Study: Unstructured EHR Data Could Help Medication Reconciliation
May 12, 2015 in News
A computerized algorithm that compares structured and unstructured electronic health record data could help improve medication reconciliation and track patient adherence, according to a study published in BMC Medical Informatics and Decision Making, FierceHealthIT reports (Durben Hirsch, FierceHealthIT, 5/8).
For the study, researchers from Cincinnati Children’s Hospital Medical Center sought to determine whether unstructured EHR data, such as free-text clinical notes, could accurately be used for medication reconciliation. The researchers gathered electronic clinical notes and patient discharge prescription lists for 271 patients enrolled in the hospital’s Complex Care Medical Home Program (Li et al., BMC Medical Informatics and Decision Making, 5/6). They then used state-of-the-art machine learning and natural language processing to create a computerized algorithm for medication reconciliation that used both the structured and unstructured data from an EHR.
The researchers found that the algorithm was effective in using both structured and unstructured data to identify matched and discrepant medications.
However, the researchers found some errors in the algorithm’s detection and matching (FierceHealthIT, 5/8). The majority of errors were related to the use of abbreviations or misspellings in unstructured EHR data and uncommon medication names (BMC Medical Informatics and Decision Making, 5/6).
Despite the errors, the researchers concluded that “automated medication discrepancy detection shows a promising outcome in assisting medication reconciliation,” adding, “[W]e hypothesize that the computerized algorithm, when transferred to the production environment, will have potential for significant impact in reduction of effort for conducting medication reconciliation in the clinical practice setting” (FierceHealthIT, 5/8).