![[ Visit Client Website ]](images/banner.gif)
Background: In November 2015, the Florida Department of Health made several changes to their sexually transmitted disease STD surveillance system, PRISM, aimed to improve data quality of key variables. These modifications included requiring sexual orientation, self-reported gender, and partner meeting location to be completed before closing; and skip patterns in the interview section. The aim of this project is to evaluate the effect of these modifications on data completeness.
Methods: All reported cases of STDs from December 2014 through February 2015, and December 2015 through February 2016, were extracted from Florida’s PRISM. Reported cases were compared between the pre-implementation data completion percentages to the post-implementation for sexual orientation, self-reported gender, partner meeting locations, and STD interview risks.
Results: Implementation of required fields reduced unknown values from pre-implementation to post-implementation for: sexual orientation from 90% (n= 25,714) to 53% (n= 16,512); self-reported gender from 85% (n=24,292) to 20% (n=6,195); and partner meeting locations from 99% (n=4,281) to 35% (n=705). With a coding fix in January 2016 partner meeting locations missing data were reduced further to 17% (n=72). Moreover, the modifications identified more significant rare events such as bisexual orientation (246 post vs. 88 pre-implementation) and transgender (37 vs. 6). Mean time from report date to field record closure date remained similar (9 days pre- vs. 8 days post-implementation). Completeness of STD risk factor questions was similar (32.8 pre- vs. 30.8 post-implementation) as was the percentage with indication of STD risk (“no” 69% pre vs. 72% post and “yes” 19% pre vs. 16% post).
Conclusions: Requiring key variables within a surveillance system can significantly improve data completeness while identifying more targeted rare events and, in this case, without lengthening the reporting process. Skip patterns will need evaluation beyond data completeness to identify the value of the modification.