We tested the performance of the original eMERGE algorithm, modified eMERGE algorithm, and modified MGB algorithm against two independent gold standard datasets derived from chart reviews, one set from MGB EMR, and one set from the UAB EMR. We selected charts for review by applying a data floor of at least one clinical note and at least three ICD9/10 codes from distinct dates, and screened for at least one ICD9/10 code for T2D. We conducted chart reviews for 208 randomly selected screen positive subjects from MGB and 198 screen positive subjects from UAB. In addition, we conducted chart reviews for 200 screen negative subjects from UAB. At MGB, charts were reviewed for diagnostic criteria for T2D by two endocrinology fellows, initially with 20 charts reviewed by both reviewers, to assess concordance. A chart review guideline was developed and used for reviewing the remaining charts, divided into two independent sets. At UAB, charts were reviewed by a trained study coordinator using the same chart review guideline.
