Table 5. Characteristic of records with discordant classification by deterministic and probabilistic record linkage.

Record characteristics	Pair by deterministic and non-pair by probabilistic linkage		Pair by probabilistic and non-pair by deterministic linkage
	(N = 1,285)		(N = 527)
			Median	CI95%
Score	-	-	24.2	20.3–32.3
	n	%	n	%
Missing sex value	0	0	0	0
Missing name value	0	0	0	0
Missing mother’s name value	68	5.3	76	14.4
Missing date of birth value	81	6.3	58	11.0
Missing address value	7	0.5	20	3.8
Combined: unknown mother’s name; or unknown date of birth; or unknown address	129	10.0	141	26.7
Link characteristics	Pair by deterministic and non-pair by probabilistic linkagea,b		Pair by probabilistic and non-pair by deterministic linkagea,b
	(N = 733)		(N = 293)
	n	%	n	%
Difference in sex	115	15.7	0	0
Similarity measure for name lower than 70,0%	160	25.9	0	0
Similarity measure for mother’s name lower than 70,0%	147	26.6	36	16.6
Similarity measure for date of birth lower than 70,0%	45	8.3	25	10.6
	Median	IC95%	Median	IC95%
Similarity measure for name x 100	81.6	69.4–94.4	100	95.0–100
Similarity measure for mother’s name x 100	91.3	68.2–100	94.4	85.7–100
Similarity measure for date of birth	100	10.0–100	87.5	75.0–100

^a Calculating Levenshtein distance and assessing the difference in sex between records required comparing the records of the record group of the same patient. Comparing discordant records: (i) when the two discordant records were identified by only one of the techniques, the calculation was done between them; (ii) when they were records identified by both techniques and only one of them was not identified by one of the techniques, the calculation was done by comparing the unidentified record with the record of highest score in the group.

^b For this calculation, the group of records that were blocked by sex or had missing information for one of the variables were excluded.