Looking at multiple patient journeys at once can be a feat due to the inherent complexity of healthcare data. By creating a graph that connects events starting when patients are admitted to the ICU to when they are discharged, including medical visits, diagnoses, labs and procedures, we can present the patient journey in a far more intuitive way.
We will take a deep dive into patient journeys through the Medical Information Mart for Intensive Care (MIMIC)-IV de-identified Electronic Medical Records (EMR) data from 2008-2019, for patients diagnosed with pneumonia.
Using Neo4j and GraphXR to effectively model and analyze clinical and physiological patient data at scale, we can enable researchers and clinical practitioners to:
* Visualize the patient journey across time
* Use graph algorithms for patient phenotyping and characterization
* Develop diagnoses and treatment recommendations based on cohort-level trends
* Advance AI-enabled patient outcome prediction
Join us for this talk to learn how to:
* Inject EMR data into a Neo4j graph database, connecting patient medical history, diagnoses, medications and procedures
* Use Neo4j Graph Data Science graph embedding algorithms to learn low-dimensional patient history representations
* Apply graph clustering algorithms to cluster patients into similar cohorts
* Use graph embeddings as input for identifying patients at high risk for 30-day readmission
* Use GraphXR to design a dashboard visualization for clinical use that allows clinicians to toggle between:
– A single patient view of their medical history
– A birds-eye view of the care path for similar patients
– An explainable AI view of the main factors informing a patient’s risk for 30-day readmission
Speakers: Ana Areias, Mengjia Kang
Format: Full Session 30-45 min
Level: Advanced
Topics: #graphdatascience , #visualization , #machinelearning , #healthcare , #advanced
Region: AMERICAS
Visit https://development.neo4j.dev/nodes-2022 learn more at https://development.neo4j.dev/developer/get-started and engage at https://community.neo4j.com