Table 4.
Performance of algorithms to identify cases and controls from EMRs for five primary phenotypes.
GHC1 | MCRF2 | Mayo3 | NU4 | VU5 | |
---|---|---|---|---|---|
Primary Phenotype | Dementia | Cataract | Peripheral Arterial Disease | Type 2 diabetes | Cardiac conduction (Quantitative Trait) |
EMR data sources to define phenotype | Diagnoses, Medications | Diagnoses, Procedures, Medications | Procedure Reports | Diagnoses, Laboratory Tests, Medications | Diagnoses, Laboratory Tests, Medications, ECG Results |
Method to validate EMR phenotype | Physician Review* | Trained Chart Reviewers | Compared to Clinical Gold Standard | Physician Review | Physician Review |
Number of Cases/Controls | 747/2043 | 2642/1322 | 1679/1657 | 756/777 | 2950 |
Biospecimen # | 2790 | 19771 | 3336 | 8161 | 81952 |
% of total biospecimen pool | 26.8% | 13.4% | 50.3% | 9.3% | 3.6% |
PPV (case/control) | 73% | 98%/98% | 94%/99% | 98%/100% | 97% |
Review team included two physicians, a psychometrician, a neuropsychologist, and a study nurse
Group Health Cooperative
Marshfield Clinic Research Foundation
Mayo Clinic
Northwestern University
Vanderbilt University