Anru Zhang, PhD, has been awarded the Data Science Distinguished Paper Award by the American Medical Informatics Association for ‘Timeline Registration for Electronic Health Records.’ Zhang and his team aimed to tackle the bias found in longitudinal electronic health records captured during disease progression.
A patient’s electronic health record (EHR) begins when they are admitted to a hospital and diagnosed with an illness. These records are widely used among health care professionals to develop treatment plans. But Zhang found it difficult to track and predict mortality outcomes using EHRs. He said that’s because the data does not reveal the true timeline of progression since patients with the same illness can be admitted to the hospital during different stages of disease. Zhang said this means patient records are not directly comparable.
To eliminate the bias in the timeline data, Zhang and his team, including David Page, PhD, chair of the Department of Biostatistics and Bioinformatics, developed a timeline registration algorithm to help determine a baseline timeframe for disease progression. The researchers used this algorithm to analyze the datasets of over 60,000 patients admitted to critical care units at one medical center maintained by the Massachusetts Institute of Technology. Then they shifted each patient’s timeline registration by an appropriate amount to align with the baseline. They analyzed the data and found it was more comparable than before.
Results showed when using their proposed registration model, accuracy of mortality prediction increased by 1-2 percent. This improved prediction accuracy will assist doctors in making crucial health care decisions for patients in the ICU. “I feel encouraged. This project is just an initial step in our research,” Zhang said.