We propose to refine and evaluate a system for automatically creating context-relevant views of electronic health records (EHR). Older patients, with multiple conditions, tend to have larger amounts of data in their EHRs. This creates an information overload problem for clinicians managing their care, requiring more time and effort to find the data they need for a particular task or context. More importantly, not all of the relevant data may be used in decision-making. The information overload problem is exacerbated for many clinicians, such as hospitalists, who may not be familiar with a patient?s health history, but must rapidly assess a patient and make decisions about managing their care. We intend to address this problem by automatically creating context-relevant views of the patient?s EHR using Loupe, a knowledge-base containing concept associations. Loupe, which was previously created by us, has over 20 million associations between coded concepts from clinical terminologies including ICD-9, ICD-10, SNOMED-CT, RXNORM, CPT and LOINC, encompassing the categories of diagnoses, findings, medications, procedures and laboratory tests. Loupe can be used to filter EHR data based on associated concepts. This can be used to create views of the EHR around a condition such as acute kidney injury, thereby reducing the information shown to the clinician. Our proposal has two aims. The first aim is to evaluate the performance of Loupe in filtering de-identified EHR data for acute kidney injury (AKI) and geriatric altered mental status (AMS). Both these conditions are significant for elderly patients. We will use traditional information retrieval measures including precision and recall to assess the performance of Loupe. The second aim is to create and evaluate context-relevant views for AKI and AMS. We will use an iterative design approach to create the views. The evaluation will compare the context-relevant view to traditional EHR views with hospitalists as subjects. The study will be performed in a laboratory setting at the University of California San Diego Hospitals. We will compare the ability of the hospitalists to recall patient information, the time taken to review the patient data, and the number of mouse clicks.
Project Terms: Acute Renal Failure with Renal Papillary Necrosis; Address; Admission activity; Affect; Algorithms; Arizona; base; California; Caring; Categories; Clinical; clinical decision-making; Code; Collaborations; Complex; Data; Data Display; data mining; Data Sources; Decision Making; design; Diagnosis; Diagnostic tests; Elderly; electronic data; Electronic Health Record; Engineering; Evaluation; experience; Fatigue; Gold; Health; Heart failure; Hospitalists; Hospitals; ICD-9; Information Retrieval; International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10); iterative design; Knowledge; knowledge base; Laboratories; Lead; Logical Observation Identifiers Names and Codes; Manuals; Measures; Medical; Medical Records; mental state; multiple chronic conditions; Mus; older patient; Outcome; Patient Care; Patient Care Management; patient safety; Patients; Performance; Pharmaceutical Preparations; Phase; Physicians; Procedures; prototype; Public Domains; PubMed; Randomized; Recording of previous events; Research; Resources; Sapphire; Sensitivity and Specificity; SNOMED Clinical Terms; Source; System; Technology; Terminology; Test Result; Testing; Time; tool; tool development; Unified Medical Language System; Universities; Update; Work;