Information Technology and Informatics
Shaun Lawicki, MBBS (he/him/his)
University Medical Center New Orleans, Louisiana, United States
While validating equipment and testing methods is a routine practice, Transfusion Services must extend these fundamental validation principles to new and emerging applications. One such technology making its way into Transfusion Services is Artificial Intelligence. It is crucial to apply regulatory validation processes to these new applications. ImmuLINK Panel ID (Werfen, Norcross, GA) is a software tool designed to assist with antibody exclusion based on AABB’s IRL Standards and Technical Manual. The purpose of our study is to document one approach for validating this software.
Study
Design/Methods:
We reviewed AABB Standards concerning validation, including the requirements for Installation Qualification, Operational Qualification, and Performance Qualification. Using these principles, we developed a validation plan for implementing Immulink Panel ID software. Validation of the new application is expected to span 30-60 days.
Results/Findings:
Key elements of the functionality were identified and incorporated into the validation plan. We reviewed resources, including the Vendor’s Validation Guide, to determine which elements to include in the validation plan and at what phase of the validation to include them. Criteria for assessing each phase were established, along with any applicable metrics (see Table 1).
Conclusions:
Risk assessment is an integral part of validation, especially for new applications such as artificial intelligence solutions designed to expedite and enhance routine practices in Transfusion Services. It is crucial to present sufficient challenges to the application to ensure that all reasonable scenarios encountered during validation are assessed. To meet this requirement, we examined the various antibody scenarios commonly encountered in our lab. Validation protocols should not be so restrictive that they take an excessively long time to complete due to waiting for rare scenarios to occur. Therefore, we focused on antibodies common to our patient population and included performance qualifications for those groups. Rarely encountered antibodies were excluded from the validation.