Hybrid Collaborative Filtering Methods for Recommending Search Terms to Clinicians
Electronic Health Records (EHR) are used in over 88% of all US medical clinics to improve care and streamline data. In addition, they enable sharing of data to multiple providers dealing with the same patient, thereby enhancing efficiency and care.
In the last decade, medical practices have implemented EHR software to manage patients and care providers. They help provide better patient care and automate various tasks for the practice. In addition, they allow the exchange of information with one another remotely and in real-time, ensuring every clinician dealing with a patient has current, complete, and accurate patient data. However, they are limited as clinicians often operate under time pressure and must invest significant effort in retrieving information, such as demographics, prior findings, and lab results. Searching through a patient record for the same or similar data on similar patients is repetitive, time-consuming, and cumbersome. Hence, new tools are needed to simplify search functions to improve accuracy and efficiency.
This technology describes the development of a tool for EHR systems that recommends search terms based on specific patient data. It provides these terms from modeling data via clinical encounters, frequently co-occurring ICD codes, and recent search terms. The inventors have developed a beta version of the software and tested it within EHR software. These experiments demonstrated that their model outperforms state-of-the-art baseline methods used in commercial products.
This technology can be used within existing EHR software systems in medical clinics and hospitals.
The technology could reduce physicians' time to search through medical data, leading to more accurate diagnoses, treatments, and care. In addition, it can be implemented within existing commercially available HER software.