Berkson's paradox

Suppose that in the general population two conditions, diabetes and inflammation of the gallbladder, have nothing to do with each other. Having one tells you nothing about your chance of having the other. Either condition on its own is serious enough that people who have it are admitted to hospital.

A researcher takes the records of every patient in the hospital and tabulates the two conditions. She finds that among these patients, those with diabetes are much less likely to have gallbladder inflammation than those without diabetes. The records are accurate and the sample is large. What does the finding show?