Table 1.

Methods used in validations of diagnoses in the General Practice Research Database (GPRD)

		Method	Description	Example
Internal	1	Diagnostic algorithm	A diagnosis was validated by the presence of codes indicating specific symptoms/signs, prescriptions for disease-specific drugs and/or confirmatory test results	Andersohn [23] validated acute myocardial infarction by choosing only cases with codes for troponin tests, treatment with fibrinolytic drugs, coronary intervention or a hospital stay of >3 days
	2	Manual review of anonymized free text on computerized records	The complete computer records (including the anonymized free text) for individuals with a diagnosis were assessed for confirmatory evidence of disease status	Yang [15] validated colorectal cancer cases by reviewing the computerized records to look for clinical events to confirm the diagnosis
	3	Sensitivity analysis	In an analytical study, a comparison of the measure of effect using a broad set of disease/therapeutic codes with that of a restricted set more likely to represent true cases	Gupta [26] varied the definition of multiple sclerosis to examine how its association with inflammatory bowel disease changed
External	4	Questionnaire to GP	A questionnaire investigating various aspects of the computerized diagnosis was sent to the GP	Garcia Rodriguez [27] validated prostate cancer by comparing the computerized diagnosis with answers to a questionnaire filled by the GP regarding the diagnosis
	5	Record request to GP	Request to the GP to provide anonymized copies of paper medical records, hospital discharge summaries or death certificates. Obtained copies were examined to validate the diagnosis, using further diagnostic criteria	Hall [28] requested medical records of lung cancer patients in order to verify the cancer diagnosis made in the computerized records
	6	Comparison of rates	Measures of disease incidence, prevalence or patterns (e.g. time trends) from GPRD data were compared with a non-GPRD, UK-based data source	Hollowell [7] compared the incidence of chicken pox, allergic rhinitis, asthma and diabetes with external rates