Developing an Automated System for Assigning Reactor Dispositions

Background: Rising numbers of cases of STDs in San Francisco have forced us to examine efficiencies of  our activities.  Protocols for follow-up of reported reactive serologic tests for syphilis ("reactors") were developed years ago when the number of syphilis cases were much lower than today, syphilis elimination was our primary concern, and our computer system was less capable of automation.  Our objective was to develop an algorithm to automatically remove biological false positives and results reported on old cases from the manual review of new reactors, leaving us with more time for other activities.

Methods: Protocols for manually assigning a disposition to each reactor were reviewed, and clinicians and health workers performing this task were interviewed.  Rules used to classify reactors as non-cases were formalized in terms of the data in each patient's computer record.  Rules were then applied retrospectively to all reactors reported in 2015, and the results of the algorithm were compared with the dispositions assigned manually.

Results: Of 12,044 reactors reported in 2015, 5324 would have been automatically closed as an old case or a BFP based on the algorithm we developed.  This included 74 reactors that had been classified as new cases by program staff.  A manual review of these cases by two independent reviewers revealed that 15 should not have been classified as new cases, and that the remainder would have been identified when treatment information was reported by the provider.

Conclusions: Using an automated algorithm will allow us to correctly disposition nearly 45 percent of reported reactors without any manual review.  Not only would we not miss many cases, using an automated system may prevent reactors from being classified incorrectly as new cases.  Automating this process will decrease workload and increase accuracy of syphilis reactor dispositions.