TABLE 2.

Outcomes captured in the Integrated Medical Database of National Taiwan University Hospital (NTUH-iMD) compared to the linked National Health Insurance Research Database (NHIRD) among oral hypoglycemic agent users with continuous 6-month baseline and 24-month follow-up data (n = 3,845).

		NHIRD		EHR-claims agreement	EHR outcome representativeness
		Yes	No
a) Cardiovascular events
NTUH-iMD	Yes	143	24	96%	51%
	No	140	3,538
b) Nephropathy-related events
NTUH-iMD	Yes	793	66	81%	55%
	No	648	2,338
c) Heart failure admission
NTUH-iMD	Yes	39	28	98%	49%
	No	40	3,738

Outcome events were assessed during the 24-month follow-up period. We calculated EHR-claims agreement by dividing the number of patients whose outcome status were coded in the same way in both the EHRs and linked claims database by the total number of patients who appeared in both databases. A patient defined as having the outcome in both the EHRs and the claims database, or not having the outcome in either database, would suggest high agreement. We calculated EHR outcome representativeness by dividing the number of patients classified as having the outcome in both the EHRs and the claims database by the number of patients classified as having the outcome in the claims database. See Supplementary Table S3 for the definitions and calculations of EHR-claims agreement and EHR representativeness.

EHR, electronic health record.