Analysis of blood donor return behaviour using survival regression methods.

The acquisition of a sufficient and safe supply of blood products is required to support modern medical care; in most countries this has meant a reliance on voluntary, nonremunerated blood donors. Recent reported shortages in the United States and elsewhere have highlighted the need for a method to measure and evaluate blood donor return behaviour. This paper describes a framework within which standard time-to-outcome methods can be used to analyse blood donor return behaviour. Survival curves and relative risk estimates derived from a proportional hazards analysis of a large administrative dataset are reported. In addition to assessing the effect of sex, age and other key donor demographic factors on the probability of a subsequent donation attempt, the analysis reveals that the relative risks are time-dependent. This suggests that the likelihood of attempting a subsequent donation may also depend on the time since the index donation attempt. The implications for blood collection agencies and transfusion researchers of this new perspective on donor behaviour are discussed.