Clustering Patients Trajectories (Autumn 2014)

Meeting 11/3-15

• Spørsmål til Bård: Send til Aslak
• Hvordan gruppere? Kompleksitetsindikator? Se på hendelser knyttet til Diabetes... Hamming-kode sekvensene (Informasjons-teori). Ignore selve sekvensen mellom hendelsene?

Meeting 10/2-15

Fra Aslak: Håvard bør ha klart noe å vise til Bård Kulseng i løpet av uka, nå som kontakten allerede er etablert. (A la skjermbildene som Ingunn og Gry presenterte for Birgit Mørch i høst).

Meeting 28/1-15

Bård Kulseng. Overlege Endokrinologisk. Kjenner diabetes veldig godt. Kan peke på noen (Magnus Strømmen, overvekt) som kan hjelpe til.

Hensikt med Clustering

• Diabetes er en stor gruppe. Hvilke forskjellige grupper finnes?
  • Hovedstrømmer / detaljer. Hvilke steder "kan slås sammen"?
  • Tegne et tre vha. automatisk clustering
• Eller gå i retning av kompleksitets-indikator. Ta en prat med Gro.
• Håvard fortsetter på høstprosjektet i ca. 2 måneder til. Sender mail til Aslak.

Clustering Patients Trtajectories (Autumn 2014)

...

PAsTAs (Patient Trajectories, or “Pasientforløp”) is a project that is analyzing what happens to chronically ill patients, as they are moved between their primary doctor, the hospital, and other services offered by the local government. The overall aim of the PAsTAs project is to “keep the patient out of hospital by improving patient trajectories in primary care”.

...

The focus of this project will be to analyze the patient trajectories of cases pertaining to chronic diseases. Task will include creating a solution for scalable visualization of such data based on different filters or scales. Two such filters could be:*

• Relation of group. Meaning you can adjust the visualization based on severity of condition, age or other distinguishable parameters.

...

• Size of institution. Meaning adjustments will change whether to include a whole hospital as one entity, or split it into different departments or wards, or even specific practices.

The task for the fall project will be to identify which parameters of the data are most relevant to filtering and how to best group clusters of patients.

Scientific goal:

Find a reasonable way to group a cluster of patients based on relevant parameters from their condition, care and outcomes.

Method:

Creating a visualization of data will require a good foundation of data. Prepared sets of data will be provided by supervisors of PAsTAs. Analyzing these sets of data will accumulate metadata about each patients appointments, which will need processing in order to yield meaningful information which can be represented within a given set of filters.

Representing patients with an equal or similar diagnosis based on the difference in care and outcomes is a point of focus.

Solution:

Focus on the solution will be in the spring semester.

Assuming this project is going to continue in the future, perhaps the most important task is to make good data models from the raw sets of data. These data models can then be used to further explore the overall patterns of patient trajectories. Creating the visualization based on these data models will demonstrate the potential of the models and will force the development in a direction of usability.

...