Table 2:

Phenotype validation mechanisms

Mechanism	Description	Example
Disease registries	Compare the phenotype cohort with those present in the registry	Comparison of a diabetes phenotype cohort with those patients present in a diabetes registry (e.g., T1D exchange)
Chart review	Compare the phenotype cohort with the patients identified by manual review of medical records	Comparison with a diabetes gold standard, produced by double manual review of patient medical records
Cross-EHR concordance	Compare percentage of cases identified by a phenotype across different sources, and identify any overlap	Comparison of the percentage of patients identified by a diabetes phenotype in primary and secondary care EHRs, and the identification of any case overlap
Risk factors	Compare the magnitude of the phenotype cohort with standard risk calculations	Comparison with the output of a Cox hazards model
Prognosis	Compare the magnitude of the phenotype cohort with external prognosis models	Comparison with a survival analysis
Genetic associations	Compare whether the presence of a patient in a phenotype cohort is consistent with their genetic profile	A patient is more likely to be a valid member of a diabetes cohort if they have the HLA-DR3 gene