9

whole blood donation is associated with a higher LOC rate compared to apheresis donation [4,5] and 2 M. Bravo et al. 2011 The Author(s) Vox Sanguinis 2011 International Society of Blood Transfusion, Vox Sanguinis (2011) because the physiology and duration of apheresis donations are different. When all data cleaning activities were completed for Periods 2, 3A and 3B, the dataset contained of 554 534 donation records of donors who gave or intended to give allogeneic whole blood during the 12-month study period. Statistical analysis The proportion of donors who fainted was calculated by dividing the number fainting by the total number of presentations or donations in each corresponding demographic group or other grouping variable. Rates per 1000 presentations or per 1000 donations depending on the period are reported. Reacting donors in previous periods were excluded from the denominators before calculations are made for subsequent periods. For each period, separate multivariable logistic models were fit to the data, comparing fainting vs. not fainting to identify factors associated with the risk of reaction in each period. The approach to model building was not the same for all four periods. Because there were a small number of fainting events in Periods 1 and 3B, we preselected age, sex, donor status, donor site and EBV in the multivariable analysis based on previous studies of aggregate reaction rates in which these factors were significant [1–4]. Variables that were significant (P £ 0Æ05) were kept in each model. For period 3B reactions, we also conducted likelihood ratio tests to assess whether each variable represented a statistically significant improvement to the overall model. We report odds ratios (OR) and 95% confidence intervals (CI) of all variables included in the final models. For Periods 2 and 3A, unadjusted (univariate) ORs and 95% CIs were initially computed for each factor and provided an assessment of the crude association of each variable with the risk of fainting. All the variables that were significant in univariate analyses were included in a single overall multivariable model for each period. The adjusted OR for each factor was derived using a multivariable modelling approach that included risk factors and confounders of the relationship between fainting and each variable. Potential factors that were no longer significant (P > 0Æ05) when included in each multivariable model were removed in a stepwise manner until only the significant factors remained. For all models, in most cases the reference group for a categorical variable was the group with the highest frequency. While our focus was on fainting in order to devise future intervention strategies, we assessed frequency of injuries and need for OMC during the three different donation periods to begin to understand the periods during which injury and disability occur from LOC. The proportion of injury and OMC was calculated by dividing the number fainting with injury and fainting with OMC events by the total number of fainting in each corresponding period. Data cleaning and statistical analysis were performed using STATA 11 SE (Stata Corporation, College Station, TX, USA), and results were graphed using Excel 2007 (Microsoft Corporation, Redmond, WA, USA). Results Frequency of fainting events For the 12-month study period, there were 956 766 donor registrations with intent to donate, and 554 534 (58%) of which were intended whole blood donations. Among registered donors, there were 43 LOC events during registration and medical health screening for a rate of 0Æ04 ⁄ 1000 registered donors (Period 1). There were 1520 LOC events in Periods 2 and 3 (LOC rate of 2Æ7 ⁄ 1000 donations). During Periods 2 and 3, 78 injuries were reported (Injury rate of 0Æ14 ⁄ 1000) of which 73 (94%) were associated with LOC. Of the 1520 fainting reactions, 39% (593), 51% (772) and 10% (155) occurred in period 2, 3A and 3B, respectively. No donor had more than one LOC reaction associated with any donation. Rates of LOC during the donation periods are reported in Table 1 and Fig. 1b. The rate of off-site reactions (Period 3B) associated with fixed site donations is low compared to LOC rates associated with donations on mobile buses and mobile inside set-ups. Low EBV is associated with a profoundly increased rate of off-site reactions when compared to donations by individuals with more than 4 or 5 l EBV. Multivariable analysis results by period For Period 1, the final model included age, sex and donor status. EBV and donor site were found to be non-significant and were not retained in the model. In Period 1, the likelihood of LOC was higher in younger donors compared to donors 25–65 years of age (Table 2 and Fig. 2). The adjusted ORs and 95% CIs were 11Æ1 (4Æ6–27Æ3), 10Æ3 (4–27) and 4Æ9 (1Æ0–23Æ8) for 17–18, 19–22 and 23– 24 years old donors

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19