Adapted from a newsletter first sent October 9, 2026.
In the fall of 2026 Feeding America and the Elevance Health Foundation released one of the largest food as medicine evaluations so far: 1.45 million patients screened for food insecurity, 161,000 households connected to food across 13 states. Hospitalizations dropped 14 percent and ED visits 11 percent. The clearest movement in HbA1c, BMI and LDL showed up in people who got consistent dietitian support.
Everyone quoted the 14 percent. The number I couldn't stop looking at was the gap between 1.45 million and 161,000. The units differ, patients and households, so it isn't a clean conversion rate. The gap is still the point.
I was trained on a sequence: assess, diagnose, intervene, monitor. We would never flag malnutrition risk, write it in the chart and stop. Yet that's how food insecurity screening runs in a lot of systems. The question gets asked at intake, the box gets checked and the "intervention" is a handout. We built the front half of the nutrition care process and called it a program.
This evaluation is useful because it shows what the back half looks like when someone builds it: a real connection to food, and a person who follows up. The outcomes belong to the people who made it through the whole sequence, not to the 1.45 million who were only asked.
Screening volume is easy to report and easy to hit. It rewards asking. I'd like to know how many systems could tell you today what share of their positive screens reach any intervention at all, and what would reward answering instead.