Wed 6 Nov, 2019 07:51 am
I am validating an automated surveillance program. It is about infection incidence on a hospital ward. I could easily check if the small list of infections produced by the automated surveillance includes all patients that really have an infection (in other words; calculate the positive predictive value). The list of patients not labeled as infected by the program is a whole lot longer. This brings me to two questions:
1. I'm not sure if I should get a sample size of the infection-negatives and check if they are indeed infection-negative. Or should I get a smaller (for example 6 month cohort with 300 patients) and check all of them?
2. Is combining methods even an option (check all of the positive-labeled patients and only a sample of the negative-labeled patients)?
Thanks in advance,