Oncology alternative payment models: lessons from commercial insurance

Many payers and clinicians are committed to advancing value-based care through the establishment of alternative payment models (APMs) that incentivize practices and clinicians to improve quality and reduce cost. A multistakeholder working group has observed that in specialty fields such as oncology, despite many attempts to design and implement APM pilots for commercial and Medicare Advantage populations, practical challenges and small numbers of episodes and patients present headwinds to viability and scalability. Despite this, some payers report emerging good practices and are optimistic about APMs. Careful and realistic consideration of the specific goals of a proposed model is warranted, as is close examination of the feasibility of transferring risk.

Impact of clinical trial enrollment on episode costs in the Oncology Care Model (OCM)

6513

Background: Clinical trials are critical for improving outcomes for patients with cancer. However, there is some concern from health insurers that clinical trial participation can increase total cost of care for cancer patients. We investigated the impact of clinical trial participation on total costs paid by Medicare during the OCM program in a large community-based practice. Methods: Tennessee Oncology (TO) is a community oncology practice comprising over 90 oncologists across 30 sites of care. We linked TO trial data and electronic medical record data with OCM data for episodes of care from 2016-2018. To assess the impact of trial participation on total cost relative to routine care, we created matched comparator groups for each OCM episode based on cancer type, metastatic status, number of comorbidities, performance status, and age. Patients with breast cancer receiving hormone therapy only were excluded. Absolute and percent cost differences between groups were calculated for episodes that had a comparator group size of five or greater. Differences in total cost for trial episodes were compared to non-trial episodes, and significance was assessed using the Mann–Whitney U test. We also studied the impact of trial participation on receipt of active treatment in the last 14 days of life (TxEOL), hospice use, and hospitalizations. Results: During the study period, 8,026 completed OCM episodes met study criteria. Patients were enrolled in a clinical trial for 459 of these episodes. On average, episodes during which patients were on trial cost $5,973 less than matched non-trial episodes (Table), independent of early versus late-phase trial. Most savings resulted from decreased drug costs. There were no differences in rates of TxEOL (15% vs. 14% p=1.0), rates of hospitalizations (31% vs. 30% p=0.54), or hospice use (52% vs. 62% p=0.08) between trial and non-trial episodes. Median difference from comparator group average cost was significantly lower for clinical trial episodes (-18% vs. -6%, p<0.01). Conclusions: In the community setting, total costs paid by Medicare for patients participating in clinical trials during OCM episodes were lower than costs for similar patients receiving routine care. Clinical trial participation did not adversely impact end-of-life care or likelihood of hospitalization. These findings suggest that patient participation in clinical trials does not increase total cost of care nor enhance financial risk to payers.[Table: see text]

Bundling cancer subtypes in value-based care: A pilot analysis of lymphoma episodes in the Oncology Care Model

e18850

Background: Value based models (VBMs) in which cancers are bundled are a growing alternative to fee for service, as in the Oncology Care Model (OCM). However, bundles in OCM may not capture the clinical granularity needed to predict resource utilization for cancer subtypes. One such bundle is lymphoma, which groups highly heterogeneous diseases with distinct treatments and differing intensity of care. Here, we compare OCM predicted episode costs (targets) to actual episode costs by lymphoma subtype. Methods: Our cohort study used OCM data from a large academic medical center (AMC) and large community oncology practice (COP). Six-month episodes of lymphoma beginning between July 2016 and June 2019 were categorized based on ICD-10 diagnoses on antineoplastic infusions and E&M visits, as well as disease and data modeling. Episodes were subdivided into follicular (FL), diffuse large B (DLBCL), small B (SBCL), mantle (MCL), Hodgkin (HL), Waldenstrom macroglobulinemia (WM), mature T/NK (T/NK), and Other. The distributional consistency of episode costs and targets for each subtype relative to the rest of the episodes was evaluated by Kolmogorov-Smirnov tests. We also compared the proportion of subtypes contributing to episodes in the AMC vs. COP. Results: A total of 1801 lymphoma episodes were identified (44% in AMC, 56% in COP). The most common subtypes (DLBCL and FL) contributed a larger proportion of episodes in the COP, while less frequent subtypes (T/NK, WM) were more prevalent at the AMC. Further, episode costs are significantly different across individual subtypes. Target variance was significantly lower than cost variance across subtypes. For example, the average target for WM was $50.4K, average costs were $40.2K, with 26% of episodes over target. In contrast, the average target for T/NK was $55.9K, average costs were $72.7K, with 64% of episodes over target. Conclusions: VBMs such as OCM currently aggregate cancer types and lack clinical granularity. Our evaluation of OCM episodes at an AMC and COP found considerable differences in lymphoma populations and in costs by subtype. Failure to account for clinical features (i.e. lymphoma history) could lead to inappropriate shifts of risk from payers to providers in VBMs.[Table: see text]

Real-world patterns of chemotherapy and immunotherapy utilization at end of life in a large community oncology network

22

Background: End-of-life anti-neoplastic treatment does not improve quality of life nor prolong survival of advanced cancer patients. It is also not cost-effective. To-date, there has been little data examining real-world patterns of chemotherapy and immunotherapy treatment at end of life. We investigated use of chemotherapy and/or immunotherapy in the last 14 days of life across a community oncology network of 5 practices, 100 sites of care, and 160 oncology providers. Methods: Using a real-time, network-wide database, we identified patients with solid tumor malignancies who died during an episode of active treatment, defined as having received intravenous (IV) chemotherapy and/or immunotherapy within 90 days of death. We then identified patients in this cohort who received IV chemotherapy and/or IV immunotherapy within 14 days of death (TxEoL). We studied TxEoL patterns by cancer type, treatment type, line of therapy, patient age, patient race, and oncology provider years in practice. Statistical significance was assessed using Pearson’s Chi-squared test. Results: 2,858 qualifying solid tumor cancer patients with dates of death between 1/1/2019 and 5/31/2020 were identified. Observed rates of TxEoL were 16.7% for immunotherapy alone vs. 19.6% for chemotherapy +/- immunotherapy (p = 0.09). We found high variation in TxEoL across 132 oncologists that had 5 or more deceased patients (range: 0% to 50%, mean: 19.2%, median: 19.6%). We found no association of TxEOL with physician years in practice, patient age or race. Rates of TxEoL in the first-line setting were significantly higher than in second-line setting or later (23.3% versus 16.4%, p < 0.01). Patients with head and neck, pancreatic, and hepatobiliary malignancies were the most likely to receive TxEoL, while patients with prostate, brain, and ovarian malignancies were the least likely to receive TxEoL. Conclusions: Our data and method identified wide variation in TxEoL patterns across a large community oncology network, suggesting room for provider-level interventions to improve treatment decisions in patients at high risk of death. Studies within our group, such as examining the impact of palliative care referrals on IV anti-cancer treatment in patients potentially facing end of life, are ongoing.

Feasibility of and associated cost savings from transitioning to therapeutic biosimilar use in a large community oncology network

9

Background: The use of biosimilar drugs in the treatment of cancer offer an opportunity for oncology providers to decrease total cost of care while preserving quality. However, it remains unclear whether providers and patients may resist biosimilar use due to concerns over safety and efficacy. Our national network of 5 practices with over 100 clinics committed to a conversion to therapeutic biosimilars for trastuzumab and bevacizumab after their introduction in July 2019. Methods: Common steps to foster therapeutic biosimilar conversion included frequent communication from medical directors to providers and staff, incorporation of biosimilars into default treatment regimen orders, providing clinical teams lists identifying candidates for conversion, and tracking reasons why biosimilar switch did not occur. Most practices prioritized converting patients initiating new treatments, then later transitioning patients receiving maintenance therapy. This phased approach was taken to ensure that prior authorization and patient consent could be obtained prior to conversion. Rates of biosimilar use were calculated by comparing the number of administrations for which a biosimilar was given to the total number of administrations for which a biosimilar could have been given. Cost savings were calculated by comparing the difference in Medicare allowed rates for each originator and biosimilar drug pair at the time of administration. Results: Biosimilar use increased over time at all practices, from 0% to an average of 67% for trastuzumab and 78% for bevacizumab. The decrease in cost attributed to the use of biosimilars in the study period totaled over $4.4 million. Challenges to biosimilar use included physician preference for the originator drug, difference in preferred agents across payers, and challenges with biosimilar drug storage. Patients rarely had concerns over efficacy and safety. Conclusions: Therapeutic biosimilar adoption in a large oncology network is feasible and can lead to significant cost savings. [Table: see text]

Impact of a built-in electronic medical record prompt on guideline-recommended prophylactic antiviral usage in patients with multiple myeloma receiving proteasome inhibitors

248

Background: Guidelines support the use of prophylactic antivirals to prevent reactivation of herpes varicella in patients with multiple myeloma (MM) on proteasome inhibitors (PI). In our network of five oncology practices spanning over 100 clinic sites, one practice has a built-in prompt for acyclovir use in patients receiving a PI, while the other four practices do not. We used this natural experiment to determine the impact of this prompt on appropriate prophylactic antiviral usage in this patient population. Methods: We retrospectively identified all patients in our network with MM beginning a regimen containing a PI between 1/1/19 and 5/28/20. Of these patients, we identified those with documentation of a prescription for acyclovir or valacyclovir before or within 2 days of the first PI dose. We compared prophylactic usage across five practices. Practice 1 had built a prompt for the prescription of acyclovir in regimens containing bortezomib or carfilzomib within the electronic medical record (EMR) which both reminded physicians and nurses and simplified the prescribing process. No other practices had similar EMR prompts. Results: We identified 583 patients with MM who received a PI during the study period. Wide variation in rates of prophylactic antiviral usage existed across the five practices (range 21%-94%). The highest rate of prophylactic antiviral usage was practice 1 (94%). This was the only practice with a built-in EMR prompt for acyclovir usage in PI regimens. We found no association between use of prophylactic antivirals and individual provider-level volume of patients with MM. Conclusions: Use of prophylactic therapy is heterogeneous across practices. A comprehensive treatment plan containing a prompt in the EMR can markedly increase appropriate utilization. We plan to add an EMR prompt and analytics-driven reminders across our network to improve utilization of all guideline-recommend, orally administered prophylactic medications. [Table: see text]

Impact of embedded palliative care providers compared to externally available palliative care services on the number of patients receiving palliative care referrals in a large community oncology practice

12103

Background: Palliative care improves quality of life and may increase overall survival in patients with solid tumor malignancies. Despite having the ability to refer patients to in-home and external palliative care services, we observed low palliative care referral rates in our practice of 90 oncologists across 30 clinics. We tested whether embedding palliative care providers directly in clinic would improve palliative care referral rates for solid tumor patients. Methods: Between 2017 and 2020, we embedded an independent palliative care provider into five clinics across middle Tennessee. Access to external palliative care services was present both before and after the intervention. Using data from our EHR and billing systems, we performed a pre-post analysis measuring palliative care referrals in the six-month periods immediately before (pre-intervention period) and after (post-intervention period) a palliative care provider was embedded in each clinic. Statistical significance was assessed using Welch’s two sample t-test. Results: 8,636 unique solid tumor patients were seen in the five clinics during the study periods (Table). Despite having the ability to refer patients to external palliative care services in the pre-intervention period, the placement of a palliative care provider into clinic increased the number of solid tumor patients that received a palliative care referral per month at all clinics (min.: 200%; max.: 990%; median: 600%). Four of the five increases were statistically significant (p-values < 0.05). Conclusions: Even when external palliative care services are available, embedding palliative care providers into community oncology clinics significantly increases the rate of palliative care referrals for solid tumor patients. [Table: see text]

Reducing ED visits by “closing the loop” for symptomatic patients

100

Background: Improving the value provided to patients and payers in our system of cancer care relies on reduction of avoidable hospital and emergency department (ED) utilization. Furthermore, recent payer pressures on community oncology practices (COPs) have resulted in an increased focus on improving care coordination (CC) under strict resource constraints. As part of the ASCO Quality Training Program, we tested several low-cost CC interventions, leveraging workflow redesign, already employed care team members, and technology already implemented, to reduce ED visits (EDV) in a single large COP clinic where more than 3,200 cancer patients were treated during 2016. Methods: Baseline EDV rates were obtained through nurse chart review during Jan.-Jun. 2016. The following CC interventions were implemented: Initiated after hours call process with access to EMR and patient access to bidirectional real-time messaging with care team members; Implemented new in-office process to “close the loop” on patient evaluations by creating follow up guidelines for symptomatic telephone triage and in-clinic patient evaluations; Implemented a standard 48 hour follow up process for all EDV and hospital admissions; Increased patient awareness of telephone triage services during and after clinic hours by: augmenting new patient education by staff, developing a magnetic reminder to call the office for non-emergent and emergency situations, and instructions for use of afterhours call system. Nurse chart reviews were conducted throughout implementation to observe effects of new CC processes on EDV. Results: We observed a 30% reduction in EDV from baseline measurement. No new FTEs added and no new technology licenses acquired for this initiative. Conclusions: Low-cost CC interventions can be implemented in COPs to avoid ED utilization. Limitations of this analysis included manual chart abstraction that could not account for EDV outside the partnering health system, illustrating data access for hospital utilization remains a major challenge for quality improvement efforts for COPs. Additional challenges have been experienced in expanding these process improvements from a single large clinic to the broader Tennessee Oncology network of more than 30 clinics.

Care coordination for oral oncolytics through pharmacy integration and cycle 1-day 1 documentation

260

Background: The growing number of oral oncolytic therapies (OOTs) necessitates a standardized EMR workflow that integrates pharmacy activities for dispense and patient management and standardizes cycle-1/day-1 (C1D1) documentation. Our practice’s treatment plans contain appropriately timed OOT follow-up activities including labs, physician follow-up visits, and pharmacy calls for toxicity and adherence checks, however complications in prescription fulfillment such as prior authorization, co-pay assistance, or inability of in-practice pharmacy to dispense limit the predictability of C1D1 dates of OOTs. Methods: An EMR query identified patients at a single clinic location of 5-medical oncologists (MDs) for whom oral oncolytic treatment plans were entered from January 1 to June 30, 2018. C1D1 date entered by the MD in the EMR was compared to the pharmacy processing system dispense date. Ten patients were identified, and 10% (1/10) had an accurate C1D1 documented within the EMR. As part of the ASCO Quality Training Program, to improve the accuracy of C1D1 documentation, a new workflow was implemented whereby: (1) a “hold” activity was added to new EMR treatment plans so that C1D1 remained pending until patients had received medication; (2) clinic checkout staff provided patients with information on the in-practice pharmacy and expectations for next steps; (3) pharmacists utilized existing reporting tools to identify newly entered treatment plans and transcribed orders into e-prescriptions sent to our practice pharmacy; (4) the pharmacy workflow ensued with pharmacy staff leading patient engagement, drug counseling; (5) pharmacists confirm C1D1, document within EMR (6) subsequent treatment plan activities were scheduled. Results: Following education and process changes within the clinic and pharmacy, accurate C1D1 documentation occurred in 90% (9/10) of patients initiating OOTs. Conclusions: Including pharmacy fulfillment time in EMR workflow can improve C1D1 documentation accuracy and associated management of OOTs. Education regarding roles and processes of prescribing MDs, pharmacy staff and clinic staff will be required to scale this process improvement throughout the organization.

Understanding the challenges for oncologists in predicting the end-of-life phase of care in cancer patients with advanced solid tumor diagnoses

281

Background: Early advanced care planning and palliative care improves outcomes during the end-of-life phase of care (EOL) for metastatic cancer patients. Identifying patients who are likely to transition to EOL is a necessary step to prioritize limited palliative care resources and is integral to success in value-based payment models. We analyzed whether physician documentation of prognosis in a clinical pathways system (CPS) could reliably predict when patients are nearing EOL for a large community oncology practice of more than 70 medical oncologists. Methods: Tennessee Oncology (TO) requires physicians to use CPS for all Medicare patients. CPS prompts physicians to answer the “prognostic question” “would you be surprised if this patient died in the next year?” for all OCM patients with advanced solid tumors at the beginning of treatment or at the time of a change in treatment plan. Prognostic question responses were compared to actual dates of death documented in the practice management system. Results: A total of 5,266 distinct patients were expected to trigger an OCM episode during 2017. The CPS prompted a response to the prognostic question for 1,228 (23%) of these OCM patients. There were 665 (54%) positive prognoses (expect patient to live more than 1 year) and 563 (46%) negative prognoses (expect patient to die within 1 year). Physicians documented accurate prognoses in 712 (58%) of cases. For patients with positive prognosis 557 (84%) were accurate. For patients with negative prognosis 155 (21.8%) were accurate. Conclusions: We found that for patients with terminal cancer, it is difficult for physicians to accurately predict prognosis. These findings support the importance of ASCO guidelines pertaining to patient access to palliative care during the entirety of cancer treatment for all patients with metastatic cancer.

[Table: see text]

The effect of guideline-concordant novel therapy use on meeting cost targets in OCM: Results from a large community oncology network

6635

Background: The Oncology Care Model (OCM) is intended to incentivize physicians to improve the quality and reduce the cost of cancer care. In OCM, providers are accountable for all costs during six month episodes of care relative to target costs (TC) derived from a baseline spending period (BSP; 2013-2015). This accountability is intended to foster care coordination to reduce preventable emergency department visits and hospitalizations (EDH). Benefits of reducing EDH may be diluted when new treatment indications for costly immunotherapies (IO) are introduced into clinical practice after BSP. Methods: We identified all non-small cell lung cancer (NSCLC) and bladder cancer (BC) OCM episodes attributed to Tennessee Oncology (TO), a large community oncology network of over 90 oncologists, during performance period 2 (PP2; the most recent PP with available data). We selected NSCLC and BC because both diseases have IO indications that became standard of care after BSP. Using claims data analytics software, we identified all NSCLC and BC episodes with spending above TC, and found a subset of these above target episodes (ATEs) without any EDH that remained above TC due to IO use. Two medical oncologists reviewed these cases in duplicate to assess guideline concordance of IO. Results: During PP2 there were 2,623 OCM episodes attributed to TO, including 240 NSCLC and 31 BC episodes. Spending was above TC in 118 (49%) and 13 (42%) of NSCLC and BC episodes, respectively. For these NSCLC and BC ATEs, EDH was prevented in 62 (53%) and 5 (38%) of cases, respectively. In NSCLC and BC ATEs without EDH, 43 (69%) and 5 (100%) of episodes included IO, respectively. Clinician review in duplicate (S.M.S.; C.A.W.) found that the use of IO was NCCN guideline concordant in 33 (77%) and 4 (80%) of these NSCLC and BC cases, respectively (K = 0.87). Conclusions: Guideline-concordant use of expensive IO as its treatment indications expand poses substantial challenges to meeting cost targets in OCM, even when practices prevent EDH. [Table: see text]