Pandemic telehealth surge created lasting but unequal shift in Medicare care delivery

Telehealth has become a lasting part of the United States' Medicare care delivery after its pandemic-era surge, but its long-term use remains sharply uneven across states and beneficiary groups, according to a new study published in the Romanian Journal of Preventive Medicine.

The study, titled AI–Driven Assessment of Pandemic Telehealth Surge and Differential Saturation Across Demographic Groups in Medicare Beneficiaries, applied supervised machine learning and explainable artificial intelligence to determine whether telehealth’s rapid expansion was a temporary emergency response or a durable shift in preventive care delivery.

The study is based on national Medicare Telehealth Trends data from 2020 through 2025.

Telehealth moved from emergency surge to stable use

The COVID-19 pandemic triggered one of the fastest care delivery shifts in the history of the Medicare program. Emergency policy changes expanded reimbursement, loosened geographic restrictions and allowed more services to be delivered remotely. Before the pandemic, telehealth use among Medicare beneficiaries was limited by tight rules and uneven infrastructure. During the crisis, it became a key channel for primary care, chronic disease follow-up, behavioral health support and medication management.

The study finds that this surge didn't vanish after the initial emergency period. Instead, telehealth use followed a clear pattern: a rapid and synchronous surge in 2020, a decline after peak pandemic disruption and then stabilization at persistent but uneven levels. The pattern suggests telehealth has become partially institutionalized within Medicare rather than returning to its pre-pandemic marginal role.

The researchers used 32,508 quarterly observations from the Medicare Telehealth Trends dataset. Each observation reflected a combination of state, quarter and beneficiary demographic group, covering age, race, sex, dual eligibility status and rurality. The primary outcome was percent telehealth utilization, defined as the share of telehealth-eligible beneficiaries using telehealth in a given quarter.

Machine learning models were used because telehealth adoption did not follow a simple linear path. The researchers compared random forest regression, gradient boosting, generalized additive models and linear models. Random forest performed best for age-specific modeling, capturing the sharp surge, post-surge decline and later stabilization with high predictive accuracy. The results showed that nonlinear models were better suited than traditional linear methods for analyzing a sudden system-wide change followed by uneven long-term adjustment.

The initial surge was broad and simultaneous. Across age groups, dual eligibility groups and racial groups, telehealth use peaked around the second quarter of 2020. That timing reflects the emergency nature of the policy and operational shift. Hospitals, clinics, providers and patients moved quickly into remote care as in-person visits were disrupted.

However, the more important finding came after the surge. Telehealth use did not collapse back to earlier levels. Instead, it entered plateau phases, indicating stable post-pandemic use. Formal validation showed that these plateaus were not random fluctuations. The study found statistically stable equilibrium levels across demographic groups, with stabilization generally emerging between mid-2022 and early 2023. This means Medicare telehealth has crossed a major threshold. It is no longer only an emergency substitute for in-person visits. It has become a durable, if uneven, part of care delivery.

For preventive medicine, that shift matters because routine follow-up, chronic disease monitoring, behavioral health access and medication adherence often depend on sustained engagement rather than one-time emergency care.

Age, income-related status and state context shape long-term use

The study found that telehealth stabilization differed meaningfully across beneficiary groups. Age was one of the clearest dividing lines. Medicare beneficiaries aged 0 to 64, many of whom qualify through disability, had the highest peak telehealth use and the highest long-term plateau. Their predicted peak reached 57.17%, and their plateau stabilized at 23.65%.

Older beneficiaries used telehealth at lower long-term levels. The predicted plateau was 10.67% for those aged 65 to 74, 9.91% for those aged 75 to 84 and 9.55% for those aged 85 and older. The gap between younger Medicare beneficiaries and the oldest group exceeded 14 percentage points during the stabilization phase, suggesting that older adults did not adopt telehealth to the same sustained extent, even after the system had adjusted to remote care.

The reasons are likely linked to digital literacy, device access, comfort with technology, disability patterns, caregiver support and care needs that may require in-person assessment. The result is not that older adults rejected telehealth entirely, but that the long-term level of use remained much lower than among younger Medicare beneficiaries.

Dual eligibility status also marked a major divide. Beneficiaries enrolled in both Medicare and Medicaid showed higher telehealth use than Medicare-only beneficiaries. During the peak period, dual eligible beneficiaries reached predicted utilization of about 54%, compared with about 43.5% among Medicare-only beneficiaries. After stabilization, dual eligible beneficiaries remained at 19.14%, compared with 10.49% for Medicare-only beneficiaries.

The finding cuts against a simple assumption that low-income or medically vulnerable groups would necessarily have lower telehealth use. Instead, the study suggests that dual eligible beneficiaries may have relied more heavily on telehealth during and after the pandemic, possibly because of higher medical need, more frequent contact with care systems or stronger dependence on remote access when transportation and in-person care were difficult.

Race-based differences were present but comparatively modest after accounting for time and geography. Black or African American beneficiaries had the highest predicted peak utilization at 49.97%, while Non-Hispanic White beneficiaries peaked at 45.75%. Plateau levels ranged from 11.78% among Asian/Pacific Islander beneficiaries to 13.75% among Black or African American beneficiaries. The spread across racial groups was about two percentage points, much smaller than the differences by age, dual eligibility or state.

The researchers’ explainable AI analysis reinforced this point. Time was the strongest driver of utilization, reflecting the dominant effect of the pandemic period. State-level factors were also important. Race had a smaller independent contribution once time and geography were included. This does not mean racial disparities are unimportant. It indicates that structural environments, including state policy, broadband access and health system organization, may be doing much of the work in shaping observed racial patterns.

Geography emerges as the largest equity fault line

Interestingly, geography outweighed demographic variation in long-term telehealth stabilization. State-level predicted plateau utilization ranged from 4.9% in the lowest-saturation areas to 29.5% in the highest. That range was larger than the plateau differences across age, dual eligibility or race.

The researchers identified three broad state-level saturation regimes: low, mid and high. Low-saturation states stabilized at an average predicted utilization of 11.2%. Mid-saturation states stabilized at 18.2%. High-saturation states stabilized at 26.7%, more than double the low-saturation level. This points to a structural divide in Medicare telehealth integration. Where beneficiaries live may matter more than who they are, at least when demographic variables are considered independently. State policy environments, broadband infrastructure, provider systems, reimbursement conditions and organizational readiness appear to shape whether telehealth becomes a routine part of preventive care.

High-saturation states included places such as California, Hawaii, Massachusetts, the District of Columbia and New York. Other states and territories stabilized below 10%. The study does not claim that geography alone causes the gap, but it shows that long-term telehealth use is strongly patterned by state-level conditions.

Policy implications

If telehealth remains uneven by geography, its benefits for preventive medicine will also be uneven. High-saturation areas may be better positioned to use telehealth for chronic disease management, behavioral health follow-up, medication review and routine preventive care. Low-saturation areas may continue to face gaps in access, especially for older adults, rural beneficiaries and patients with limited provider availability.

The study argues that telehealth should be evaluated as preventive infrastructure, not just a crisis technology. Sustained access can reduce missed follow-ups, improve continuity and support patients who face transportation, mobility or local provider barriers. But those gains depend on stable reimbursement, broadband access, digital literacy support and health system capacity.

The findings also show why simply allowing telehealth is not enough. Emergency policy changes opened the door, but state-level stabilization depended on deeper structural conditions. Without deliberate intervention, telehealth may remain embedded at high levels in some areas and weakly integrated in others.

The study acknowledges some limitations. It utilized aggregated administrative data, which cannot capture individual patient preferences, clinical outcomes, visit quality, provider behavior or variation within states. It also measured telehealth use as a broad percentage, without separating video from audio visits or distinguishing preventive, behavioral health, primary care and specialty services. The observational design cannot prove causality, and unmeasured state policy or health system factors may have influenced the patterns.

The next phase of telehealth policy should focus less on whether remote care should exist and more on whether it can be made stable, high-quality and equitable. Strengthening broadband, preserving reimbursement clarity, supporting digital literacy and aligning state-level policy will be central to ensuring that telehealth functions as a durable part of preventive care rather than a fragmented legacy of the pandemic.