Clinical Reasoning as a Knowledge Based Search

Posted: 22 Nov 2021

Date Written: September 21, 2021

Abstract

Search is usually based on syntactical filtering and TF/IDF relevance ranking. This may not meet a clinician's expectations for patients with complex medical needs. The searchable knowledge space contains both invariant (not specific to patients) and variant (specific to patients) data. Our solution leverages a growing medical knowledge base for the choosing what to present to benefit the patient.

Typical question: What is the best treatment for my 87 year old male obese patient complaining of XXX with High blood pressure, diabetes mellitus 2, taking a long list of meds and herbals?

To address this a semantical (knowledge graph based is leveraged) and there may be possible interactions between the various conditions/diseases/meds/herbals. Clinicians mentally weigh the risk/benefit for each patient they encounter.

To accurately diagnose or treat a patient requires accounting for multiple factors including the patient's demographic info, age, fragility, treatment pathways for each condition/disease, past medical history, comorbidities, current medications, social history and family history (including genetics). We leverage the Clinical Reasoning Engine (CRE) as our "non-linear" reasoning symbolic AI engine executing complex clinical logic over an ever changing/growing knowledge graph of data from multiple data sources and from multiple contexts. An effective clinician accomplishes this task by using a combination of experience, skill, and data.

With medical data and knowledge doubling approximately every 73 days, the CRE is proving to be a valuable aide.

Keywords: Knowledge Graphs

Suggested Citation

Corkum, Matt and Macfarlane, Colin and Cox, Jessica and Snyder, Paul and Dunlop, Robert, Clinical Reasoning as a Knowledge Based Search (September 21, 2021). Proceedings of the 5th Annual RELX Search Summit, Available at SSRN: https://ssrn.com/abstract=3967160 or http://dx.doi.org/10.2139/ssrn.3967160

Colin Macfarlane

Elsevier ( email )

Jessica Cox

Elsevier ( email )

Paul Snyder

Elsevier ( email )

Robert Dunlop

Elsevier ( email )

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
124
PlumX Metrics