Changes in Health Care Utilization During the First 2 Years of Massachusetts Medicaid Accountable Care Organizations

Abstract

On March 1, 2018, the Massachusetts Medicaid and Children's Health Insurance Program (MassHealth) launched an ambitious accountable care organization (ACO) program that sought to integrate care across the physical, behavioral, functional, and social services continuum while holding ACOs accountable for cost and quality. The study objective was to describe changes in health care utilization among MassHealth members during the pre-ACO baseline (2015–2017) and post-implementation periods (2018 and 2019). Using MassHealth administrative data, the authors conducted a repeated cross-sectional study of MassHealth members enrolled in ACOs during 2015–2019. Rates of primary care visits, all-cause and primary-care sensitive emergency department (ED) visits, ED boarding, hospitalizations, acute unplanned admissions, and readmissions were reported during the baseline period (2015–2017) and year 1 (2018) and year 2 (2019). Primary care visit rates increased for adult members throughout the study period from a baseline mean of 7.2–9.2 per member per year (observed-to-expected [O:E]: 1.16) in 2019. Observed all-cause hospitalization rates fell below expected values with O:E ratios of 0.96 among adults and 0.79 among children in 2018, and 0.96 and 0.92 among adults and children, respectively, in 2019. All-cause ED visit rates increased slightly, and rates of pediatric asthma-related admissions, unplanned admissions for adults with ambulatory care sensitive conditions, and unplanned admissions and ED boarding for adults with substance use disorder and serious mental illness all declined for the study period. These findings are suggestive of utilization shifts to higher-value, lower-cost care under Massachusetts's innovative and comprehensive ACO model.

Introduction

Since the passage of the Patient Protection and Affordable Care Act, accountable care organizations (ACOs) have emerged as a prominent policy vehicle for payment and delivery system reforms that seek to moderate rising cost trends while improving care quality and patient experience.^1,2 In general, payers hold ACOs to cost and quality benchmarks, then share in savings and losses when performance exceeds or falls below these benchmarks.³

Drawing insights from the accumulating evidence on diverse approaches and mixed results of early Medicare ACO programs, the Centers for Medicare and Medicaid Innovation (CMMI) (a federal policy laboratory within the Centers for Medicare and Medicaid Services) has established accountable care models as the top objective in its strategic roadmap for the next decade.⁴ Increasing accountability in state Medicaid programs alongside reformed Medicare ACO initiatives is central to fulfilling this objective.

As of 2018, 12 states had active Medicaid ACO programs; at least 10 others were pursuing such programs. In March of 2018, MassHealth (ie, the Massachusetts Medicaid and Children's Health Insurance Program) shifted over two thirds of its >1-million eligible members from traditional managed care delivery systems, including managed care organizations (MCOs) into 17 new Medicaid ACOs. As in several other states, time-limited federal investments catalyzed delivery and payment system reforms in Massachusetts. Specifically, for a 5-year period under Massachusetts' 2017–2022 1115 Demonstration agreement extension, $1.8 billion in federal funding supported infrastructure and capacity building for MassHealth, its new ACOs, and community-based organizations expected to partner with ACOs to achieve the program goal of fully integrated team-based care.⁵

MassHealth ACOs were built on a foundation of primary care, with expectations for care to be well-coordinated across the continuum of a member's physical, behavioral, functional, and social needs.⁶ Novel components of the program include requirements for ACOs to collaborate with community-based organizations known as Community Partners (CPs) to coordinate care for members with complex behavioral health and long-term services and supports needs, and expectations for ACOs to partner with social services organizations to implement newly designed Flexible Services programs to address health-related social needs.

In addition to quality and cost accountability (with downside risk) at the ACO level, MassHealth set contractual requirements for ACOs to engage frontline primary care practices with value-based payments tied to cost and quality performance. The combination of investments in new services and supports for ACO members, 2-sided financial risk-sharing, and a slate of ACO quality measures sought to improve care coordination, increase high-value preventative care, and decrease high-cost acute and emergency service utilization.^6,7

As the primary cost-containment and value-enhancing mechanism available to policymakers at scale, and the one at the heart of the long-term CMMI strategic vision, it is imperative to study innovative ACO initiatives. Relatively few Medicaid enrollees in ACOs nationally, inconsistent programs implemented across states to date, and heterogenous evidence from prior initiatives underscore the need for research to guide future implementation and refinement of accountable care initiatives.

With the MassHealth ACO program in its infancy during the study period, the authors performed an exploratory study to describe changes in health care utilization among MassHealth members during the pre-ACO and early post-implementation periods. To examine what secular trends may have looked like in Massachusetts absent the ACO program implementation, changes in health care utilization were also assessed in the MCO-enrolled members during the same period. Although the effects of ACOs on outcomes are expected to build over time, the results can be interpreted as preliminary signals of whether changes were directionally consistent with the ACO program's goals while also highlighting potential targets for policy refinement and further research.

Methods

Study design and data

The authors conducted an annual repeated cross-sectional study using administrative data from MassHealth from 2015 through 2019. Data were organized into a baseline period (2015, 2016, and 2017), program year 1 (2018), and program year 2 (2019).⁸ The MassHealth administrative data used for this study included member enrollment, provider characteristics, claims, and encounter files.⁸

This study was reviewed and determined as nonhuman subjects research by the University of Massachusetts Chan Medical School Institutional Review Board.

Study population

The study population included MassHealth members aged 2–64 years enrolled with MassHealth ACOs or MCOs for most of the calendar year (specifically, at least 320 days). Members who were cared for by ACO-affiliated providers are defined as the ACO group, and members cared for by providers that were enrolled in traditional MCOs were included in the MCO group. The ACO baseline (2015–2017) population was defined as MassHealth members who would have been enrolled with ACOs had the program been in effect at that time based on their primary care provider's (PCP) ACO affiliation at the time of ACO launch.^8,9

The MCO baseline population includes MassHealth members being cared for by providers that were in traditional MCOs in the post-implementation years (2018 and 2019). Adults (ages 18–64), children (ages 2–17), adults with behavioral health conditions (ie, serious mental illness [SMI] or substance use disorder [SUD]; Supplementary Appendix SA1), adults with diabetes (Supplementary Appendix SA2), and children with asthma (Supplementary Appendix SA3) were identified as subpopulations of interest consistent with program priorities and accountability metrics.⁷ Diagnoses of diabetes, SMI, SUD, and asthma are associated with high utilization rates but are thought to be responsive to primary care intervention and care coordination.¹⁰

Exposure

Members were considered unexposed during the baseline period and exposed during 2018 and 2019. Exposure to ACOs in 2018 was incomplete with the program launching on March 1 and because members had the option to change their PCP and plan (ie, either switching to another ACO or a non-ACO plan type) assignment immediately preceding and for 90 days after the program was initiated.^8,9 At ACO program initiation, primary care practices that cared for MassHealth members were given the choice to affiliate with an ACO (PCPs could only affiliate exclusively with 1 ACO) or to remain in the previous managed care system (eg, MCO). Changes to primary care practices' affiliations could occur annually; few changes occurred from 2018 to 2019.

Study outcomes

Study outcomes included counts of primary care visits per member per year, emergency department (ED) visits per 1000 persons, unplanned hospital admissions per 1000 persons, and hospital readmissions per 100 hospital discharges. Primary care visits were defined as visits with any family doctor, advanced practice provider, pediatrician, or obstetrics and gynecology specialist, consistent with quality measure specifications.⁸

ED visits, unplanned hospital admissions, and readmissions were also identified consistent with ACO Healthcare Effectiveness Data and Information Set (HEDIS) quality measure specifications.¹¹ ED boarding, or spending 12 or more hours from the time of arrival to discharge, was also reviewed for the SMI/SUD population. Studies have found ED boarding to be a consequence of inadequate outpatient and inpatient supply of behavioral health services.¹² The definition of primary care sensitive ED visits was taken from the literature.¹³ Acute and chronic ambulatory sensitive conditions were defined using Agency for Healthcare Research and Quality specifications.¹⁴

Numerators of all utilization metrics included events with service dates between January 1 and December 31st of the measurement year(s).⁸ The denominator of each measure included attributed members for that measurement year with the exception of readmissions, which used a denominator of hospital discharges.⁸ In addition to rates, ratios of utilization metrics in program years 1 and 2 as compared with the baseline period are also reported.

Covariates

Several variables in the administrative data were used as predictors in multivariable models. These included age, sex, the Diagnostic Cost Group (DxCG) medical morbidity summary score (summary score developed from a regression model that is based on coded clinical condition categories and incorporates age and gender), disability (client of the Massachusetts Department of Mental Health, Department of Disability Services, or entitled to MassHealth due to disability), and housing problems (International Classification of Diseases, Tenth Revision code of Z59.0–59.1, Z59.81, or >3 addresses within the year).^15–17

Statistical analysis

Characteristics of the study population in each study period (baseline, 2018, and 2019) were first summarized as percentages for categorical variables and means with standard deviations for continuous variables. The authors then reported crude rates for each utilization measure and calendar year overall, and subgroups of interest.⁸ To assess if the changes in health care utilization during 2018 and 2019 were associated with fluctuations in characteristics of the population, the team built multivariable models with the covariates described earlier using baseline (2015–2017) data and used them to generate expected (ie, predicted) values during 2018 and 2019.^8,15

Poisson or negative binomial models, as appropriate, were used to calculate expected values. To improve the model's performance over time in a setting of changing enrollment, the model was built using all MassHealth members eligible for enrollment in ACOs or MCOs in the baseline period. Observed-to-expected (O:E) ratios compared crude rates with model expected rates for each program period.⁸ The O:E ratio varies around 1.0, with values <1 indicating lower than expected utilization and values >1 indicating greater than expected utilization for the measure.⁸ A higher O:E ratio value for the utilization outcome is preferred and consistent with program goals for metrics such as primary care visits, whereas an O:E ratio <1 is preferred for values such as hospitalizations.⁸

Results

Population characteristics

The population of interest included 2.1 million MassHealth members for the 5-year period; ∼700,000 (1,867,933 person-years) would-be ACO and about 120,000 (368,668 person-years) MCO members at baseline, 652,655 ACO and 110,563 MCO members in 2018, and 677,101 ACO and 80,261 MCO members in 2019. Population characteristics are summarized in Table 1.

Table 1.

Population Characteristics

Characteristic	ACO population					MCO population
	Baseline			Program year 1	Program year 2	Baseline			Program year 1	Program year 2
	2015	2016	2017	2018	2019	2015	2016	2017	2018	2019
Population size (person-years)	658,306	545,017	664,610	652,655	677,101	137,305	111,209	120,154	110,563	80,261
Female sex, %	55.1	53.1	53.5	53.6	53.5	53.7	52.2	52.3	50.7	50.7
Age in years, mean (SD)	26.4 (18.0)	27.2 (18.2)	25.9 (18.2)	26.0 (18.2)	26.1 (18.3)	28.7 (17.9)	28.9 (18.0)	28.4 (18.3)	29.2 (17.5)	31.4 (17.3)
Race/ethnicity,a %
Non-Hispanic White	36.7	45.2	35.1	32.9	33.2	48.5	58.9	47.9	46.5	47.4
Non-Hispanic Black	11.0	5.6	10.6	10.8	10.8	7.1	3.6	6.9	7.6	7.6
Hispanic	9.1	3.5	8.1	8.2	8.3	9.1	3.3	7.7	7.4	7.2
Other	6.3	3.5	6.1	6.2	5.9	3.9	1.6	4.4	4.7	4.5
Unknown	36.9	42.3	40.2	41.9	41.8	31.4	32.6	33.1	33.8	33.4
Housing instability, %	9.9	12.2	12.7	11.6	10.9	9.2	11.5	11.6	11.4	9.7
Any disability, %	13.0	14.5	14.2	13.8	13.9	9.7	14.1	11.2	9.9	10.8
NSS, mean (SD)a	0.19 (2.1)	0.14 (2.1)	0.09 (1.0)	0.11 (1.0)	0.09 (1.0)	−0.3 (2.0)	−0.4 (1.9)	−0.3 (1.0)	−0.2 (0.9)	−0.2 (1.0)
DxCG	1.0 (2.1)	1.0 (2.0	1.0 (2.1)	1.1 (2.2)	1.2 (2.4)	1.1 (2.0)	1.1 (2.1)	1.0 (2.1)	1.1 (2.1)	1.2 (2.4)

Notes: Housing instability includes members who were noted to have 3 or more addresses or coded as homeless during the study period. Any disability indicates the member is a client of the Department of Mental Health, the Department of Developmental Services, or entitled to MassHealth due to a disability. The NSS is a census-based measure of area-level socioeconomic stress.¹⁷ The NSS is standardized to have mean 0 and SD = 1 among all individuals eligible to enroll in ACOs during the baseline period. The DxCG relative risk score is a measure of medical morbidity with mean set to 1 in the baseline (2015–2017) population.

^a Caution should be exercised when interpreting race distributions; >30% reported as unknown.

ACO, accountable care organization; DxCG, Diagnostic Cost Group; MCO, managed care organization; NSS, Neighborhood Stress Score; SD, standard deviation.

Approximately 58% of the baseline ACO and 65% of the MCO population were adults, among whom there was an average age of 39.1–39.6 years; the majority were women (54.1%–59.5%; Supplementary Appendix SA4). Approximately 1 in every 9 adult ACO members experienced housing instability and 1 in 5 had a disability, whereas 1 in every 10 adult MCO members experienced housing instability and 1 in 7 had a disability; these figures fluctuated modestly for the study period.

The Neighborhood Stress Score (NSS) in the adult population was 0.1 (standard deviation [SD]: 1.8) in ACO and −0.3 (SD: 1.8) in MCO members at baseline and 0.1 (SD: 1.0) in ACO and −0.3 (SD: 1.0) in the MCO adults in 2019. The DxCG relative risk score increased in ACO-enrolled adults from a baseline value (standard deviation) of 1.5 (2.4) in 2015 to 1.8 (2.9) in 2019. However, the DxCG did not fluctuate in MCO-enrolled adults for the same time period.

Characteristics of the ACO- and MCO-enrolled pediatric members were relatively stable throughout the study period (Supplementary Appendix SA4). The average age of pediatric members was 9.0–9.3 years at baseline, with slightly more boys than girls. The proportion of pediatric ACO-enrolled members with reported housing insecurity slightly decreased from an average of ∼12% in the baseline period to 9.4%–9.8% in 2019. The proportion of pediatric members reporting disability averaged ∼6% throughout the study; NSS and DxCG summary scores also remained relatively consistent over time.

Overall health care utilization

For the overall population, primary care visit rates increased throughout the study period for ACO-enrolled members from a 6.2 per member per year in the baseline period to 6.9, or an 11.9% increase in 2018, and up to 7.2, or 17.7% increase over baseline, in 2019. Primary care visit rates declined among the MCO-enrolled adults, from 5.3 per member per year in the baseline period to 4.1 in 2018 and 4.6 in 2019. Observed primary care visit rates exceeded the expected rates in both 2018 (O:E: 1.08) and 2019 (O:E: 1.12) in ACO-enrolled members and fell below expected rates in 2018 (O:E: 0.67) and 2019 (O:E: 0.70) among MCO-enrolled members.

Primary-care sensitive ED visit rates among ACO-enrolled adults were 283.5 per 1000 members at baseline, up to 302.3 (6.6% increase over baseline) in 2018, and back down to 283.9 in 2019. Similar changes in primary-care sensitive ED visits in MCO-enrolled members were observed. Observed rates for primary-care sensitive ED visits were higher than expected in 2018 in the ACO- (O:E: 1.07) and MCO-enrolled (O:E: 1.05) members, but at or below expected in 2019 among both the ACO- (O:E: 0.99) and MCO-enrolled (O:E: 0.95) members.

Adult health care utilization

Adult primary care visits increased in the ACO-enrolled population from a baseline of 7.3 per member per year to 8.5, 16.6% increase from baseline (O:E: 1.11), in program year 1 and 9.2, (26.2% increase; O:E: 1.16) in program year 2 (Table 2). Primary care visits in MCO-enrolled adults decreased from a baseline of 6.0 per member per year to 4.6 (O:E: 0.66) in program year 1 and 5.1 (O:E: 0.71) in program year 2. Throughout the study period, primary care visit rates for adults with SMI or SUD increased by ∼17% in the ACO-enrolled adults and decreased by ∼14% in MCO-enrolled adults.

Table 2.

Adult Utilization Before and After Accountable Care Organization Implementation

Variable	ACO population					MCO population
	Baseline			Program year 1	Program year 2	Baseline			Program year 1	Program year 2
	2015	2016	2017	2018	2019	2015	2016	2017	2018	2019
Population size (person-years)	385,572	328,678	371,041	354,074	364,263	90,883	72,805	75,718	73,164	57,918
Primary care visits
Number of visits per member per year	7.2	7.1	7.5	8.5	9.2	5.8	6.3	6.0	4.6	5.1
O:E ratio	1.00			1.11	1.16	0.88			0.66	0.71
ED visits (all-cause)
Number of visits per 1000 members	759.8	711.3	854.0	859.5	864.7	689.2	724.6	782.8	792.4	741.5
O:E ratio	1.00			1.07	1.04	0.98			1.03	0.96
Hospitalizations (all-cause)
Number of visits per 1000 members	154.9	139.5	162.3	158.0	169.7	141.4	140.0	154.7	153.4	156.4
O:E ratio	0.99			0.96	0.96	1.03			1.03	0.99
Acute unplanned admissions for chronic ACSCs
Number of visits per 1000 members	2.8	2.3	1.9	2.2	1.6	2.4	1.5	1.9	0.9	1.0
O:E ratio	0.97			0.83	0.54	0.93			0.43	0.44
Acute unplanned admissions for acute ACSCs
Number of visits per 1000 members	0.9	0.8	1.0	0.9	0.7	0.4	0.5	0.4	0.1	0.2
O:E ratio	1.04			0.93	0.71	0.57			0.19	0.21
Readmissions
Number of readmissions per 100 discharges	20.3	20.8	22.3	22.4	22.7	18.8	20.8	20.0	21.5	22.0
O:E ratio	1.01			0.95	0.90	1.01			0.97	0.92

Notes: For the O:E ratio, observed is the calculated outcome for the quality measure; expected is calculated based on a model that accounts for changes in member characteristics (age and sex, disability, housing instability, and medical morbidity) between baseline (2015–2017) and 2018 and 2019. The ratio of O:E values varies around 1.0, with values <1 indicating lower than expected outcomes and values of >1 indicating greater than expected outcomes for the measure.

ACO, accountable care organization; ACSCs, ambulatory care sensitive conditions; ED, emergency department; MCO, managed care organization; O:E, observed-to-expected.

Observed primary care visit rates in ACO-enrolled adults with SMI/SUD exceeded expected rates in both program year 1 (O:E: 1.05) and year 2 (O:E: 1.10); observed rates were below expected rates in the MCO-enrolled adults (O:E: 0.66 and 0.72, respectively). Although observed ED visits for adults with SMI or SUD conditions were closer to expected rates in both 2018 and 2019 in the ACO and MCO population, ED boarding substantially decreased throughout the study period in both groups.

ED boarding rates for ACO-enrolled adults with SMI or SUD decreased from 567.8 (O:E: 1.00) per 1000 members at baseline to 348.1 (38.7% decrease from baseline; O:E: 0.57) in 2018 and 318.2 (44.0% decrease from baseline; O:E: 0.46) in 2019. Decreases were also noted in the MCO population, but to a lesser magnitude. The observed acute unplanned admissions for both ACO- and MCO-enrolled adults with SMI or SUD were slightly below expected in 2018 (ACO O:E: 0.95; MCO O:E: 0.98) and 2019 (ACO O:E: 0.92; MCO O:E: 0.97).

For ACO-enrolled adults with diabetes, the rate of primary care visits increased from a baseline of 15.1 visits per member per year to 16.6 (9.8% increase from baseline; O:E: 1.07) in program year 1 and into program year 2 at 17.7 (16.9% increase from baseline; O:E: 1.11). The rate of primary care visits decreased for MCO-enrolled adults with diabetes from 10.4 (O:E: 0.82) to 8.7 (O:E: 0.6) in year 1 and into program year 2 at 10.7 (O:E: 0.71).

Acute unplanned admissions for adults with diabetes decreased from a baseline of 404.5 to 393.8 (2.6% decrease from baseline; O:E: 0.87) in 2019 in ACO-enrolled adults, and slightly increased from 383.8 to 408.1 (O:E: 0.93) in MCO-enrolled adults. A summary of health care utilization in adults with SMI or SUD and adults with diabetes is located in Supplementary Appendix SA5.

Pediatric health care utilization

Observed PCP visit rates in ACO-enrolled pediatric members slightly exceeded expected rates in 2018 (4.8% increase; O:E: 1.03) and were consistent with baseline in 2019 (0.2% increase from baseline; O:E: 0.99), whereas all-cause hospitalizations in the same members were below expected in 2018 (3.6% decrease; O:E: 0.79) and 2019 (6.3% decrease from baseline; O:E: 0.92).

In MCO-enrolled pediatric members, observed PCP visit rates were below expected rates (O:E: 0.85) at baseline, were at expected rates in 2018 (O:E: 1.00), and slightly above expected rates in 2019 (1.04), and all-cause hospitalizations also were below expected rates in 2018 (14.7% decrease; O:E: 0.55) and 2019 (6.8% decrease; O:E: 0.46) in MCO-enrolled pediatric members. Table 3 describes health care utilization in the pediatric population.

Table 3.

Pediatric Utilization Before and After Accountable Care Organization Implementation

Variable	ACO population					MCO population
	Baseline			Program year 1	Program year 2	Baseline			Program year 1	Program year 2
	2015	2016	2017	2018	2019	2015	2016	2017	2018	2019
Population size (person-years)	272,734	216,339	293,569	298,581	312,838	46,422	38,404	44,436	37,399	22,343
Primary care visits
Number of visits per member per year	4.7	4.7	4.8	5.0	4.7	3.8	4.7	4.6	4.7	4.9
O:E ratio	1.00			1.03	0.99	0.93			1.00	1.04
ED visits
Number of visits per 1000 members	450.7	427.6	456.6	493.3	442.8	380.3	390.9	421.5	434.8	370.3
O:E ratio	1.01			1.11	1.00	0.91			0.99	0.83
Hospitalizations
Number of admissions per 1000 members	23.9	21.3	21.7	21.5	20.9	18.8	17.3	17.0	15.1	16.5
O:E ratio	1.05			0.79	0.92	0.86			0.55	0.46
Readmissions
Number of readmissions per 100 discharges	6.9	6.9	8.0	7.7	9.5	7.5	6.2	5.0	6.2	7.1
O:E ratio	0.99			0.90	0.98	0.93			0.73	0.67

Notes: For the O:E ratio, observed is the calculated outcome for the quality measure; expected is calculated based on a model that accounts for changes in member characteristics (age and sex, disability, housing instability, and medical morbidity) between baseline (2015–2017) and 2018 and 2019. The ratio of O:E values varies around 1.0, with values <1 indicating lower than expected outcomes and values of >1 indicating greater than expected outcomes for the measure.

ACO, accountable care organization; ED, emergency department; MCO, managed care organization; O:E, observed-to-expected.

Pediatric asthma admissions decreased in both the ACO- and MCO-enrolled members. In ACO-enrolled pediatric members, asthma admissions decreased from 14.5 per 1000 enrolled ACO members at baseline to 9.7 (33.1% decrease; O:E: 0.67) in 2018 and 7.9 (45.5% decrease from baseline; O:E: 0.54) in 2019. In MCO-enrolled pediatric members, asthma admissions decreased from 10.8 per 1000 enrolled ACO members at baseline to 6.6 (38.9% decrease; O:E: 0.44) in 2019 and 4.7 (56.5% decrease; O:E: 0.29) in 2019. Observed rates of asthma-related admissions in pediatric members were well below expected rates in both groups in program years 1 and 2 (Supplementary Appendix SA5).

Discussion

In the first 2 years of an ambitious restructuring of the health care delivery system in Massachusetts, utilization trends among members of new Medicaid ACOs were consistent with the goal of shifting away from high-cost acute service utilization and toward primary care. The transition from legacy fee-for-service driven and volume-oriented care to accountable health and social systems is incremental, and the effects of reforms are hypothesized to accumulate over time.

Although in some cases improvements on some measures among MCO enrollees suggest improvements among ACO enrollees may have been part of statewide secular trends, spillover effects of ACO-program implementation investments may have contributed to improvements for MCO enrollees. As such, policymakers and their stakeholder communities must resist the inclination to pre-judge programs on the basis of narrow sets of metrics during early implementation.

However, it is encouraging that even in the earlier phases of ACO implementation, adult primary care utilization increased in the ACO-enrolled members and decreased in the MCO-enrolled members from the pre-ACO period, including for subgroups with chronic physical and behavioral health conditions. Visits to the ED rose in the program's 1st year before returning to baseline in both groups.

Hospital admissions for ambulatory care sensitive conditions (ACSCs) declined among adults in ACO- and MCO-enrolled adults, whereas hospital admissions for children with asthma also decreased from baseline in ACO- and MCO-enrolled pediatric members. Taken together, these findings suggest increases in primary care engagement and effective outpatient management of costly chronic health conditions, consistent with ACO programmatic goals of improved care coordination.

It is unclear whether the declines the authors observed in hospitalization rates were sufficient to offset increases in costs associated with higher rates of primary care utilization among MassHealth members and whether the use of other outpatient care (eg, specialists, laboratory services) increased. A systematic review examining ACO effects found mixed evidence regarding whether ACOs decrease health care costs.^3,18 In settings with a high prevalence of unmet needs and inadequate patient engagement with the health care system, some observers have questioned whether cost reduction is a reasonable expectation of programs seeking better care coordination,¹⁸ since better connection to services seems likely, at least in the short term, to lead to new utilization for previously unmet needs.

Others have sought to explain heterogeneity in estimates of ACO effects on costs in light of program and ACO characteristics. The Medicare Shared Savings Program (SSP), a federal ACO initiative that is the largest to date, has achieved considerable savings despite weak incentives for long-term cost reduction.¹⁹ The most successful Medicare SSP ACOs have been those anchored by physician-led organizations rather than hospital-based ACOs, likely due in part to conflicting interests facing hospital-led ACOs.

Namely, even when 1 payer relationship is value-based, reductions in hospital service volume may be a net negative for the organization when a majority of all-payer revenue continues to flow through volume-based payment arrangements.¹⁸ The MassHealth ACO program, with 3 distinct ACO models, extensive flexibility in programmatic design, and diverse ACO characteristics, will provide ample opportunity for future studies exploring sources of ACO performance heterogeneity.

Established in 2010, the Centers for Medicare & Medicaid Services Innovation Center has developed a strategy to support the transformation of the health care delivery system to focus on health equity, value-based payment arrangements (including the development of ACOs), and person-centered care.⁴ As of 2018, at least 12 state Medicaid programs were engaged in alternative payment strategies through the development of ACOs.²⁰ Despite common features such as shared savings and a focus on improving care coordination and quality performance, there is considerable variation in the design and implementation of these ACO programs.²⁰

In previously published reports from Medicaid ACOs in Oregon, Maine, and Vermont, improvements in several acute care utilization metrics, such as decreases in hospitalizations, were noted in the early-ACO implementation years, analogous with MassHealth's reported trends.^21,22 Contrasting with the findings in this study, these programs did not report increases in primary care utilization, possibly due to a shift of care toward other service lines such as case management; however, they reported improvements in multiple other HEDIS access measures suggesting that access to care was maintained.²¹

Built on a foundation of primary care, MassHealth's ACO and Delivery System Reform Incentive Payment programs heavily emphasized patient engagement, needs assessment, and fully integrated care capable of addressing the full spectrum of identified needs. Therefore, the authors have interpreted early increases in primary care visit rates in ACO-enrolled members as a positive indicator, consistent with program design.

The ACO model is anticipated to cause larger acute care utilization shifts in populations with chronic conditions, where care coordination and primary care interventions may prevent serious adverse health outcomes and costly acute care interventions.⁵ In a previous study of privately insured youths and adults treated for SUD under an ACO model, it was reported that SUD service use and spending did not decline in the early phase of ACO implementation.²³

In contrast, the authors found substantial reductions in ED boarding rates for ACO-enrolled adults with SMI or SUD, although broader statewide efforts targeting ED boarding were underway, and the reduction cannot be unambiguously attributed to ACO implementation. In addition, modest reductions in acute unplanned admissions and increases in primary care visits were exhibited for ACO-enrolled adults with SMI or SUD and adults with diabetes.^3,24 These notable reductions in ED boarding and acute unplanned admission rates, in the setting of increases in primary care visits, suggest better outpatient management for members with these conditions. Whether quality indicators specific to these populations also improved merits further investigation.

It is hypothesized that utilization of low-value services and costs will decrease under an effective ACO model.²⁵ Although accountable care may be 1 partial solution to rising health care costs, cost reduction is only 1 of the potential benefits of MassHealth's program, which uses a combination of contractual requirements and incentives to forge stronger connections between ACOs and community-based organizations.^26,27

With >1 in 10 ACO-enrolled members reported to be housing unstable and >1 in 8 with a recorded disability, the integration of social services into the health care delivery framework to address health-related social needs, such as access to safe and affordable housing and access to sufficient and nutritious food supplies, should help improve quality of life and may lead to positive health outcomes among ACO members.²⁸

Furthermore, solidified partnerships with CP programs, primarily responsible for care coordination of members with complex health and social needs, may improve member navigation of the health care system by leveraging the expertise of existing community organizations.⁶ By promoting integrated care teams, decreases in care fragmentation, and patient-centered care delivery by linking clinical and social services, health care providers are hypothesized to provide higher care quality and improved care continuity, leading to better health outcomes.²⁹ Improvements in member-reported care quality is also hoped for due to decreased fragmentation and potentially increased member engagement with health care services.³⁰

Limitations

This study has limitations. The authors used a repeated cross-sectional design rather than contemporaneous comparisons to describe trends in utilization. Although the authors compared model predicted values with observed utilization rates to examine whether utilization trends were explained by changes in population characteristics, such analyses are susceptible to time-varying biases such as changes in coding practices.

Adult medical morbidity appeared to increase throughout the study period, which may reflect a combination of rising medical morbidity, changes in coding practices, and member migration on and off MassHealth. Decreases in crude rates of hospital admissions for ACSCs despite increases in coded medical morbidity are encouraging. The authors did not perform statistical testing to compare pre- and post-values; such statistical tests would likely return exclusively significant results because of the large sample sizes.

Comparative quasi-experimental studies, which will be conducted once data are available for later years of the program, will better address questions of causality and costs. Second, race and ethnicity were missing from the data set for a third of the population studied. Furthermore, the authors required members to be enrolled for at least 320 days within a given calendar year, consistent with the standard approach for most quality measures.

However, utilization patterns during the study period may have differed for individuals with shorter enrollment periods. Finally, the baseline pre-ACO members were identified using MassHealth's PCP attribution algorithm, which may have misattributed members (eg, those without a PCP or with multiple PCPs) to ACO or non-ACO providers.

Conclusion

In this foundational study of nearly 700,000 members transitioning from traditional Medicaid managed care into newly implemented Medicaid ACOs, increases in primary care visits and decreased utilization of many acute and emergency services, especially in adults with chronic health conditions, were observed in members enrolled in the ACO program. These findings appear to be early indications of promising utilization shifts to higher-value lower-cost care under Massachusetts's innovative and comprehensive ACO model.

Authors' Contributions

Conceptualization, project administration, writing—original draft, and writing—review and editing by M.S. Data curation, methodology, validation, and software by E.O.M. Visualization and writing—review and editing by A.S.A. and J.H. Conceptualization, methodology, writing—review and editing, and supervision by M.J.A.

Author Disclosure Statement

The listed authors certify that they do not have any interests, funding, or employment that may inappropriately influence or affect the integrity of the submission.

Funding Information

This study was conducted as part of the Independent Evaluation the Massachusetts 1115 Demonstration Extension (2017–2022), for which UMass Chan Medical School is the Independent Evaluator; this project partially funds effort from Meagan Sabatino, Dr. Alcusky, Dr. Ash, Dr. Mick, and Dr. Himmelstein. M.S. is also funded by the NIH Training Grant titled “Prevention and Control of Cancer: Training for Change in Individuals and Systems” (Grant No. 5T32CA172009-08).

Supplementary Material

Supplementary Appendix SA1

Supplementary Appendix SA2

Supplementary Appendix SA3

Supplementary Appendix SA4

Supplementary Appendix SA5