Design: Retrospective cohort, claims linked to laboratory results
Population: 27,650 US adults with laboratory-confirmed diabetes and no prior record
Primary outcome: Diagnostic coding and treatment initiation within 1 year
Effect: 66.6% coded and treated; 19.7% coded only; 4.1% treated only; 9.6% neither

A US retrospective cohort linked commercial and Medicare claims from 2017 to 2023 with laboratory results, and followed 27,650 adults whose results met diabetes criteria but who had no prior diabetes or prediabetes record.

Within a year, 66.6% had both a diagnosis code and a glucose-lowering prescription, 19.7% a code only, 4.1% treatment only and 9.6% neither. Higher HbA1c made coding and treatment more likely. A heavier comorbidity burden made both less likely — the patients with the most to lose were the ones most often left alone.

Claims data cannot see lifestyle advice or a deliberate decision not to treat, so some of the gap is appropriate. Still, a result in the chart that nobody names as diabetes is a common failure in any system, including Indian clinics where results come back after the patient has gone.

Act on an HbA1c of 6.5% or above, even when it is not the reason for the visit.
Write the diagnosis into the problem list so the next clinician sees it.
Be alert in multimorbid patients, where other problems tend to crowd diabetes out.
Set up a way to recall patients whose abnormal results arrive after they leave.

Why it matters

Multimorbid patients, at the highest risk, were the ones whose diabetes was most often left unrecorded.

The statistics, in plain English

These are proportions from claims records. A missing code does not always mean a missed diagnosis — some care is never billed — so the true gap may be smaller than 33%, but it is unlikely to be trivial.