Background:
California and Virginia are developing and evaluating the use of automated aberration detection systems in sexually transmitted disease (STD) surveillance. Recent attention on public health aberration detection systems has focused on syndromic surveillance and rare bio-terrorism agents. These systems may also be applicable to detection of aberrations in STD surveillance systems. The parameters of such systems need to be optimized to provide reasonable determination of an unusual occurrence of events.

Objective:
To describe general statistical issues, issues specific to systems used for disease surveillance, evaluation issues, and specific statistical algorithms.

Method:
In California, three statistical algorithms were evaluated to determine operational parameters, such as detection sensitivity, in a STD surveillance system: Historic Limits, Poisson regression, and CUSUM. In Virginia, PROC MACONTROL in SAS was used with a aberration detection point set at two standard deviations (SD) from the long-term historical weekly mean.

Result:
In California, Historic Limits (“MMWR”) algorithm with 3 SD cut point and elimination of county periods with < 3 cases still results in 101 flags, but only 5 jurisdictions with more than 5 flags. Too many flags were generated using “standard” parameter values (e.g. 95% CI; inclusion of small case counts) for available resources.

Conclusion:
STDs represent a large proportion of reportable diseases, making automated systems essential for rapid detection of outbreaks and aberrations in reporting. Initial analysis of algorithms and parameter choices indicates values need to be set conservatively or too many flags will be generated for the systems to be useful.

Implications:
These types of systems may help provide a safety net to detect outbreaks that have been overlooked by other routine activities and/or detect outbreaks at one geographic level that are not detected by systems at another level. They are also critical for the detection and remediation of problems in reporting and surveillance.