Abstract
In this paper, we present a smart disease surveillance system that reveals insights into disease trajectories indicating risks of subsequent diseases starting from the current health conditions, for individual patients or populations. Using a pattern mining algorithm, we extract disease trajectory patterns from temporally modeled encounters of 17 million patients in the medical knowledge graph and develop a disease surveillance system on 477,933 mined patterns of disease progression. The system predicts future disease trajectory of individual patients and facilitates in-depth exploration into disease mechanisms, root causes and future disease progression at a patient cohort level thereby enabling early interventions for complex diseases and promoting an evidence based precision medicine approach for healthcare providers.
Key findings
- 1Surveillance that looks forward: the system shows disease trajectories that indicate the risk of subsequent diseases, starting from a patient's current health conditions, for individual patients or whole populations.
- 2Built on 17 million patients: their encounters are modeled in time order inside a medical knowledge graph.
- 3Patterns mined, not hand-coded: a pattern mining algorithm extracts disease trajectory patterns from those encounters, and the surveillance system runs on the 477,933 patterns of disease progression it found.
- 4Predicts the patient's next diagnoses: for an individual patient, the system predicts the future disease trajectory.
- 5Explains progression across a cohort: for a patient cohort, it supports in-depth exploration of disease mechanisms, root causes and future disease progression.
- 6Built for early intervention: the aim is earlier intervention in complex diseases and an evidence-based precision medicine approach for healthcare providers.
Why this matters
Most care responds to the diagnosis in front of the clinician. A trajectory view asks a different question: for patients with this history, what tends to come next? Answered from the time-ordered encounters of 17 million patients, it gives a provider a reason to look for the next likely condition before it presents, which is where early intervention in complex disease starts.
The same patterns serve two users. A clinician sees one patient's likely next diagnoses; a population health or research team explores root causes and progression across a cohort. The abstract reports the system's scale rather than its predictive accuracy; the knowledge graph model, the mining algorithm and a worked patient example are in the full paper.
Keywords
- Disease Trajectories
- Knowledge Graph
- Pattern mining
- Data Science
Editors and affiliations
Joseph Finkelstein
University of Utah, Salt Lake City, UT, USA
Robert Moskovitch
Ben-Gurion University of the Negev, Beer Sheva, Israel
Enea Parimbelli
University of Pavia, Pavia, Italy
References
- 1.Aspland, E., Gartner, D., Harper, P.: Clinical pathway modelling: a literature review. Health Syst. 10(1), 1 to 23 (2021). doi:10.1080/20476965.2019.1652547
- 2.Hidalgo, C.A., Blumm, N., Barabási, A.L., Christakis, N.A.: A dynamic network approach for the study of human phenotypes. PLoS Comput. Biol. 5(4), e1000353 (2009). doi:10.1371/journal.pcbi.1000353
- 3.Jensen, A.B., et al.: Temporal disease trajectories condensed from population-wide registry data covering 6.2 million patients. Nature Commun. 5(1), 4022 (2014).
- 4.Murtagh, F.E., Murphy, E., Sheerin, N.S.: Illness trajectories: an important concept in the management of kidney failure (2008).
Paper copyright 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG.
