Axios’ Bob Herman reports this morning on the saga of a heart pump called Impella RP and what it says about the data the FDA has to work with when it evaluates medical devices.

Details: The FDA approved Impella RP in 2017, relying on studies showed 73% of patients who got the pump were still alive within a month.

• But those studies had some major limitations — the combined sample size was just 60 patients, and trials were not randomized.
• So the FDA required Abiomed, maker of the device, to track patients who receive the device.

The results aren't good. The FDA warned providers this week that survival rates have dropped precipitously — just 4 of 23 patients (17%) who got the Impella RP were alive within a month.

• Since 2015, there have been 21 adverse events linked to the Impella RP, including 5 deaths, according to federal data analyzed by Madris Tomes, a former FDA regulator who founded the software firm Device Events.

Between the lines: The FDA and Abiomed both said patients in the post-approval study were a lot sicker than people who were in the original studies.

• But if patients were critically ill or already close to death, using the heart pump would be like "throwing a Hail Mary" — and wouldn't tell us much about whether the patient benefited from the device, said Robert Yeh, an interventional cardiologist in Boston who has used the Impella.

The big picture: "We don't have great randomized trials ... for many of the devices like Impella. As a result, we are left to a data-free zone," Yeh said.

Flashback: The International Consortium of Investigative Journalists published a series that explained why devices have a lower bar for clinical safety and effectiveness than drugs.

The Trump administration is approving Medicaid work requirements but isn't requiring states to assess the impacts of those policies on their programs, the Los Angeles Times' Noam Levey reports. That appears to violate Obama-era rules that govern the program.

The big picture: Of the 17 states that have sought federal permission to add work requirements to their Medicaid programs, 9 have not included estimates of how many people would lose their coverage as a result.

• And in the 8 states where work requirements have already been approved, none has a plan in place to track whether Medicaid recipients find jobs, per the L.A. Times — even though that's the stated goal of this entire endeavor.
• That would seem to contravene 2012 rules that said waiver requests like these need to include "an estimate of the expected increase or decrease in annual enrollment," though the federal Centers for Medicare & Medicaid Services told the LAT that no such rules apply.

Quick take: This could complicate the Trump administration's job as it defends work requirements in court, where it has to show that the new restrictions are consistent with Medicaid's goals as a health care program.

The number of underinsured Americans — meaning, people whose out-of-pocket costs eat up a significant percentage of their income — is steadily climbing, according to a new survey by the Commonwealth Fund.

• The increase has been sharpest among those who get insurance through their employers, rising from 17% of insured workers in 2010 to 28% last year.
• In the individual insurance market, the rate of underinsured consumers has gone from 37% to 42% in the same time frame.

Why it matters: A person is considered underinsured if their out-of-pocket spending exceeds 5-10% of their income, or if their deductible is more than 5% of their income.

• By definition, those are patients who have a harder time paying their medical bills.
• And patients with high deductibles are also more exposed to the full cost of health care services, contributing to the public outcry over health care costs.

Go deeper: Workers' health care costs just keep rising

An increasing number of patients suffering from rosacea are self-medicating with a product that was designed for horses. And dermatologists want them to stop.

• The veterinary product uses the same active ingredient as an FDA-approved gel for rosacea treatment, but has never been tested on humans for safety or effectiveness.
• But the FDA-approved treatment is expensive.
• "Horse paste, by contrast, can be purchased from Amazon and vet supply stores for as little as $4, no prescription necessary," Insider writes in an exposé of this whole weird thing.