Where the medications ran out

Adapted from a newsletter first sent May 28, 2026.

A produce prescription study published in the spring of 2026 did something rare. It pulled outcomes straight from the electronic health record, across four programs and 2,218 patients. The headline was a half-point drop in HbA1c on average, which is fine. Then you look at who dropped the most.

Patients who started above 9 percent, the ones whose diabetes was truly uncontrolled, dropped nearly two points. That’s about four times the average.

This didn’t surprise me. Those patients rarely show up at a produce prescription desk because of a wellness flyer. Usually a clinician has run out of pharmacologic levers and decided to try something else. They are tired of being told they’re failing. They aren’t looking for a salad lecture. I think they’re looking for the first thing in a year that didn’t come with a side effect.

What bothers me is that programs often don’t enroll those patients. They enroll people at mild risk, because mild-risk patients graduate nicely, look good in a year-one report and don’t wreck the budget if attrition climbs. The result is a respectable average and a biggest effect left on the table.

I’ve sat with patients at 10.4. They weren’t noncompliant. They were under-supported, and food is one of the few interventions they hadn’t been failed by yet.

So I wonder how many programs would look better if they enrolled fewer people, and harder ones. And how many “too sick to engage” labels are really descriptions of the system that was supposed to do the engaging.

Previous
Previous

Points are not a prescription

Next
Next

The meal in the clamshell