|
Abstract:
|
HIV incidence is an important measure of the HIV epidemic. However, it is difficult to know the annual numbers of new infections because HIV is a silent disease with a long, asymptomatic, incubation time period. The number of diagnoses could be very different from the number of infections in a given time period due to diagnosis delay (time from infection to diagnosis). If HIV diagnosis delay is relatively short, then the number of diagnoses could be used as an approximation of incidence. A recent publication (Xia et al. JAIDS 2017) proposed an estimator based on the number of diagnoses in the year of interest and adjacent years. To evaluate the performance of this incidence estimator, we conducted a simulation study. We generated data under various conditions and estimated HIV incidence using several estimators based on the number of diagnoses including the estimator proposed by Xia et al. Bias and the confidence-interval coverage are used to evaluate the performance of each estimator. Results show that under the assumption of constant incidence and testing rate, all estimators are unbiased, but Xia's estimator is the most unstable.
|
