Methodologic considerations in the design and analysis of nested case-control studies: association between cytokines and postoperative delirium

Table 3 Performance of Greedy and Optimal Algorithm at Different Levels of Caliper for a 1:2 Match

Match Algorithm	Age Caliper	GCP Caliper	Number of Pairs	Number Cases/Controls^b	Mean Distance	Sum of Distance
Greedy	1	1	8	7/6/1	1.02	8.15
Optimal	1	1	8	7/6/1	1.02	8.15
Greedy	2	2	28	21/14/7	1.90	53.29
Optimal	2	2	28	21/14/7	1.90	53.29
Greedy	3	3	38	26/14/12	2.28	86.54
Optimal	3	3	38	26/14/12	2.28	86.63
Greedy	4	4	48	32/16/16	2.87	137.9
Optimal	4	4	49	33/17/16	2.97	145.7
Greedy	5^a	5	56	34/12/22	3.47	194.4
Optimal	5^a	5	60	36/12/24	3.91	234.4
Greedy	6	6	62	37/12/25	3.98	247.1
Optimal	6	6	67	39/11/28	4.27	286.1

^aCaliper is defined to be the minimum allowable difference between the case and the control. In our study, we used a caliper of five for age and GCP, and zero for the other four categorical variables
GCP general cognitive performance
^bThere are three numbers listed in this column, for example 7/6/1 means seven cases were matched, six cases matched to 1:1, so six controls; and one case matched to 1:2, so two controls. This gives a total of eight matched pairs

ISSN: 1471-2288