Premium
Estimating the incidence of HIV infection in repeat blood donors with low average donation frequency
Author(s) -
Brambilla Donald J.,
Busch Michael P.,
Glynn Simone A.,
Kleinman Steven H.
Publication year - 2021
Publication title -
transfusion
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.045
H-Index - 132
eISSN - 1537-2995
pISSN - 0041-1132
DOI - 10.1111/trf.16144
Subject(s) - incidence (geometry) , confidence interval , donation , medicine , blood donations , interval (graph theory) , blood donor , human immunodeficiency virus (hiv) , estimation , demography , statistics , immunology , mathematics , geometry , management , combinatorics , sociology , economics , economic growth
Background The standard approach to estimating HIV incidence in repeat blood donors includes only donors who made two or more donations in an estimation interval. In China and some other countries, large proportions of repeat donors donate only once in a 1‐ or 2‐year interval. The standard approach may not represent risk among all repeat donors in these areas. Two approaches to including all repeat donors in the incidence estimate were evaluated in a simulation study. Study Design and Methods Under one approach, a donor infected at the first donation contributes a partial case to incidence that equals the proportion of time since the preceding donation that is in the estimation interval. Under the other, that donor contributes a full case if at least half the time since the previous donation is in the estimation interval and nothing otherwise. Infections identified at the second or subsequent donations in the interval contribute full cases as usual. The simulations involved proportions with single donations of 11% to 65% combined with a variety of patterns of rising, falling, or constant incidence. Results The partial‐case approach was unbiased under more test conditions than the whole‐case approach and exhibited smaller bias when both were biased. Under both approaches, bias >10% occurred only when rates of single donations >50% were combined with large changes in incidence over time. Conclusion The partial‐case approach performed better than the whole‐case approach. The conditions producing bias >10% are so extreme that they are unlikely to be encountered in the field.