Making risk statements stick.

Abstract

Estimates of risk associated with blood transfusion are reported from a variety of sources using different numerical constructs. These data must be judged for validity and generalizability to facilitate decisions for interventions and to estimate potential benefits of interventions. Risk estimates reported in consistent terms, such as occurrences per million units transfused, will assist in comparisons of risks and the expected effect observed at the practitioner level. Use of the estimated number needed to treat puts the effect of an intervention in perspective for the individual practitioner and for national health authorities. We re-evaluated data reported from several recent studies of transfusion risk to highlight this approach. In the USA, the number needed to treat estimated to prevent one HIV transmission is 4.3 million (mini-pool NAT); to prevent one death from bacterial sepsis is 21 thousand (conversion to single donor platelets), and 16 thousand (bacterial screening of platelet concentrates). As interventions are continuing to drive infectious disease transmission rates lower and lower, expressing residual risk as the number needed to treat demonstrates that further improvements in safety are unlikely to be recognized at the local level even though the overall impact at the national level is significant.