New and novel treatments often pose learning curves for clinicians, such that, as postmarket safety surveillance data begin to accrue, regulators are left to disentangle poor outcomes attributable to a new device or procedure from so-called learning effects, where unfamiliarity and human fallibility can sometimes meet to derail patient outcomes.
Sharon Davis, PhD, MS, Michael Matheny, MD, MS, and colleagues propose simulating learning effects using synthetic patients. They report a generalizable framework for such simulations in BMC Medical Research Methodology. Synthesized patients are assigned to notional providers and institutions having varying levels of experience with new treatments; based on the sorts of features reflected in medical records, patients are assigned novel or standard treatment; outcomes are based on patient risk, treatment-associated risks, and learning effects at both provider and institutional levels.
The authors say their highly customizable framework can aid development and testing of electronic algorithms for distinguishing learning and treatment effects, thereby helping to identify training opportunities and hasten treatment improvements.
Other researchers on the study from Vanderbilt University Medical Center include Dax Westerman, MS, and Theodore Speroff, PhD. The VUMC researchers were joined by researchers from six other institutions. The research was supported by the National Institutes of Health (HL149948).