Modelling blood safety

Abstract

This thesis describes the development and application of methods and models to support decision making on safety measures aimed at preventing the transmission of infections by blood donors. Safety measures refer to screening tests for blood donors, quarantine periods for blood plasma, or methods for detecting or eliminating contaminations during processing. Chapter 1 describes the outline of the PROTON study. In this study a random sample of 20 Dutch hospitals was taken, covering 28% of all blood transfused in the Netherlands between 1996 and 2006. This data allows the construction of descriptive statistics of the Dutch blood transfusion recipient in terms of gender, disease and survival per type of blood component. In Chapter 2 a detailed study is performed on the estimation of recipient survival after transfusion. It is shown that the standard Kaplan-Meier method can be applied to calculate survival after transfusion despite the fact that patients can receive multiple transfusions. Variance estimates, however, will generally be biased downward and have to be determined by bootstrapping transfusion recipients. In Chapter 3 the cost-effectiveness of bacterial culturing (BCU) is compared to that of pathogen inactivation (PRT) for detecting bacterial contamination in platelets. Despite the fact that the individual cost-effectiveness ratio estimates are highly uncertain, it can be concluded that BCU is without doubt more cost-effective than PRT. In chapter 4 the cost-effectiveness of additional hepatitis B screening is determined. Despite a higher effectiveness, testing individual donations is less cost-effective than testing pooled donations. Chapter 5 describes an enhancement of the value of information (VOI) concept. This provides a decision maker with information on clinical outcomes as well as a value in monetary terms. As this attributable VOI can be obtained by merely reorganizing data already available, its calculation is recommended for all VOI analyses. Chapter 6 delineates a generic probabilistic model for assessing risks of plasma derived medicinal products. This model renders improved estimates as well as new opportunities for analysis as compared to models used in the past. Chapter 7 describes the development and application of a regression model dedicated to virus validation robustness studies. The new model provides stronger p-values to the relevant covariates and its use is recommended for future studies. Chapter 8 describes statistical tools and methods for signalling changes or outliership of incidence rates in specific blood collection centres. The tools can be applied by any blood establishment or plasma fractionation institute. All of the models described in this thesis aim to enhance insights and are able to deliver, be it at a price: they force reality into a rigid mathematical structure. However, when forcing focus on the essential aspects of a system they still allow deriving valid conclusions. They offer the opportunity to develop novel insights, explore alternatives and view outcomes from new perspectives. Models therefore deliver an irreplaceable and unequivocal contribution to the quality of decision making within blood safety.