SP4-19 Use of logistic regression and receiver operating characteristic curves to discriminate between blood donors and non-donors among North American medical students in Grenada, West Indies

IntroductionThe WHO estimates that 50 units of blood per 1000 inhabitants per year is necessary to meet the transfusion needs of a country. In Grenada 8.7 units are donated per 1000 inhabitants yearly.MethodsWe surveyed 452 St. George's University second year medical school students (>95% North American) regarding blood donation attitudes and experiences before and after arrival in Grenada. This analysis of the 363 responses, determined if age, gender, knowledge of blood group, time since first arrival in Grenada and donating blood in the year preceding arrival would predict who donated (14%) since arrival. In SPSS, we used logistic regression to estimate probabilities of donating blood since arrival as a function of each covariate. From these probabilities we generated Receiver Operating Characteristic curves with the area under each curve estimating the covariate's ability to predict blood donation after arrival in Grenada.Results68% of the time [95% CI 60% to 76%] students who had donated blood the year before arrival were more likely to be donors after arrival than those who had not. This percentage was slightly higher (72%, 95% CI 63% to 80%) if, compared to the same group, these students had been in Grenada for a longer time and knew their blood group. Neither gender nor age showed predictive ability.ConclusionBlood drive promotion should first target students who were not donors the year before arriving in Grenada as they are less likely to donate, without prompting, upon arrival.