Ateev Mehrotra is a professor at Harvard Medical School whose research focuses on consumerism in health care and innovations in delivery such as telemedicine. Ateev is a member of the 2022 Tradeoffs Research Council.

We’ve all heard about the pandemic-related surge in telemedicine. However, what has received little attention is how this shift has given researchers new types of data on how patients and providers interact. 

Take the longstanding problem of overprescribing of antibiotics. Research shows 30% of antibiotic prescriptions are unnecessary, and overprescribing can cause unnecessary side effects and contribute to antibiotic resistance. 

Enter new data from telemedicine. Three articles by Kathryn Martinez, Michael Rothberg and colleagues from the Cleveland Clinic used data from Amwell, one of the largest private telemedicine companies, to assess several challenges doctors have raised around reducing antibiotic prescriptions.

One concern is whether patients would be upset if doctors didn’t prescribe them antibiotics. Some telemedicine companies have patients rate each visit on a 5-star scale, and using this data, the researchers found patients who got antibiotics were more likely to give their doctors a 5-star rating (91%) than those who didn’t (71%).

Another concern is that explaining to patients why they don’t need antibiotics will add time to visits. The length of a telemedicine visit is automatically recorded, and the researchers used this to discover that visits that ended without an antibiotic prescription lasted about a minute longer on average than those with a prescription.

Finally, comparing providers’ prescribing rates for the same condition can be a good way to identify overprescribers. But that only works if diagnosing is consistent. Since telemedicine patients are effectively randomized to whichever provider is available next, we would expect all providers to get roughly the same number of patients with conditions that require antibiotics, like sinusitis. But the researchers found physicians who most often prescribed antibiotics were more likely to diagnose patients with sinusitis, suggesting a bias toward overprescribing that would otherwise be obscured by their diagnoses.

These findings illustrate some of the challenges we face in cutting antibiotic overprescribing (even though the data only capture telemedicine visits from one company). But what excites me is that these new insights were made possible by the shift to telemedicine and data that are harder to capture in an in-person visit.

And this is just the tip of the iceberg. One could apply natural language processing to transcriptions of visits to capture whether the tone used by a physician impacts quality. There is also great potential to use telemedicine visits to test different ways of improving quality. And because of the enormous volume of some telemedicine vendors (for example, Teladoc can have almost 15,000 visits per day), these studies could be done extremely quickly.