Back

Association Between Telehealth Use and 30-day Medicare Spending

Ellimoottil, C.; Kulkarni, A. J.; Zhu, Z.; Dunn, R. L.; Chang, C.-H.; Chun, E.; Hou, H.; Lee, J. D.; McCullough, J. S.; Thompson, M. P.

2025-02-02 health policy
10.1101/2025.01.31.25321423 medRxiv
Show abstract

ObjectivesThe objective of this study was to investigate whether healthcare visits initiated by telehealth had higher or lower 30-day spending compared to in-person-initiated visits. The study compared the overall spending, rates of return visits, laboratory tests, and imaging procedures within 30 days for Medicare fee-for-service patients who underwent in-person and telehealth evaluations between July 1, 2020 and December 31, 2022. Study DesignA large-scale retrospective cohort study using propensity score matching. MethodsThis study included 100% of Medicare fee-for-service beneficiaries who are aged >65 years while excluding those with Medicare Advantage coverage and those without continuous Medicare Parts A and B. We identified patients with no prior visits and 30-day episodes of care initiated by outpatient telehealth or in-person visits from July 1, 2020 through December 31, 2022. We then compared adjusted 30-day total Medicare spending, rates of return visits, laboratory tests, and imaging utilization between propensity-matched index visits initiated by telehealth versus in-person. ResultsTelehealth-initiated visits were associated with lower 30-day spending ($260 vs. $342; net: -$82), though return visit rates were higher for telehealth (16.1% vs. 14.1%). Both lab test rates (7.8% vs. 24.2%) and imaging rates (3.5% vs 7.8%) were lower for telehealth-initiated episodes compared to in-person-initiated episodes. ConclusionsTelehealth-initiated episodes of care were associated with lower 30-day Medicare spending and reduced utilization of labs and imaging. These findings suggest that telehealth, when used as a substitute for office visits, may reduce overall Medicare spending.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.