Development and performance characteristics of Platelet Virtual Crossmatch (PLT VXM), a software application for the evaluation and management of platelet transfusion-refractory patients.

BACKGROUND Platelet (PLT) transfusion refractoriness increases bleeding complications, hospital stays, and PLT inventory usage. Immune-mediated refractoriness can be evaluated for using a physical PLT crossmatch with ABO-compatible inventory and, if positive, managed with HLA-compatible PLT inventory and donors. Manual completion of these complex tasks can be time-consuming and potentially error-prone. This study was conducted to determine if a Web-based software application could improve process efficiency and accuracy. STUDY DESIGN AND METHODS Workflow analysis was performed to identify process, data, and analytic requirements for a software application for three PLT transfusion-refractoriness associated tasks: (a) physical PLT crossmatch inventory selection, (b) HLA-compatible inventory selection, and (c) HLA-compatible donor selection. After software application development, a comparison study was performed over 10 consecutive days, with each task performed manually and with the software application (Platelet Virtual Crossmatch [PLT VXM]) for a different unique immune-mediated PLT transfusion-refractory recipient. Task completion time, number of incompatible units/donors presented, and number of documentation errors were compared. RESULTS PLT VXM is a Web-based software application developed using R and the Shiny Web application framework. PLT VXM significantly reduced median task completion times by 4.5 (49%), 11.2 (79%), and 59.1 minutes (94%), respectively. PLT VXM did not present any incompatible PLT units or donors for user consideration. PLT VXM also had a lower number of documentation errors than the manual process, and none of these documentation errors were software generated. CONCLUSION Computer-aided evaluation and management of immune-mediated PLT transfusion-refractory recipients can significantly improve workflow and reduce manual errors in this complex process.