Transforming Clinical Practice Initiative Boosted Participation in Medicare Alternative Payment Models : The Journal of Ambulatory Care Management

Secondary Logo

Journal Logo

Original Articles

Transforming Clinical Practice Initiative Boosted Participation in Medicare Alternative Payment Models

Zurovac, Jelena PhD, MS; Barna, Michael MA; Merrill, Angela PhD; Zickafoose, Joseph S. MD, MS; Felt-Lisk, Sue MPA; Vollmer Forrow, Lauren MA; Gallegos, Karen Frederick BA; Everhart, Damian PhD, RN; Flemming, Robert PhD

Author Information
Journal of Ambulatory Care Management: July/September 2022 - Volume 45 - Issue 3 - p 150-160
doi: 10.1097/JAC.0000000000000419

Abstract

OVER the last decade, the Centers for Medicare & Medicaid Services (CMS) has made increasing efforts to implement or promote value-based payment programs for Medicare, Medicaid, and commercial payers, including Alternative Payment Models (APMs) with meaningful upside potential and downside risk (Muhlestein et al., 2017; U.S. Department of Health and Human Services [DHHS], 2015; VanLare et al., 2012). Tying payment to value and using incentives to improve care are key elements of the CMS Quality Strategy, designed to improve health services in the United States (CMS, 2019). CMS' efforts accelerated with the creation of the Quality Payment Program (QPP) to encourage clinicians who serve Medicare beneficiaries to switch from fee-for-service to value-based payment programs, including APMs (DHHS, n.d.; Medicare Access and CHIP Reauthorization Act of 2015). However, many clinicians and their practices struggle to transform their care in ways that meet the goals of value-based payment and enroll in Medicare APMs, especially smaller, rural, independent, and specialty practices (Aviki et al., 2019; Casalino, 2017; U.S. Government Accountability Office [GAO], 2016).

Moving practices into APMs may be a critical step toward improving patients' outcomes, reducing waste in the health care system, and sustaining better results through investment in value and optimal care delivery. With an increasing emphasis on value over volume, it is important to identify and study interventions that boost participation in APMs. CMS funded the Transforming Clinical Practice Initiative (TCPi) to help clinicians in outpatient practices of any specialty type improve health care quality (CMS, 2020). A key goal of TCPi was to advance the movement to value-based payment arrangements and encourage practices to participate in APMs (Abraham & McGann, 2019). We analyzed whether practices that joined TCPi were more likely than well-matched comparison practices to newly enroll in Medicare APMs.

DESCRIPTION OF TCPi

To date, TCPi was the largest national-scale model to support outpatient practice transformation through collaborative and peer-based learning without an associated payment mo-del. From September 29, 2015, to September 30, 2019, TCPi enrolled clinicians from 18 344 primary and specialty care practices represented by tax identification numbers (TINs) or CMS certification numbers (CCNs). All physician practices except those already enrolled in a Medicare APM were eligible to participate in TCPi. TCPi practices were diverse in terms of practice size, geographic area, patients served (covered by Medicare, Medicaid, or commercial payers), and practice type (primary care, specialty, mixed primary and specialty care, practice size, urban, or rural). Furthermore, TCPi encouraged enrollment of and emphasized supports for rural practices and practices in medically underserved communities.

TCPi's goals included improving health outcomes; reducing potentially avoidable hospitalizations, emergency department visits, health care spending, and testing and procedures; increasing practices' new participation in APMs; and building the evidence base for practice transformation. To accomplish these goals, TCPi awarded $685 million to 29 Practice Transformation Networks (PTNs), 10 Support and Alignment Networks (SANs), and 2 SAN 2.0s to engage practices and clinicians in practice transformation. Practice Transformation Networks and SAN 2.0s included large health care systems and quality improvement collaboratives, which worked with participating practices on patient-centered care design, continuous data-driven quality improvement, and sustainable business operations that were considered primary drivers for value-based payment (CMS, 2020). Support and Alignment Networks included national and regional professional associations and public-private partnerships, which developed or expanded tools that support transformation (such as practice registries and decision support tools) and spread successful efforts to their broader networks and professional sectors. TCPi developed a change package based on the experience of high-performing practices. TCPi published various transformation resources, including the change package, on CMS' Web site (CMS, 2020).

CMS, PTNs, and SANs expected that these transformation efforts would better prepare practices to join APMs (Abraham & McGann, 2019). Practice Transformation Networks and SANs offered many educational activities to support practices in joining an APM. TCPi allowed PTNs and SANs to determine their own content and approach to the assistance, considering the specific needs and environments of their diverse enrolled practices and professional sectors. TCPi put more emphasis on APM participation in 2018 and 2019, the last 2 years of the model.

METHODS

We analyzed the effect of TCPi on new enrollment into Medicare APMs using 6958 physician practices enrolled in TCPi through September 30, 2018, and a closely matched comparison group of 6958 practices. We estimated effects as the difference in enrollment between TCPi and comparison group at follow-up (after TCPi practices joined TCPi). This approach is analogous to a difference-in-differences analysis because the preintervention APM enrollment is identical (zero) for the 2 groups. To compare the 2 groups on the same time line, we started following each comparison practice at the same time we started following its matched TCPi practice, 90 days after its enrollment into TCPi. We followed all practices through January 2020, 3 months after TCPi ended.

Data sources

Our quantitative analysis included practices that joined TCPi from the model's commencement through September 30, 2018, a year before the model ended. To assess which practices subsequently started participating in Medicare APMs, we used the February 7, 2020, extract of CMS' Master Data Management file for enrollment into 6 Medicare APMs: Medicare Shared Savings Program (MSSP), Next Generation Accountable Care Organization (ACO) Model, Comprehensive Primary Care Plus, Comprehensive End-Stage Renal Disease (ESRD) Care Model, Maryland Total Cost of Care Model, and Vermont ACO model. We separately obtained enrollment data through January 2020 from CMS for 2 other models: Oncology Care Model (OCM) and Bundled Payments for Care Improvement Advanced (BPCI-A) Model. Using Medicare APM enrollment data through early 2020 provided a range of nearly a year and a half to more than 4 years of follow-up, depending on when practices joined TCPi. We used several other data sources for practice and beneficiary characteristics (see Supplemental Digital Content Tables 1.1 and 1.2, available at: https://links.lww.com/JACM/A116, and Supplemental Digital Content Figure 1.2, available at: https://links.lww.com/JACM/A116).

To understand PTNs' capacity to help practices join APMs, we analyzed several sources of survey and qualitative interview data. TCPi's Development, Management, and Improvement contractor surveyed PTNs in November 2018 and conducted semistructured follow-up interviews from January to March 2019 focused on activities and challenges with APM enrollment. To learn about practices' use of PTN assistance and their readiness for APMs, TCPi's Evaluation and Analysis contractor surveyed a stratified random sample of TCPi practices in 2018 (562 respondents) and another sample in 2019 (555 respondents). These responses represented 49% and 41% of the sampled practices, respectively. Responses were weighted for nonresponse. The Evaluation and Analysis contractor also interviewed 69 TCPi practices in summer 2017 and reviewed PTNs' September 2019 reports. This study was exempt from institutional review board review.

Study population

Starting with 18 022 practices (defined by TIN or CCN) enrolled in TCPi through September 2018, we first removed from the sample practices for which Medicare APM enrollment was not relevant. Because TCPi enrolled practices regardless of which payers cover their patients, Medicare APMs are not relevant for all TCPi practices, and we had data only on enrollment into Medicare APMs. Furthermore, because most of the Medicare APMs in our data are geared toward ambulatory care practices, retaining practices with at least 1 attributed beneficiary enabled us to focus on that group.

To that end, we attributed Medicare fee-for-service beneficiaries to practices that provided the largest share of each beneficiary's outpatient evaluation and management visits using an approach similar to the one used by the MSSP and other models. One modification was that we included a broader group of specialists and specialty care evaluation and management visits to encompass the diverse practices that participated in TCPi and reflect the inclusion of enrollment data for 2 specialist-focused APMs (OCM and BPCI-A). We excluded 6187 practices that either did not have any attributed Medicare fee-for-service beneficiaries because they provided little or no care to these beneficiaries or that were not in Medicare practice data. The second-largest exclusions were of 2249 practices in Puerto Rico or an undetermined location. Because a large proportion of practices in Puerto Rico participated in TCPi, very few good comparisons were available. We also excluded 1046 practices that were already participating in a Medicare APM before enrolling in TCPi. After applying all exclusion criteria, we arrived at a sample of 6958 practices (see Supplemental Digital Content 1, available at: https://links.lww.com/JACM/A116). Of those, 62% were composed primarily of specialists (Table).

Table. - Select Baseline Characteristics for TCPi and Comparison Practicesa,b
Characteristic Mean for TCPi Practices (n = 6958) Mean for Comparison Practices (n = 6958) Standardized Difference
Exact-match characteristicsc (% of practices)
Practice specialtyd,e
Primary care 25.3 25.3 0.00
Specialty care 62.0 62.0 0.00
Mixed primary/specialty care 12.7 12.7 0.00
Small practice: 3 or fewer clinicians 69.1 69.1 0.00
Located in a rural county (2013) 13.2 13.2 0.00
Fewer than 40 attributed beneficiaries per quarter during baseline 44.3 44.3 0.00
Practice characteristics (% of practices)
Average number of attributed beneficiaries per quarter during baseline 364.4 320.9 0.02
Practice sizef 14.5 13.3 0.01
Ownership typeg,h (2017)
Hospital-owned 27.5 23.8 0.09
Physician-owned 72.5 76.2 0.09
Proportion of clinicians in medical homes in year before TCPi enrollmenti
All 3.2 2.5 0.04
Some but not all 4.0 3.1 0.05
None 92.7 94.4 0.06
At least 1 clinician received meaningful use payment for EHRs before TCPi enrollment 59.5 53.4 0.12
At least 1 clinician participated in PQRS before TCPi enrollment 99.6 99.6 0.00
Accepts Medicaidg,h (2017) 68.9 64.8 0.10
In a whole county health professional shortage area (year before TCPi enrollment)
Primary care 3.8 3.3 0.03
Mental health 26.7 26.5 0.01
Average beneficiary outcomes
Medicare Part A and B expenditures ($/beneficiary) 2803 2746 0.05
Outpatient ED visits (#/1000 beneficiaries) 133.4 133.8 0.01
All-cause inpatient admissions (#/1000 beneficiaries) 76.1 75.5 0.02
Abbreviations: AAAHC, Accreditation Association for Ambulatory Health Care; CMS, Centers for Medicare & Medicaid Services; ED, emergency department; EHRs, electronic health records; MD-PPAS, Medicare Data on Provider Practice and Specialty; NCQA, National Committee for Quality Assurance; NPPES, National Plan and Provider Enumeration System; PCMH, patient-centered medical home; PQRS, Physician Quality Reporting System; PTN, Practice Transformation Network; SAN, Support and Alignment Network; TCPi, Transforming Clinical Practice Initiative; TIN, Tax Identification Number; TJC, The Joint Commission.
aEnrollment data from PTNs and SAN 2.0s, MD-PPAS, and the NPPES; meaningful use and PQRS participation data from CMS; Medicaid acceptance and ownership from SK&A; and medical home participation from NCQA PCMH Recognition Program data, TJC, AAAHC, and 10 state certification programs.
bCharacteristics were measured as of the year before a practice enrolled, unless otherwise noted.
cEach TCPi practice was matched to a comparison practice that shared each exact-match characteristic. For example, a primary care TCPi practice was always matched to another primary care practice.
dPrimary care practices had more than 90% of primary care clinicians; specialty care practices had fewer than 10%, and mixed primary/specialty practices had between 10% and 90% primary care clinicians.
eWe used the CMS definition of primary care, which includes general practice, family medicine, internal medicine, and geriatric medicine specialties (CMS, 2015).
fClinicians in each TIN were those who billed the plurality of their Medicare Part B charges to that TIN in the year before the practice's enrollment.
gThese characteristics were defined for each practice location. Because practice locations might share a TIN, we assigned to each TIN the proportion of practice locations within a TIN that had the characteristic, weighted by the number of clinicians at each location.
hData for these characteristics were available for more than three-quarters of primary care practices compared with less than two-thirds of mixed primary/specialty care practices and less than 10% of specialty care practices.
iMost practices recognized as medical homes gained this status through NCQA's PCMH Recognition Program.

Comparison group

To reduce the likelihood of selection bias and obtain rigorous effect estimates, we constructed a comparison group of practices similar to the TCPi group by using an algorithm called GroupMatch, an extension of optimal propensity score matching developed to address rolling enrollment of practices (Pimentel et al., 2020; Rosenbaum, 1989). We selected one comparison for each TCPi practice from a pool of all nonparticipating practices in the country using matching without replacement. We matched the 2 groups on practice characteristics, including specialty; size; ownership; baseline participation in programs that might affect the likelihood of joining TCPi and subsequent Medicare APM participation; and attributed beneficiaries' characteristics, such as patient demographics, prevalence of chronic conditions, claims-based health care expenditures, and health care use. We also matched on area characteristics such as poverty, education, and availability of health care resources (Table). Whenever possible, we drew comparisons from the same hospital referral region (HRR) or state: 72% of comparisons were from the same HRR as TCPi practices, 22% were not from the same HRR but were from the same state, and only 6% were out of HRR and out of state.

The matched comparison group was closely similar to TCPi practices on all measured characteristics; nearly all were within 0.10 standard deviations (Table; see Supplemental Digital Content Tables 1.1 and 1.2, available at: https://links.lww.com/JACM/A116, and Supplemental Digital Content Figure 1.2, available at: https://links.lww.com/JACM/A116). Therefore, balance on all measured characteristics was much better than the target of 0.25 standard deviations for key variables, which is considered adequate to proceed with an impact analysis if using regression adjustment to account for differences that persist after matching (Stuart, 2010).

Outcomes

Our primary outcome was the percentage of practices (TINs) that newly joined a Medicare APM between 90 days after joining TCPi and January 2020. The Medicare Shared Savings Program, the largest Medicare APM in terms of enrollment, enrolls entire TINs. For the remainder of Medicare APMs in our sample, which enroll individual clinicians or groups of clinicians, we regarded a practice as having participated if at least one of its clinicians was enrolled. We also evaluated participation in Medicare Advanced APMs, which typically require participants to bear a significant financial risk (CMS, n.d.).

Statistical analysis

We estimated effects as the difference in Medicare APM participation between the TCPi and comparison groups at follow-up (after TCPi practices joined TCPi). This approach is analogous to a difference-in-differences analysis because the preintervention APM enrollment is identical for the 2 groups (zero). By construction, because we are assessing new Medicare APM enrollment, TCPi or comparison practices did not participate in Medicare APMs at baseline. To compare the 2 groups on the same time line, we started following each comparison practice at the same time we started following its matched TCPi practice, 90 days after its enrollment into TCPi. In regression analyses, we controlled for the characteristics used in matching (see Supplemental Digital Content 2, available at: https://links.lww.com/JACM/A116).

We estimated the effect of TCPi on several subgroups: primary care, specialty care, mixed primary and specialty care, rural, and small practices (ie, those with 3 or fewer clinicians). We defined primary care practices as those with more than 90% of clinicians in family practice, internal medicine, general practice, or geriatric medicine (CMS, 2015). Specialty care practices had less than 10% of primary care clinicians. We defined other practices as mixed primary and specialty care practices.

RESULTS

Effect of TCPi on Medicare APM participation

For all TCPi practices in our sample and for each analyzed subgroup, we found statistically significantly larger new Medicare APM enrollment among TCPi versus comparison practices (Figure 1). Before regression adjustment, we found that twice as many TCPi practices newly joined Medicare APMs than comparison practices: 21.1% of TCPi practices and 10.8% of comparisons were participating in a Medicare APM, a difference of 10.3 percentage points. After regression adjustment, the difference was 8.9 percentage points (P < .01). Estimated effects on Medicare Advanced APM participation were also favorable: 3.8 percentage points overall (P < .01). To compare the size of estimated impacts for the 2 outcomes, we divided the estimated impacts by the standard deviation for each outcome and found that the estimated impact on Medicare APM participation was 0.24 standard deviations and 0.14 standard deviations for Advanced Medicare APM participation.

F1
Figure 1.:
Participation rates and estimated effects of TCPi on Medicare APM participation through January 2020, for all practices and by subgroup. We assessed whether practices started participating in Medicare APMs at least 90 days after joining TCPi. The number of TCPi practices in each subgroup is shown within parentheses. We present the effects as differences relative to the counterfactual, the APM participation for TCPi practices in the absence of the intervention. The counterfactual is the unadjusted value of APM participation for TCPi practices minus the estimated effect. CI indicates confidence interval; TCPi, Transforming Clinical Practice Initiative. From TCPi model enrollment data; Master Data Management participation data from February 2020; Oncology Care Model and Bundled Payment for Care Improvement Advanced Model APM participation data from January 2020; and administrative claims data from October 1, 2011, through December 31, 2018.

The estimated effects were favorable and statistically significant for every analyzed subgroup. For overall Medicare APM participation, they ranged from 3.8 percentage points for specialty care practices to 20.2 percentage points for primary care practices (all P values < .01). For Medicare Advanced APM enrollment, estimated effects were favorable and statistically significant for most subgroups, ranging from 0.5 percentage points for specialty care practices (P = .37) to 11.6 percentage points for primary care practices (P < .01) (Figure 1).

Of the 1468 TCPi practices that started participating in a Medicare APM, 1023 (69.7%) started participating in the MSSP, and 351 (23.9%) started in the Next Generation ACO model (Figure 2). Of the TCPi practices that joined the MSSP, 671 (65.6%) were in tra-ck 1, which had no downside risk (Figure 2). Nearly all practices that joined Medicare APMs joined in 2017 or later (Figure 3). On average, TCPi practices started participating in a Medicare APM a year and a half after enrolling in TCPi.

F2
Figure 2.:
Practice-level Medicare APM participation through January 2020, by APM (counts and percentages participating). The figure shows the counts of practices that started participating in each APM at least 90 days after enrolling in TCPi. Because practices could participate in multiple APMs, the totals across APMs may not sum to the total number of practices that participated in an APM. No practices in our sample participated in the Vermont All-Payer ACO model. On July 1, 2019, CMS ended enrollment in Tracks 1, 1+, 2, and 3, which were predominant for practices in our study period, and began enrollment in the BASIC and ENHANCED tracks with different levels of risk.a ACO indicates accountable care organization; APM, Alternative Payment Model; ESRD, end-stage renal disease; TCPi, Transforming Clinical Practice Initiative. From TCPi model enrollment data; Master Data Management participation data from February 2020; Oncology Care Model and Bundled Payment for Care Improvement Advanced Model APM participation data from January 2020; and administrative claims data from October 1, 2011, through December 31, 2018.
F3
Figure 3.:
Practice-level Medicare APM participation through January 2020, cumulative by year. We assessed whether practices started participating in Medicare APMs at least 90 days after joining TCPi. The parentheses show the percentage of TCPi practices by the end of the given year that also started participating in an APM by the end of that year. On July 1, 2019, CMS ended enrollment in Tracks 1, 1+, 2, and 3, which were predominant for practices in our study period, and began enrollment in the BASIC and ENHANCED tracks with different levels of risk.a APM indicates Alternative Payment Model; BPCI-A, Bundled Payment for Care Improvement Advanced Model; CEC, Comprehensive End-Stage Renal Disease Care Model; CPC+, Comprehensive Primary Care Plus; MD TCOC, Maryland Total Cost of Care; MSSP (1 and 2), Medicare Shared Savings Program Tracks 1 and 2; MSSP (1+ and 3), Medicare Shared Savings Program Tracks 1+ and 3; MSSP (B/E Tracks), MSSP BASIC and ENHANCED tracks; Next Gen, Next Generation Accountable Care Organization; OCM, Oncology Care Model; TCPi, Transforming Clinical Practice Initiative; VT ACO, Vermont All-Payer Accountable Care Organization Model. From TCPi model enrollment data; Master Data Management participation data from February 2020; Oncology Care Model and Bundled Payment for Care Improvement Advanced Model APM participation data from January 2020; and administrative claims data from October 1, 2011, through December 31, 2018.

TCPi support for APM participation

Survey and interview responses support the findings that TCPi boosted APM participation. Information from interviews, surveys, and PTN September 2019 reports indicate that TCPi helped practices prepare for value-based arrangements. The PTN survey about readiness, confidence, and need for transitioning practices into APMs and the follow-up interviews revealed consistent themes:

Sharing learning opportunities: Most PTNs shared information about local and national learning opportunities related to APMs with affiliated practices.

Educational sessions: Several PTNs developed educational sessions and materials to prepare practices for APMs. These sessions addressed understanding of APM structures and necessary capabilities, developing a road map that included payer engagement, and training practices to implement strategies that would help them demonstrate value in APM arrangements.

Engaging with local ACOs and independent practice associations: Several PTNs conducted extensive network inventories to understand the arrangements in which their providers were engaged. Practice Transformation Networks found that many clinicians had joined diverse, commercial value-based arrangements (such as employer plans, independent practice associations, or ACOs), which provided experience and confidence for participating in Medicare arrangements.

The 2018 practice survey revealed that practices that received more PTN assistance were more likely to report readiness for value-based arrangements. The 2019 practice survey showed that 38% of respondent practices took different or quicker action toward transformation because of TCPi assistance related to the QPP. Practices reported that PTNs helped them implement a change package that provided strategies for a set of 15 change concepts, including engaged and committed leadership, population health management, patient and family engagement, transparent measurement and monitoring, and other components related to readiness for value-based care. In addition, 4 of the 10 SANs reported offering educational activities to support health professionals' readiness for and movement toward APMs.

DISCUSSION

TCPi practices had a higher Medicare APM participation relative to comparison practices overall and across all subgroups, which suggests that TCPi provided effective and ongoing support for APM participation across a wide variety of TCPi practices.

Findings from the PTN survey and interviews and the practice survey support these favorable findings. Practice Transformation Networks offered ample support to TCPi practices for joining APMs. Because of the timing of the 2 initiatives, the incentives possible through the QPP and the TCPi assistance to advance readiness for APMs might have had a synergistic effect.

TCPi practices had a higher Medicare APM participation among specialty care practices (by 3.8 percentage points) and small practices (by 8.5 percentage points) relative to comparison practices. However, the levels of participation among these practices were much lower than for other practice types in our sample, consistent with the concern in the literature that such practices face major challenges to joining an APM (Aviki et al., 2019; Casalino, 2017; GAO, 2016). Our findings are also consistent with the fact that specialist-focused Medicare APMs are still being developed. Only 3 APMs in our data were focused on specialists (the Comprehensive ESRD Care Model, OCM, and BPCI-A Model). Furthermore, the existing specialty-focused Medicare APMs address only certain types of non–hospital-based specialty care. Practice Transformation Networks highlighted in interviews that small practices often did not meet APM attribution thresholds, which is a key reason for their lower APM participation.

Rural practices face many challenges with joining APMs (Aviki et al., 2019; Casalino, 2017; GAO, 2016). However, participation in TCPi greatly boosted their Medicare APM enrollment relative to comparison practices (by 10.0 percentage points), and rural practices' levels of Medicare APM enrollment became comparable to levels of enrollment across all TCPi practices in our sample. This is likely due to TCPi's focus on enrolling rural practices into TCPi and supporting their transformation.

Among practices that started participating in a Medicare APM, nearly all joined in 2017 and later. This finding is consistent with TCPi's heavy emphasis on increasing APM participation during 2018 and 2019 and the initial years of the QPP.

Limitations

As with all nonexperimental designs, the main limitation of our study is that some practices might have joined APMs regardless of TCPi, perhaps because they were highly motivated. Although it is impossible to know to what extent greater motivation among TCPi practices versus comparison practices contributed to favorable findings, there are several mitigating factors. One is that the TCPi and comparison practices were closely similar in terms of prior participation in several programs that reflect motivation and affinity for practice transformation, including medical home initiatives, the CMS Physician Quality Reporting System, and meaningful use of electronic health records. In addition, we used a local comparison group whenever possible (72% of comparison practices were in the same HRR as TCPi practices), incorporated the preintervention outcome by limiting the sample to practices that did not participate in APMs at baseline, and matched on and controlled for dozens of diverse characteristics that might influence practices' decision to join TCPi or participate in APMs.

Although it is possible that selection bias had partially inflated the impact estimates, research shows that the risk that the true impact of TCPi is zero or unfavorable is very small, given the use of a largely local comparison group, preintervention outcome, and a rich set of covariates and given large estimated impacts (Cook et al., 2008; Zurovac et al., 2021). Furthermore, qualitative evidence shows that practices that received more TCPi assistance were more likely to report readiness for APMs, suggesting that TCPi helped increase APM participation.

Another limitation is that SANs also served practices outside of TCPi, which means that their efforts might have helped some comparison practices transition to APMs. If so, our results underestimate the true effect of TCPi. But the risk of such bias is small because only 4 out of 10 SANs had activities that supported APM transitions. There were no data about which practices were supported by SANs. Finally, we could not evaluate the impact on TCPi on Medicaid APM enrollment because of lack of data.

Future work

In the future, we plan to expand our analysis. In addition to increasing Medicare APM participation, TCPi was designed to affect patients' health and health care use outcomes. Expanding upon research about early effects of TCPi (Dai et al., 2022), our research team is working to evaluate the effect of TCPi on these outcomes for people enrolled in Medicare and Medicaid for the full duration of the initiative and beyond.

Furthermore, additional research is needed to understand whether practices' participation in APMs leads to better outcomes or lower cost, and how to structure incentives to achieve these goals (de Brantes et al., 2021; Hassol et al., 2018; McWilliams et al., 2017; McWilliams, Hatfield, et al., 2019; McWilliams, Landon, et al., 2019).

CONCLUSIONS

TCPi practices had a higher Medicare APM enrollment relative to comparison practices overall and across all subgroups, which suggests that a combination of network-based and locally implemented support can be an effective means to increase APM participation across a wide variety of practices.

REFERENCES

Abraham M. R., McGann P. (2019). Contribution of the Transforming Clinical Practice Initiative in advancing the movement to value-based care. The Annals of Family Medicine, 17(Suppl. 1), S6–S8. doi:10.1370/afm.2425
Aviki E. M., Schleicher S. M., Mullangi S. (2019). The Oncology Care Model and other value-based payment models in cancer care. JAMA Oncology, 5(3), 298–299. doi:10.1001/jamaoncol.2018.5735
Casalino L. P. (2017). The Medicare Access and CHIP Reauthorization Act and the corporate transformation of American medicine. Health Affairs (Millwood), 36(5), 865–869. doi:10.1377/hlthaff.2016.1536
Centers for Medicare & Medicaid Services. (2015). Two-step attribution for measures included in the value modifier. Retrieved December 15, 2021, from https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/PhysicianFeedbackProgram/Downloads/Attribution-Fact-Sheet.pdf
Centers for Medicare & Medicaid Services. (2019). CMS' quality strategy. Retrieved December 15, 2021, from https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/Value-Based-Programs/CMS-Quality-Strategy
Centers for Medicare & Medicaid Services. (2020). Transforming clinical practice initiative. Retrieved December 15, 2021, from https://innovation.cms.gov/initiatives/Transforming-Clinical-Practices/
Centers for Medicare & Medicaid Services. (n.d.). Advanced Alternative Payment Models (APMs). Retrieved December 15, 2021, from https://qpp.cms.gov/apms/advanced-apms
Cook T. D., Shadish W. R., Wong V. C. (2008). Three conditions under which experiments and observational studies produce comparable causal estimates: New findings from within-study comparisons. Journal of Policy Analysis and Management, 27(4), 724–750.
Dai M., Chung Y., Peterson L. E., Petterson S., Phillips R. L Jr. (2022). Family practices in Transforming Clinical Practice Initiative showed no changes in Medicare costs or utilization. Medical Care, 60(1), 50–55.
de Brantes F., Rutledge V., Bozic K. (2021). Medicare APMs at a crossroads. American Journal of Managed Care, 27(9), e290–e292. doi:10.37765/ajmc.2021.88624
Hassol A., Gu Q., Christian T. J., Newes-Adeyi G., McClellan S., Norton G., Hsu V. D. (2018). Evaluation of the Oncology Care Model: Performance period one. Abt Associates. Retrieved December 15, 2021, from https://innovation.cms.gov/Files/reports/ocm-secondannualeval-pp1.pdf
McWilliams J. M., Chernew M. E., Landon B. E. (2017). Medicare ACO program savings not tied to preventable hospitalizations or concentrated among high-risk patients. Health Affairs (Millwood), 36(12), 2085–2093. doi:10.1377/hlthaff.2017.0814
McWilliams J. M., Hatfield L. A., Landon B. E., Chernew M. E. (2019). Spending reductions in the Medicare Shared Savings Program: Selection or savings? [Working paper 26403]. Cambridge, MA: National Bureau of Economic Research. Retrieved December 15, 2021, from https://www.nber.org/papers/w26403
McWilliams J. M., Landon B. E., Rathi V. K., Chernew M. E. (2019). Getting more savings from ACOs—can the pace be pushed? New England Journal of Medicine, 380(23), 2190–2192. doi:10.1056/NEJMp1900537
Medicare Access and CHIP Reauthorization Act of 2015, HR 2, 114th Cong. (2015). Retrieved December 15, 2021, from https://www.congress.gov/bill/114th-congress/house-bill/2/text?overview=closed
Muhlestein D., Burton N., Winfield L. (2017, February 3). The changing payment landscape of future CMS payment models foreshadows future plans. Health Affairs Blog. Retrieved December 15, 2021, from https://www.healthaffairs.org/do/10.1377/hblog20170203.058589/full/
Pimentel S. D., Vollmer Forrow L., Gellar J., Li J. (2020). Optimal matching approaches in health policy evaluations under rolling enrollment. Journal of the Royal Statistical Society, Series A, 183(4), 1411–1435.
Rosenbaum. P. R. (1989). Optimal matching for observational studies. Journal of the American Statistical Association, 84(408), 1024–1032.
Stuart E. A. (2010). Matching methods for causal inference: A review and a look forward. Statistical Science, 25(1), 1–21. doi:10.1214/09-STS313
U.S. Department of Health & Human Services. (2015). Better, smarter, healthier: Delivery system reform. Retrieved December 15, 2021, from https://www.healthit.gov/sites/default/files/facas/Joint_DSR_slides_2015-10-06.pdf
U.S. Department of Health & Human Services. (n.d.). Quality payment program overview. Retrieved December 15, 2021, from https://qpp.cms.gov/about/qpp-overview
U.S. Government Accountability Office. (2016). Medicare Value-Based Payment Models: Participation challenges and available assistance for small and rural practices. Retrieved December 15, 2021, from https://www.gao.gov/assets/690/681541.pdf
VanLare J. M., Blum J. D., Conway P. H. (2012). Linking performance with payment: Implementing the Physician Value-Based Payment Modifier. JAMA, 308(20), 2089–2090.
Zurovac J., Cook T. D., Deke J., Finucane M., Chaplin D., Coopersmith J., Forrow Vollmer L. (2021). Absolute and relative bias in eight common observational study designs: Evidence from a meta-analysis [Working paper]. Retrieved December 15, 2021, from https://arxiv.org/abs/2111.06941v2
Keywords:

accountable care organizations; ambulatory care; health care reform; health policy; Medicare; reimbursement mechanisms; value-based purchasing

Supplemental Digital Content