An individual is considered to be at-risk if they are over 75 years of age and they have at least 2 risk factors from the categories associated with falls listed below.
- Demographic - age, housebound status, living arrangements
- Physical deficits - cognitive impairment, reduced vision, difficulty rising from a chair, foot problems, neurologic changes, hearing impairments
- Historical factors - use of cane or walker, previous falls, acute illness, medications
- Environmental hazards - loose or ripped carpeting, missing hand rails, poor lighting and uneven access surfaces
When the risk factors are combined with drugs and drug families known to affect balance, the risk of falling increases. Is it possible for our healthcare system to apply the principals of Content Intelligence to their available information to identify at risk individuals to improve patient care and reduce costs?
Information such as patient medical records, at home caregiver and nurse’s notes and social worker observations contain a wealth of information but they’re not connected in any way. Each organization has their own system; the data formats are different and the terminology is based upon their own domain.
Smartlogic worked with a customer in the senior health care space to help them solve this problem.
The organization leveraged the U.S. National Library of Medicine’s Medical Subject Headings (MeSH) public ontology as a starting point. Using Semaphore Ontology Manager, they built out additional detail for their area of interest, adding terms that corresponded to potential incidents and environmental hazards. They built associations between drugs and drug families and the risk of someone falling using a specific “risk increasing drugs” relationship. Similarly, they associated environmental hazards such as loose carpet, steep stairs and missing handrails with the risk of falling.
After publishing this detailed model Semaphore’s Classification Server extracted facts from unstructured, semi-structured and structured content including doctors’ records, social worker reports and comments from care givers and visiting nurses. Sophisticated Natural Language Processing techniques within Semaphore ensured a high degree of accuracy in the created metadata, which was expressed as RDF triples; the semantic web standard for expressing facts and relationships. The model and triples were ingested into a graph database, and data analysts created queries to analyze and the data, discover the content and identify patterns and connections between risk factors and individual patients.
With a list of patients at risk, the senior care provider could take a pro-active approach to managing care. They could reach out and talk with doctors about alternative medications, suggest home improvements and repairs, work with local authorities to improve access points and provide prescreening by at home caregivers and social workers.
Using Content Intelligence to these healthcare providers were empowered to make data driven business decisions to improve quality of care and patient outcomes while reducing cost.
