With advances in natural language processing (NLP), machine learning (ML) and artificial intelligence (AI), there are new opportunities for improving findability among existing public-facing resources. This project seeks to inform findability, especially for multiple chronic condition (MCC) resources, by describing current search capabilities and limitations across several of AHRQ’s publicly available domains and by identifying and piloting a novel NLP/ML approach to make suggested improvements. This work intentionally engages with the overlap of numerous disciplines including information extraction, information retrieval, data and text mining, knowledge management, and best practices in health care. We are looking to apply this work across all domains but will start by focusing on specific AHRQ domains. Given limited API access, we scraped the content of digital.ahrq.gov and the patient centered medical home (PCMH) resources and performed automated search using a set of related terms that align with an MCC scenario: hypertension, osteoarthritis, and chronic kidney disease. We obtained results confirming the limitations of existing search.
Marcial, Laura Haak; Santini, Silas; Kery, Caroline; Brown, Stephen; Chew, Rob & Blumenfeld, Barry
The UNT Digital Library