Despite the potential incentives for medical oncologists to reduce radiation therapy utilization under the Oncology Care Model, we find no evidence that such reduction occurred.
ABSTRACT
Objectives: CMS created the Oncology Care Model (OCM) to increase the delivery of cost-efficient cancer care, but in linking medical oncologist compensation to total costs of care, the model also prompted concerns about reductions in radiation therapy utilization. We compare practices that participated in the model with those that did not through its launch to estimate whether radiation therapy utilization was reduced under the OCM.
Study Design: Retrospective analysis of a secondary claims-based data set.
Methods: We used 5 years of reimbursement claims data from a large community oncology network in which approximately half of the practices participated in the OCM to measure the relative change in utilization following OCM participation compared with practices that did not participate in the OCM. We evaluated use of radiation therapy for all cancer diagnoses and, more specifically, bone metastases, lung cancer, and breast cancer to assess whether effects varied by setting using 3 quasi-experimental estimation techniques (difference-in-differences, event study, and triple differences regressions).
Results: We found no evidence of reductions in radiation therapy utilization associated with the OCM between participant and nonparticipant practices in any of the specifications or subpopulations analyzed.
Conclusions: Despite the potential incentives for medical oncologists to reduce radiation therapy utilization, we found no evidence that such reduction occurred.
Am J Manag Care. 2022;28(10):515-519. https://doi.org/10.37765/ajmc.2022.89249
Takeaway Points
CMS created the Oncology Care Model (OCM) under the Patient Protection and Affordable Care Act given the trajectory of Medicare cancer spending.1,2 The model sought to drive care quality under the traditional fee-for-service system by adding an additional $160 Monthly Enhanced Oncology Services payment per chemotherapy-initiated episode to drive care coordination and an additional Performance-Based Payment (PBP) to drive cost efficiency within each episode.3
Although total-cost-of-care models such as the OCM have the appeal of cost predictability and containment through active utilization management, they also raise the possibility that medical oncologist gatekeepers may view the services of other specialists as a source of costs needing containment. The PBP incentives in the model underscore this concern. There is limited research concerning the OCM and radiation therapy, but episodes that involve radiation therapy within the OCM have been found to be generally more costly than those not involving radiation therapy.4 The emerging literature on the OCM also suggests that the model led to modest reductions in care and costs. These effects include relative reductions in office-based care,5 emergency department visits, intensive care unit episodes, end-of-life care,6 and hospitalizations.7-16 These reductions are consistent with similar models such as the Medicare Shared Savings Program,17-19 the 2-sided Pioneer program,20 the accountable care organization (ACO) Investment Model,21 and other ACO models.22,23
In this study, we assess the extent to which radiation oncology referrals were affected by whether medical oncologists were participating in the OCM.
METHODS AND MATERIALS
Overall Research Approach
We combined reimbursement claims data from a large community oncology network in which approximately half of the practices participated in the OCM with 3 quasi-experimental regression specifications (difference-in-differences [DID], event study, and triple differences [DDD] methods) to evaluate whether the OCM was associated with relative reductions in radiation therapy use. This allowed us to measure relative changes in care patterns before and after the OCM between contemporaneous participant and nonparticipant practices while accounting for changes common to all providers (eg, scientific advances, practice guideline changes) that are unrelated to the OCM. In each specification, we controlled for mean differences between OCM and non-OCM practices and individual providers (which controls for any group differences in behaviors such as referral patterns), the month of care, mean patient characteristics such as age and sex, mean share of patients treated with chemotherapy, number of Medicare patients and overall patients treated, and the number of physicians at a given practice. We also included practice-level trends.
Many oncologists follow accepted guidelines concerning radiation therapy,24 but some cancers have less clear guidance, which allows for more individual provider discretion in whether to refer patients for radiation therapy.25 Therefore, in addition to measuring potential effects of the OCM across all patients with cancer, we also looked at specific subgroups with relatively higher levels of discretion: those with bone metastases, lung cancer, and breast cancer. Cancer with bone metastases is a setting with relatively higher radiation therapy utilization, whereas breast cancer and lung cancer are settings in which radiation can be considered discretionary. For example, in an older woman with early breast cancer who is a candidate for breast conservation, radiation can be omitted.
Data and Study Population
The reimbursement claims data used in this study provide utilization measures from a large community oncology network that includes 14 practices that participated in the OCM and 16 practices that did not. In this setting, patients with advanced disease are largely under the primary care of a medical oncologist across the sample. As part of the same national network, these practices also share other similarities, such as financial integration between medical and radiation specialties, purchasing, staffing, pathways, and electronic health record and decision support systems. The data contain provider-by-month totals or means and span 5 years for care given between July 1, 2014, and June 30, 2019. One large outlier practice was excluded because of distinct preperiod patterns that would render the pre-OCM samples incomparable. This study was approved by the governing institutional review board under the exemption criteria.
Data Construction
The key outcome measure is utilization of radiation therapy services. This variable was defined by the presence of any 1 of the following Current Procedural Terminology codes: 77261, 77262, 77263. These codes identify treatment planning and represent a 1-time charge per course of therapy for varying levels of planning difficulty (ie, simple or “clearly defined,” intermediate or “moderate level of planning difficulty,” and complex treatment planning, respectively).26 The unit for this outcome variable is the monthly mean number of unique billed radiation therapy services for a given physician. This is not whether a patient has ever received radiation care but rather the share of a physician’s patients who received radiation care planning services in each month.
We analyzed potential changes in care patterns associated with the OCM on 4 subpopulations: all cancer diagnoses, bone metastases (defined as any of the following International Classification of Diseases, Ninth Revision [ICD-9] or Tenth Revision [ICD-10] codes: 198.5, C79.50-C79.52), lung cancer (defined as any of the following ICD-9 or ICD-10 codes: 162.3-162.9, C34.0-C34.9), and breast cancer (defined as any of the following ICD-9 or ICD-10 codes: 174.0-174.9, C50.0-C50.9). All cancers were included to provide visibility into the overall effects of the model, acknowledging that the OCM may lead to participants selecting different patients and biasing the comparison.7 We then assessed potential care differences within the bone metastases, lung cancer, and breast cancer subpopulations both to avoid this potential selection bias and to identify potentially stronger effects given the relatively higher levels of discretion for radiation therapy among these subpopulations.
Statistical Analyses
We utilized a DID model to estimate the average association of providing care at OCM-participating practices compared with nonparticipating practices across 2 years of preimplementation and 3 years of postimplementation periods. We also included 2 variations to the DID model: (1) an event study and (2) a DDD model. In all specifications, we controlled for mean differences between practices, providers, month, and monthly mean values for average patient age and age squared, mean share of female patients, mean share of chemotherapy-treated patients, total monthly number of patients and Medicare patients, and total number of physicians at a given practice. DID regressions also included group-specific linear time trends. Exact specifications are specified in the eAppendix (available at ajmc.com).
Whereas a DID model considers the average relative postlaunch vs prelaunch differences between OCM and non-OCM practices, an event study considers the relative monthly differences between OCM and non-OCM practices. Intuitively, the event study improves upon a simple plot of means by subtracting the non-OCM value from the OCM value for each period and controlling for potential confounders through a multivariate regression. The estimates are then plotted with 95% CIs (as in the Figure) where statistically significant differences between OCM and non-OCM practices can be seen when a point estimate for a given period, along with its CI lower and upper limit values, are either all positive (for statistically significant positive estimates) or all negative (for statistically significant negative estimates). This allows us to assess 2 important considerations: (1) whether the OCM and non-OCM practices trended similarly in the preperiod (and thus may be suitable comparators) and (2) whether the OCM took time to induce practice pattern differences. In contrast, our DDD specification alters the DID model to assess the degree to which providers who see more Medicare patients may experience greater effects compared with providers with fewer Medicare patients. It is analogous to a dose-effect response in which more Medicare exposure corresponds with more OCM incentive exposure and induces stronger effects.
Finally, while the event study provides insights into potential differences in preperiod trends, we also assessed the extent to which the OCM is associated with different preperiod levels in the observable control characteristics listed above through a simple linear regression model (also specified in the eAppendix).
RESULTS
Selection Bias Considerations
For each specific event study analysis of the cancer subgroups (ie, bone metastases, lung cancer, and breast cancer), most preperiod estimates are not statistically significant at the 90% confidence level and the plotted estimates are relatively flat, suggesting similar preperiod trends when accounting for mean patient characteristic differences (Figure). For the analysis of all cancer types, 7 of the 23 preperiod estimates are negative and statistically significant. However, this is driven by noise in the month prior to OCM launch in which the OCM mean plot increases and the non-OCM mean plot decreases from preperiod values. By instead omitting the preperiod variable for 2 months prior to the OCM where the OCM and non-OCM plots do not uncharacteristically diverge, only 3 preperiod (and 3 postperiod) estimates are statistically significant, supporting this noisy data concern. A similar pattern occurs for lung cancer that again disappears with changing the omitted period to 2 months prelaunch. These results are available upon request but can also be visualized by their flat plots in the Figure.
Finally, among the 11 observable control variables included in the regression, 2 estimates are statistically significant (OCM-participating providers saw patients who were on average 0.024 years older [95% CI, 0.001-0.046; P = .04] and provided 0.263 more units of radiation therapy services [95% CI, –0.010 to 0.536; P = .06]) (Table 1). At a 90% confidence level, one would expect slightly more than (but approximately) 1 statistically significant estimate.
Event Study
Given the 4 regression populations (including the populations with all cancer and lung cancer with 2 months prior to launch as the omitted reference variable), each with 36 postlaunch estimates (eg, a total of 144 estimates), only 7 estimates are statistically significant. This is about what one expects at random given a 95% confidence level. Although the practices themselves decided whether to participate in the OCM, their preperiod characteristics and trends appear to be comparable and the plots of these estimates also show little meaningful relative deviations from their respective comparison practices after OCM launch (Figure).
DID
Given that there are few statistically significant estimates in the event study specification, and that the DID model estimates average prelaunch vs postlaunch differences, there are no statistically significant estimates at the 90% confidence level. These results again hold when focusing only on observations with at least 1 unit of radiation therapy billed (results available upon request).
DDD
There are no statistically significant estimates in the DDD specification at the 90% confidence level—for full samples and samples only with at least 1 unit of radiation therapy billed—suggesting that variation in shares of Medicare patients (and thus OCM exposure) is not associated with different levels of radiation therapy use (Table 2).
DISCUSSION
Across all 4 patient populations, we found no evidence of relative reductions in radiation therapy use among participants in comparison with nonparticipant practices, suggesting that the OCM had little effect on in-network referred radiation therapy care. These results provide timely insights for providers and policy makers who continue to evaluate participation in and design of these models. Following the COVID-19 public health emergency, the Center for Medicare & Medicaid Innovation extended the model through June 2022 and included several flexibilities (eg, providing the option to forgo 2-sided risk arrangements, removal of COVID-19 episodes from PBP calculations, relaxed quality reporting).27 CMS has further decided to indefinitely delay implementation of the Radiation Oncology Model, following stakeholder feedback.28 This model sought to assess whether various payment models (ie, prospective, bundled, site-neutral, modality agnostic, or episodic models) reduce Medicare radiation therapy costs.29-31 Other models that involve PBPs for both medical and radiation oncologists within the same model may yield different utilization patterns. However, our results suggest that medical oncologists in the OCM did not meaningfully alter their referral patterns.
Limitations
This research has several limitations. For one, we were unable to precisely identify more specific subpopulations with cancer in which discretion over radiation therapy is at its highest or in which radiation therapy is most prevalent (eg, we assess breast cancer generally rather than stage I-III disease specifically). Thus, we relied upon the ICD-9 and ICD-10 systems to identify our subpopulations but recognize their shortcomings as coding systems. We attempted to identify the margins in which these effects would be most pronounced (eg, by assessing providers who see relatively more Medicare patients; by focusing on bone metastases, lung cancer, and breast cancer subpopulations; and by conditioning our analyses only on observations with at least 1 unit of radiation therapy billed). In all scenarios, we found no statistically significant effects. However, we recognize that there may be smaller subpopulations inaccessible to us in which the effects are most pronounced and measurable but otherwise muted among all other observations. Finally, we were unable to identify referrals occurring outside of the network. If this was the case most of the time or was more subject to reductions due to in-network vs out-of-network referral preferences, then our empirical approach would miss these important effects due to data limitations.
CONCLUSIONS
Overall, despite the potential incentives for medical oncologists to cut back on radiation therapy services, we found no evidence suggesting that the OCM was associated with reduced levels of radiation therapy. These estimates hold across several model specifications, and our results suggest that medical oncologists did not meaningfully alter radiation oncology services during the first few years of the OCM.
Acknowledgments
In-kind support by way of data was provided by McKesson and The US Oncology Network through their Health Services Research Committee, led by Beatrice Mautner, MSN. Publication of study results was not contingent on approval or censorship of the manuscript. Mittsy Mosshart, BBA; Beth Alvarez, MS; Pat Cayot; and Allen Kowalczyk, BS, prepared and extracted the data used in this study.
Author Affiliations: Tulane University (BW), New Orleans, LA; McKesson (VK, LW, NR), The Woodlands, TX.
Source of Funding: None.
Author Disclosures: Dr Walker was employed by McKesson while writing this article. Drs Kavadi, Wilfong, and Robert are employed by McKesson.
Authorship Information: Concept and design (BW, VK, LW, NR); acquisition of data (BW); analysis and interpretation of data (BW, VK, LW, NR); drafting of the manuscript (BW, VK, LW, NR); critical revision of the manuscript for important intellectual content (BW, VK, LW); statistical analysis (BW); administrative, technical, or logistic support (LW); and supervision (BW, VK).
Address Correspondence to: Brigham Walker, PhD, Tulane University, 1440 Canal St, Ste 1937, New Orleans, LA 70112. Email: bwalker6@tulane.edu.
REFERENCES
1. Oncology Care Model overview. CMS Innovation Center. February 2021. Accessed February 19, 2021.
https://innovation.cms.gov/files/slides/ocm-overview-slides.pdf
2. Kline RM, Bazell C, Smith E, Schumacher H, Rajkumar R, Conway PH. Centers for Medicare and Medicaid Services: using an episode-based payment model to improve oncology care. J Oncol Pract. 2015;11(2):114-116. doi:10.1200/JOP.2014.002337
3. Oncology Care Model. CMS Innovation Center. Accessed February 19, 2021. https://innovation.cms.gov/innovation-models/oncology-care
4. Hurwitz MD, Minetola J, Csik VP. Evaluation of impact of radiation therapy on OCM episodes: lessons applicable to promoting value based care in the RO-APM. Int J Radiat Oncol Biol Phys. 2019;105(1)(suppl):S178-S179. doi:10.1016/j.ijrobp.2019.06.217
5. Walker B, Frytak J, Hayes J, Neubauer M, Robert N, Wilfong L. Evaluation of practice patterns among oncologists participating in the Oncology Care Model. JAMA Netw Open. 2020;3(5):e205165. doi:10.1001/jamanetworkopen.2020.5165
6. Brooks GA, Jhatakia S, Tripp A, et al. Early findings from the Oncology Care Model evaluation. J Oncol Pract. 2019;15(10):e888-e896. doi:10.1200/JOP.19.00265
7. Shenolikar R, Ryan K, Shand B, Kane R. Impact of Oncology Care Model (OCM) on episode costs and performance revenues: considerations for oncology practices. J Clin Oncol. 2018;36(suppl 30):102. doi:10.1200/JCO.2018.36.30_suppl.102
8. Mendenhall MA, Dyehouse K, Hayes J, et al. Practice transformation: early impact of the Oncology Care Model on hospital admissions. J Oncol Pract. 2018;14(12):e739-e745. doi:10.1200/JOP.18.00409
9. Schleicher SM, Chaudhry B, Waynick CA, et al. The effect of guideline-concordant novel therapy use on meeting cost targets in OCM: results from a large community oncology network. J Clin Oncol. 2019;37(suppl 15):6635. doi:10.1200/JCO.2019.37.15_suppl.6635
10. Song A, Csik VP, Leader A, Maio V. The Oncology Care Model: oncology’s first foray away from volume and toward value-based care. Am J Med Qual. 2019;34(4):321-323. doi:10.1177/1062860618824016
11. Parikh RB, Bekelman JE, Huang Q, Martinez J, Emanuel EJ, Navathe AS. Characteristics of medical oncologists participating in the Oncology Care Model. J Clin Oncol. 2019;37(suppl 15):e18017. doi:10.1200/JCO.2019.37.15_suppl.e18017
12. Ennis RD, Parikh AB, Sanderson M, Liu M, Isola L. Interpreting oncology care model data to drive value-based care: a prostate cancer analysis. J Oncol Pract. 2019;15(3):e238-e246. doi:10.1200/JOP.18.00336
13. Li S, Peng Y, Li S, et al. Variations in hospitalization and emergency department or observation (ED/OB) stays using the oncology care model (OCM) methodology in Medicare data. J Clin Oncol. 2018;36(suppl 30):112. doi:10.1200/JCO.2018.36.30_suppl.112
14. McInnes S, Carrino CM, Shoemaker L. Frontline oncology care team primary palliative symptom guideline education, the Oncology Care Model, and emergency department visits. J Clin Oncol. 2018;36(suppl 34):143. doi:10.1200/JCO.2018.36.34_suppl.143
15. Hoverman JR, Taniguchi CB, Hayes J, Eagye K, Mann BB, Neubauer MA. Unraveling the high cost of end-of-life care: an Oncology Care Model experience. J Clin Oncol. 2019;37(suppl 15):11534. doi:10.1200/JCO.2019.37.15_suppl.11534
16. Perry M, Rudy-Tomczak K, Hines S. A process for improving patient survey scores in the Oncology Care Model (OCM). J Clin Oncol. 2018;36(suppl 30):222. doi:10.1200/JCO.2018.36.30_suppl.222
17. McWilliams JM, Hatfield LA, Chernew ME, Landon BE, Schwartz AL. Early performance of accountable care organizations in Medicare. N Engl J Med. 2016;374(24):2357-2366. doi:10.1056/NEJMsa1600142
18. McWilliams JM, Hatfield LA, Landon BE, Hamed P, Chernew ME. Medicare spending after 3 years of the Medicare Shared Savings Program. N Engl J Med. 2018;379(12):1139-1149. doi:10.1056/NEJMsa1803388
19. McWilliams JM, Chernew ME, Landon BE. Medicare ACO program savings not tied to preventable hospitalizations or concentrated among high-risk patients. Health Aff (Millwood). 2017;36(12):2085-2093. doi:10.1377/hlthaff.2017.0814
20. McWilliams JM, Chernew ME, Landon BE, Schwartz AL. Performance differences in year 1 of Pioneer accountable care organizations. N Engl J Med. 2015;372(20):1927-1936. doi:10.1056/NEJMsa1414929
21. Trombley MJ, Fout B, Brodsky S, McWilliams JM, Nyweide DJ, Morefield B. Early effects of an accountable care organization model for underserved areas. N Engl J Med. 2019;381(6):543-551. doi:10.1056/NEJMsa1816660
22. Kaufman BG, Spivack BS, Stearns SC, Song PH, O’Brien EC. Impact of accountable care organizations on utilization, care, and outcomes: a systematic review. Med Care Res Rev. 2019;76(3):255-290. doi:10.1177/1077558717745916
23. Navathe AS, Bain AM, Werner RM. Do changes in post-acute care use at hospitals participating in an accountable care organization spillover to all Medicare beneficiaries? J Gen Intern Med. 2018;33(6):831-838. doi:10.1007/s11606-018-4368-z
24. Gebhardt BJ, Heron DE, Beriwal S. A peer review process as part of the implementation of clinical pathways in radiation oncology: does it improve compliance? Pract Radiat Oncol. 2017;7(5):332-338. doi:10.1016/j.prro.2017.01.006
25. Gradishar WJ, Anderson BO, Abraham J, et al. Breast cancer, version 3.2020, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2020;18(4):452-478. doi:10.6004/jnccn.2020.0016
26. Billing and coding guidelines for radiation oncology including intensity modulated radiation therapy (IMRT). CMS. Updated November 1, 2015. Accessed February 19, 2021. https://downloads.cms.gov/medicare-coverage-database/lcd_attachments/34652_13/L34652_RAD014_BCG.pdf
27. CMMI extends OCM and provides important flexibilities. Community Oncology Alliance. June 4, 2020. Accessed February 19, 2021. https://www.communityoncology.org/cmmi-extends-ocm-and-provides-
important-flexibilities/
28. Radiation Oncology Model. CMS Innovation Center. Updated August 29, 2022. Accessed September 7, 2022. https://innovation.cms.gov/innovation-models/radiation-oncology-model
29. Howard DH, Torres MA. Alternative payment for radiation oncology. JAMA. 2019;322(19):1859-1860. doi:10.1001/jama.2019.15888
30. Thaker NG, Rewari A, Hubbard A. Future of alternative payment models and big data analytics in the post–COVID-19 era: implications for radiation oncology. Int J Radiat Oncol Biol Phys. 2020;108(2):353-355. doi:10.1016/j.ijrobp.2020.06.046
31. Radiation Oncology Model. CMS. Updated June 29, 2022. Accessed February 19, 2021. https://www.innovation.cms.gov/innovation-models/radiation-oncology-model
Exploring Racial, Ethnic Disparities in Cancer Care Prior Authorization Decisions
October 24th 2024On this episode of Managed Care Cast, we're talking with the author of a study published in the October 2024 issue of The American Journal of Managed Care® that explored prior authorization decisions in cancer care by race and ethnicity for commercially insured patients.
Listen
Examining Low-Value Cancer Care Trends Amidst the COVID-19 Pandemic
April 25th 2024On this episode of Managed Care Cast, we're talking with the authors of a study published in the April 2024 issue of The American Journal of Managed Care® about their findings on the rates of low-value cancer care services throughout the COVID-19 pandemic.
Listen
FDA Approves Danziten for Chronic Myeloid Leukemia Without Mealtime Restrictions
November 14th 2024The FDA has granted approval to Azurity Pharmaceuticals' nilotinib tablets (Danziten), a novel version of the tyrosine kinase inhibitor for chronic myeloid leukemia that can be taken without mealtime restrictions.
Read More