Table 2: Performance matching 2 real use case datasets.

From: Probabilistic record linkage of de-identified research datasets with discrepancies using diagnosis codes

Data time span Matching method Number of matches TPRa PPVa Computing time
6 years 0.5 cutoff 4,369 0.93 0.81 96 sb
6 years 0.9 cutoff 4,179 0.91 0.84 96 sb
6 years F-S blocked 2,594,443 0.81 <0.01 49 minb
6 years F-S blocked 1-1 5,696 0.38 0.26 49 minb
6 years F-S  > 4 daysc
6 years F-S 1-1  > 4 daysc
11 years 0.5 cutoff 4,043 0.84 0.80 96 sb
11 years 0.9 cutoff 3,625 0.80 0.84 96 sb
11 years F-S blocked 2,898,367 0.80 <0.01 62 minb
11 years F-S blocked 1-1 6,356 0.29 0.17 62 minb
11 years F-S  > 4 daysc
11 years F-S 1-1  > 4 daysc
  1. aBased on the 3,831 silver standard true matches.
  2. bUsing a 3.5 GHz Intel Core i7 processor with 32 GB of memory available.
  3. cUsing a 3.6 GHz Intel Xeon 5600 series processor with 96 GB of memory available.