Surveillance is not a study, but we still need a study design, cohort or cross-sectional, to analyze surveillance data. Here is another example that a wrong study design is used when analyzing surveillance data.

academic.oup.com/aje/article-...

Thank you for your response. I am posting two more papers on this topic.

journals.lww.com/epidem/fullt...

www.tandfonline.com/doi/10.1080/...

The study design should be a cohort study, not a cross-sectional study. The proportion of late diagnoses is not only affected by HIV testing, but also by HIV incidence. A decreasing HIV incidence can drive the proportion of late diagnoses up.

The proportion of late diagnoses is a wrong metric. The denominator should be the number of infections in a year, not the number of diagnoses in a year.
journals.lww.com/aidsonline/c...

en.wikipedia.org/wiki/The_Bab...

Grok 3?

bsky.app/profile/nata...