Making eligibility decisions based on trends seen in past cases, rather than explicit criteria, is a feasible approach to clinical trial recruitment that is both compatible with data structures found in electronic health records and has the potential for automation, according to a study in BMC Medical Informatics & Decision Making.
Researchers developed a prototype predictive modeling system by using existing data from manually assessed eligible and ineligible patients to develop a prediction model. The model's performance was evaluated retroactively across three clinical trials, using manually selected patients as controls.
The correlation of the model's patient selection and the manual patient selection ranged from 0.88 to 0.99 for the three trials, all with high confidence intervals. The results suggest predictive modeling to be a feasible approach to clinical trial selection, without the restrictions of traditional eligibility criteria and with the potential to be automated in the future.
More Articles on Predictive Modeling:
RWJF: 3 Steps to Eliminating Super-Utilization
4 Technology Strategies Required for Population Health Management Success
Successful Implementation of a New Care Delivery Model