Associations Between ACO Enrollment Status and Drug and Nondrug Costs Among Older Adults Newly Diagnosed With ADRD

Abstract

Objective:

Despite the promising role of accountable care organizations (ACOs) in improving care coordination and lowering total costs of care, it is unclear whether ACO enrollment leads to reduced drug costs among individuals with Alzheimer’s disease and related dementias (ADRD). This longitudinal cohort study examined the long-term effects of ACO enrollment on drug and nondrug Medicare payments among Medicare beneficiaries with ADRD, focusing on their comorbidity burden.

Setting, Participants, and Measurements:

This study tracked annual Part D prescription drug and nondrug payments of 111,235 fee-for-service Medicare beneficiaries newly diagnosed with ADRD one year prior and five years after the ADRD diagnosis year (2017) by their ACO enrollment and comorbidity burden status. Comorbidity burden was assessed based on the number of comorbidities ADRD patients had, categorized into no multiple chronic conditions (No MCC; 0 or 1), low MCC (2–4), and multisystem morbidity (MM; 5≤) groups.

Results:

ACO enrollees spent significantly less on both drug and nondrug payments than non-ACO enrollees, with greater payment differences observed among those with higher comorbidity burden. Drug payment differences ranged from $55 to $)3 among those with low MCC and $70–$11) among those with MM. Nondrug payment differences were the greatest during the diagnosis year, with $1,GGG for those with MCC and $3,G70 for those with MM.

Conclusions:

The findings of this study underscore meaningful differences in prescription drug and nondrug payments associated with ACO enrollment among Medicare beneficiaries with ADRD, particularly among those with greater coexisting comorbidity burden.

INTRODUCTION

While medications are cost-effective treatments to manage chronic conditions and save downstream medical costs,¹ Medicare prescription drug spending is substantial, with an estimated $137 billion in Part D spending in 2025.² People with Alzheimer’s disease and related dementias (ADRD) often have complex medication regimens, incurring 1.5 times higher average payments for prescription drugs than non-ADRD beneficiaries.³

Over 90% of Medicare beneficiaries with ADRD have at least one other chronic condition, and they are about three times more likely to have four or more chronic conditions.³ The co-occurrence of ADRD and other chronic conditions often leads to polypharmacy and potentially inappropriate medication use,⁴ possibly requiring more drugs and increasing per-person costs.^3,5

Prior studies have suggested the promising role of Accountable Care Organizations (ACOs) in managing chronic conditions,^1,6–8 including ADRD.^9,10 While much research has focused on the effects of ACOs on total Medicare payments and suggests ACOs’ cost-saving potential,^1,11,12 less has explored their effects on prescription drug costs, possibly because Part D drug payments have not been a part of the calculation for ACO-shared savings and risks.^1,13,14 Previous studies on Part D prescription drug payments found mixed spending differences between ACO and non-ACO enrollees among non-ADRD beneficiaries,^1,7,13 and there is a lack of empirical evidence for beneficiaries with ADRD.

It is unclear whether ACO enrollment is associated with reduced or increased prescription drug payments among patients with ADRD. On the one hand, providers in the ACOs may pursue less intensive treatment plans in the setting of ADRD due to limited expected benefits and greater risks for harm with polypharmacy and intensive treatment in this population, potentially lowering drug spending.^15,16 On the other hand, ACOs can improve medication adherence through better coordinated and integrated care, which could increase drug use but reduce costly downstream utilization.^1,7 In addition, ACOs may also affect patients’ medication payments by encouraging the substitution of brand-name prescription drugs for generics, involving pharmacists directly in care, and educating patients on therapeutic alternatives.^7,17 The overall effect of these mechanisms on prescription drug spending and their implications for overall Medicare payments remains unknown.⁷

The objective of this study is to examine the effects of ACO enrollment on Part D prescription drug and nondrug payments among Medicare beneficiaries newly diagnosed with ADRD, and how these associations vary by the number of coexisting comorbidities. Management of coexisting chronic conditions is particularly important for newly diagnosed ADRD patients, as poor management can have a detrimental impact on their overall health and disease progression.^16,18

Given that ADRD is a progressive disease that worsens over time,³ we hypothesize that prescription drug payments would increase as more medications are needed to manage the complex health needs of ADRD patients. Based on the evidence of the trends in total Medicare payments,¹¹ nondrug Medicare payments would peak at their diagnosis year, fall back, and then steadily increase over the subsequent years. It is critical to explore the relationship between drug and nondrug payments and whether increases in medication spending are associated with lower overall Medicare payments among ADRD patients. Increasing medication payments may reduce nondrug payments by preventing costly hospitalizations and ED visits, or it can incur more via improved treatment adherence.¹⁹ For ADRD patients with complex health needs, this evidence can be more relevant to policymakers. We further hypothesize that ADRD patients with more chronic conditions would incur higher spending for both Part D and nondrug Medicare payments and that ACO enrollees would have lower spending than non-ACO enrollees.

METHODS

Data Sources and Study Sample

The Medicare Beneficiary Summary File (MBSF), MBSF Cost and Use file, and Medicare Shared Savings Program (MSSP) ACO beneficiary-level file were combined using unique beneficiary IDs to create the 2016–2022 longitudinal retrospective cohort dataset. The study followed Part D and nondrug Medicare payments in the year preceding the diagnosis (2016) and the subsequent five years (2018–2022). Those with ADRD were identified based on the CMS Chronic Conditions Data Warehouse (CCW) Algorithms.^11,20,21 The 2022 Social Vulnerability Index (SVI)²² was then added to the combined Medicare claims and ACO data using state and county FIPS codes.

The study sample included Medicare beneficiaries aged 65 or older, continuously enrolled in fee-for-service (FFS) and Part D, with no prior ADRD diagnosis in and before 2016 but in 2017. The study restricted its sample to FFS beneficiaries, as this study focused on MSSP ACOs serving about 11 million Medicare beneficiaries.²³

Measures

Prescription drug and nondrug costs

Using the MBSF cost and use file, annual prescription drug and nondrug payments were calculated for Medicare beneficiaries with ADRD by ACO enrollment status. Part D Medicare payments were the sum of the plan payments for covered prescription drugs and the low-income cost-sharing subsidy amount, excluding any applicable rebate amounts or other discounts.²⁴ Nondrug payments were calculated by subtracting Part D Medicare payments from total Medicare payments, calculated as the sum of Medicare payments on key healthcare services consistent with the 2021 Alzheimer’s Disease Facts and Figures Special Report.²⁵ Part D payments were adjusted by the Consumer Price Index (CPI)–Prescription Drug rates and total Medicare payments were adjusted by the CPI–Medical Care rates using 2022 as the base year.²⁶

ACO enrollment status

ACO enrollees were defined as Medicare beneficiaries enrolled in MSSP ACO for all four quarters (i.e., full year). No ACO group included beneficiaries who were never enrolled or had partial enrollment (i.e., any-quarter enrollment) in MSSP ACO in a given year.

Counts of comorbidities

Comorbidities used in the study included 12 broad categories of chronic conditions that are prevalent among older adults with ADRD and available in the MBSF CCW data files: diabetes, heart-related disease, depression, hypertension, cancer, asthma, arthritis, osteoporosis, hyperlipidemia, chronic obstructive pulmonary disease, chronic kidney disease, and hypothyroidism.^18,27,28 Having each condition was identified by CCW algorithms.²⁰ The number of having these chronic conditions was calculated and categorized into three groups: none or one (no MCC), two to four (low MCC), and five or more chronic conditions (MM; multisystem morbidity) in addition to ADRD.^29,30 Having five or more chronic conditions has been associated with increased complexity in care compared with low MCC³⁰ and worse disability in instrumental activities of daily living (IADL) than those with low or no MCC.²⁹

Covariates

Other independent variables in the analyses included age, sex, race and ethnicity, and neighborhood disadvantage represented by the county-level SVI. Given that ACOs are not equitably distributed and those enrolled in ACOs are more likely to have favorable social and health status,⁹ the SVI was included as a covariate to account for county-level social vulnerabilities and geographic variations.^9,22 Overall SVI scores were categorized into quartiles, with the first quartile (Q1) representing the lowest level of social vulnerability and the fourth quartile (Q4) representing the highest. Year dummy variable was also included.

Statistical analysis

The sample characteristics at the base year (2016) were summarized by ACO enrollment status. T-tests and chi-squared tests were used to compare individual and neighborhood-level characteristics between ACO enrollees and non-ACO beneficiaries. Next, generalized estimating equation (GEE) models were used to estimate the population-averaged effects of ACO enrollment and comorbidities on prescription drug payments among the newly ADRD-diagnosed cohort. GEE models with gamma distribution and log-link function controlling for individual and neighborhood characteristics were implemented to account for right-skewed payment outcomes.^31,32 An unstructured correlation structure was chosen.³² Interaction terms between ACO enrollment status and counts of comorbidities were included in the model to examine whether the effect of ACO enrollment on drug payments varies by the number of coexisting comorbidities beneficiaries had. Lastly, marginal effects were estimated. The same analyses were applied to nondrug payments. This study was approved by the University of Maryland’s Institutional Review Board. All analyses were conducted using SAS 9.4 and Stata version 18.

RESULTS

The study included 111,235 FFS beneficiaries newly diagnosed with ADRD in 2017 and continuously enrolled in Medicare Part D during the study period. The descriptive characteristics of the study cohort by ACO enrollment status at baseline (2016) are presented in Table 1. 23,753 (21.4%) beneficiaries were enrolled in ACOs. ACO enrollees had greater proportions of older adults with two or more coexisting chronic conditions (87.5% versus 81.2%), non-Hispanic White (85.4% versus 80.1%), and residents in neighborhoods with the least social vulnerabilities (19.6% versus 13.3%). On the other hand, non-ACO beneficiaries were relatively healthier with higher proportions of people with no or one chronic condition other than ADRD (18.8% versus 12.5%), more likely to be racial and ethnic minorities (19.8% versus 14.6%), and live in neighborhoods with the highest social vulnerabilities (34.7% versus 28.2%).

TABLE 1.

Baseline Characteristics of FFS Medicare Beneficiaries Newly Diagnosed With ADRD by ACO Enrollment Status in 2016

	ACO N = 23,753 (21.4) Mean (SD)/No.(%)	Non-ACO N = 87,482 (78.6) Mean (SD)/%	p
No. of chronic conditions			< 0.001
No MCC (0–1 CC)	2,972 (12.5)	16,428 (18.8)
MCC (2–4 CC)	12,341 (52.0)	42,891 (49.0)
MM (5+ CC)	8,440 (35.5)	28,163 (32.2)
Age	77.6 (7.0)	77.1 (7.1)	< 0.001
Gender			0.578
Male	8,036 (33.8)	29,765 (34.0)
Female	15,717 (66.2)	57,717 (66.0)
Race/Ethnicity			< 0.001
Non-Hispanic White	20,281 (85.4)	70,108 (80.1)
Black	1,301 (5.5)	5,594 (6.4)
Hispanic	973 (4.1)	5,634 (6.4)
Other	1,198 (5.0)	6,146 (7.0)
SVI			< 0.001
Q1	4,643 (19.6)	11,628 (13.3)
Q2	5,799 (24.4)	20,140 (23.0)
Q3	6,603 (27.8)	25,395 (29.0)
Q4	6,708 (28.2)	30,319 (34.7)

Notes: Values represent means (standard deviations) or frequencies (proportions) of population characteristics. Multiple chronic conditions indicate having two to four chronic conditions and multisystem morbidity indicates having five or more chronic conditions among 12 prevalent coexisting conditions among ADRD patients. Abbreviations: ACO, accountable care organizations; ADRD, Alzheimer’s disease and related dementias; MCC, multiple chronic condition; MM, multisystem morbidity; SVI, social vulnerability index; Q, quartile.

Figure 1 shows the unadjusted trends of prescription drug and nondrug payments across the three comorbidity groups during the study period. In general, Part D payments slightly decreased in the ADRD diagnosis year (2017) and steadily increased, while nondrug payments doubled in the diagnosis year and decreased over time until the COVID-19 pandemic period. Across all comorbidity groups, the unadjusted payment trends show that ACO enrollees encountered lower payments than non-ACO enrollees, except ADRD patients without MCC. Among ADRD patients with MCC, ACO enrollment was associated with lower drug payments, with differences ranging from $173 to $332, and lower nondrug payments ranging from $872 to $2,054. Among those with MM, beneficiaries enrolled in ACOs exhibited increasingly larger spending differences over time following ADRD diagnosis, with drug payment differences ranging from $1,135 to $1,359, and nondrug payment differences ranging from $2,526 to $4,367 compared with non-ACO beneficiaries. Among ADRD patients with no MCC, ACO enrollees spent more than non-ACO beneficiaries on prescription drugs in 2022. However, the ACO enrollees with no MCC incurred substantially lower nondrug payments in 2022, leading to a net difference of $1,934 in 2022.

FIGURE 1.

Unadjusted prescription drug and nondrug payments per person per year by ACO enrollment and comorbidity status for Medicare FFS beneficiaries with new diagnosis of ADRD in 2017. Notes: Figures represent unadjusted prescription drug and nondrug payments per person per year by ACO enrollment and comorbidity status among fee-for-service beneficiaries newly diagnosed with ADRD in 2017. Part D Medicare payments were adjusted by the CPI – prescription drug rates. Nondrug Medicare payments were calculated by total Medicare payments minus Part D payments. Total Medicare payments were adjusted by the CPI – Medical Care rates using 2022 as the base year (AHRQ, 2024). Abbreviations: ACO: accountable care organizations; MCC: multiple chronic conditions; MM: multisystem morbidity.

The results from the adjusted GEE regression models estimating the effects of ACO enrollment and multiple comorbidities on drug and nondrug Medicare payments among FFS beneficiaries with Part D enrollment are presented in Table 2. Beneficiaries enrolled in ACO for a full year spent significantly more on prescription drugs than those who were never enrolled in ACO or any quarter in a given year (exp(coef.) = 1.10; 95% CI, 1.06–1.15). Having a greater number of chronic conditions increased drug payments (1.13; 1.10–1.15 for MCC and 1.22; 1.19–1.26 for MM). However, the interaction terms between ACO enrollment and MCC or MM were significantly negative (0.89; 0.84–0.93 and 0.88; 0.84–0.93, respectively), indicating that increases in drug payments associated with ACO enrollment were smaller among those with MCC or MM.

TABLE 2.

Adjusted GEE Regression Estimates of the Effects of ACO Enrollment and Multiple Comorbidities on Drug and Nondrug Medicare Payments Among Beneficiaries Newly Diagnosed With ADRD

	Drug Payments		Nondrug Payments
	Exp(b) (95% CI)	p	Exp(b) (95% CI)	p
ACO full year enrollment (ref: Never/any quarter)	1.10 (1.06–1.15)	< 0.001	0.95 (0.92–0.99)	0.005
No. of chronic conditions (ref: No MCC)
MCC	1.13 (1.10–1.15)	< 0.001	2.45 (2.39–2.50)	< 0.001
MM	1.22 (1.19–1.26)	< 0.001	6.10 (5.96–6.25)	< 0.001
ACO × No. of CC (ref: No ACO & No MCC)
ACO × MCC	0.89 (0.84–0.93)	< 0.001	0.95 (0.92–0.99)	0.011
ACO × MM	0.88 (0.84–0.93)	< 0.001	0.96 (0.93–1.00)	0.059
Race/Ethnicity (ref: Non-Hispanic White)
Black	1.31 (1.24–1.39)	< 0.001	1.06 (1.03–1.09)	< 0.001
Hispanic	1.56 (1.48–1.64)	< 0.001	0.93 (0.91–0.95)	< 0.001
Other	1.90 (1.82–1.99)	< 0.001	0.82 (0.80–0.84)	< 0.001
Sex (ref: Male)
Female	1.02 (0.99–1.06)	0.138	0.92 (0.90–0.93)	< 0.001
Age (Years)	0.96 (0.96–0.96)	< 0.001	0.99 (0.99–1.00)	< 0.001
SVI (ref: Q1)
Q2	1.01 (0.96–1.06)	0.669	1.02 (1.00–1.04)	0.035
Q3	1.06 (1.01–1.10)	0.015	0.99 (0.97–1.01)	0.220
Q4	1.27 (1.21–1.33)	< 0.001	1.03 (1.02–1.05)	< 0.001
Year (ref: 2016)
2017	1.12 (1.10–1.14)	< 0.001	1.99 (1.97–2.02)	< 0.001
2018	1.33 (1.30–1.36)	< 0.001	1.19 (1.17–1.20)	< 0.001
2019	1.52 (1.49–1.56)	< 0.001	1.15 (1.13–1.16)	< 0.001
2020	1.68 (1.64–1.73)	< 0.001	1.13 (1.12–1.15)	< 0.001
2021	1.97 (1.92–2.03)	< 0.001	1.32 (1.30–1.34)	< 0.001
2022	2.17 (2.11–2.23)	< 0.001	1.44 (1.42–1.46)	< 0.001

Notes: Values represent exponentiated coefficients from adjusted GEE model with log link and gamma distribution and are interpreted as multiplicative effects on the mean drug and nondrug (i.e., total Medicare payments minus Part D Medicare payments) payments (n = 111,144). Abbreviations: ACO: accountable care organizations; CC: chronic conditions; MCC: multiple chronic conditions (2–4 cc); MM: multisystem morbidity (5+ cc); No.: number; SVI: social vulnerability index; Q: quartile.

In contrast to drug payments, ACO enrollment was associated with approximately 5% lower nondrug Medicare payments (0.95; 0.92–0.99). Compared with beneficiaries without MCC, the presence of MCC or MM was associated with substantially higher nondrug payments, with approximately 2.5-fold higher payments among those with MCC and 6-fold higher payments among those with MM. Having MCC or MM had a more pronounced impact in reducing nondrug payments among ACO enrollees compared to non-ACO beneficiaries with no MCC (p = 0.023 in a joint Wald test).

Figure 2 shows predicted drug and nondrug payments per person per year by ACO enrollment status across the three comorbidity groups. Prescription drug payments continuously increased throughout the study period, while nondrug payments peaked in the diagnosis year and decreased until they hit the pandemic period. ACO enrollees with no MCC had consistently higher predicted drug payments than non-ACO beneficiaries with no MCC, with payment differences ranging from $249 to $422. In contrast, ACO enrollees with MCC or MM encountered slightly lower drug payments than non-ACO beneficiaries with the same range of chronic condition counts. The payment differences ranged from $55 to $93 among those with MCC and $70 to $119 among those with MM.

FIGURE 2.

Estimated prescription drug and nondrug payments per person per year by ACO enrollment and comorbidity status for Medicare FFS beneficiaries with new diagnosis of ADRD in 2017. Notes: Figures represent postestimated predicted mean prescription drug and nondrug payments per person per year by ACO enrollment and comorbidity status among fee-for-service beneficiaries newly diagnosed with ADRD in 2017 after GEE model. Part D Medicare payments were adjusted by the CPI – prescription drug rates. Nondrug Medicare payments were calculated by total Medicare payments minus Part D payments. Total Medicare payments were adjusted by the CPI – Medical Care rates using 2022 as the base year (AHRQ, 2024). Abbreviations: ACO: accountable care organizations; MCC: multiple chronic conditions (2–4 cc); MM: multisystem morbidity (5+ cc).

In nondrug payments, ACO enrollees consistently spent less than non-ACO beneficiaries, regardless of the presence of MCC or MM. The payment differences were the highest among ADRD patients with MM, followed by those with MCC, and those without MCC. Differences in nondrug payments between ACO and non-ACO enrollees were largest in the diagnosis year, amounting to $359 for ADRD patients with no MCC, $1,666 for those with MCC, and $3,670 for those with MM. These payment differences also widened during 2021 and 2022. The yearly drug and nondrug payment differences are available in the Supplemental Material.

DISCUSSION

To our knowledge, this paper is the first to examine the long-term effect of ACOs on Part D and nondrug spending among ADRD patients. This study used a unique retrospective longitudinal dataset among older FFS beneficiaries newly diagnosed with ADRD and tracked their spending on drug and nondrug Medicare payments one year prior and five years after the ADRD diagnosis. Despite greater comorbidity burden among ACO enrollees at baseline, we found that ACO enrollment was consistently associated with lower drug and nondrug payments in newly diagnosed ADRD patients even after adjusting for sociodemographic and clinical factors, suggesting more efficient delivery of care from ACOs. The findings add to the limited empirical evidence on the associations between ACO enrollment and Medicare payments among older adults with ADRD, who have disproportionately high healthcare spending.

In particular, our results showed that prescription drug payments for newly diagnosed ADRD patients increased over time throughout the study period after controlling for the presence of multicomorbidities, race/ethnicity, gender, age, and neighborhood social vulnerabilities. This finding is consistent with the earlier hypothesis that more medications will be needed to care for the deteriorating health conditions of older adults. Increased prescription drug costs are not necessarily harmful if they can be justified by reducing overall costs of care.^19,33,34 For example, Roebuck et al.³³ found that increased drug costs from improved medication adherence generated substantial total savings due to reductions in hospitalizations and ED visits among individuals with four chronic vascular diseases. Our study cannot suggest that higher drug payments are beneficial, and more research is needed to link this pattern to other utilization and health outcomes.

The trends of predicted nondrug payments were also consistent with the prior hypothesis that the spending would reach the highest point in the ADRD diagnosis year due to expensive diagnostic tests and increased ambulatory visits.^3,35 ADRD patients incurred temporal and large incremental Medicare spending in the year of clinical diagnosis, regardless of sociodemographic characteristics, due to multiple diagnostic testing involving costly neuroimaging tests, such as CT, MRI, or PET scans, unplanned hospitalizations, and postacute care services.^36,37 The predicted nondrug payments varied substantially across different comorbidity groups. ADRD is frequently tested for and diagnosed during costly acute care or postacute care from adverse events.³⁶ Healthcare providers would often confirm the diagnosis once cognitive decline progresses to significantly interfere with patients’ self-care, leading to sudden flare-ups of conditions that would otherwise be manageable in an outpatient care setting.³⁶ Thus, the higher payments borne by those with MCC or MM in the prediagnosis year may reflect more severe health conditions.

Overall, ACO enrollees were found to be associated with lower Part D drug and nondrug payments across comorbidity groups, except for beneficiaries without MCC. Differences in nondrug payments between ACO and non-ACO enrollees were the greatest in the diagnosis year, while differences in prescription drug payments grew over time among those with MCC or MM. In contrast, ACO enrollees with no MCC consistently spent more than non-ACO counterparts throughout the study period. The observed divergence between unadjusted and adjusted drug payment trends was shown among ADRD patients with no MCC. While ACO enrollees with no MCC exhibited lower drug payments than their non-ACO counterparts in the unadjusted comparisons, predicted drug payments were higher among ACO enrollees. This pattern likely reflects differences in baseline risk profiles and care processes. Unadjusted estimates may capture favorable baseline characteristics among ACO enrollees, whereas adjusted estimates compare clinically similar individuals and may reflect greater use of guideline-concordant pharmacologic therapy and more proactive medication management within ACOs. Among ADRD patients with MCC or MM, medication use is likely already extensive, and risk adjustment may attenuate observable differences in drug spending. Together, these findings suggest that ACOs perform better in reducing both drug and nondrug payments for patients with more complex healthcare needs, such as ADRD patients with coexisting MCC or MM. The results are encouraging in that ACO providers may prioritize their resources in care for more complex patients.¹

Although there is no cure for ADRD and the new disease-modifying drugs are administered by providers in Part B, Part D still covers a broad range of prescription medications that help manage ADRD symptoms, such as donepezil and memantine.³ Although MSSP ACOs are primarily responsible for managing Medicare Part A and Part B payments,^1,7 more and more ACOs have taken proactive steps to improve prescribing practices,^7,8,17,38 potentially reducing drug-related payments. Such efforts include the provision of prescribing guidelines that highlight lower-cost generic medications to participating clinicians, improving treatment adherence through better care coordination, and having clinical pharmacists review complex medication regimens for high-risk patients, reducing duplications or switching to effective generic drugs.^1,7,13,17,39 ACOs also employ medication reconciliation at the transition of care to prevent unnecessary prescriptions and adverse drug events.^17,38,40 In addition, many of the ACO quality metrics involve medication optimization, such as for diabetes and cardiovascular care, and meeting these measures indirectly encourages more efficient, cost-effective drug use.^17,39,41 All of these initiatives could potentially have contributed to lower prescription drug payments observed among ACO enrollees.

Beyond medication management, MSSP ACOs may also reduce nondrug Medicare payments through improved care coordination and transitional care, which can help prevent avoidable hospitalizations and readmissions which are key drivers of inpatient spending.^1,6,10 Enhanced coordination across care settings may reduce duplicative or low-value diagnostic testing, particularly during transitions between outpatient, inpatient, and postacute care.^6,10,11 Additionally, proactive management of comorbidities may reduce the need for costly inpatient and diagnostic services.^1,6–8

The COVID-19 pandemic represented a major shock to Medicare utilization and spending and may have influenced the long-term patterns observed in this study. Pandemic-related disruptions, such as reductions in elective care, changes in ambulatory service use, shifts in medication utilization, and rapid expansion of telehealth, likely affected both Part D and nondrug payments.^42,43 Predicted payment trends showed increases in both drug and nondrug spending during the pandemic period. Notably, however, ACO enrollees consistently exhibited lower predicted drug and nondrug payments than non-ACO beneficiaries throughout the COVID-19 period, suggesting that care coordination and population health management infrastructure within MSSP ACOs may have supported more efficient care delivery even under pandemic-related constraints.^42,44

There are several limitations in this study. First, the study sample was limited to Medicare beneficiaries continuously enrolled in FFS. The results are not generalizable to Medicare advantage (MA) enrollees with ADRD and other types of ACOs. ACOs may vary widely in care efficiency across participating providers,⁴⁵ and such differences must be accounted for in future studies. Next, our study is subject to limitations inherent in large administrative and claims-based datasets. Given the large sample size, small absolute differences in sample characteristics may achieve statistical significance without being clinically meaningful. Accordingly, some statistically significant findings should be interpreted with caution. In addition, the comorbidity groups were categorized based on the number of conditions each ADRD patient had among 12 broad, prevalent conditions in older adults with ADRD.^18,27,28 No weights for the severity of conditions were used in this study. In addition, we were unable to distinguish subtypes of dementia or assess dementia severity. Differences in clinical trajectories across dementia subtypes may have influenced patterns of healthcare use and spending over the follow-up period. Moreover, though the study controlled for overall social vulnerabilities at the county level, the study could not control for individual-level socioeconomic differences that could result in unobserved heterogeneity issues. Lastly, while MSSP ACO attribution is determined by CMS based on primary care utilization rather than beneficiary choice, we could not fully account for potential provider-level selection or unobserved differences in provider characteristics associated with ACO participation.

The observational period of this study predates the FDA’s approval and early adoption of new dementia drugs that treat the underlying causes of Alzheimer’s disease. In addition to the annual drug price of $26,500, use of new dementia drugs will increase frequency of diagnostic testing, brain scans, and other safety monitoring services, incurring $82,500 per year.⁴⁶ As these treatments become more widely implemented, future studies are needed to examine how emerging disease-modifying treatments influence drug and nondrug Medicare spending patterns and interact with ACOs. As the prevalence of ADRD continues to rise, optimizing care models to enhance cost efficiency without compromising quality is essential. This study highlights the potential of ACOs to curb healthcare spending while potentially improving care coordination for this clinically vulnerable population. Unfortunately, ACOs are less likely to enroll and retain ADRD patients than other patients.⁹ Further incentives may be needed to encourage ACOs to take a more active role in managing prescription drug use, particularly given its significant contribution to overall healthcare costs.^19,33,34

Supplementary Material

Supplementary material associated with this article can be found in the online version at https://doi.org/10.1016/j.jagp.2026.01.014.

Highlights.

What is the primary question addressed by this study?

The study examined whether enrollment in Accountable Care Organizations (ACOs) is associated with differences in prescription drug and nondrug Medicare payments among older adults newly diagnosed with ADRD.

What is the main finding of this study?

Beneficiaries with ADRD enrolled in ACOs had significantly lower drug and nondrug Medicare payments, with the largest payment differences observed among those with multiple chronic conditions.

What is the meaning of the finding?

The findings of this study highlight the potential of ACOs in lowering healthcare spending for clinically vulnerable population and the need for direct incentives for ACOs to manage prescription drug use.

DATA SHARING STATEMENT

The Medicare data used in this study cannot be shared by the authors but may be obtained from CMS through a formal data use agreement.