In the evaluation of suveillance systems for infectious conditions, two attributes of interest are sensitivity and Positive Predictive Value. That is, the proportion of all true cases detected, and the proportion of all reported cases that are true cases. Calculation of these values relies on a laboratory or other gold standard.
For most behavioral conditions (e.g., suicide) surveillance is done using a survey of self-reported behaviors. I am wondering how we adapt sensitivity and PPV to this situation.
Clearly reliability, and validity are relevant. Has anyone had experience with this sort of evaluation?
------------------------------
Donna Stroup
Director
Data for Solutions, Inc.
------------------------------