The Takeaways: Week 9 of 2022

Three selections this week, all healthcare. I am a little backlogged on newsletters.


Susan Kelly, Healthcare Dive. Anesthesiology practices get paid more when backed by private equity, study finds (March 2, 2022)

A study of more than 2.2 million privately insured patients found that prices paid to anesthesia practitioners were higher when hospital outpatient departments and ambulatory surgery centers contracted with a physician management company, which are commonly referred to as staffing companies. Prices rose even more if the staffing firm had private equity backing, according to researchers at Columbia University's Mailman School of Public Health and Weill Cornell Medical College.

Elise Reuter, Healthcare Dive. CMS code seen as major step toward reimbursement for digital therapeutics (March 1, 2022)

Pointing to the number of digital health companies that are seeking out employers as customers, [Pear Therapeutics CEO Corey McCann] said, "What you have today is a whole host of unregulated products that target relatively healthy normals and that do so for the service of self-insured employer groups ... It's a relatively small section of healthcare and has become a relatively congested portion of the healthcare market."

McCann added that while there's nothing wrong with that model, Pear would need an FDA label because the company sought to provide treatment for more high-acuity conditions. Many of these patients also are covered by traditional commercial insurance plans or state Medicaid plans.

Casey Ross, STAT. AI gone astray: How subtle shifts in patient data send popular algorithms reeling, undermining patient safety (February 28, 2022)

Its downfall was perhaps foreseeable given its reliance on data susceptible to time-related changes. ICD codes have been updated 10 times since the initial edition was created in 1929, and practices surrounding blood work and orders for antibiotics shift even more frequently. Machine learning experts said relying on them essentially means sacrificing an algorithm’s long-term reliability for a short-term boost in accuracy that may make it more attractive to potential buyers.

“Operational behaviors [of caregivers] change much more than human physiology in a span of five years,” said Nigam Shah, a professor of biomedical informatics at Stanford University. “So when you’re using operational features, it’s no surprise you see a bunch of degradation in model performance.”