Abstract: Data quality concerns hamper secondary use of Electronic Health Records (EHR). Evaluating plausibility of data values, as a measure for truthfulness of EHR data, is commonly conducted through manual rule-based procedures. DQe-p presents an alternative algorithmic solution to identifying implausible clinical observations by applying unsupervised learning techniques. We are improving the specificity and sensitivity of DQe-p in characterizing implausible clinical observations, and exploring potentials for expanding its utility as a patient characterization tool.
Learning Objective 1: Explore use of unsupervised learning for evaluating plausibility in data from electronic health/medical record systems.
Hossein Estiri (Presenter)
Harvard Medical School
Jeffrey Klann, Harvard Medical School
Kavishwar Wagholikar, Harvard Medical School
Shawn Murphy, Harvard Medical School