Skip to main content
. 2016 May 12;23(6):1166–1173. doi: 10.1093/jamia/ocw028

Table 2:

Performance assessed using a manually reviewed evaluation set for T2DM and MI

T2DM Cases Ctrls Acc PPV
PheKB definition 152 152 0.92 0.96
Noisy labeling process 152 152 0.89 0.81
XPRESS 152 152 0.89 0.90
MI Cases Ctrls Acc PPV
OMOP definition 94 94 0.87 0.84
Noisy labeling process 94 94 0.85 0.80
XPRESS 94 94 0.89 0.86

Numbers for “PheKB definition” show the performance of the rule-based definitions for identifying T2DM cases (authored by Vanderbilt University) applied to our database.

Numbers for “OMOP definition” show the performance of the rule-based definitions for identifying MI cases (broad definition with hospitalization) applied to our database.

Numbers in the “Noisy labeling process” row show the performance of the keywords based label assignment.

Numbers against “XPRESS” show the performance of the XPRESS models for identifying cases.