Project 1: Reference Interval Tool (LabRI Tool) - 2021 to 2024
The “Laboratory Reference Interval” (LabRI) method employs a multi-criteria automated approach, integrating various algorithms throughout different stages to identify the optimal reference distribution (related to the reference subpopulation) and calculate reference limits (RLs). The pipeline sequentially arranges algorithms and each step’s output is the next step’s input. The method was developed through an analysis of publications on recommendations, limitations, and findings in biostatistics applied to laboratory results and RIs. It emphases data distribution analysis, outlier detection, mixture deconvolution, data transformation, and distribution truncation.
Example:
Project 2: Development of the Educational Chatbot “LabRI Chatbot” for Clinical Laboratories - 2024 to Present
Accreditation standards harmonized with ISO 15189 establish requirements for defining and verifying Reference Intervals (RIs) used in laboratories. To comply with these standards, laboratories must periodically verify pre-established RIs, confirming their suitability for the served population. To facilitate the implementation of these requirements in Brazil, our Working Group has launched several initiatives:
All these resources are freely accessible. However, we identified the need for greater professional literacy in this area. Therefore, the objective of this project is to develop a chatbot to educate clinical laboratory professionals about estimating, transferring, and verifying reference intervals.
To develop a Generative Pre-trained Transformer (GPT) called “LabRI Chatbot” to educate clinical laboratory professionals about reference intervals, using six instructional parameters to guide its actions and behavior.
The alpha version of the GPT “LabRI Chatbot” was internally tested for response validation on topics such as direct and indirect methods, estimation, verification, transfer, and sample size. The chatbot answered accurately and educationally to questions with answers contained in the uploaded document. For questions not addressed in the document, he answered generically but without delirium.