Association Between Broadband Internet Availability and Telemedicine Use

Methods

This study was approved by the Harvard Medical School Institutional Review Board, which waived the requirement for informed patient consent because of deidentified data. Per capita rates of telemedicine visits for each US county in 2016 were calculated using a nationally representative 20% sample of Medicare fee-for-service beneficiaries and data from OptumLabs Data Warehouse, a deidentified claims database for privately insured enrollees and Medicare Advantage enrollees in a large, private US health plan. Telemedicine visits were defined as health professional claims with either a telemedicine modifier (GT, GQ, 95) or a telemedicine-specific code (G0425-7, G0406-8, or G0459). Counts of visits were based on the beneficiary’s county of residence. Because Medicare does not cover telemedicine for nonrural beneficiaries living in metropolitan counties, we did not include these beneficiaries in our per capita rates. We removed counties without rural residents, without both commercial and Medicare beneficiaries, and outlier counties in the top percentile of per capita telemedicine use.

We measured broadband availability at the county level using the FCC’s Fixed Broadband Deployment Data.⁴ Broadband access was defined as the availability of wired internet download speeds of at least 25 megabytes per second and upload speeds of at least 3 megabytes per second.⁵ Counties were categorized based on the share of the county’s population that could purchase broadband: low availability (0%-40%), medium availability (>40%-70%), or high availability (>70%). In each county, we also determined the number of census blocks targeted by the FCC’s Connect America Fund.

We split our sample of counties (N = 2785) into 3 groups based on their level of rurality using 2013 Department of Agriculture’s Rural-Urban Continuum Codes,⁶ including metropolitan counties with rural residents (codes 1-3), nonmetropolitan counties (4-7) with smaller towns/cities (≥2500 urban residents), and nonmetropolitan counties (8-9) without urban areas (ie, fully rural). In each subsample of counties, we predicted per 1000 beneficiary telemedicine rates at each broadband category using negative binomial regressions, controlling for county level socioeconomic characteristics, per capita hospitals, health centers and psychiatrists, and state level differences in commercial reimbursement regulations for telemedicine (Table). Sensitivity analyses examined the Medicare and commercial population separately. P values <.05 were considered significant. Analysis was conducted using Stata, version 14 (StataCorp LLC).

Results

Our sample consisted of 869 metropolitan counties with at least some rural residents, 1317 nonmetropolitan counties with smaller towns/cities, and 599 fully rural counties.⁶ In fully rural counties, greater broadband access was associated with greater telemedicine use: counties with low broadband availability had 34% fewer visits per capita compared with counties with high broadband availability (13.4 per 1000 vs 20.4, P = .004) (Figure). In metropolitan counties with rural residents and nonmetropolitan counties with smaller towns/cities, greater broadband access was not associated with differences in telemedicine use. For example, in metropolitan counties with rural residents, counties with low vs high broadband availability had 2.4% more visits per capita (4.2 vs 4.1 per 1000; P > .99). Results were consistent when we modeled commercial and Medicare rates separately or used different cutoffs for broadband availability. Of the FCC’s Connect America Fund investment, 16.1% was to fully rural counties.

Discussion

Broadband availability was associated with greater telemedicine use, but only in fully rural counties. The FCC’s Connect America Fund has awarded billions of dollars in subsidies to expand broadband, but most of that funding is in counties where we did not observe an association between broadband availability and telemedicine use. More targeted funding to fully rural counties, where wired broadband may be a critical barrier, could help to alleviate disparities in access to health specialists.

Our analysis was limited to wired broadband availability and may not represent all connectivity options available and our measure of telemedicine only captures when health care professionals bill for telemedicine visits.