Pattern of Audio Telehealth Among Medicare Beneficiaries with Alzheimer's Disease and Related Dementias

Abstract

Objectives.

The Medicare telehealth expansion increased telehealth use, but the utilization patterns of audio-only telehealth remain unclear among those with Alzheimer’s Disease and Related Dementias (ADRD). We aimed to describe 2020-22 trends in audio-only telehealth utilization and provider specialty, and to examine variations in utilization patterns by patient characteristics within the ADRD population.

Design.

Cohort study.

Setting and Participants.

We identified 5,021,632 older Medicare beneficiaries with a diagnosis of ADRD enrolled in Medicare fee-for-service (FFS) plans between 2020 and 2022.

Methods.

Using Medicare FFS claims data from 2020 to 2022, we describe the modality of evaluation and management (E&M) visits. We calculated the proportion of E&M visits delivered via telehealth and audio-only and described the provider specialty. To further explore the pattern of telehealth use during the later stages of the pandemic, we used 2022 data to classify telehealth users into audio-only, supplementary, and video-only categories. We examined its association with individual race, ethnicity, socioeconomic status, and rurality using a linear probability model and a multinomial logistic regression. Average marginal effects were estimated.

Results.

Overall, E&M telehealth visits among beneficiaries with ADRD declined from 21% in 2020 to 8% in 2022, while the audio-only share remained stable at around 20%. Over half of audio-only visits were furnished by primary care providers. Among telehealth users, traditionally underserved populations were significantly more likely to rely exclusively on audio-only visits and less likely to use them as supplemental care. For example, higher adjusted probabilities of exclusive audio-only use were observed among Black individuals (+0.4 percentage points [pp], p<0.01) and rural beneficiaries (+0.8 pp, p<0.01).

Conclusions and Implications.

Audio-only telehealth services were consistently used by Medicare beneficiaries with ADRD during the study period. Continued coverage for audio-only services could be a potential approach to promote digital health equity for older adults with ADRD.

Brief summary

Racially and ethnically minoritized, dual-eligible beneficiaries, and those living in rural or deprived areas, were more likely to rely solely on audio-only visits and less likely to use them as a supplement to other care modes.

INTRODUCTION

The utilization of telehealth among Medicare beneficiaries has increased substantially in recent years, primarily driven by the COVID-19 pandemic and the expansion of Medicare telehealth coverage.¹ One of the significant coverage changes was the expansion of Medicare coverage to include audio-only telehealth.² Before the pandemic, Medicare regulations required both audio and video components for telehealth and imposed geographic restrictions, limiting telehealth access to beneficiaries residing in rural areas and through specific healthcare facilities (e.g., federally qualified health centers). ² With these restrictions lifted during the pandemic, audio-only telehealth became an accessible modality,³ serving as an alternative option for patients and their family caregivers who could not or preferred not to access video telehealth or, in general, in-person visits.⁴

Despite its potential benefits, relatively few studies have examined patterns of audio-only telehealth use. Early evidence suggests that such services may be disproportionately used by patients who are older, racially and ethnically minoritized, dually eligible for Medicare and Medicaid, or living in areas with limited broadband access.^5-9 For example, a study conducted within a large health system found that Black and dual-eligible beneficiaries were more likely to rely on audio-only visits.⁶ Another study using national Medicare data observed similar trends during 2020–2022.⁷ However, these studies were either limited to a single health system, primarily descriptive in nature, or did not specifically focus on high-need populations such as older adults with Alzheimer’s Disease and Related Dementias (ADRD).

Access to adequate healthcare remains a challenge for the ADRD population due to their physical and cognitive impairments.¹⁰ However, the impact of ADRD and comorbidities often increases healthcare utilization.¹¹ To date, it remains unclear to what extent audio-only telehealth was used among the ADRD population. Little is known about the contexts in which these visits occurred, the types of providers delivering audio-only care, and which subgroups of patients with ADRD were most reliant on audio-only services. We are particularly interested in the relationship between audio-only telehealth utilization and traditionally underserved populations, including racial and ethnic minority populations, Medicare-Medicaid dual eligibles, individuals living in resource-deprived communities, and those residing in non-metropolitan areas. Addressing these gaps is critical, as cognitive impairment, digital literacy barriers, and geographic location may uniquely shape telehealth access in this population.

To address these gaps, this study aimed to: (1) describe trends in audio-only telehealth utilization among Medicare beneficiaries with ADRD from 2020 through 2022; (2) identify the provider types that commonly utilized audio-only telehealth for the ADRD population; and (3) examine patient characteristics associated with the pattern of audio-only telehealth.

METHODS

Data and cohort

This study utilized the Master Beneficiary Summary File (MBSF) and Medicare Fee-for-Service (FFS) claims data, including the Carrier files, from 2020 through 2022. We used the Master Beneficiary Summary File (MBSF) chronic condition segment to identify beneficiaries with an ever diagnosis of ADRD.¹² We excluded beneficiaries enrolled in Medicare Advantage due to incomplete claims capture in the Carrier files. The MBSF also provides core demographic data, including race, ethnicity, and dual-eligibility status. We utilized the Minimum Data Set (MDS 3.0) to identify nursing home stays and classify beneficiaries as long-stayers (continuous stays of 100 days or more) or not, which was subsequently controlled for in the regression models. We further merged the cohort with publicly available datasets at the zip-code or county level: Area Deprivation Index (ADI) to define community deprivation (zip-code level),¹³ American Community Survey (ACS) for community characteristics (zip-code level),¹⁴ Rural Urban Commuting Area (RUCA) scores for rurality (zip-code level, categorized as metropolitan, micropolitan, small town, and rural),¹⁵ and the Area Health Resource File (AHRF) for county-level healthcare supply.¹⁶

Variables

We identified both in-person and telehealth outpatient Evaluation and Management (E&M) visits in the Medicare Carrier files using the Restructured BETOS classification system.¹⁷ Telehealth E&M visits were identified using CPT codes and modifiers consistent with CMS guidelines and classified into two modalities: audio-only and video-based (detailed in Appendix Table S1).¹⁸ We constructed two outcomes of interest. First, we created a binary outcome for any telehealth use (beneficiaries with >=1 telehealth visit versus in-person-only). Second, among beneficiaries with any telehealth use, we defined the pattern of audio-telehealth use as follows: audio-only (audio-only telehealth visit), supplementary audio (audio-only telehealth visit in addition to video and/or in-person visit), and video-only (video-only telehealth visit). Key independent variables included: race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Others), Medicare-Medicaid dual eligibility status, community deprivation status (defined as living in a deprived community with an ADI >= 85),¹⁹ and rurality (categorized as metropolitan, micropolitan, small town, and rural).¹⁵

Statistical analysis

To describe the utilization of audio-only telehealth in outpatient E&M visits, we began our analysis at the visit level by calculating the annual percentage of outpatient E&M visits delivered via telehealth and the percentage of audio-only telehealth among all telehealth visits for the study period (2020–2022). We also examined the distribution of provider specialties for audio-only visits, identified using the CMS provider specialty code variable in the Carrier file.

Next, to examine the factors associated with audio telehealth use during the later stages of the pandemic, we estimated a set of regression models using 2022 data at the beneficiary level. A linear probability model was first used to examine patient factors associated with the likelihood of any telehealth use (versus in-person only). Next, among those with any telehealth use, a multinomial logistic regression was conducted to evaluate the relative likelihood of being an audio-only user or a supplementary audio user, compared to video-only users. All models were adjusted for beneficiary characteristics and community characteristics. Average marginal effects were calculated from the multinomial logistic regression to facilitate the interpretation of results. The study has been approved by the University of Rochester Institutional Review Board.

RESULTS

Visit-level analysis

Figure 1 displays the percent distribution of E&M visits delivered via telehealth and the proportion of telehealth visits that were audio-only among Medicare beneficiaries with ADRD. In general, we observed a decrease in the percentage of telehealth among overall E&M visits, from 21% in 2020 to 8% in 2022. However, the percentage of audio visits among telehealth was consistent across the study years, accounting for 20% of telehealth visits annually (20.4% in 2020 to 19.7% in 2022; see Appendix Table S2). By examining provider specialties, we found that over half of the specialty codes fell into primary care (e.g., internal medicine, family practice, and nursing practitioner/physician assistant (NP/PA)) (see Appendix Table S3).

Figure 1.

Distributions of E&M visits by the mode

Beneficiary-level analysis

Table 1 presents the regression results using only 2022 data, and Table S4 shows the characteristics of the analytical cohort. As the coefficients in Model 1 were estimated using a linear probability model, they represent average marginal effects. We found that several sociodemographic characteristics were significantly associated with telehealth utilization. For instance, individuals identified as Black had a 2.1 percentage points lower probability of using telehealth than their White counterparts (β = −0.021, p < 0.01). Similarly, individuals living in deprived communities (β = −0.005, p < 0.01) and non-metro areas, such as small towns (β = −0.019, p < 0.01), were significantly less likely to use telehealth than those in advantaged areas. In contrast, Hispanic individuals (β = 0.044, p < 0.01), Asian individuals (β = 0.087, p < 0.01), and dual-eligible beneficiaries (β = 0.022, p < 0.01) were more likely to utilize telehealth.

Table 1.

Regression results

	Model 1	Model 2
VARIABLES	Any telehealth user (REF: exclusive in- person)	Audio-only (REF: video users w/o audio)	Supplementary audio users  (REF: Video users w/o audio)
Race/Ethnicity (Ref: non-Hispanic Whites)
Blacks	−0.021***	0.361***	−0.040***
Hispanic	0.044***	0.295***	−0.113***
Asians	0.087***	0.281***	−0.039***
Others	0.065***	0.218***	0.149***
Duals	0.022***	0.238***	−0.047***
Deprived community	−0.005***	0.072	−0.051***
Rurality (Ref: metropolitan)
Micropolitan	−0.018***	0.286***	−0.013
Small town	−0.019***	0.449***	−0.017
Rural area	−0.008***	0.706***	0.038***
Age	−0.005***	0.032***	0.005***
Female	−0.019***	−0.056***	0.005
NH resident	−0.117***	−2.348***	−1.020***
# of CC >=10	0.118***	−0.825***	0.185***
Primary care shortage (Ref: no shortage)
Full shortage	0.033***	0.056	−0.100***
Partial shortage	0.040***	0.143***	0.033***
ACS covariates
% of HH with personal computer	0.004***	−0.023***	−0.008***
% of HH with internet	0.001***	−0.000	−0.000
# of MD per 1000 older adults	0.001***	0.002***	0.004***
Constant	0.286***	−4.893***	−0.533
Observations	3,641,523	1,210,651	1,210,651

Model 1 was estimated by a linear probability model to model the probability of having at least one telehealth (any telehealth), while Model 2 examined the relative odds of being audio-only users or supplementary audio users among those any telehealth users. NH: nursing home; CC: chronic condition; ACS: American Community Survey; HH: household; MD: medical doctors;

^***< 0.01

^**< 0.05.

In Model 2, we examined the characteristics associated with the pattern of audio telehealth utilization. The results of the multinomial logistic regression suggested that, except for people living in resource-deprived communities (β = 0.072, p > 0.05), individuals identified as Blacks (β = 0.361, p < 0.01), Hispanic (β = 0.295, p < 0.01), duals (β = 0.238, p < 0.01), and people living in non-metro areas (e.g., rural areas (β = 0.706, p < 0.01), were all significantly more likely to be audio-only users compared to their counterparts. On the other hand, those populations mentioned above were, at the same time, significantly less likely to be supplementary audio users, except for people from the most rural areas (β = 0.005, p < 0.01).

To facilitate the interpretation of the multinomial logistic regression, we reported average marginal effects to quantify differences in the adjusted probability of being an audio-only user or a supplementary audio user across groups (Table 2). Consistent with the regression results, all underserved populations were significantly more likely to be audio-only users. For example, Black individuals had a 0.4 percentage-point (pp) higher probability of being audio-only users. Similarly, the differences in the adjusted probability of being an audio-only user increased as areas became more rural: compared with those in metropolitan areas, individuals in micropolitan, small town, and rural areas were 0.3, 0.5, and 0.8 percentage points more likely to use only audio-telehealth, respectively. On the other hand, underserved populations were significantly less likely to use audio visits as their supplementary approach. For example, duals and individuals residing in deprived communities were both 1.1 pp (p < 0.01) less likely to be supplementary audio users.

Table 2.

Adjusted differences in the probability of the audio telehealth pattern by subgroup

Percentage points (pp)	Audio-only	Supplementary Audio user
Race and Ethnicity (Ref: non-Hispanic White)
Black	0.4***	−1.0***
Hispanic	0.3***	−2.5***
Asian	0.3***	−1.0***
Others	0.1***	3.3***
Dual status (Ref: non-duals)
Dual	0.2***	−1.1***
Community deprivation status (Ref: non-deprived community)
Deprived community	0.1**	−1.1***
Rurality (Ref: Metropolitan)
Micropolitan	0.3***	−0.4**
Small towns	0.5***	−0.5**
Rural	0.8***	0.5

The numbers presented above are the average marginal effects, x 100, estimated from the multinomial logistic regression, in the unit of percentage point.

^***< 0.01

^**< 0.05.

DISCUSSION

Using 2020-2022 national data, this study described the utilization of audio-only telehealth among Medicare beneficiaries with ADRD and how it varied across subgroups. Although overall telehealth use declined during the pandemic, the relative share of audio-only visits remained consistent from 2020 to 2022. During the study period, over half of the audio telehealth visits were delivered by primary care providers. While the regression results showed that individuals who were Hispanic and duals were more likely to use any telehealth than their counterparts, we found that among those who used telehealth, underserved populations, including racially and ethnically minoritized, dual-eligible beneficiaries, and those living in rural or deprived areas, were more likely to rely solely on audio-only visits and less likely to use them as a supplement to other care modes.

Though Medicare covered telehealth before the COVID-19 pandemic, coverage was limited to video-audio telehealth.² Hence, the coverage of audio-only telehealth following the telehealth expansion policy represents an alternative approach for individuals to access care. When examining provider types, we found that half fell into the primary care categories (e.g., internal medicine and family practice). While we cannot directly determine whether audio-only use reflects limited access to video or in-person care, patient or caregiver preference, and/or provider practice patterns, our findings regarding the volume of audio-only telehealth furnished by primary care providers add to the current literature on audio-only telehealth as a potential pathway to primary care.²⁰ Future studies could examine the relationship between the expansion of audio telehealth coverage and access to care, especially for vulnerable populations.

While our results are consistent with a prior paper that Hispanic and dual-eligible individuals were more likely to use any telehealth,²¹ we further found that these groups, along with other traditionally underserved populations, were significantly more likely to be audio-only users and less likely to use telehealth as a supplementary service. Hence, it is possible that traditionally underserved populations benefited from the expansion of telehealth, though they were more likely to be audio-only users. With the recent policy change, Medicare coverage of audio-only telehealth services is now extended through December 2027.²² Ensuring continued access to audio-only telehealth could be a potential way to promote digital health equity for older adults with cognitive impairment.

Limitations

This study has a few limitations. First, because we used claims data, we were unable to distinguish whether the observed differences stemmed from access barriers, preferences, or provider behaviors. Second, our analysis may be subject to residual confounding, such as caregiver availability and support, as well as digital literacy, which could also significantly affect how beneficiaries use different telehealth modalities. Third, focusing on telehealth users allowed us to account for unmeasured factors, such as telehealth preferences, though this decision may be subject to selection bias. Lastly, our analysis focused on Medicare FFS beneficiaries; therefore, the results may not be generalizable to the entire Medicare population. Similarly, although our study period extended through 2022, allowing for telehealth utilization to stabilize following the initial pandemic shock, the observed trends may continue to evolve.

CONCLUSIONS

Following the expansion of Medicare telehealth coverage, audio-only telehealth services were consistently utilized among Medicare beneficiaries with ADRD, maintaining a consistent share of overall telehealth use from 2020 through 2022. We found that over half of the audio-only visits were furnished by primary care providers, highlighting their potential role in providing essential care to the ADRD population. Crucially, traditionally underserved populations, including racial and ethnic minoritized individuals, dual-eligible beneficiaries, and those living in resource-deprived and non-metropolitan areas, were significantly more likely to rely exclusively on audio-only visits and less likely to use this modality as a supplement to other care. Continued coverage for audio-only services could be a potential approach to promote digital health equity for older adults with ADRD.