Table 2. Sensitivity and Specificity of Tuberculosis Recent Transmission (RT) Estimates Using State-Based, SaTScan-Based, and County-Based Clustering Methods and Accuracy of Estimates Across a Range of Plausible Hypothetical RT Prevalence Rates (n = 1,085), Arkansas, Maryland, and Massachusetts, January 1, 1998–September 30, 2000.
| RT Estimation Method | No. of True-Positive Cases | No. of False-Positive Cases | No. of True-Negative Cases | No. of False-Negative Cases | Sensitivity, % | Specificity, % | Accuracy of Estimate by Hypothetical RT Prevalence, % | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||||
| 10% | 15% | 20% | 25% | 30% | 50% | |||||||
| State-based clustering | 72 | 129 | 884 | 0 | 100.0 | 87.3 | 88.5 | 89.2 | 89.8 | 90.5 | 91.1 | 93.6 |
| SaTScan-based clusteringa | 58 | 53 | 960 | 14 | 80.6 | 94.8 | 93.3 | 92.6 | 91.9 | 91.2 | 90.5 | 87.7 |
| County-based clustering | 58 | 68 | 945 | 14 | 80.6 | 93.3 | 92.0 | 91.4 | 90.7 | 90.1 | 89.5 | 86.9 |
Abbreviation: RT, recent transmission.
a
SaTScan was developed by Dr. Martin Kulldorff (Harvard Medical School, Boston, Massachusetts) in conjunction with Information Management Services, Inc. (Calverton, Maryland) (35).