The promise of record linkage for assessing the uptake of health services in resource constrained settings: a pilot study from South Africa

Table 1 Linkage scenarios by identifiers and string comparison techniques applied to names

		String comparison techniques applied to first and surnames
		Exact	JW ≥ 0.7	JW ≥ 0.9	DM	Soundex	JW ≥ 0.9 or DM or soundex
Identifiers used	Routinely collected identifiers*	S1	S2	S3	S4	S5	S6
	Routinely collected identifiers + household member first name	S7	S8	S9	S10	S11	S12
	Routinely collected identifiers + household member first name and surname			S13	S14	S15
	Deterministic linkage on National ID Number or telephone number followed by best of S1-S15**						S16
	S16 + clerical review of 5%, 10%, 15%, and 20% of record pairs above and below the threshold value above which record pairs are automatically accepted as matches						S17-S20

*Routinely collected identifiers = first name, last name, sex, day of birth, month of birth, year of birth and village; JW = Jaro-Winkler; DM = double metaphone code.
**The best of the 15 probabilistic linkage scenarios is the one that yields the maximum sensitivity and PPV.

ISSN: 1471-2288