Table 3.

The performance of different algorithms for identifying when antibiotic prophylaxis has been prescribed in large administrative data sets.

VARIABLE	AP* LIKELY VERSUS AP POSSIBLE OR AP UNLIKELY	AP LIKELY OR AP POSSIBLE VERSUS AP UNLIKELY
Algorithm A
Sensitivity, % (confidence interval)	25.4 (19.4 to 32.1)	18.0 (13.6 to 23.0)
Specificity, % (confidence interval)	100 (99.7 to 100)	100 (99.7 to 100)
PPV,† %	100	100
NPV,‡ % (confidence interval)	90.6 (89.8 to 91.2)	85.3 (84.6 to 86.0)
Positive likelihood ratio	NA§	NA
Negative likelihood ratio (confidence interval)	0.75 (0.69 to 0.81)	0.82 (0.78 to 19.31)
Algorithm B
Sensitivity, % (confidence interval)	53.9 (46.6 to 61.1)	42.5 (36.6 to 48.6)
Specificity, % (confidence interval)	99.0 (98.3 to 99.4)	99.9 (99.5 to 100)
PPV, % (confidence interval)	88.1 (81.3 to 92.7)	98.3 (93.5 to 99.6)
NPV, % (confidence interval)	93.9 (93.0 to 94.7)	89.2 (88.2 to 90.2)
Positive likelihood ratio (confidence interval)	53.15 (31.09 to 90.96)	276.40 (68.73 to 111.63)
Negative likelihood ratio (confidence interval)	0.47 (0.40 to 0.54)	0.58 (0.52 to 0.64)
Algorithm C
Sensitivity, % (confidence interval)	87.6 (82.1 to 91.9)	75.8 (70.3 to 80.8)
Specificity, % (confidence interval)	95.5 (94.3 to 96.5)	98.2 (97.3 to 98.8)
PPV, % (confidence interval)	73.2 (68.0 to 77.8)	89.6 (85.2 to 92.8)
NPV, % (confidence interval)	98.2 (97.4 to 98.8)	95.1 (94.0 to 96.0)
Positive likelihood ratio (confidence interval)	19.50 (15.21 to 25.02)	41.10 (27.50 to 61.44)
Negative likelihood ratio (confidence interval)	0.13 (0.09 to 0.19)	0.25 (0.20 to 0.30)

^*AP: Antibiotic prophylaxis.

^†PPV: Positive predictive value.

^‡NPV: Negative predictive value.

^§NA: Not applicable.