RESUMO
BACKGROUND: There continues to be uncertainty about the optimal approach to documenting bleeding data in platelet transfusion trials, with a desire to apply a common assessment tool across all trials. With this in mind, a consensus bleeding assessment tool (BAT) has been developed by the Biomedical Excellence for Safer Transfusion (BEST) collaborative, based on review of data collection forms used in published randomized trials and following content validation with a range of healthcare professionals at seven haematology centres through BEST members. This study aimed to evaluate reliability and reproducibility of the consensus BAT. METHODS: Replicated clinical assessments of bleeding were undertaken by participants with haematological malignancies recruited at four haematology centres in an international, multicentred, observational study. Concordance of repeat assessments was calculated for agreement in site and grade of bleeding observed. RESULTS: Forty patients consented to participate, and 13 trained bleeding assessors collected these data. Bleeding assessments were carried out on 113 separate days. Of all 225 bleeding assessments, 204 were compared for grade concordance, and 160 were compared for site concordance. There was very good grade concordance (83%, 95% confidence interval 74-93%) and good bleeding site concordance (69%, 95% confidence interval 57-79%) in observations of bleeding. Discordance was primarily in relation to assessing skin bleeding. CONCLUSIONS: Alongside a structured training programme, levels of concordance for a consensus BAT were high. Researchers using assessment tools for bleeding need to balance comprehensive data collection against potential loss of accuracy for some types of bleeding, such as skin findings.