Skip to main content
. 2014 Oct 31;22(1):65–75. doi: 10.1136/amiajnl-2013-002577

Table 2:

Ranking performance

Rank P1 (using ontology) Cumulative P1 P2 (Lucene matching) Cumulative P2
1 63.9% (n = 122) 63.9% (n = 122) 51.3% (n = 98) 51.3% (n = 98)
2 14.1% (n = 27) 78.0% (n = 149) 12.0% (n = 23) 63.4% (n = 121)
3 8.40% (n = 16) 86.4% (n = 165) 8.37% (n = 16) 71.7% (n = 137)
4 3.10% (n = 6) 89.5% (n = 171) 4.18% (n = 8) 75.9% (n = 145)
5 3.70% (n = 7) 93.2% (n = 178) 5.23% (n = 10) 81.2% (n = 155)
6 3.10% (n = 6) 96.3% (n = 184) 1.04% (n = 2) 82.2% (n = 157)
7 0.00% (n = 6) 96.3% (n = 184) 0.00% (n = 0) 82.2% (n = 157)
8 1.50% (n = 3) 97.8% (n = 187) 1.04% (n = 2) 83.2% (n = 159)
9 0.60% (n = 1) 98.4% (n = 188) 2.09% (n = 4) 85.3% (n = 163)
10 0.00% (n = 0) 98.4% (n = 188) 0.52% (n = 1) 85.6% (n = 164)
≥10 0.00% (n = 0) 98.4% (n = 188) 3.66% (n = 7) 89.5% (n = 171)
Not found 1.60% (n = 3) 10.5% (n = 20)
Total 100% (n = 191) 100% (n = 191)

P1,2 shows the rank of 191 expert selected ‘best’ matches within the automatically produced lists of relevant matches, using ontology annotations of the desired data elements or Lucene matching only, respectively. BiobankConnect predicted ‘best’ matches as first choice (rank 1) in 63.9% of cases and within the ‘top 10’ in 98.4% of cases.