Remote Patient Monitoring Adoption for Hypertension Management Among Medicare Beneficiaries

This cohort study compares adoption of remote patient monitoring among Medicare beneficiaries who continued a fee-for-service plan vs those who switched to a Medicare Advantage plan.

Key Points

Question

Among Medicare beneficiaries with hypertension who switched from Medicare fee-for-service to Medicare Advantage, was switching associated with changes in remote patient monitoring adoption, clinician continuity, and hypertension-related acute care use?

Findings

In this cohort study using data from 281 620 Medicare beneficiaries from 2016 to 2022, matched beneficiaries who switched to Medicare Advantage had lower remote patient monitoring adoption, greater clinician loss without replacement, and higher hypertension-related emergency department visits and hospitalizations compared with matched beneficiaries who remained in fee-for-service.

Meaning

These findings suggest that transitions into Medicare Advantage may require continuity safeguards and clearer incentives to support remote monitoring and hypertension management.

Abstract

Importance

Remote patient monitoring (RPM), including self-measured blood pressure monitoring with clinician review and telehealth-supported feedback, can support hypertension management. However, RPM use and related care continuity after switching from Medicare fee-for-service (FFS) to Medicare Advantage (MA) remain unclear.

Objective

To compare RPM adoption, clinician continuity, and hypertension-related acute care utilization among beneficiaries who remained in Medicare FFS vs switched to MA plans, categorized as value-based contract (VBC) proxy or non-VBC.

Design, Setting, and Participants

This cohort study with an observational difference-in-differences design with propensity score matching used data from 2016 to 2022 Medicare enrollment, FFS claims, and MA encounter data. Beneficiaries were aged 65 years or older with prevalent diagnosed hypertension in 2018 and continuous enrollment in Parts A and B in 2018. Treated groups switched from FFS to MA in January 2019 and remained enrolled through 2022; comparators remained in FFS. Follow-up extended from January 1, 2019, through December 31, 2022. Data analysis was conducted from April to July 2025.

Exposure

Switching from Medicare FFS to MA-VBC proxy or MA non-VBC in 2019.

Main Outcomes and Measures

The primary outcome was annual RPM adoption during hypertension-related visits; secondary outcomes included clinician loss without replacement, clinician switching or substitution, and hypertension-related emergency department (ED) visits and hospitalizations.

Results

Matched samples included 281 620 beneficiaries, with 46 833 MA-VBC proxy plan switchers and 46 833 FFS comparators (27 920 [59.6%] aged 71 years or older and 27 685 female [59.1%] in each group) and 93 977 MA non-VBC switchers and 93 977 FFS comparators (67 188 [71.5%] aged 71 years or older; 53 122 female [56.5%] in each group). Common comorbidities included diabetes, chronic kidney disease, and heart failure. Switching to MA was associated with lower RPM adoption in 2022 (MA-VBC proxy: odds ratio [OR], 0.55; 95% CI, 0.42-0.72; −0.63 percentage points; non-VBC: OR, 0.73; 95% CI, 0.54-0.99; −0.52 percentage points), greater clinician loss without replacement (MA-VBC proxy: OR, 1.27; 95% CI, 1.23-1.32; 3.41 percentage points; MA non-VBC: OR, 1.09; 95% CI, 1.06-1.12; 0.83 percentage points), and higher hypertension-related hospitalizations (MA-VBC proxy: OR, 1.75; 95% CI, 1.48-2.06; MA non-VBC: OR, 1.94; 95% CI 1.71-2.19; 1.56 percentage points in both comparisons). Event-study analyses showed postswitch divergence through 2022.

Conclusions and Relevance

In this cohort study of older Medicare beneficiaries with hypertension, switching from FFS to MA was associated with lower RPM adoption, greater clinician discontinuity, and higher hypertension-related acute care use. These findings suggest that continuity safeguards and clearer payment or quality incentives during MA transitions may support remote monitoring and clinician follow-up for hypertension.

Introduction

High blood pressure affects 119.9 million US adults and is a leading preventable risk factor for cardiovascular disease (CVD).^1,2 Remote patient monitoring (RPM), including self-measured blood pressure monitoring with clinician review and telehealth-supported feedback, has been associated with improved hypertension control and reduced clinic visits.^3,4,5,6,7,8 The Centers for Medicare & Medicaid Services (CMS) reimbursement policies created a new pathway for wider use of digital health technologies in Medicare.^9,10,11 RPM billing generally requires an eligible Medicare beneficiary, a condition warranting physiologic monitoring, patient consent, use of a device that automatically collects and transmits physiologic data, regular data collection during a billing period, and clinician management with interactive communication when required. Because RPM often depends on an established clinician-patient relationship, device setup, follow-up, and workflow integration, insurance transitions that disrupt clinician continuity may affect whether beneficiaries receive RPM.

During the same period, Medicare Advantage (MA) enrollment increased substantially.¹² Switching from Medicare fee-for-service (FFS) to MA may alter access to clinicians because MA plans commonly use managed networks and prior authorization tools.¹³ These transitions may require beneficiaries to reestablish in-network care, which could delay or interrupt hypertension management and reduce opportunities to initiate RPM.¹⁴ Payment incentives may also operate in competing directions.¹⁵ Per-service FFS reimbursement may make RPM billing more direct,¹⁴ whereas MA and value-based arrangements may encourage RPM if it reduces high-cost acute care but may discourage adoption if RPM is treated as an operational expense or requires additional administrative steps.¹⁶ Thus, the expected association of MA plan type with RPM adoption is uncertain.

Prior studies comparing MA and Medicare FFS report mixed findings on quality, service use, and outcomes,^17,18 and evidence on RPM adoption across payment models remains limited.^19,20 Moreover, the impact of switching plans on clinician continuity, a known determinant of chronic disease management, has not been examined in the context of digital care. In this study, beneficiaries were continuously enrolled in Medicare FFS in 2018 before switching to MA or remaining in FFS in January 2019. Leveraging a difference-in-differences (DiD) approach with propensity score matching, we assessed the association of switching from Medicare FFS to MA with RPM use, clinician continuity, and acute care utilization among people with prevalent diagnosed hypertension.

Methods

Data and Sample

This cohort study was determined to be non–human participants research by New York University Langone Health institutional review board; as such, informed consent was not required. Reporting follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Using 2016 to 2022 Medicare data from the CMS Chronic Conditions Warehouse Virtual Research Data Center, we analyzed Medicare enrollment files, FFS claims (Parts A and B), and MA encounter data (Part C).²¹ We defined a 2018 baseline cohort of beneficiaries with prevalent diagnosed hypertension using International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes I10 to I15, identified by 1 or more inpatient, skilled nursing facility, or home health claims, or 2 or more outpatient or carrier claims at least 30 days apart in 2018 claims. Inclusion criteria were being aged 65 years or older on January 1, 2018; residence in the 50 US states or District of Columbia; no death date through December 31, 2022; and continuous enrollment in Medicare FFS Parts A and B in 2018, ensuring a common preswitch baseline. We further restricted the cohort to individuals continuously enrolled in FFS or MA plans throughout 2019 to 2022 (eFigure 1 in Supplement 1). Kidney failure or end-stage kidney disease was not an exclusion criterion; kidney disease indicators were included as baseline covariates. Race and ethnicity were obtained from Medicare enrollment files, classified administratively rather than by study investigators, and included to describe the cohort and support exact matching; race and ethnicity categories included Asian American or Pacific Islander, non-Hispanic Black, Hispanic, non-Hispanic White, and other or unknown (beneficiaries coded as American Indian or Alaska Native, other, or unknown).

We defined 3 mutually exclusive study groups based on Medicare enrollment status as of January 1, 2019: (1) MA value-based contracting (VBC)–proxy, defined as beneficiaries who transitioned from Medicare FFS in January 2019 to an MA plan classified using a proxy for VBC; (2) MA non-VBC, defined as beneficiaries who transitioned from Medicare FFS in January 2019 to an MA plan not classified as VBC by this proxy; and (3) Medicare FFS (comparison group), defined as beneficiaries who remained continuously enrolled in Medicare FFS throughout the study period. MA value-based arrangements could not be observed directly because clinician-level contract data were unavailable in the claims and encounter files. We therefore used monthly MA plan type codes from the Medicare Beneficiary Summary File as a plan-type proxy for MA products more likely to operate with risk-based or quality-incentive arrangements.²¹ Plans coded as Health Maintenance Organization, Health Maintenance Organization Point-of-Service, Medicare-Medicaid Plan, Programs of All-Inclusive Care for the Elderly, or Regional Preferred Provider Organization were classified as the value-based plan-type proxy; local preferred provider organization and other nonclassified MA products were categorized as non–value-based plan-type proxy.^12,22 Because plan type does not perfectly identify the payment contract between a plan and its clinicians, results by VBC category should be interpreted as exploratory and subject to misclassification.

Outcomes

The primary outcome was RPM adoption, defined as a binary indicator of whether a beneficiary received at least 1 RPM service for hypertension management during a hypertension-related outpatient visit in a calendar year from 2019 through 2022. RPM services were identified using CMS-recognized Current Procedural Terminology and Healthcare Common Procedure Coding Systemcodes²³ and a telehealth identification algorithm (eTable 1 in Supplement 1). To enhance clinical specificity, the RPM claim was required to occur on the same date as a visit with a primary diagnosis of hypertension (ICD-10-CM codes I10-I15). Therefore, RPM use was captured in the context of hypertension management rather than other physiologic monitoring.

Secondary outcomes included clinician continuity and acute care utilization. Clinician continuity was assessed relative to each beneficiary’s 2018 observed clinician panel, defined using individual-level National Provider Identifiers (NPIs) from claims and encounter records. This measure reflects observed treating clinicians who were responsible for hypertension care rather than the full contracted network and was not restricted to a single specialty or a single assigned primary care clinician. For 2019 to 2022, we classified continuity into clinician loss without replacement, defined as the absence of at least one 2018 clinician with no new clinician added, and clinician switching or substitution, defined as the loss of at least one 2018 clinician accompanied by the addition of at least 1 new clinician. We did not measure plan network size because network directory or contracted-clinician data were not available in the analytic files.

Acute care utilization included ED visits and hospitalizations with a primary diagnosis of hypertension (ICD-10-CM codes I10-I15) or major CVD (ICD-10-CM codes I00-I78), identified from inpatient and outpatient claims or encounters. Restricting outcomes to the primary diagnosis was intended to capture encounters for which hypertension or CVD was the principal reason for acute care,^24,25 and to reduce differential coding concerns across payment models. Baseline acute hospitalization and emergency department (ED) visit variables used in matching were all-cause utilization measures.

Statistical Analysis

Because beneficiaries who switch to MA may differ systematically from those who remain in Medicare FFS, we used propensity score matching followed by DiD models. All analyses were conducted from April to July 2025 in the CMS Virtual Research Data Center using SAS software version 9.4 (SAS Institute Inc), in a secure deidentified research environment.

Propensity Score Matching

We estimated the probability of switching from FFS in 2018 to either a MA-VBC proxy plan or a MA non-VBC plan in 2019 using separate multivariable logistic regression models. The models included baseline 2018 covariates: age in 5-year bands, sex, race and ethnicity, dual Medicare-Medicaid eligibility, urbanicity, US Census division, comorbidities derived from 2017-2018 claims,²⁶ baseline all-cause ED visits, and acute hospitalizations. We performed exact matching on age, sex, race and ethnicity, and urbanicity, followed by 1:1 nearest-neighbor matching within exact-match strata using a caliper width of 0.2 SDs of the logit-transformed propensity score. Any unmatched beneficiaries who did not switch from FFS to MA plans were excluded. Common support for matching and its balance were assessed using logit-transformed propensity score density plots before and after matching and standardized mean differences with love plots.²⁷

Difference-in-Differences Estimation

Using the matched cohorts, we implemented a DiD model to estimate the impact of switching to MA on the outcomes of interest:

where y_it is the binary outcome for individual i in year t, α is the intercept, Treatment_i indicates if individual i switched to MA VBC or non-VBC plans in 2019 (β is its coefficient). Year_t are year fixed effects (2019-2022; 2018 is the reference), and γ_t are their coefficients, ƍ_t are DiD estimates for each year, X′_i is a vector of baseline covariates (Ǿ is the coefficient), and ε_it is the error term, with robust SEs clustered at the beneficiary level to account for within-person correlation over time.

We reported both odds ratios (ORs) and average marginal effects (ie, perceived differences in percentage points) to improve interpretability, with the latter showing the absolute change in the probability of RPM adoption. To assess the parallel-trends assumption, we also estimated an event-study version of the model (2016-2022; 2018 is the reference) and conducted joint Wald test that the preperiod treatment-by-year coefficients (2016-2017) equals 0 for outcomes other than provide change. Two-sided P values are reported, with P < .05 considered significant.

Dynamic Effects and Robustness Checks

To evaluate the timing and persistence of effects, we constructed event-study plots covering 2016 to 2022, using 2018 as the reference year. These plots display year-specific estimates from DiD models. For outcomes with sufficient preintervention variation, such as clinician continuity and acute care utilization, we tested for parallel pretrends (2016-2018) to assess the validity of the identifying assumptions, by t test (clinician continuity) or joint Wald test (acute care utilization), that the preintervention treatment-by-year coefficients equal 0. Two-sided P values were reported, with P < .05 considered significant. For RPM, pre-2019 billing activity was minimal due to the absence of dedicated procedure codes, limiting pretrend testing.

We conducted multiple sensitivity analyses and robustness checks. These included alternative continuity definitions, such as clinician gain without loss and any clinician change, to assess whether findings reflected network disruption or broader care instability. We repeated analyses using organization-level NPIs. We also examined preswitch trends in hypertension-related and CVD-related ED visits and hospitalizations from 2016 through 2018. To assess sensitivity to unmeasured confounding, we calculated E-values for the primary 2022 RPM associations and selected acute care outcomes. E-values quantify the minimum magnitude of association that an unmeasured confounder would need to have with both MA switching and the outcome, conditional on measured covariates, to explain away the observed association.²⁸

Results

Sample Characteristics

Before matching, the 3 groups differed in baseline characteristics. After 1:1 propensity score matching, the matched analytic samples included 281 620 beneficiaries, with 46 833 beneficiaries who switched to the MA-VBC proxy and 46 833 matched FFS beneficiaries (27 920 [59.6%] aged 71 years or older and 27 685 female [59.1%] in each group; 2 beneficiaries were unmatched and excluded) and 93 977 beneficiaries who switched to the MA non-VBC proxy and 93 977 matched FFS beneficiaries (67 188 [71.5%] aged 71 years or older; 53 122 female [56.5%] in each group; 1 beneficiary was unmatched and excluded). Across matched groups, common comorbidities included diabetes, chronic kidney disease, and heart failure. Propensity score density plots indicated adequate common support before the matching and improved overlap within the region of common support after matching for both comparisons. Matching improved balance across observed characteristics, with all standardized mean differences less than 0.10 (Table 1; eTable 2 and eFigures 2-3 in Supplement 1).

Table 1. Baseline Characteristics of Propensity Score–Matched Beneficiaries in 2018 by Study Group^a.

Characteristic	Beneficiaries, No. (%)b		SMDc	Beneficiaries, No. (%)b		SMDc
	FFS matched to MA-VBC (n = 46 833)	MA-VBC (n = 46 833)		FFS matched to MA non-VBC (n = 93 977)	MA non-VBC (n = 93 977)
Age, y
≤70	18 913 (40.4)	18 913 (40.4)	0.000	26 789 (28.5)	26 789 (28.5)	0.000
71-75	12 876 (27.5)	12 876 (27.5)	0.000	27 784 (29.6)	27 784 (29.6)	0.000
76-80	8491 (18.1)	8491 (18.1)	0.000	19 870 (21.1)	19 870 (21.1)	0.000
81-85	4515 (9.6)	4515 (9.6)	0.000	11 922 (12.7)	11 922 (12.7)	0.000
>85	2038 (4.4)	2038 (4.4)	0.000	7612 (8.1)	7612 (8.1)	0.000
Sex
Male	19 148 (40.9)	19 148 (40.9)	0.000	40 855 (43.5)	40 855 (43.5)	0.000
Female	27 685 (59.1)	27 685 (59.1)	0.000	53 122 (56.5)	53 122 (56.5)	0.000
Race and ethnicityd
Non-Hispanic Asian American or Pacific Islander	1861 (4.0)	1861 (4.0)	0.000	1169 (1.2)	1169 (1.2)	0.000
Non-Hispanic Black	6969 (14.9)	6969 (14.9)	0.000	7944 (8.5)	7944 (8.5)	0.000
Hispanic	4767 (10.2)	4767 (10.2)	0.000	3162 (3.4)	3162 (3.4)	0.000
Non-Hispanic White	32 109 (68.6)	32 109 (68.6)	0.000	79 222 (84.3)	79 222 (84.3)	0.000
Other or unknown	1127 (2.4)	1127 (2.4)	0.000	2480 (2.6)	2480 (2.6)	0.000
Dual Medicare and Medicaid coverage
No dual enrollment	35 686 (76.2)	35 106 (75.0)	−0.032	89 990 (95.8)	90 528 (96.3)	0.021
Partial or full dual enrollment	11 147 (23.8)	11 727 (25.0)	0.032	3987 (4.2)	3449 (3.7)	−0.021
Urbanicity of residence
Metropolitan area	38 637 (82.5)	38 637 (82.5)	0.000	78 635 (83.7)	78 635 (83.7)	0.000
Micropolitan area	4974 (10.6)	4974 (10.6)	0.000	9133 (9.7)	9133 (9.7)	0.000
Rural area	3222 (6.9)	3222 (6.9)	0.000	6209 (6.6)	6209 (6.6)	0.000
US Census division
New England	2032 (4.3)	2008 (4.3)	−0.002	5692 (6.1)	5867 (6.2)	0.008
Mid-Atlantic	4274 (9.1)	4347 (9.3)	0.005	29 040 (30.9)	27 796 (29.6)	−0.033
South Atlantic	8902 (19.0)	8991 (19.2)	0.005	14 705 (15.6)	14 642 (15.6)	−0.002
East North Central	8490 (18.1)	8355 (17.8)	−0.008	11 466 (12.2)	11 571 (12.3)	0.003
East South Central	4226 (9.0)	4313 (9.2)	0.007	4148 (4.4)	4129 (4.4)	−0.001
West North Central	2298 (4.9)	2533 (5.4)	0.021	8444 (9.0)	8438 (9.0)	−0.000
West South Central	6913 (14.8)	7121 (15.2)	0.013	4308 (4.6)	4333 (4.6)	0.001
Mountain	3128 (6.7)	3047 (6.5)	−0.007	10 781 (11.5)	11 496 (12.2)	0.027
Pacific	6570 (14.0)	6118 (13.1)	−0.029	5393 (5.7)	5705 (6.1)	0.012
Comorbidity conditions
Acute myocardial infarction	436 (0.9)	532 (1.1)	0.019	862 (0.9)	987 (1.1)	0.012
Alzheimer disease	519 (1.1)	527 (1.1)	0.002	898 (1.0)	1012 (1.1)	0.011
Anemia	9366 (20.0)	9656 (20.6)	0.015	20 298 (21.6)	20 801 (22.1)	0.013
Asthma	3792 (8.1)	4239 (9.1)	0.033	7842 (8.3)	8706 (9.3)	0.032
Atrial fibrillation and flutter	6334 (13.5)	6619 (14.1)	0.016	16 127 (17.2)	16 906 (18.0)	0.021
Benign prostatic hyperplasia	5785 (12.4)	5860 (12.5)	0.005	15 097 (16.1)	15 290 (16.3)	0.006
Cancer	4689 (10.0)	4893 (10.4)	0.013	12 912 (13.7)	13 380 (14.2)	0.014
Cataract	13 681 (29.2)	13 933 (29.8)	0.011	34 417 (36.6)	34 548 (36.8)	0.003
Chronic kidney disease	10 163 (21.7)	10 348 (22.1)	0.010	18 522 (19.7)	19 324 (20.6)	0.021
Chronic obstructive pulmonary disease	7305 (15.6)	7832 (16.7)	0.031	11 714 (12.5)	12 672 (13.5)	0.029
Depression	8855 (18.9)	9070 (19.4)	0.012	13 653 (14.5)	14 644 (15.6)	0.028
Diabetes	20 826 (44.5)	21 002 (44.8)	0.008	34 817 (37.0)	35 004 (37.2)	0.004
Glaucoma	6902 (14.7)	7023 (15.0)	0.007	17 612 (18.7)	17 779 (18.9)	0.005
Heart failure	5796 (12.4)	6318 (13.5)	0.032	10 557 (11.2)	11 708 (12.5)	0.036
Hip or pelvic fracture	297 (0.6)	313 (0.7)	0.004	584 (0.6)	736 (0.8)	0.017
Hyperlipidemia	38 510 (82.2)	38 526 (82.3)	0.001	79 893 (85.0)	79 772 (84.9)	−0.004
Hypothyroidism	10 468 (22.4)	10 520 (22.5)	0.003	24 186 (25.7)	24 781 (26.4)	0.014
Ischemic heart disease	13 431 (28.7)	14 075 (30.1)	0.030	29 162 (31.0)	29 634 (31.5)	0.011
Non-Alzheimer dementia	1536 (3.3)	1661 (3.5)	0.014	2479 (2.6)	2833 (3.0)	0.020
Osteoporosis with/without pathological fracture	4360 (9.3)	4445 (9.5)	0.006	9768 (10.4)	10 252 (10.9)	0.016
Parkinson disease and secondary parkinsonism	312 (0.7)	405 (0.9)	0.019	825 (0.9)	946 (1.0)	0.012
Pneumonia, all-cause	1388 (3.0)	1623 (3.5)	0.026	2552 (2.7)	2959 (3.1)	0.023
Rheumatoid arthritis or osteoarthritis	18 897 (40.3)	19 251 (41.1)	0.015	40 393 (43.0)	41 120 (43.8)	0.016
Stroke or transient ischemic attack	2885 (6.2)	3068 (6.6)	0.015	5924 (6.3)	6591 (7.0)	0.027
Health care utilization
All-cause hospitalization	6584 (14.1)	7386 (15.8)	0.045	14 232 (15.1)	15 236 (16.2)	0.028
All-cause emergency department visit	12 606 (26.9)	13 204 (28.2)	0.028	22 543 (24.0)	23 570 (25.1)	0.025

Abbreviations: FFS, fee-for-service; MA, Medicare Advantage; SMD, standardized mean difference; VBC, value-based contract.

^a Propensity scores were estimated using logistic regression including age, sex, race and ethnicity, US region, dual eligibility, urbanicity, comorbidities, baseline all-cause acute hospitalization and emergency department visits, and baseline utilization and medications measured in 2018; a 1:1 nearest-neighbor matching was performed within a 0.2 SD caliper of the logit propensity score.

^b The matched baseline cohort is defined in 2018.

^c SMDs less than 0.100 indicates adequate balance.

^d Race and ethnicity were obtained from Medicare enrollment files. Other or unknown includes beneficiaries coded as other, unknown, or American Indian or Alaska Native.

RPM Adoption

RPM use was negligible in the preswitch year (2018); among 7 676 950 beneficiaries with diagnosed hypertension, 22 666 (0.3%) received any RPM services across all groups, consistent with the absence of dedicated CMS billing codes before 2019. After code implementation, RPM use increased in all cohorts but remained low and a persistent gap emerged between FFS and MA enrollees. By 2022, 813 FFS beneficiaries (1.7%) had received RPM during a hypertension-related visit compared with 560 (1.2%) in the MA-VBC proxy comparison; in the non–value-based plan-type proxy comparison, 1368 FFS beneficiaries (1.5%) vs 882 MA beneficiaries (0.9%) received RPM. In 2022, DiD estimates showed lower odds of RPM uptake after switching to either MA type (MA-VBC proxy vs FFS: OR, 0.55; 95% CI, 0.42-0.72; −0.63 percentage points; MA non-VBC vs FFS: OR, 0.73; 95% CI, 0.54-0.99; −0.52 percentage points) (Figure 1 and Table 2).

Figure 1. Event-Study Difference-in-Differences Estimates for Remote Patient Monitoring Adoption in Hypertension Care, Medicare Advantage (MA) vs Fee-for-Service (FFS), 2018-2022.

Remote patient monitoring codes were introduced in 2019; no preperiod event-study was possible. Odds ratios are represented with dots, while error bars represent 95% CIs. The blue shading for the year 2018 indicates the preswitch reference year. VBC indicates value-based contract.

Table 2. Adjusted DiD Outcomes by Year, MA (VBC and Non-VBC) vs FFS, 2019-2022.

Outcome by year	MA-VBC vs FFS, OR (95% CI)a	P value	DiD, percentage points	MA non-VBC vs FFS, OR (95% CI)a	P value	DiD, percentage points
Primary outcome: RPM adoption for primary hypertension careb
2019	0.79 (0.61 to 1.02)	.07	−0.07	0.62 (0.45 to 0.85)	.003	−0.08
2020	0.66 (0.50 to 0.87)	.003	−0.21	0.63 (0.46 to 0.87)	.004	−0.24
2021	0.60 (0.46 to 0.79)	<.001	−0.40	0.77 (0.57 to 1.05)	.09	−0.31
2022	0.55 (0.42 to 0.72)	<.001	−0.63	0.73 (0.54 to 0.99)	.04	−0.52
Secondary outcomes
Clinician loss without replacementc
2019	1.45 (1.40 to 1.50)	<.001	4.75	1.22 (1.19 to 1.25)	<.001	2.21
2020	1.30 (1.26 to 1.34)	<.001	3.84	1.12 (1.09 to 1.14)	<.001	1.30
2021	1.28 (1.24 to 1.33)	<.001	3.47	1.10 (1.08 to 1.13)	<.001	1.04
2022	1.27 (1.23 to 1.32)	<.001	3.41	1.09 (1.06 to 1.12)	<.001	0.83
Clinician switching or substitution
2019	0.84 (0.77 to 0.92)	<.001	−0.62	0.94 (0.88 to 1.01)	.11	−0.19
2020	0.85 (0.78 to 0.92)	<.001	−0.63	0.96 (0.89 to 1.02)	.18	−0.23
2021	0.79 (0.72 to 0.85)	<.001	−1.08	0.97 (0.91 to 1.03)	.35	−0.30
2022	0.77 (0.71 to 0.83)	<.001	−1.32	0.94 (0.88 to 1.00)	.05	−0.53
Hypertension-related ED visitd
2019	1.62 (1.44 to 1.82)	<.001	1.20	1.43 (1.30 to 1.56)	<.001	0.75
2020	1.82 (1.61 to 2.05)	<.001	1.40	1.41 (1.29 to 1.55)	<.001	0.68
2021	1.77 (1.58 to 1.99)	<.001	1.69	1.48 (1.36 to 1.62)	<.001	0.99
2022	1.74 (1.56 to 1.95)	<.001	2.03	1.56 (1.43 to 1.70)	<.001	1.49
CVD-related ED visitd
2019	1.16 (1.07 to 1.25)	<.001	0.88	1.10 (1.04 to 1.17)	<.001	0.54
2020	1.27 (1.17 to 1.37)	<.001	1.28	1.05 (0.99 to 1.11)	.08	0.23
2021	1.29 (1.20 to 1.40)	<.001	1.78	1.12 (1.06 to 1.19)	<.001	0.74
2022	1.24 (1.15 to 1.33)	<.001	1.86	1.19 (1.13 to 1.26)	<.001	1.41
Hypertension-related hospitalizationd
2019	2.53 (2.12 to 3.02)	<.001	1.53	3.11 (2.72 to 3.54)	<.001	1.62
2020	2.33 (1.95 to 2.79)	<.001	1.40	2.51 (2.19 to 2.87)	<.001	1.28
2021	2.29 (1.93 to 2.71)	<.001	1.85	2.41 (2.12 to 2.74)	<.001	1.65
2022	1.75 (1.48 to 2.06)	<.001	1.56	1.94 (1.71 to 2.19)	<.001	1.56
CVD-related hospitalizationd
2019	1.22 (1.12 to 1.33)	<.001	1.02	1.33 (1.25 to 1.41)	<.001	1.44
2020	1.25 (1.14 to 1.36)	<.001	1.06	1.21 (1.14 to 1.29)	<.001	0.89
2021	1.23 (1.13 to 1.34)	<.001	1.33	1.25 (1.18 to 1.33)	<.001	1.29
2022	1.12 (1.03 to 1.21)	.01	1.00	1.20 (1.13 to 1.27)	<.001	1.23

Abbreviations: CVD, cardiovascular disease; DiD, difference-in-differences; ED, emergency department; FFS, fee-for-service; MA, Medicare Advantage; OR, odds ratio; RPM, remote patient monitoring; VBC, value-based contract.

^a ORs and 95% CIs are from DiD logit models with robust SEs clustered at the beneficiary level. Percentage-point DiD estimates are average marginal effects from the same models. P values are 2-sided with α = .05.

^b The unit of analysis is the beneficiary-year. RPM is defined by relevant Current Procedural Terminology and Healthcare Common Procedure Coding System codes plus a telehealth identification algorithm (full code list in eTable 1 in Supplement 1).

^c Clinician outcomes are defined relative to each beneficiary 2018 National Provider Identifier–based provider panel.

^d ED and hospitalization outcomes are identified when hypertension or CVD is the primary diagnosis on the encounter claim.

Clinician Continuity

In 2022, beneficiaries who transitioned to the MA-VBC proxy were more likely to experience clinician loss without replacement than those who remained in Medicare FFS (OR, 1.27; 95% CI, 1.23-1.32; 3.41 percentage points). A smaller increase was observed among beneficiaries who switched to the MA non-VBC (OR, 1.09; 95% CI, 1.06-1.12; 0.83 percentage points). MA-VBC proxy switchers were also less likely to experience clinician switching or substitution, defined as losing at least 1 prior clinician while gaining at least 1 new clinician (OR, 0.77; 95% CI, 0.71-0.83; −1.32 percentage points) (Figure 2 and Table 2).

Figure 2. Event-Study Difference-in-Differences Estimates for Clinician Continuity, Medicare Advantage (MA) vs Traditional Medicare Fee-for-Service (FFS), 2016-2022.

Odds ratios are represented with dots, while error bars represent 95% CIs. The blue shading for the years 2017 and 2018 indicates the pooled preswitch reference period. VBC indicates value-based contract.

Acute Care Utilization

MA enrollees were more likely to use acute care after switching from Medicare FFS. In 2022, switching to the MA-VBC proxy or the MA non-VBC was associated with a 1.56–percentage-point increase in hypertension-related hospitalizations compared with remaining in Medicare FFS (MA-VBC: OR, 1.75; 95% CI, 1.48-2.06; MA non-VBC: OR, 1.94; 95% CI, 1.71-2.19). Similar patterns were observed for ED visits; switching to the MA-VBC proxy was associated with a 2.03–percentage-point increase in hypertension-related ED visits (OR, 1.74; 95% CI, 1.56-1.95) and a 1.86–percentage-point increase in CVD-related ED visits (OR, 1.24; 95% CI, 1.15-1.33), while switching to the MA non-VBC was associated with increases of 1.49 percentage points (OR, 1.56; 95% CI, 1.43-1.70) and 1.41 percentage points (OR, 1.19; 95% CI, 1.13-1.26) for hypertension-related and CVD-related ED visits, respectively (Figure 3 and Table 2).

Figure 3. Event-Study Difference-in-Differences Estimates for Hypertension-Related Utilization, Medicare Advantage (MA) vs Traditional Medicare Fee-for-Service (FFS), 2016-2022.

Odds ratios are represented with dots, while error bars represent 95% CIs. The blue shading for the year 2018 indicates the preswitch reference year. ED indicates emergency department; VBC, value-based contract.

Temporal Dynamics

For hypertension-related ED visits, joint Wald tests of the 2016 and 2017 interaction terms were not significant for MA-VBC proxy vs FFS (χ²₂ = 0.21; P = .90) or MA non-VBC vs FFS (χ²₂ = 4.21; P = .12). For hypertension-related hospitalizations, corresponding tests were also not significant for MA-VBC proxy vs FFS (χ²₂ = 0.88; P = .64) or MA non-VBC vs FFS (χ²₂ = 4.75; P = .09), consistent with the parallel-trends assumption for DiD design (Figure 2 and Figure 3). For RPM adoption, trends were uniformly near 0 across all groups before 2019 due to the absence of dedicated CMS billing codes, which limited formal evaluation of the parallel trends assumption for this outcome.

Sensitivity Analyses

Unadjusted outcome trends (eTable 3 in Supplement 1) were consistent in direction with adjusted DiD estimates. Sensitivity analyses using alternative clinician continuity definitions and organization-level NPIs were consistent with the main findings (eTables 4-5 in Supplement 1). E-value analyses suggested that an unmeasured confounder would need to be associated with both switching to MA and 2022 RPM adoption by risk ratio (E-value) magnitudes of 3.011 (E-value CI bound, 2.107) for the MA-VBC proxy comparison and 2.095 (E-value CI bound, 1.128) for the MA non-VBC comparison, above and beyond measured covariates, to fully explain away the observed associations. Corresponding E-values for 2022 hypertension-related hospitalization were 2.892 (E-value CI bound, 2.328) and 3.280 (E-value CI bound, 2.815), respectively (eTable 6 in Supplement 1).

Discussion

In this national cohort study of older adults with hypertension, switching from Medicare FFS to MA was associated with lower likelihood of receiving RPM for hypertension management, greater clinician discontinuity, and higher hypertension-related acute care use. These associations were observed for both MA plan-type categories, and findings should be interpreted as associations rather than causal effects.

The plan-type proxy for value-based MA products did not show higher RPM uptake than FFS. This finding should not be interpreted as evidence that all value-based contracts reduce RPM adoption because plan type is an imperfect proxy for clinician-level payment arrangements. Rather, the results suggest that current MA product and payment structures may not consistently translate broad incentives for hypertension control into RPM adoption.²⁹ The expected direction is theoretically ambiguous; MA organizations and clinicians may have incentives to adopt RPM if it prevents costly ED visits or hospitalizations,³⁰ but they may also face weaker direct reimbursement, implementation costs, or administrative barriers that slow adoption.³¹

Implementation complexity may also pose a barrier to RPM adoption. RPM requires device procurement, patient onboarding, consent documentation, data transmission, staff workflows, clinician review, and billing or reporting infrastructure.^32,33 COVID-19 pandemic–era telehealth flexibilities and rapid digital health expansion may have increased RPM awareness, but COVID-19 also disrupted routine care, staffing, and patient access.³¹ Year fixed effects captured secular shocks during this period, but residual differential pandemic effects across FFS and MA settings may remain.³⁴

A central implication of the findings is that switching itself may disrupt hypertension management. After switching to MA, beneficiaries were more likely to lose prior clinicians without observed replacement and less likely to substitute new clinicians in the same year. This pattern is consistent with the need to reestablish in-network care after moving into managed-care networks,³⁵ although our claims-based data measured observed clinician continuity rather than full network size. Because RPM initiation often depends on a sustained clinician-patient relationship for device setup, troubleshooting, data review, and feedback, clinician discontinuity may reduce opportunities to initiate or maintain RPM during the transition period.

The higher hypertension-related acute care use among MA switchers may reflect several nonexclusive mechanisms,³⁶ including selection into MA among beneficiaries experiencing changing care needs, disruption in established care relationships, delays in reestablishing in-network hypertension management, and lower RPM uptake. Coding differences across MA and FFS remain an important methodological consideration. MA plans receive risk-adjusted capitated payments and may have stronger incentives to document comorbidities, whereas FFS claims are primarily used for service-level billing. Therefore, we focused on primary diagnosis acute care events to capture the principal reason for care and to reduce bias from secondary diagnosis coding differences.

These findings suggest opportunities to better align MA incentives with high-value digitally enabled care delivery during insurance transitions.¹¹ Quality rating programs could consider measures that capture continuity of care and engagement with evidence-based digital tools that support blood pressure control. Targeted payment approaches, implementation support, or carve-outs for RPM within capitated arrangements may help offset upfront costs that deter adoption.³⁷ Future studies should link claims with MA contract characteristics, plan star ratings, insurer characteristics, clinician network files, and community-level social vulnerability measures to identify which plan and market features are associated with better continuity and RPM adoption.

Limitations

Several limitations warrant consideration. First, despite propensity score matching followed by DiD analysis, residual confounding may persist from unmeasured clinician-level factors, such as technology readiness or billing practices, and patient-level factors, such as health literacy or willingness to use digital devices. E-values were added to quantify how large in magnitude unmeasured confounding would need to be to explain away selected associations, but they do not eliminate the possibility of residual confounding or identify the source of bias. Second, MA plan type was used as a proxy for value-based arrangements because clinician-level contract data were unavailable; this may misclassify actual payment arrangements and attenuate or obscure differences by contract type. Third, the clinician continuity measure was based on observed NPIs in claims and encounter data and did not measure plan network size, contracted-network breadth, or primary care practitioner–specific continuity. Fourth, we could not directly test the parallel trends assumption for RPM because dedicated RPM billing codes were not available before 2019, although event-study plots for clinician continuity and acute care showed stable preperiod trends. Fifth, claims data lack direct clinical measures such as blood pressure values, which limits assessment of hypertension control. Sixth, COVID-19 pandemic–era disruptions may have influenced both RPM adoption and acute care use across MA and FFS settings. Sixth, MA encounter data completeness varies across plans, and some plans may be more likely to underdocument RPM adoption. We did not link plan characteristics, star ratings, insurer identifiers, or community social vulnerability measures; future work should examine these modifiers.

Conclusions

In this cohort study of older Medicare beneficiaries with hypertension, switching from Medicare FFS to MA was associated with lower RPM adoption, greater clinician discontinuity, and higher hypertension-related acute care use over 1 to 4 years of follow-up. These patterns were similar across MA plan-type proxy categories. Continuity safeguards and clearer payment or quality incentives during transitions into MA may improve remote monitoring and clinician follow-up for hypertension.

Supplement 1.

eFigure 1. Sample Selection Diagram

eTable 1. Outcome & Code Definitions (RPM CPTs; HTN/CVD ICD-10 Ranges; Provider Lost/Changed/Gain Definitions; Organization-Level Definitions)

eTable 2. Baseline Characteristics of All Beneficiaries in 2018 Before Matching, by Study Group

eFigure 2. Propensity Score Density Plots Before and After Matching

eFigure 3. Love Plot of Pre- and Post-Matching Covariate Balance

eTable 3. Unadjusted Outcomes by Year, Medicare Advantage (VBC-Proxy and Non-VBC) vs Fee-For-Service, 2016-2022

eTable 4. Adjusted Difference-in-Differences in Provider Gain or Provider Lost-or-Gain (Change) by Year, Medicare Advantage (VBC and non-VBC) vs Fee-For-Service, 2019-2022

eTable 5. Adjusted Difference-In-Differences in Organization-Level Providers Gain or Loss by Year, Medicare Advantage (VBC and non-VBC) vs Fee-For-Service, 2019-2022

eTable 6. E-Values for Selected Primary and Secondary Outcome Associations

Supplement 2.

Data Sharing Statement