Volume 15, Number 4—April 2009
Research
Enhancing Time-Series Detection Algorithms for Automated Biosurveillance
Table 3
Sensitivity for detection of additional counts, by method and dataset, for selected BioSense data used in algorithm modification study*
Minimum SD | Stratified baseline | Baseline duration, d | Sensitivity |
||||
---|---|---|---|---|---|---|---|
Department of Defense |
Hospital emergency department |
||||||
Count | Rate | Count | Rate | ||||
0.2 | No | 7 | 40.6† | 43.9 | 40.2† | 39.1 | |
1.0 | No | 7 | 52.3 | 70.8 | 50.4 | 53.6 | |
1.0 | No | 14 | 58.6 | 76.8 | 58.7 | 60.9 | |
1.0 | No | 28 | 62.0 | 79.4 | 62.8 | 64.8‡ | |
1.0 | Yes | 7 | 64.9 | 75.7 | 50.2 | 53.8 | |
1.0 | Yes | 14 | 75.1 | 80.4 | 57.6 | 60.1 | |
1.0 | Yes | 28 | 77.0 | 82.0‡ | 60.5 | 62.1 |
*All facility–syndrome days were included in calculations. The number of additional counts varied according to categories of average count for each facility–syndrome (0.5–<2, 2–<4, 4–<6, 6–<8, 8–<10, 10–<20, 20–<40, and >40) to produce 40% sensitivity for the initial method. For the Department of Defense, the additional counts were 5.0, 9.1, 11.7, 13.6, 16.0, 20.9, 30.4, and 40.0 for the average count categories, respectively. For the hospital emergency departments, the additional counts were 4.3, 6.3, 8.2, 9.5, 10.4, 12.9, 18.7, and 28.2, respectively.
†Initial method.
‡Best method for the dataset.