Natural Language Processing for Mimicking Clinical Trial Recruitment in Critical Care: A Semi-automated Simulation Based on the LeoPARDS Trial

Clinical trials often fail on recruiting an adequate number of appropriate patients. Identifying eligible trial participants is a resource-intensive task when relying on manual review of clinical notes, particularly in critical care settings where the time window is short. Automated review of electronic health records has been explored as a way of identifying trial participants, but much of the information is in unstructured free text rather than a computable form. We developed an electronic health record pipeline that combines structured electronic health record data with free text in order to simulate recruitment into the LeoPARDS trial. We applied an algorithm to identify eligible patients using a moving 1-hour time window, and compared the set of patients identified by our approach with those actually screened and recruited for the trial. We manually reviewed clinical records for a random sample of additional patients identified by the algorithm but not identified for screening in the original trial. Our approach identified 308 patients, of whom 208 were screened in the actual trial. We identified all 40 patients with CCHIC data available who were actually recruited to LeoPARDS in our centre. The algorithm identified 96 patients on the same day as manual screening and 62 patients one or two days earlier. Analysis of electronic health records incorporating natural language processing tools could effectively replicate recruitment in a critical care trial, and identify some eligible patients at an earlier stage. If implemented in real-time this could improve the efficiency of clinical trial recruitment.