The FDA on Wednesday finalized guidance from 2021 that aims to help pharma companies understand when they may use data from electronic health records (EHRs) or medical claims as part of drug approval packages.
The guidance expands on the regulator’s work from 2013, and the FDA has now provided more details about the selection of data sources and how best to evaluate the relevance and reliability of EHR and medical claims data for use in a clinical study.
“Given that existing electronic health care data were not developed for research purposes or to support regulatory submissions to FDA, it is important to understand their potential limitations,” the FDA said in the 35-page guidance.
The agency noted that EHR data are often limited to what was captured by one hospital system or network, and may not include care plans or other data obtained in different facilities, which is why “EHR data may not accurately reflect the presence, characteristics or severity of a particular disease.”
The guidance also points to AI as potentially allowing more rapid processing of unstructured electronic health care data. While saying it does not endorse any particular AI tech, the agency said that advances in natural language processing, machine learning and particularly deep learning can help to “extract data elements from unstructured text in EHRs; (2) develop computer algorithms that identify outcomes; or (3) evaluate images or laboratory results.”
Other changes from the draft guidance include clarifying that recommending the use of quantitative bias analyses, “either a priori for feasibility assessment, or to facilitate interpretation of study results, or for both purposes, to demonstrate whether and how misclassification, if present, might impact study findings,” the FDA said in the Federal Register.