The assumption of no long reporting delays may result in underestimates of US AIDS incidence

Abstract

OBJECTIVE: To evaluate the effect of the assumption of no long reporting delays on estimates of AIDS incidence. DESIGN: Reported AIDS cases must be adjusted for reporting delays to estimate AIDS incidence. We compared the adjustments supplied with the Centers for Disease Control and Prevention (CDC) AIDS Public Information Data Set with a set of adjustments that differ with respect to CDC assumption of no long delays. Both methods assume that no changes in delay have occurred throughout the reporting period. METHODS: Probability distributions of reporting delays were calculated from the delay adjustment weights supplied by CDC, and from an alternative method that estimates the probability of long delays from the surveillance data. AIDS incidence estimates from these two approaches were calculated and compared. RESULTS: The CDC adjustments assume that there will be no reporting delays longer than 61 months, whereas the alternative method estimates that 5.9% of case reports will be delayed longer than 61 months. The CDC adjustments result in lower estimates of AIDS incidence and a flattening of the epidemic. CONCLUSIONS: In addition to a 6.2% reduction in total estimated AIDS incidence, the CDC assumption changes the shape of the estimated epidemic. These may result in as much as 4-21% reductions in model estimates of HIV incidence and prevalence