Accountable Care Organizations and Spending for Patients Undergoing Long-Term Dialysis

Visual Abstract

Abstract

Background and objectives

Despite representing 1% of the population, beneficiaries on long-term dialysis account for over 7% of Medicare’s fee-for-service spending. Because of their focus on care coordination, Accountable Care Organizations may be an effective model to reduce spending inefficiencies for this population. We analyzed Medicare data to examine time trends in long-term dialysis beneficiary alignment to Accountable Care Organizations and differences in spending for those who were Accountable Care Organization aligned versus nonaligned.

Design, setting, participants, & measurements

In this retrospective cohort study, beneficiaries on long-term dialysis between 2009 and 2016 were identified using a 20% random sample of Medicare beneficiaries. Trends in alignment to an Accountable Care Organization were compared with alignment of the general Medicare population from 2012 to 2016. Using an interrupted time series approach, we examined the association between Accountable Care Organization alignment and the primary outcome of total spending for long-term dialysis beneficiaries from prior to Accountable Care Organization implementation (2009–2011) through implementation of the Comprehensive ESRD Care model in October 2015. We fit linear regression models with generalized estimating equations to adjust for patient characteristics.

Results

During the study period, 135,152 beneficiaries on long-term dialysis were identified. The percentage of long-term dialysis beneficiaries aligned to an Accountable Care Organization increased from 6% to 23% from 2012 to 2016. In the time series analysis, spending on Accountable Care Organization–aligned beneficiaries was $143 (95% confidence interval, $5 to $282) less per beneficiary-quarter than spending for nonaligned beneficiaries. In analyses stratified by whether beneficiaries received care from a primary care physician, savings by Accountable Care Organization–aligned beneficiaries were limited to those with care by a primary care physician ($235; 95% confidence interval, $73 to $397).

Conclusions

There was a substantial increase in the percentage of long-term dialysis beneficiaries aligned to an Accountable Care Organization from 2012 to 2016. Moreover, in adjusted models, Accountable Care Organization alignment was associated with modest cost savings among long-term dialysis beneficiaries with care by a primary care physician.

Introduction

Representing only 1% of the population, beneficiaries with kidney failure on long-term dialysis nonetheless account for >7% of Medicare’s traditional fee-for-service spending (1). Notably, only a minority of these costs are directly attributable to provision of dialysis and dialysis-related drugs, such as erythropoiesis-stimulating agents, especially following the introduction of bundled dialysis payments with the ESRD Prospective Payment System in 2011. Instead, most of the expense of this population stems from the management of a constellation of comorbidities (e.g., diabetes mellitus and heart failure) and associated complications afflicting the typical patient on long-term dialysis (2,3). Although the costs of dialysis and dialysis-related drugs are relatively fixed, those related to the management of comorbidities are more variable and potentially amenable to efforts at cost containment (3). In attempts to reduce costs of care for medically complex beneficiaries, the Centers for Medicare and Medicaid Services (CMS) enacted the Accountable Care Organization (ACO) model of care in 2012.

Medicare Shared Savings Program (MSSP) ACOs aim to accomplish improvements in quality of care and reduced spending though the provision of highly coordinated care, generally directed by primary care physicians (PCPs). It is unclear, however, whether ACOs would be an appropriate model of care for long-term dialysis beneficiaries. On the one hand, ACOs may be an effective model given their focus on coordinated care, which would be expected to be applicable to long-term dialysis beneficiaries who often have multiple comorbidities for which they are frequently hospitalized (4). Previous research has shown that participation in early ACOs has resulted in cost savings for clinically vulnerable beneficiaries in the general Medicare population (5). On the other hand, PCPs may lack specialized knowledge and experience in caring for the medical issues of long-term dialysis beneficiaries or may duplicate care provided by nephrologists who already necessarily follow such beneficiaries regularly. This may result in care inefficiencies, potentially leading to increased spending (6). For example, PCPs may not be aware that for many long-term dialysis beneficiaries, cancer screening is not indicated due to limited life expectancy (6). In addition, ACO-aligned PCPs may choose to avoid caring for costly long-term dialysis beneficiaries (7).

In this context, we analyzed Medicare data to examine the relationship between the MSSP ACO model and spending for long-term dialysis beneficiaries. We examined trends in the proportion of long-term dialysis beneficiaries receiving care in an MSSP ACO, as well as investigated differences in spending for long-term dialysis beneficiaries in the MSSP ACO model versus those in the traditional Medicare fee-for-service model.

Materials and Methods

Data Sources

This retrospective cohort study was on the basis of national Medicare claims from a 20% random sample of beneficiaries, including data from the Carrier, Denominator, Medicare Provider Analysis and Review, and Outpatient research identifiable files. The Shared Savings Program Beneficiary–level research identifiable file was used to attribute beneficiaries to ACOs during each study quarter after the start of an MSSP contract. The Shared Savings Program Provider–level research identifiable file was used to determine which providers were aligned to an ACO and to characterize each MSSP ACO (i.e., ACO start date). Finally, the organizational structure (i.e., ACO leadership) of the MSSP ACO was determined through the Leavitt Partners ACO Database. This validated database contained 839 Medicare, Medicaid, and commercial ACOs at the time of the analysis (8).

Study Population

All Medicare fee-for-service beneficiaries 18 years of age or older who were diagnosed with kidney failure and were on long-term dialysis between January 1, 2009 and December 31, 2016 were identified. Beneficiaries were identified as being on long-term dialysis if they had at least one dialysis claim during the calendar year of interest in the Outpatient and Carrier files (Healthcare Common Procedure Coding System codes 90951–90970). It was also confirmed that these beneficiaries had a kidney failure diagnosis using International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification diagnosis codes (International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes: 585.6, v45.1, v56.1, v56.2, v56.3, and v56.31; International Classification of Diseases, Tenth Revision, Clinical Modification diagnosis codes: N18.6).

We determined whether a beneficiary was aligned to an ACO on the basis of the assignment in the Shared Savings Program beneficiary-level research identifiable file. Beneficiaries were assigned to an ACO according to CMS rules on the basis of where they received the largest share of their primary care services (9). For comparability between ACO-aligned and nonaligned populations, all beneficiaries were required to have at least one primary care service (Healthcare Common Procedure Coding System codes 99201–99215, 99304–99350, G0402, G0438, and G0439). Furthermore, we required that, for beneficiaries nonaligned during the calendar year of interest, the primary care service was furnished by a provider who qualifies to be an ACO-aligned professional as defined by CMS (i.e., by provider codes) (9). These ACO-aligned or qualifying providers were not necessarily PCPs. Finally, a requirement for continuous enrollment in Medicare Parts A and B during the calendar year of interest and 1 year prior was also imposed (for purposes of comorbidity adjustment).

Study Outcome

The primary outcome of this study was total quarterly spending per beneficiary. This was on the basis of the sum of all payments for Parts A and B claims filed on behalf of the beneficiary. All payments were price standardized to account for geographic payment differences and add-on payments for indirect medical education and disproportionate share hospitals using methods described by the Medicare Payment Advisory Commission (10). In addition, all payments were inflation adjusted to 2015 US dollars, the last year of data for the spending analyses, using the Consumer Price Index.

Statistical Analyses

Main Analyses.

First, the trend in the percentage of long-term dialysis beneficiaries participating in an MSSP ACO from 2012 to 2016 was plotted, with a comparison with the trend among the overall Medicare population (Figure 1). Then, all subsequent analyses were truncated to September 30, 2015 because implementation of the Comprehensive ESRD Care (CEC) Model may have affected interpretation of findings occurring after its October 1 start date (1).

Figure 1.

Time trends in percentage of long-term dialysis population versus overall Medicare population alignment to Accountable Care Organization (ACO), 2009–2016.

Differences in characteristics among long-term dialysis beneficiaries, by time period and ACO alignment status, were examined. Characteristics included beneficiary age, sex, race/ethnicity, dual eligibility, level of comorbidity (on the basis of the hierarchical condition category method [11]), and receipt of at least one primary care service from a PCP, defined as physicians with a specialty of general practice, family practice, internal medicine, or geriatric medicine (provider specialty codes: 01, 08, 11, and 38). Chi-squared tests were used to compare categorical variables, and one-way ANOVA tests were used to compare continuous variables. If significant overall differences were found across the groups, the Bonferroni method was used to make post hoc pairwise comparisons between ACO-aligned versus nonaligned long-term dialysis beneficiaries in the post-ACO contract period.

Crude mean total quarterly spending for long-term dialysis beneficiaries was plotted in the pre-ACO contract period (January 2009 through December 2011) and for beneficiaries aligned to MSSP ACOs versus those nonaligned during the post-ACO contract period (January 2012 to October 2015) (Figure 2). Then, to further assess the relationship between MSSP ACO participation and spending for long-term dialysis beneficiaries, an interrupted time series approach at the beneficiary-quarter level was used. Linear regression models with generalized estimating equations were estimated to account for repeated measures at the beneficiary level. The main exposure was a binary, time-varying indicator for beneficiary alignment to an MSSP ACO (set to one during the quarter when the beneficiary was aligned to an ACO and zero otherwise). The models were adjusted for beneficiary age, sex, race/ethnicity, dual eligibility, indicators for the 79 hierarchical condition categories, and receipt of at least one primary care service from a PCP, as well as year and quarter fixed effects. The rationale for adjusting for the presence of service from a PCP for the main model was as follows. Given Medicare rules governing attribution of a patient to an ACO on the basis of care provided by an ACO-participating physician prioritize PCPs, we expected to find a higher percentage of patients with PCPs among ACO-aligned beneficiaries. As patients followed by PCPs may be systematically different than those without PCPs, adjusting for the presence of a PCP allows assessment of the independent effect of the ACO, which is predominantly mediated through its promotion of population health management, such as through care navigators. Furthermore, as ACO strategies may be primarily directed at or work through PCPs, we also examined for the presence of an interaction between the ACO effect and the presence of a PCP, finding this term to be statistically significant. As such, we also present stratified models among patients with and without a PCP to examine the ACO effect within each group.

Figure 2.

Overall decrease in mean total quarterly spending among long-term dialysis beneficiaries. Mean total quarterly spending for long-term dialysis beneficiaries before and after ACO program initiation, by ACO alignment. The black line denotes spending among non–ACO-aligned beneficiaries, and the gray line denotes spending among ACO-aligned beneficiaries. The vertical dashed line indicates the initiation of the ACO program by CMS.

In additional analyses designed to further examine sources of spending, separate models were performed with outcomes of dialysis-related spending (on the basis of claims for outpatient dialysis procedures and dialysis-related vascular access procedures), nondialysis-related spending (all Medicare claims not included under dialysis spending claims), outpatient (all claims in outpatient files and carrier file claims in the outpatient context) spending, and inpatient (all claims in inpatient files and carrier file claims in the inpatient setting) spending.

Sensitivity Analyses.

To determine whether the association between alignment to an ACO and quarterly beneficiary spending differed between early and late adopters of the ACO model, separate models were created for beneficiaries who were aligned to ACOs with contract start dates in 2012, 2013, 2014, and 2015. To investigate if the effects of savings differed between hospital- and physician-led ACOs and hospital-physician partnerships, a model was fit with an interaction term between the ACO indicator and the ACO leadership structure. To examine whether nephrologist participation in the ACO modified the effects on spending, a model was fit with an interaction term between the ACO indicator and an indicator variable representing whether the ACO had a participating nephrologist. Finally, because spending may change with increased ACO experience, a model was fit with an interaction term between the ACO indicator and the number of years the ACO was in the program during the year the beneficiary was aligned to the ACO.

All analyses were performed using SAS Version 9.4 (Cary, NC). Tests were two tailed and with the probability of type 1 error at 0.05. Our institution’s Health Sciences Institutional Review Board deemed this study to be exempt from its oversight because we used a limited dataset.

Results

Time Trends of Long-Term Dialysis Population versus Overall Medicare Population Alignment to Accountable Care Organizations, 2009–2016

A total of 135,152 unique beneficiaries on long-term dialysis were identified during the extended time period from 2009 to 2016. Examining trends in ACO alignment, the long-term dialysis and overall Medicare populations tracked closely to one another in terms of percentages of their beneficiaries participating in an ACO from 2012 to 2014, both rising from just below 6% to around 18% (Figure 1). However, in 2015, coinciding with the launch of the CEC model, the percentages started to diverge, with a larger percentage of general Medicare beneficiaries attributed to an ACO than the percentage of long-term dialysis beneficiaries. By 2016, 27% of general Medicare beneficiaries were attributed to an ACO, whereas 23% of long-term dialysis beneficiaries were attributed to an ACO.

Beneficiary Characteristics of Patients on Long-Term Dialysis by Accountable Care Organization Alignment in Pre- and Postcontract Periods, 2009 to October 2015

After subsequently restricting the sample to 2009 to October 2015 (due to the start of the CEC model in October 2015), a total of 118,255 unique beneficiaries on long-term dialysis were identified. During the ACO postcontract period from 2012 to October 2015, there were 26,694 beneficiary-years of alignment to an ACO and 167,817 beneficiary-years not aligned (Table 1). In the precontract period of 2009 through 2011, there were 133,234 beneficiary-years. Although statistically significant, overall differences in the distribution of characteristics across age, sex, race, dual eligibility, and level of comorbid illness were very modest. The percentage of beneficiaries receiving primary care services by a PCP, however, was notably greater (82% versus 73%) for those with alignment to an ACO compared with those not aligned to an ACO. Pairwise differences for beneficiaries assigned versus not assigned to an ACO in the postcontract period were also statistically significant, except for sex.

Table 1.

Beneficiary characteristics for long-term dialysis beneficiaries by Accountable Care Organization alignment in pre- and postcontract periods

Beneficiary Characteristics	Precontract Beneficiaries 2009–2011, n=133,234	Postcontract Non–Accountable Care Organization Beneficiaries 2012 to October 2015, n=167,817	Postcontract Accountable Care Organization Beneficiaries 2012 to October 2015, n=26,694
Age, mean (SD), yr	63 (15)	63 (14)	64 (14)
Women, n (%)	61,595 (46)	76,319 (45)	12,160 (46)
Race, n (%)
White	58,435 (44)	68,716 (41)	11,551 (43)
Black	48,047 (36)	62,422 (37)	9614 (36)
Other	26,752 (20)	36,679 (22)	5529 (21)
Dual eligible, n (%)	59,801 (45)	78,704 (47)	11,492 (43)
No. HCCs, mean (SD)	5.8 (2.9)	5.8 (2.9)	5.9 (2.8)
Receipt of any primary care services by a primary care physician, n (%)	96,980 (73)	122,372 (73)	21,889 (82)

HCC, hierarchical condition categories.

Relationship between Accountable Care Organization Alignment and Mean Quarterly Spending for Long-Term Dialysis Beneficiaries, 2009 to October 2015

From 2009 to October 2015, there was an overall decrease in total inflation-adjusted spending among long-term dialysis beneficiaries (Figure 2). During the post-ACO contract period from 2012 to October 2015, crude spending for beneficiaries assigned to an ACO versus those not assigned to an ACO appeared similar. However, in time series analysis after adjustment for receipt of primary care services from a PCP, year, quarter, sex, race, age, dual eligibility, and comorbidities (Supplemental Table 1), beneficiary alignment to an ACO was associated with savings of $143 (95% confidence interval [95% CI], 5 to 282) per beneficiary-quarter (Table 2). In additional models examining subsets of spending, ACO alignment was associated with savings of $170 (95% CI, 34 to 306) per beneficiary-quarter for nondialysis spending but no significant savings for dialysis spending. Similarly, ACO alignment was associated with inpatient spending savings of $81 (95% CI, 26 to 136) per beneficiary-quarter but no significant savings for outpatient spending.

Table 2.

Multivariable model examining the relationship between alignment to the Accountable Care Organization and mean quarterly spending for long-term dialysis beneficiaries

Parametera	Estimate, $	95% Confidence Interval, $
ACO alignment
Nonaligned	Referent
ACO aligned	−143	−282 to −5
Receives primary care services from PCP
No primary care service from PCP	Referent
Has primary care service from PCP	3044	2964 to 3125
Year
2009	Referent
2010	−305	−432 to −177
2011	−1824	−1954 to −1693
2012	−1351	−1480 to −1222
2013	−1770	−1899 to −1641
2014	−2212	−2345 to −2079
2015	−1516	−1659 to −1372
Quarter
1	Referent
2	1274	1209 to 1338
3	1908	1837 to 1978
4	2104	2031 to 2178
Sex
Women	Referent
Men	−214	−307 to −120
Race
White	Referent
Black	456	344 to 569
Other/unknown	−1219	−1346 to −1091
Age, yr
18–65	Referent
66–69	285	150 to 420
70–74	547	409 to 684
75–79	536	389 to 684
80–84	402	243 to 561
85+	257	69 to 445
Dual eligibility
Not dual eligible	Referent
Dual eligible	1864	1765 to 1964

ACO, Accountable Care Organization; PCP, primary care physician.

^aEstimates for individual hierarchical condition categories not shown.

Given the presence of a significant interaction between PCP involvement and ACO alignment on spending (P<0.001), we conducted stratified analyses to explore this further. In an analysis restricted to patients who received a primary care service by a PCP, ACO alignment was associated with a significant savings of $235 (95% CI, 73 to 397) per beneficiary-quarter. In contrast, in an analysis restricted to patients who do not receive a primary care service by a PCP, ACO alignment was associated with additional spending of $295 (95% CI, 45 to 545) per beneficiary-quarter.

Predicted Total Spending per Beneficiary-Quarter on the Basis of Accountable Care Organization Contract Start Date, Organizational Type, Nephrologist Participation, and Year in Program

Sensitivity analyses showed no significant differences in spending for long-term dialysis beneficiaries across the factors of ACO contract start date, organizational type, nephrologist participation, or years of experience (Figure 3).

Figure 3.

No differences in predicted total spending per beneficiary-quarter for long-term dialysis beneficiaries on the basis of ACO contract start date, organizational type, nephrologist participation, and year in program. Predicted total spending per beneficiary-quarter on the basis of ACO contract start date, organizational type, nephrologist participation, and year in program. 95% CI, 95% confidence interval.

Discussion

This study found that the percentage of long-term dialysis beneficiaries aligned to an MSSP ACO increased to about 23% as of 2016. After adjustment, ACO alignment was associated with a savings of $143, or 0.9%, of spending per beneficiary-quarter for long-term dialysis beneficiaries. Notably, stratified analyses demonstrated that savings associated with ACO alignment were limited to long-term dialysis beneficiaries receiving primary care services by a PCP at $235, or 1.5%, of spending per beneficiary-quarter.

The close alignment in the growing percentage of long-term dialysis beneficiaries attributed to an ACO over the period from 2012 to 2014 with that seen in the overall Medicare population suggests that most ACOs were not making explicit efforts to avoid high-cost long-term dialysis beneficiaries. Through the choice of participating physicians, ACOs can theoretically exercise some control over which beneficiaries are aligned to them (7). The divergent trend in the percentage of long-term dialysis beneficiaries attributed to MSSP ACOs starting in 2015 likely represents alignment of a portion of such beneficiaries to the CEC model, which was first implemented in the last quarter of 2015 and is scheduled to continue through 2020. Beneficiaries aligned to an ESRD Seamless Care Organization (ESCO), the care coordination entity established by the CEC model, generally cannot be aligned to any other Shared Savings Program. Therefore, this decrease in attribution of long-term dialysis beneficiaries compared with Medicare beneficiaries is likely due to the initiation of the CEC model.

The spending savings of $143 per beneficiary-quarter associated with ACO-aligned long-term dialysis beneficiaries is relatively modest. Nevertheless, this level of savings exceeds in absolute terms that of early ACOs for the general population of Medicare beneficiaries, at $34–$66 per beneficiary-quarter, and is similar to the amount of savings for clinically vulnerable beneficiaries, defined as beneficiaries who have an average of three Hierarchical Condition Categories, at $147 per beneficiary-quarter (4,12,13). Savings may be greater in absolute terms for more medically complex beneficiaries given the higher baseline costs. In relative terms, however, the reduction of 0.9% (or even 1.5%, when restricted to beneficiaries with care by a PCP) of spending we observed is smaller than that observed either for clinically vulnerable beneficiaries (2.3% of spending) or in ESCOs, which have achieved savings amounting to 1.8%–2.1% of spending for long-term dialysis beneficiaries in their first 2 performance years (14,15). However, the vast majority of ESCOs have a two-sided risk, whereas most MSSP ACOs during the study period were still in a one-sided risk model.

How may ACOs have achieved savings for long-term dialysis beneficiaries? One possible mechanism is through targeted interventions, such as ACOs partnering directly with dialysis providers and nephrologists to deliver improved management of care for long-term dialysis beneficiaries (5). Alternatively, the broader population health management strategies deployed by ACOs, such as increased resources for after-hours health care access, advanced care planning, and palliative care, may have been particularly effective for the long-term dialysis population (16). Our finding that savings among ACO-aligned beneficiaries were limited to those with care by a PCP is more consistent with this second possibility, given that these care management strategies may be particularly directed at, or work through, PCPs. For example, PCPs may provide better access than nephrologists to address urgent, new health concerns outside of the context of dialysis for those aligned to an ACO, and after-hours access to PCPs has been shown to reduce emergency department visits and hospitalizations (17,18). There is a potential note of caution, however, as we did observe substantially higher spending in beneficiaries who received primary care services from a PCP. Indeed, in our own data, crude spending was similar for ACO-aligned and nonaligned beneficiaries (Figure 2) driven by the higher percentage of patients with PCP involvement in the ACO group.

Our work should be interpreted in the context of its limitations. First, because of the voluntary nature of the MSSP ACO model, there may be some selection bias present in the results. However, patient characteristics were fairly well balanced between those aligned to an ACO and those not aligned, with most differences, with the exception of care by a PCP, not being clinically meaningful even if statistically significant. Second, our results were limited to evaluation of MSSP ACOs and may not be generalizable to other types of ACOs, such as Pioneer, Next Generation, and Commercial ACOs. Nevertheless, the vast majority of providers participating in an ACO program are doing so via the MSSP. Third, our analysis of spending was limited to Medicare claims and did not account for other sources of expenses, such as patient copays, additional Medicare payments to provider for participation in an advanced payment model, or administrative costs for ACOs. Finally, although our study examined the relationship between ACO alignment on spending for long-term dialysis beneficiaries, further research is necessary to understand the association between ACO alignment and health outcomes for this population.

These limitations notwithstanding, this is a novel national study of spending for long-term dialysis beneficiaries aligned to MSSP ACOs and has potentially important implications. Participation in an ACO may be an effective method of reducing spending for these historically expensive beneficiaries for those who already receive care by a PCP. Current and upcoming models of care specific to this population, such as the CEC Model and the newer models as part of the Advancing American Kidney Health Executive Order, share conceptual similarities to ACOs, with their focus on improving quality of care while reducing spending (19). In contrast to ACOs, however, these models place nephrologists and dialysis providers in a central role for care coordination and management. Additional research focused on a better understanding of how the more typically primary care–centric ACOs achieved savings for long-term dialysis beneficiaries may provide useful insights to inform future models of care delivery.

Disclosures

All authors have nothing to disclose.

Funding

This work was supported by Agency for Healthcare Research and Quality grants R01HS026908 (to R. Hirth, J. Hollingsworth, and V. Shahinian), R01HS024728 (to J. Hollingsworth), and R01HS024525 (to J. Hollingsworth).

Supplemental Material

This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.02150220/-/DCSupplemental.

Supplemental Table 1. Comorbidities adjusted for in main analysis.