Table 1 Three statistical methods for detection of tuberculosis outbreaks, 2008-2009
From: Using statistical methods and genotyping to detect tuberculosis outbreaks
Method | Description | Geographic unit | Time period | Parameters for alert threshold |
|---|---|---|---|---|
County-based LLR (current TB GIMS method) | Measures difference between the observed and expected geospatial concentration of cases | County | 2-year window and 3-year window | Currently designated in TB GIMS as a county-based LLR of ≥ 5 |
CUSUM | Calculates a monthly cumulative sum of variations between the observed and expected case counts; indicates an aberrant event above a threshold | County | Cumulative by month; 2-year moving window used to determine background rate | Threshold determined by background rate (based on 6th lowest case count among all 2 year windows), alternative rate (background rate + 3), null average run length (expect, on average, one false alarm every 100 months) |
SaTScan | Applies spatial scan statistic to determine areas with significantly higher geospatial concentration of cases | None assumed; a circular area is flexibly determined by algorithm; cases aggregated by zip code | 2-year window and 3-year window | P value to determine location of clusters (< 0.05) using maximum search radius of 20, 50, and 100 km for the circular scanning |