Empirical aspects of record linkage across multiple data sets using statistical linkage keys: the experience of the PIAC cohort study

Karmel, Rosemary; Anderson, Phil; Gibson, Diane; Peut, Ann; Duckett, Stephen; Wells, Yvonne

doi:10.1186/1472-6963-10-41

Table 4 Criteria for selecting match keys for PIAC stage 3 (examples for 60 keys)

From: Empirical aspects of record linkage across multiple data sets using statistical linkage keys: the experience of the PIAC cohort study

Key no.	Linkage key	Joint. unique key rate (measure A)	^(a)Est. number of links	Est. FMR (measure B)	^(b)Comparison key	Marginal true: false (measure C)	^(c)Est. 'worst case' FMR
1	s3g2\|dmYOB\|s\|pc	99.999	55631	0.00	701	>1000	0.04
2	s3g2\|dmYOB\|_\|pc	99.957	56120	0.00	702	>1000	0.09
3	s3g2\|dm_ob\|s\|pc	99.878	57047	0.01	703	>1000	0.82
4	s3g2\|dmYOB\|s\|pc2	99.993	63788	0.01	704	>1000	0.55
5	s3_\|dmYOB\|s\|pc	99.896	56819	0.01	705	925.9	0.48
6	s3g2\|dm_ob\|_\|pc	99.878	57547	0.02	706	578.7	1.63
7	s3g2\|dmYOB\|_\|pc2	99.934	64338	0.02	707	592.1	1.09
8	s3_\|dmYOB\|_\|pc	99.896	57326	0.03	708	466.2	0.95
9	s3g2\|dmYOB\|s\|st	99.981	67206	0.04	709	317.7	1.93
10	s3g2\|dmYOB\|_\|st	99.897	67781	0.08	710	159.5	3.82
11	s3g2\|__YOB\|s\|pc	99.715	58484	0.12	711	103.9	15.40
12	_g2\|dmYOB\|s\|pc	99.797	56031	0.14	712	88.2	3.17
13	s3g2\|dmYOB\|s\|_	99.792	67743	0.17	713	80.7	5.74
14	s3g2\|__YOB\|_\|pc	99.613	59012	0.23	714	51.9	30.52
15	_g2\|dmYOB\|_\|pc	99.707	56541	0.27	715	44.0	6.28
16	s3g2\|dmYOB\|_\|_	99.650	68327	0.29	716	44.9	10.23
17	s3g2\|dm_ob\|s\|pc2	99.647	65447	0.34	717	36.9	10.16
18	s3_\|dm_ob\|s\|pc	99.478	58319	0.41	718	28.9	8.84
19	s3_\|dmYOB\|s\|pc2	99.583	65185	0.43	719	29.5	5.90
20	*s3g2\|dm_ob\|_\|pc2*	*99.496*	*66024*	*0.67*	*720*	*18.1*	*20.14*
601	s3g2\|dmYOB\|s\|pc	100.000	44977	0.00	. .	. .	0.00
602	s3g2\|dmYOB\|_\|pc	99.998	45392	0.00	601	>1000	0.00
603	s3g2\|dm_ob\|s\|pc	99.998	46105	0.00	601	>1000	0.01
604	s3g2\|dmYOB\|s\|pc2	100.000	51170	0.00	601	>1000	0.00
605	s3_\|dmYOB\|s\|pc	99.992	45855	0.00	601	>1000	0.00
606	s3g2\|dm_ob\|_\|pc	99.998	46529	0.00	603	>1000	0.01
607	s3g2\|dmYOB\|_\|pc2	99.998	51629	0.00	604	>1000	0.01
608	s3_\|dmYOB\|_\|pc	99.992	46276	0.00	602	>1000	0.01
609	s3g2\|dmYOB\|s\|st	100.000	53592	0.00	604	>1000	0.02
610	s3g2\|dmYOB\|_\|st	99.998	54071	0.00	609	>1000	0.03
611	s3g2\|__YOB\|s\|pc	99.976	47166	0.00	601	>1000	0.12
612	_g2\|dmYOB\|s\|pc	99.978	45258	0.00	601	>1000	0.02
613	s3g2\|dmYOB\|s\|_	100.000	53901	0.00	609	>1000	0.04
614	s3g2\|__YOB\|_\|pc	99.962	47607	0.00	602	>1000	0.23
615	_g2\|dmYOB\|_\|pc	99.976	45678	0.00	612	>1000	0.05
616	s3g2\|dmYOB\|_\|_	99.998	54382	0.00	613	>1000	0.09
617	s3g2\|dm_ob\|s\|pc2	99.994	52466	0.00	604	>1000	0.08
618	s3_\|dm_ob\|s\|pc	99.986	47016	0.00	606	776.8	0.07
619	s3_\|dmYOB\|s\|pc2	99.968	52178	0.00	604	>1000	0.05
620	s3g2\|dm_ob\|_\|pc2	99.992	52936	0.00	617	772.5	0.16
701	s3g2\|dmYOB\|s\|pc	100.000	49060	0.00	. .	. .	0.00
702	s3g2\|dmYOB\|_\|pc	99.984	49502	0.00	701	>1000	0.00
703	s3g2\|dm_ob\|s\|pc	99.957	50305	0.00	701	>1000	0.04
704	s3g2\|dmYOB\|s\|pc2	99.996	55840	0.00	701	>1000	0.03
705	s3_\|dmYOB\|s\|pc	99.952	50034	0.00	701	>1000	0.02
706	s3g2\|dm_ob\|_\|pc	99.957	50757	0.00	703	>1000	0.07
707	s3g2\|dmYOB\|_\|pc2	99.977	56333	0.00	704	>1000	0.05
708	s3_\|dmYOB\|_\|pc	99.952	50486	0.00	702	>1000	0.04
709	s3g2\|dmYOB\|s\|st	99.989	58515	0.00	704	>1000	0.09
710	s3g2\|dmYOB\|_\|st	99.965	59029	0.00	709	>1000	0.18
711	s3g2\|__YOB\|s\|pc	99.847	51479	0.00	701	>1000	0.70
712	_g2\|dmYOB\|s\|pc	99.869	49369	0.00	701	152.5	0.14
713	s3g2\|dmYOB\|s\|_	99.944	58836	0.01	709	144.2	0.26
714	s3g2\|__YOB\|_\|pc	99.792	51951	0.01	702	679.1	1.39
715	_g2\|dmYOB\|_\|pc	99.822	49816	0.01	712	220.5	0.29
716	s3g2\|dmYOB\|_\|_	99.892	59352	0.01	713	174.0	0.52
717	s3g2\|dm_ob\|s\|pc2	99.855	57266	0.01	704	251.2	0.46
718	s3_\|dm_ob\|s\|pc	99.715	51314	0.01	706	91.6	0.40
719	s3_\|dmYOB\|s\|pc2	99.803	56960	0.01	704	156.5	0.27
720	s3g2\|dm_ob\|_\|pc2	99.796	57771	0.02	717	85.5	0.92

(a) Estimated number of links was derived from simple deterministic matching on the key (retaining only one occurrence of duplicates).
(b) Comparative linkage key is one which is slightly more detailed and includes all the match key elements of the current key. There is not a strict hierarchy for the linkage keys, so in some cases there may be more than one appropriate key for the comparison.
(c) 'Worst case' FMR is estimated assuming that the number of categories within a key element is equal to that implied by the most common category (s3: 72, g2: 11, dmob: 182, yob: 19, s: 2, st: 3, pc2: 11, pc: 156, aged care assessment date: 161, assessment team identifier: 25).
Note: See note to Table 3 for definition of keys; '600' series include assessment date; '700' series linkage keys include assessment team identifier. Table only includes keys that were expected to have fewer than four times as many people with non-unique match keys as SLK-581. This equates to key 20 if client region is the only additional match data, key 64 if aged care assessment date and region are included and key 46 if assessment team identifier and region are included. Keys in bold italicsare those identified as not selected for use. Table showing all tested keys is available from the authors on request.

Back to article page

ISSN: 1472-6963

Contact us

General enquiries: journalsubmissions@springernature.com

BMC Health Services Research

Contact us