gynecologic oncologists serve as the primary consultants for patients undergoing hysterectomy who are medically complex, surgically challenging, or have a cancer diagnosis, this would require vetting of a sophisticated risk adjustment strategy to ensure that gynecologic oncologists would not bear excess financial risk in this model. In addition, collaboration and agreement with key stakeholders from benign gynecology and other gynecologic subspecialties would be paramount.

Risk adjustment: How do we account for our complex patient population?

As discussed, the patients of gynecologic oncology surgeons are an inherently high-risk case mix, as the subspecialty serves as a consulting service for patients who are diagnosed with malignancy, who need complex surgical procedures, and who have high comorbidity scores. This patient population can be highly variable in terms of risk of complications and costs. Therefore, for bundled payment models to remain viable from the surgeon’s perspective, there needs to be an accurate methodology that accounts for this variation. We identified three main strategies to approach risk calculation: database modeling (i.e., using historical ACS NSQIP, MarketScan, or Medicare data), proprietary software program modeling, or homegrown data modeling. We found that simplified risk stratification, such as that used by the CMS-approved Comprehensive Joint Replacement surgical APM, was one strategy to set different target prices based on indication for the procedure and the presence of major surgical complications [46]. Alternatively, a strategy utilized in APM pilots negotiated with private payers has been to use narrow inclusion criteria, resulting in homogenous patient populations that enter into the APM episode, which results in less variation in complications and costs. This approach naturally simplifies the risk adjustment process and limits the inclusion of outliers.

Quality outcome measurement: How do we select clinically meaningful quality metrics that are also measurable?

As one of the key tenets of value-based payment models is to improve patient outcomes, quality metrics are an essential component of APMs. Selected quality outcomes must be clinically meaningful to the patient within the defined episode time period, which includes the preoperative, surgical, and postoperative phases. One of the biggest challenges is to ensure that the metric is easy to measure in order to limit the administrative burden on providers and their practices. We identified potential barriers with using certain quality measures, such as extracting documentation of counseling from the electronic medical record. We also noted that certain measures were more relevant in specific populations, which could add to the complexity of measurement. For example, nearly all Medicare patients are ≥65 years old; thus, discussion of ovarian preservation or fertility-sparing options would not apply. We also identified factors that would enhance the use of certain quality measures, such as utilizing metrics that are commonly reported by hospitals (i.e., complication rate, readmission rate) or those already included in other APMs. The Hospital CAHPS is a PRO measure commonly used for hospital reporting but is not specific to surgery, whereas the CAHPS Surgical Care Survey has been created for surgical patients but has not been widely implemented [54, 56].

Stakeholder engagement: Who are the key players needed to successfully develop an APM?

Successful APM development requires institutional buy-in and investment of resources to develop necessary analytics and measurement tools. An ideal team would include a project manager and diverse representatives from hospital leadership, billing/finance to provide cost information, information technology to assist in data measurement, data analysts, and healthcare provider champions from the department associated with the disease entering the bundle. The most challenging aspect of APM development is understanding an institution’s current state and how it compares to the base or ideal state the payer would be comparing the institution’s performance to. This requires team members from billing and analytics to test various scenarios on historic patients from the institution. To address risk adjustment, institutions may choose to develop an internal risk prediction tool, which would rely heavily on an analytics team, or to engage private risk management consultants to customize a model for the institution. Infrastructure to provide timely measurement of quality metrics are needed to ensure the ability to report on required quality metrics. Ultimately, there must be evidence to the payer and the institution that there is an opportunity for cost savings and quality improvement that would be mutually beneficial. After entering into a contract, a project team would need to continually assess quality and cost performance to ensure the contract is benefiting patients, payers, and the institution.

Conclusions and future directions

The PPRTF’s previous efforts focused upon developing an ECAP model for submission to the CMS. Since then, we have identified challenges to implementing the ECAP based on feedback from the CMS and have worked to refine the major components of attribution, risk adjustment, and quality outcome measurement within the model design. Meanwhile, several institutions have successfully negotiated with private payers to initiate novel, bundled payments for specific surgical episodes. Future data demonstrating whether practices, institutions, and payers who are engaged in APMs are able to achieve cost savings, without sacrificing quality, will be critical to determine their overall value.

Gynecologic oncologists practice across a diverse set of healthcare organizations, and even within individual institutions there may be a wide variety of practice structures (i.e., mix of hospital-employed vs. privately employed services) that make financial alignment more challenging; therefore, a comprehensive APM (such as one administered through the CMS) would need to factor in several important observations. First, there is great variation in the range of reimbursement costs across the country, including at the local and regional level. Second, the methodology used to select the patient cohort (attribution) may greatly impact the potential outcomes. Third, there is no clear data on what analytical techniques are most effective or least costly to inform stratification (risk adjustment) to account for heterogeneity of the patient population. Fourth, some practices may already be near their optimal thresholds for modifiable factors, such as route of surgery, and desired postoperative outcomes (quality outcome measurement) with limited margin for improvement or error.

With the rapidly changing landscape of value-based payment models, we believe this paper will provide valuable information to the SGO membership as they are increasingly asked to join similar payment models. With the recent proposal of the mandatory radiation oncology bundle by the CMS, it is now more important than ever that gynecologic oncologists familiarize themselves with APMs [59]. As this program matures, the PPRTF will actively advocate for gynecologic cancer patients and providers. We continue to collaborate and discuss APM development efforts with other medical specialty organizations, including the ACS, the American College of Obstetricians and Gynecologists (ACOG), the American Society for Radiation Oncology (ASTRO), and the American Society for Clinical Oncology (ASCO). In the meantime, we are continuing our dialogue with private payers who have already implemented a variety of hysterectomy-related APMs around the country. We believe this paper, along with web-based resources and quarterly newsletters by the PPRTF, will provide practical tools and educational materials for SGO members to use in their own practices as they discuss and engage in payment reform efforts.

Highlights.

• Gynecologic oncologists are experimenting with a variety of different alternative payment models (APMs)

• There are many challenges associated with developing new APMs in gynecologic oncology

• Implementation challenges include stakeholder engagement, attribution, risk adjustment, and quality measurement