Table 3

Summary of performance of outbreak detection algorithms for RRV data from Brisbane and Townsville.

Method

LGA


Brisbane

Townsville


TP

FP

FN

(n = 19)

Peak FN

(n = 9)

TP

FP

FN

(n = 23)

Peak FN

(n = 11)


HLM-A

40.0

10.0

54.5

62.5

12.9

19.4

50.0

60.0

(n = 11)

(n = 8)

(n = 8)

(n = 5)

HLM-B

42.1

5.3

54.5

50.0

20.0

24.0

50.0

60.0

(n = 11)

(n = 8)

(n = 8)

(n = 5)

C1-A

41.2

41.2

68.4

44.4

23.3

37.2

56.5

36.4

C1-B

46.2

30.8

68.4

44.4

35.0

45.0

69.6

45.5

C2-A

53.3

33.3

57.9

33.3

36.4

54.6

65.2

36.4

C2-B

58.3

16.7

63.2

22.2

31.6

36.8

73.9

45.5

C2-C

60.0

40.0

63.2

22.2

27.8

44.4

73.9

45.5

C2-D

60.0

30.0

68.4

33.3

41.7

33.3

78.3

54.5

NBC-A

81.8

9.1

52.6

22.2

57.1

19.1

56.5

18.1

NBC-B

73.3

13.3

42.1

11.1

57.9

31.6

56.5

18.1

NBC-C

88.9

11.1

52.6

0.0

57.1

42.9

60.9

18.1

NBC-D

72.7

27.3

47.4

33.3

53.3

40.0

60.9

18.1

NBC-E

40.6

34.4

47.4

33.3

44.8

13.8

47.8

36.4

NBC-F

40.6

37.5

47.4

33.3

46.9

21.9

39.1

27.3

POD-A

61.5

23.1

36.4

12.5

69.2

23.1

37.5

0.0

(n = 11)

(n = 8)

(n = 8)

(n = 5)

POD-B

70.0

20.0

36.4

12.5

75.0

25.0

37.5

0.0

(n = 11)

(n = 8)

(n = 8)

(n = 5)

SaTScan

62.5

25.0

45.5

25.0

44.4

44.4

50.0

20.0

(n = 11)

(n = 8)

(n = 8)

(n = 5)


True positive (TP) values represent the percentage of signals that overlapped with a DSP, false positives (FP) are categorised as the percentage of signals lasting ≥2 weeks which do not overlap with a DSP, false negatives (FN) are the percentage of all DSPs that were not detected by the algorithm, and peak FN is the percentage of DSPs starting between 1 December and 30 April that were not detected by the algorithm. Algorithm codes are the same as detailed in Figure 1.

Pelecanos et al. BMC Medical Informatics and Decision Making 2010 10:74   doi:10.1186/1472-6947-10-74

Open Data