ÌÇÐÄvlog

Key Points

QuestionÌý Was the Centers for Medicare & Medicaid Services Oncology Care Model (OCM), an alternative payment model for cancer patients undergoing chemotherapy, associated with differences in Medicare spending, utilization, quality, and patient experience over the model’s first 3 years?

FindingsÌý In this exploratory difference-in-differences study of Medicare fee-for-service beneficiaries with cancer undergoing chemotherapy (483 310 beneficiaries with 987 332 episodes treated at 201 OCM participating practices and 557 354 beneficiaries with 1 122 597 episodes treated at 534 comparison practices), OCM was associated with a statistically significant relative decrease in total episode payments of $297 that was not sufficient to cover the costs of care coordination or performance-based payments. There were no statistically significant differences in most measures of utilization, quality, or patient experiences.

MeaningÌý In its first 3 years, the OCM was significantly associated with modestly lower Medicare episode payments that did not offset model payments to participating practices, and there were no significant differences in most utilization, quality, or patient experience outcomes.

ImportanceÌý In 2016, the US Centers for Medicare & Medicaid Services initiated the Oncology Care Model (OCM), an alternative payment model designed to improve the value of care delivered to Medicare beneficiaries with cancer.

ObjectiveÌý To assess the association of the OCM with changes in Medicare spending, utilization, quality, and patient experience during the OCM’s first 3 years.

Design, Setting, and ParticipantsÌý Exploratory difference-in-differences study comparing care during 6-month chemotherapy episodes in OCM participating practices and propensity-matched comparison practices initiated before (January 2014 through June 2015) and after (July 2016 through December 2018) the start of the OCM. Participants included Medicare fee-for-service beneficiaries with cancer treated at these practices through June 2019.

ExposuresÌý OCM participation.

Main Outcomes and MeasuresÌý Total episode payments (Medicare spending for Parts A, B, and D, not including monthly payments for enhanced oncology services); utilization and payments for hospitalizations, emergency department (ED) visits, office visits, chemotherapy, supportive care, and imaging; quality (chemotherapy-associated hospitalizations and ED visits, timely chemotherapy, end-of-life care, and survival); and patient experiences.

ResultsÌý Among Medicare fee-for-service beneficiaries with cancer undergoing chemotherapy, 483 319 beneficiaries (mean age, 73.0 [SD, 8.7] years; 60.1% women; 987 332 episodes) were treated at 201 OCM participating practices, and 557 354 beneficiaries (mean age, 72.9 [SD, 9.0] years; 57.4% women; 1 122 597 episodes) were treated at 534 comparison practices. From the baseline period, total episode payments increased from $28 681 for OCM episodes and $28 421 for comparison episodes to $33 211 for OCM episodes and $33 249 for comparison episodes during the intervention period (difference in differences, −$297; 90% CI, −$504 to −$91), less than the mean $704 Monthly Enhanced Oncology Services payments. Relative decreases in total episode payments were primarily for Part B nonchemotherapy drug payments (difference in differences, −$145; 90% CI, −$218 to −$72), especially supportive care drugs (difference in differences, −$150; 90% CI, −$216 to −$84). The OCM was associated with statistically significant relative reductions in total episode payments among higher-risk episodes (difference in differences, −$503; 90% CI, −$802 to −$204) and statistically significant relative increases in total episode payments among lower-risk episodes (difference in differences, $151; 90% CI, $39-$264). The OCM was not significantly associated with differences in hospitalizations, ED visits, or survival. Of 22 measures of utilization, 10 measures of quality, and 7 measures of care experiences, only 5 were significantly different.

Conclusions and RelevanceÌý In this exploratory analysis, the OCM was significantly associated with modest payment reductions during 6-month episodes for Medicare beneficiaries receiving chemotherapy for cancer in the first 3 years of the OCM that did not offset the monthly payments for enhanced oncology services. There were no statistically significant differences for most utilization, quality, and patient experience outcomes.

Cancer is the second leading cause of death in the United States, resulting in nearly 600 000 deaths annually.¹ It is estimated that more than 1.8 million individuals will be diagnosed with cancer in 2021.¹ The median age of patients diagnosed with cancer is 66 years,² and a substantial portion of the care patients with cancer receive is covered by Medicare. Medical spending on cancer treatment was estimated at $200 billion in 2020.³ Inpatient hospital care accounted for the largest single component of cancer spending in 2004-2010⁴; however, chemotherapy may now be the most costly treatment component, with many newly approved chemotherapy treatments priced at more than $10 000 per month of therapy.⁵ In this rapidly evolving environment, there are increasing calls to reorganize cancer care delivery and payment to improve the quality and value of cancer care.

Quiz Ref IDThe Centers for Medicare & Medicaid Services’ (CMS) Oncology Care Model (OCM) is the largest and most important alternative payment model addressing value-based payment for cancer care, with more than 3200 oncologists participating.⁶ The OCM seeks to improve the quality and coordination of cancer care at the same or lower cost as Medicare.^7,8 Under the OCM, approximately 200 physician practices voluntarily entered agreements with financial and performance accountability for 6-month episodes of care for beneficiaries with cancer undergoing chemotherapy (including cytotoxic, hormonal, and biologic therapies).

This study used a difference-in-differences framework to estimate the association of the OCM with changes in Medicare spending, health service utilization, health outcomes, quality, and care experiences during the first 3 model years. Changes over time for OCM episodes were compared with changes for fee-for-service Medicare episodes among a comparison group of practices not participating in the OCM.

Quiz Ref IDParticipating OCM practices are paid under Medicare’s fee-for-service billing rules and may bill Medicare for a $160 Monthly Enhanced Oncology Services payment for each month of an episode. Practices earn performance-based payments if they meet total cost of care and quality goals (based on the OCM Performance-Based Payment Quality Score). Practices could elect 1-sided or 2-sided risk arrangements; all participating practices elected 1-sided risk through the first 30 months of the model.

The OCM requires participating practices to enhance care coordination and patient engagement; provide core functions of patient navigation; use certified electronic health record technology; offer 24-hour, 7-day access to an appropriate clinician with real-time access to the practice’s medical records; use data for continuous quality improvement; create a care plan for each OCM beneficiary with information such as prognosis, treatment goals, advance care plans, estimated out-of-pocket costs, and psychosocial and survivorship plans; and treat Medicare beneficiaries with therapies according to nationally recognized guidelines. The combination of the OCM’s target benchmarks for total costs of care and performance-based payments that incorporate quality assessments incentivize practices to reduce Medicare payments while maintaining or improving quality.

We used a difference-in-differences design to estimate the association of the OCM with changes in Medicare payments, utilization, health outcomes, quality, and care experiences during the first 3 model years. We used Medicare fee-for-service Parts A and B claims, Part D Prescription Drug Event, and Medicare enrollment and coverage data from January 1, 2014, through June 30, 2019, supplemented with CMS Health Professional Shortage Area and Area Health Resource data; National Plan and Provider Enumeration System data; a proprietary Office-Based Physician File; and academic medical school affiliation data.^9,10 The study protocol was approved by the Abt Associates Institutional Review Board, which waived informed consent.

The study population included Medicare fee-for-service beneficiaries with any type of cancer who had a qualifying chemotherapy episode.¹¹ Episodes were identified using outpatient, carrier, and durable medical equipment claims. Episodes triggered by Part D chemotherapy were required to have a Part B claim with a cancer diagnosis on or in the preceding 59 days. All episodes had an evaluation and management visit with a cancer diagnosis.

Cancer diagnoses were categorized as 1 of 24 cancer types per OCM rules. For some analyses, we categorized low-risk breast cancer (breast cancer episodes treated with hormonal therapy only), low-intensity prostate cancer (prostate cancer episodes treated with hormonal therapy only), and low-risk bladder cancer (bladder cancer episodes treated with intravesicular therapy only) as lower-risk episodes; all other episodes were higher-risk episodes.

Episodes were included if the beneficiaries were continuously enrolled in fee-for-service Medicare Parts A and B, had Medicare as the primary payer, and were not Medicare end-stage renal disease coverage beneficiaries for all 6 months of their episode (or until death). Episodes were attributed to the oncology practice (based on tax identification number) that provided the plurality of cancer-related evaluation and management visits during the episode.

We used propensity score matching to select non-OCM comparison practices—and their attributed episodes—that were similar to the 201 OCM practices in the pre-OCM baseline period.¹² We estimated the propensity for practice participation in the OCM based on episode, practice, and market factors and matched each OCM practice to up to 10 non-OCM practices (see eAppendix A and eTable 1 in Supplement 1). The comparison group provided a counterfactual to measure the effect of the OCM.

We examined various prespecified outcomes. Outcome measures were calculated at the beneficiary episode level, except end-of-life and survival measures, which were calculated at the beneficiary level.

The primary outcome assessing Medicare spending was total episode payments, defined as the sum of Medicare Parts A, B, and D payments for all cancer and noncancer services received during an episode, whether delivered by the attributed practice or not. Total episode payments exclude Monthly Enhanced Oncology Services payments and performance-based payments. We also evaluated subcomponents of spending (eAppendix B1 in Supplement 1).

Utilization measures included hospitalizations in acute care hospitals, emergency department (ED) visits, and select Part B outpatient services (eg, evaluation and management visits, imaging, radiation therapy). Hospitalizations and ED visits are included in the OCM Performance-Based Payment Quality Score.

We identified all chemotherapy agents (except hormonal therapies) received from day 1 of an episode through day 8 to characterize the episode-initiating treatment regimen (eAppendix B2 in Supplement 1).

Novel therapies are defined for the OCM as oncology drugs approved by the US Food and Drug Administration (FDA) for a particular cancer type within the prior 2 years; benchmark prices may be adjusted to reflect situations in which a practice has a higher proportion of expenditures for such novel therapies than is reflected in the trended baseline prices.¹¹ Because the specific indication for a drug cannot be reliably ascertained with claims data (eg, histology, stage, and genetic markers are not available), a drug is considered novel if used for the cancer type(s) for which it was approved.

We examined use of checkpoint inhibitor immunotherapies during treatment episodes for cancer types for which FDA approvals or clinical guidelines (through 2018) supported the use of immunotherapy agents.

Distinct from the OCM measures of ED visits and hospitalizations described above, we adapted the CMS measure OP-35 (endorsed by the National Quality Forum) of chemotherapy-associated hospitalizations and ED visits¹³ (eAppendix B3 in Supplement 1).

We adapted measures from the American Society of Clinical Oncology Quality Oncology Practice Initiative¹⁴ to assess timeliness of adjuvant chemotherapy, defined as chemotherapy initiation within 2 months after surgery for patients with colon cancer or breast cancer (eAppendix B4 and eTable 2 in Supplement 1).

We examined national measures of care quality, including hospitalizations and 2 or more ED visits in the last 30 days of life, Part B chemotherapy in the last 14 days of life, and hospice enrollment 3 or more days (and not more than 180 days) before death.^14,15 Enrollment in hospice 3 or more days before death is included in the OCM Performance-Based Payment Quality Score.

We examined restricted mean survival time¹⁶ through 18 months for beneficiaries who were likely treated for newly diagnosed or newly recurrent or progressive cancer and had no episode in the prior 12 months (eAppendix B5 and eFigure 1 in Supplement 1). We included beneficiaries with 1 of 7 cancer types with high prevalence and at least moderately high expected mortality (acute leukemia, high-risk breast cancer, chronic leukemia, colorectal cancer, lung cancer, lymphoma, and pancreatic cancer).

Patient experiences, which contributed to the OCM Performance-Based Payment Quality Score, were assessed using a survey instrument adapted from the CAHPS Cancer Care Survey¹⁷ (eAppendix B6 and eTables 3-4 in Supplement 1). Beneficiaries in the OCM and comparison practices provided an overall rating of care and answered questions about access, effective communication, exchange of information, symptom management, shared decision-making, patient self-management, end-of-life care, and out-of-pocket spending.

We examined spending, care delivery, and outcomes for 6-month episodes initiated during the pre-OCM baseline period (July 2, 2014, through January 1, 2016) and the first 3 years of the model (6-month performance periods 1-5, including episodes initiated and completed between July 1, 2016, and June 30, 2019).

We used difference-in-differences regression analyses¹⁸ to estimate the effects of the OCM, adjusting for observable factors unrelated to the OCM that could influence outcomes. Difference-in-differences analysis is a statistical technique that compares changes in an outcome between the baseline period and the intervention period for the treatment group (OCM episodes) relative to a comparison group (comparison episodes). We adjusted our difference-in-differences estimates for episode-, practice-, and market-level factors to control for time-varying changes and influences that affected both the OCM and comparison groups (eAppendix C in Supplement 1).

Episode adjustment variables included beneficiary race and ethnicity because of additional plans to assess whether OCM effects differed by race and ethnicity. We used the Medicare RTI race code variable (eAppendix A in Supplement 1), categorized as non-Hispanic Black, Hispanic, non-Hispanic White, and other.

For each claims-based measure, we tested the null hypothesis that OCM and comparison episodes had parallel trends during the 18-month baseline period (eTable 5 in Supplement 1). Difference-in-differences results are not reported for outcome measures for which we rejected the parallel trends assumption (α = .05).^12,19

We present difference-in-differences results as point estimates with 90% confidence intervals. A 2-sided P < .10 was used for the OCM evaluation to reduce the likelihood of not identifying model effects. Because of this and our examination of numerous outcomes, study findings should be considered exploratory.

Because payments, clinical status and severity, treatments, and potential for savings can vary considerably among the 24 types of cancer, we conducted payment and utilization analyses by subgroups of episodes. We categorized episodes as lower-risk episodes if the primary cancer type was low-risk breast cancer, low-intensity prostate cancer, or low-risk bladder cancer. As defined by the OCM, low-risk breast cancer and low-intensity prostate cancer episodes are treated with hormonal therapies only, and low-risk bladder cancer episodes are treated with intravesicular therapies only (local therapies instilled into the bladder). Episodes for all other cancer types were categorized as higher-risk episodes.¹¹

Because total episode payments do not include the model’s Monthly Enhanced Oncology Services payments or performance-based payments, we estimated the overall net financial effects of OCM incorporating these payments, along with the relative decreases in total episode payments, through the first 4 performance periods (data for the fifth performance period was not yet available because practices often bill for the monthly payments after an episode ends and because performance-based payments are calculated several months after each performance period ends).

Analyses were conducted using SAS Enterprise Guide version 7.1 (SAS Institute Inc) and Stata/MP versions 14.2 and 15 (StataCorp).

We examined 987 332 OCM episodes attributed to 201 participating practices for 569 771 beneficiaries and 1 122 597 comparison episodes attributed to 534 nonparticipating practices for 657 137 beneficiaries in the baseline and intervention periods. Episode-level characteristics of the study population are shown in Table 1 and in eTable 6 in Supplement 1.

Demographic characteristics were similar in OCM and comparison episodes in both the baseline and intervention periods. Overall, 60% of OCM episodes and 57% to 58% of comparison episodes were for female beneficiaries. Comparison episodes had a higher proportion of beneficiaries who were dually eligible for Medicare and Medicaid (approximately 16%) relative to OCM episodes (approximately 14%).

Thirty-two percent to 36% of episodes were lower-risk episodes. Low-risk breast cancer episodes were the most common episode cancer type (22%-24%). The distribution of cancer types was similar for OCM and comparison episodes; one exception was low-intensity prostate cancer episodes, which represented a higher proportion of comparison episodes (11%) than OCM episodes (8%) in the baseline and intervention periods.

The OCM practices were, on average, larger than comparison practices based on both episode volume and number of physicians (eTable 6 in Supplement 1). The OCM practices also had more nurse practitioners and physician assistants than comparison practices, but the mix of oncology subspecialties (hematologist/oncologist, surgical oncologist, etc) was similar. The OCM practices were more likely than comparison practices to be affiliated with an academic medical center but less likely to be part of a health system/hospital.

In the baseline period, mean total episode payments were $28 681 for OCM episodes and $28 421 for comparison episodes. In the first 3 years of the OCM, mean total episode payments increased to $33 211 for OCM episodes and $33 249 for comparison episodes (Table 2; eFigure 2 in Supplement 1). The largest increases were in payments for Part B chemotherapy drugs and Part D drugs.

The OCM was associated with a statistically significant $297 decrease in total episode payments relative to comparison episodes (90% CI, −$504 to −$91; P = .02). This $297 represents 1% of the mean OCM baseline value of $28 681. These results do not include Monthly Enhanced Oncology Services payments or performance-based payments. Relative decreases in total episode payments attributable to the OCM were not sufficient in any performance period to offset the averaged billed Monthly Enhanced Oncology Services payments of $704 per episode.

Episode payments varied considerably for different types of cancer, particularly between higher- and lower-risk episodes. For both OCM and comparison episodes, the mean total episode payments were approximately $46 700 for higher-risk episodes and $7500 for lower-risk episodes in the intervention period (Table 2).

The association of the OCM with total episode payments differed significantly for higher-risk and lower-risk episodes (P < .001 for interaction). Among higher-risk episodes, the OCM was associated with a statistically significant decrease in total episode payments of $503 in OCM episodes relative to comparisons (90% CI, −$802 to −$204; P = .006), representing 1.3% of the mean OCM baseline value of $39 934. This relative decrease was statistically significant for 4 common cancer types: lung cancer, lymphoma, colorectal cancer, and high-risk breast cancer (eFigure 3 in Supplement 1).

In contrast, for lower-risk episodes, the OCM was associated with a statistically significant $151 increase in payments for OCM episodes relative to comparison episodes (90% CI, $39-$264; P = .03), representing 2.1% of the mean OCM baseline value of $7226.

The OCM was associated with a statistically significant $114 (90% CI, −$20 to −$25) decrease in Part A payments (for hospitalizations, hospice, and postacute care) among OCM episodes relative to comparisons (Table 2). The mean Part A payments in the baseline period were approximately $6000 (20%-21% of total episode payments) and decreased more for the OCM than for comparison episodes (eFigure 2 in Supplement 1).

Part B payments comprised 60% of total episode payments. The OCM was associated with a statistically significant $175 (90% CI, −$340 to −$9) relative decrease in Part B payments (Table 2), which represented 1% of the mean OCM baseline Part B payments ($17 080). The OCM was not significantly associated with changes in Part B chemotherapy drug payments, which comprised 30% of total episode payments (eFigure 2 in Supplement 1). Rather, statistically significant relative decreases were observed for nonchemotherapy drugs (difference in differences, −$145; 90% CI, −$218 to −$72) (Table 2), especially supportive care drugs used to mitigate chemotherapy adverse effects (difference in differences, −$150; 90% CI, −$216 to −$84).

The OCM had no statistically significant association with payments for radiation therapy, outpatient visits, or laboratory services (Table 2). There was a statistically significant association of the OCM with lower payments for imaging services (difference in differences, −$18; 90% CI, −$29 to −$8).

Part D payments increased substantially from baseline to the intervention period in both OCM and comparison episodes, but there was no statistically significant association of the OCM (Table 2). Part D payments accounted for approximately 23% of total episode payments in the baseline period and 30% of total episode payments in the intervention period (eFigure 2 in Supplement 1).

There was no statistically significant association of the OCM with key utilization measures (Table 3). Hospitalizations in acute care hospitals and ED visits declined between the baseline and intervention periods similarly for OCM and comparison episodes. The OCM was not significantly associated with outpatient visits or radiation therapy services. The OCM was associated with 46 (90% CI, 6-86) fewer standard and other imaging services per 1000 episodes. Additional utilization measures are shown in eTable 7 in Supplement 1.

Episode-initiating chemotherapy regimens for lung cancer, colorectal cancer, high-risk breast cancer, and high-intensity prostate cancer were very similar in OCM and comparison episodes, in both the baseline and intervention periods (eFigures 4-7 in Supplement 1). While the most frequent episode-initiating chemotherapy regimens changed substantially over time for some cancer types, changes were similar in OCM and comparison episodes. The OCM and comparison episodes also had similar use of novel chemotherapy drugs, with no statistically significant association with the OCM (Table 4).

Among episodes for cancer types in which checkpoint inhibitor immunotherapy drugs were approved and used in at least 5% of episodes in the baseline period, there were large increases in use of immunotherapies (Table 4) that were significantly greater for OCM episodes than for comparisons for lung cancer and melanoma episodes. For cancer types with little or no use of immunotherapy in the baseline period, in which difference-in-differences analyses were not performed, there was similar or greater use of immunotherapy in OCM episodes than in comparison episodes (eTable 8 in Supplement 1).

The OCM had no statistically significant association with chemotherapy-associated hospitalizations but was associated with a statistically significant relative decrease in the proportion of episodes with a chemotherapy-associated ED visit (difference in differences, −0.3%; 90% CI, −0.6% to 0.0%) (Table 5) of unclear clinical importance. The OCM had no statistically significant association with the timeliness of initiating adjuvant chemotherapy following surgery for breast or colorectal cancers, chemotherapy in the last 14 days of life, occurrence of 2 or more ED visits in the last 30 days of life, or hospice enrollment or timing (Table 5). The OCM was associated with a statistically significant decrease in the proportion of beneficiaries hospitalized in the last 30 days of life (difference in differences, −1.1%; 90% CI, −1.9% to −0.4%) (Table 5).

Among patients with episodes for 7 types of high-prevalence and high-mortality cancers, mean survival time through 18 months at baseline was 428 days (14.3 months) for OCM beneficiaries and 430 days (14.3 months) for comparison beneficiaries. Mean survival time increased to 434 days (14.5 months) and 438 days (14.6 months) for OCM and comparison beneficiaries, respectively. The OCM had no statistically significant association with survival (difference in differences, −2.2 days; 90% CI, −4.6 to 0.3 days) (Table 5). The interaction of OCM by cancer type was not statistically significant (P = .42) (eTable 9 in Supplement 1).

Overall ratings of care were high (mean rating, 9.3 on a 10-point scale). The OCM had no statistically significant association with overall ratings of care (difference in differences, 0.0; 90% CI, −0.1 to 0.1) or any of the composite survey measures (Table 5). There was no statistically significant association of the OCM on beneficiary-reported out-of-pocket spending (eFigure 8 in Supplement 1).

Through the end of the first 4 OCM performance periods (2.5 years), considering reductions in total episode payments, Monthly Enhanced Oncology Services payments, and performance-based payments, the OCM resulted in net losses to Medicare of $315.6 million (eFigure 9 in Supplement 1).

Quiz Ref IDIn its first 3 years, the OCM was associated with a statistically significant $297 per-episode relative decrease in Medicare total episode payments that was insufficient to offset other Medicare outlays for the OCM, including per-beneficiary per-month payments for enhanced oncology services and performance-based payments made to participating practices. Accordingly, the OCM has thus far led to net losses for the Medicare program. Quality of care did not differ significantly for OCM vs comparison episodes—including for 3 measures used in the OCM Performance-Based Payment Quality Score—excepting 2 measures in which the OCM led to small relative declines in hospitalizations in the last 30 days of life and in chemotherapy-associated ED visits. Patient experiences did not change despite inclusion in the Performance-Based Payment Quality Score and program requirements for patient navigation and care plans that offered opportunities to improve patient-clinician communication.

Quiz Ref IDCancer is a highly heterogeneous disease with various treatments, and lower-risk episodes were substantially less costly than higher-risk episodes. There were important differences in the association of the OCM with total episode payments for lower- vs higher-risk episodes. The OCM led to relative decreases in total episode payments for higher-risk episodes but relative increases in total episode payments for lower-risk episodes—for which fewer opportunities for savings exist. This suggests that future alternative payment models in oncology could focus on episodes with costly and intensive chemotherapy treatments, for which care delivery redesign has more potential to reduce payments and improve care quality. CMS’s current description of the Oncology Care First model, a potential successor to the OCM, excludes episodes with only endocrine or hormonal therapies from its performance-based payment episodes.²⁰

The largest relative decrease in payments due to the OCM was for Part B drugs. This decrease was not for chemotherapy drugs (which comprised more than 30% of total episode payments during the OCM) but for nonchemotherapy drugs (which comprised 8% of total episode payments). The lack of association of the OCM with Part B chemotherapy payments (or Part D drug payments) is consistent with the very similar chemotherapy use in OCM and comparison episodes and no association of the OCM with use of costly novel therapies.

In contrast, the OCM was associated with a statistically significant relative decrease in payments for supportive care drugs, the largest category of Part B nonchemotherapy drugs. The $150 decrease in payments for supportive care drugs represents more than half of the $297 relative decrease in total episode payments. Clinicians may perceive more opportunities for substituting supportive care drugs without negatively affecting cancer treatment outcomes. Alternatively, practices may find it easier to change supportive care drug use, which is often protocolized, than to influence highly individualized decisions about use of chemotherapy drugs.

Quiz Ref IDMany OCM practices¹⁹ had expected that greater emphasis on care coordination, symptom management, and advance care planning (facilitated by Monthly Enhanced Oncology Services payments) would lead to meaningful reductions in hospital-based care. Acute hospitalizations declined by approximately 8% in both OCM and comparison episodes, perhaps reflecting the wider cross-payer environment of health care delivery reform that affects OCM participants and nonparticipants alike. Declining trends in hospital-based care during cancer treatment episodes may also reflect the evolution of more effective, less toxic cancer therapies.

Several studies have examined effects of alternative payment models in oncology, but none approached the scale or scope of the OCM.²¹ A small UnitedHealthcare voluntary alternative payment model (5 practices treating 810 patients with breast, colon, or lung cancer) during 2009-2012 reported substantial reductions in spending without measurable effects on quality.²²

Other studies have evaluated outcomes of cancer patients in accountable care organization (ACO) initiatives. A study examining the Medicare Physician Group Practice Demonstration ACO program was associated with a $721 reduction in annual spending for patients with cancer—a reduction attributable to lower hospitalization rates.²³

Other studies found no effect of ACOs on overall spending for patients with cancer²⁴ and no or minimal effect of ACOs on end-of-life care outcomes^25,26 or surgical care quality.²⁷ The scope of the OCM makes it especially critical that the lessons of the OCM are carefully explored and considered in the design of future oncology payment models.

This evaluation has several limitations. First, practices were not randomized to participate in the OCM, and those volunteering to participate may differ from those that did not. To minimize this concern, comparison practices were selected based on a comprehensive list of beneficiary-, practice-, and market-level characteristics that might influence the outcomes of interest, and a difference-in-differences design was used to account for secular trends.

Second, the difference-in-differences design relies on the assumption of nondifferential trends over time. For all measures reported herein, there was no evidence for differential trends in the baseline period,^12,19 although the baseline period included only 6 quarters.

Third, advances in cancer treatments during the study led to changing patterns of care unrelated to the OCM, although the comparison group and difference-in-differences design help to minimize bias.

Fourth, clinical information, such as cancer stage and histology that may have allowed for more precise risk adjustment was lacking. Fifth, adjustment for multiple testing was not performed and findings were considered statistically significant if P < .10 so as not to miss program effects, increasing the likelihood of finding results due to chance.

In this exploratory analysis, the OCM was significantly associated with modest payment decreases during 6-month episodes for Medicare beneficiaries receiving chemotherapy for cancer during the first 3 years of the OCM that did not offset the monthly payments for enhanced oncology services. There were no statistically significant differences for most utilization, quality, and patient experience outcomes.

Corresponding Author: Nancy L. Keating, MD, MPH, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (keating@hcp.med.harvard.edu).

Accepted for Publication: September 17, 2021.

Author Contributions: Mss Jhatakia and Hassol had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Dr Keating and Ms Jhatakia are co–first authors. Dr Simon and Ms Hassol are co–senior authors.

Concept and design: Keating, Jhatakia, Brooks, Landrum, Kummet, Woodman, Simon, Hassol.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Keating, Jhatakia, Brooks, Cintina, Kummet, Simon, Hassol.

Critical revision of the manuscript for important intellectual content: Keating, Jhatakia, Brooks, Tripp, Landrum, Zheng, Christian, Glass, Hsu, Kummet, Woodman, Simon, Hassol.

Statistical analysis: Keating, Jhatakia, Tripp, Cintina, Landrum, Zheng, Christian, Glass, Kummet, Simon.

Obtained funding: Keating, Simon, Hassol.

Administrative, technical, or material support: Keating, Jhatakia, Hsu, Simon, Hassol.

Supervision: Keating, Jhatakia, Woodman, Simon, Hassol.

Conflict of Interest Disclosures: Dr Brooks reported receiving personal fees from Ipsen Biopharmaceuticals and UnitedHealthcare and payments to his institution for clinical trials from Taiho Pharmaceuticals, Hoffmann-La Roche, and Incyte Corporation. Ms Glass reported being employed by Abt Associates, the firm hired by the Centers for Medicare & Medicaid Services to evaluate the Oncology Care Model. Dr Simon reported being employed by UnitedHealth Group. Ms Hassol reported a contract with the Centers for Medicare & Medicaid Services outside the submitted work. No other disclosures were reported.

Funding/Support: The authors are members of an evaluation team contracted by the Centers for Medicare & Medicaid Services to evaluate the Oncology Care Model. The analyses on which this publication is based were performed under contract HHSM-500-2014-00026I sponsored by the Centers for Medicare & Medicaid Services, Department of Health and Human Services.

Role of the Funder/Sponsor: The sponsor contributed to the design and conduct of the study; collection, management, analysis, and interpretation of the data; review and approval of the manuscript; and decision to submit the manuscript for publication.

Group Information: The members of the Oncology Care Model Evaluation Team are listed in Supplement 2.

Meeting Presentation: This work was presented on June 17, 2021, at the AcademyHealth Annual Research Meeting (virtual meeting).