Predicting a donor's likelihood of donating within a preselected time interval

The procurement of some advanced blood components, like quarantined plasma units, depends critically on retesting the donor within a fixed time frame. For health care systems, such as that in Germany, with mandatory retesting of donors before plasma release, the reliable identification of donors who are more likely to return in time has an immense practical implication, be cause their blood components could be preferably selected for quarantine pu rposes. The donation histories of about 760 000 donors with 4910 000 donati on attempts were analysed. We developed a logistic regression model to calc ulate a probability of donation, p(Dts-te), within a preselected time frame (t(s)-t(e)). The donation history was compounded in a score and shown to b e very useful for determining p(Dts-te). A logistic regression model was de veloped with score and donor status as parameters; different regression coe fficients applied to first-time-donors (ftd) and to repeat donors (intercep t, int, and score factor, scf ). This model allowed us to determine the pro bability of donation, p(Dts-te), within a preselected time interval, e.g. 6 -9 months after an index donation. The p(Dts-te) can be calculated for any donor of blood services. The p(D170-275 days) ranged from about 22% to 86% for any index donation in 1996/97. First-time donors had a p(D170-275 days) of 33% and were more likely to return within the time interval than certai n subsets of repeat donors who can be defined by our model. We provided a t echnical procedure to increase the rate of plasma unit release after quaran tine storage and showed the usefulness of our procedure for blood component management, if quarantine storage is required. By applying the model to ou r current plasma quarantine programme we could retrieve about 30% more unit s, which would represent about 30 000 units per year, without incurring add itional costs. General implications for blood collection, like planning blo od drives, were discussed. The whole demand of a health care system for sin gle plasma units may be met by quarantine plasma and their cost-efficiency can be improved.