semantic social networks for integrated healthcare · cardiorenal disease and comorbidities ......

Semantic Social Networks for

Integrated Healthcare

E. Kaldoudi, N. Dovrolis

Democritus University of Thrace, Greece

[email protected]

PRO-VE 2013, 14th Working Conference on Virtual Enterprises, Dresden, Germany, 30 Sep - 2 Oct,2013


integrated healthcare

re-unite parts of a whole…

different healthcare settings, i.e. from primary to tertiary level

different types of care, i.e. hospitalization, long-term care, rehabilitation, self-care, social care, etc.

management of different conditions co-presenting in the same patient (comorbidity)

…

2

https://sites.google.com/a/uninova.pt/prove-13/


comorbidity

the presence of one or more disorders in addition to a

primary disease or disorder (either independently, or as a

consequence of the primary condition or otherwise related)

½ of all chronic patients present comorbidities

only a few overall management guidelines exist

patients receive fragmented, disease specific care

3



chronic cardiorenal disease

simultaneous (causal) dysfunction of kidney and heart

‒ diabetes and/or hypertension underlying causes

‒ a number of other comorbidities often present

‒ deterioration to end stage

renal/heart disease is

life threatening, irreversible & expensive to manage

4



some numbers..

hypertension: 1/3 of adults

diabetes: 8% of overall population

44% of chronic kidney disease is due to diabetes

chronic kidney disease: 9-16% of overall population

most chronic kidney disease patients develop heart disease

chronic heart failure: 1-2% of healthcare costs

end-stage renal disease: >2% of healthcare costs

5



cardiorenal disease and comorbidities

of major importance

early detection

aggressive management

preventive progression to end-stage cardiorenal disease

via

lifestyle and diet management

public health education

monitoring and adherence to therapy for: hypertension, diabetes, proteinuria, dyslipidaemia, smoking

integrated management of comorbidities

6


healthcare professionals

educational resources

sensor data

medical evidence

patients carers

procedures medical

conditions

events

interventions

people

medical practice

information

social, personal, economical, etc.

aspects


our approach

blending of concepts of

social networks and semantic technologies

in order to

develop comprehensive personalized dynamic models of

chronic comorbid patients, their environment and

healthcare related issues, procedures, etc.

so as to

support meaningful and relevant

patient empowerment and decision support services

8



conventional social media

connections and

relationships among

humans

CarePages

FacetoFace Health

DocCom (for doctors)

information & medical practice concepts & digital

objects



object centered social networks

people acting on a common social object

PatientsLikeMe

(disease progress, findings)

CureTogether

(treatments)

SmartPatients

(clinical trials)

CareCloud

(medical data)

Wellaho

(health record)

…..


objects



social networks so far

simple human social networks &

object-centered social networks

both rely only on human action, agency and perception

the digital object

– the fundamental element of the web –

has only an enabling role



procedures medical

conditions

healthcare professionals

sensor data

medical evidence

patients carers

events

interventions

people

medical practice

information

social, personal, economical, etc.

aspects

multiple, complex, heterogeneous networks

of people, concepts, information content, etc.

‘collaborating’ towards a better health status


our approach

use notions and concepts from Actor-Network Theory (ANT)

sociological theory

1980’s by Latour, Callon, Law

to construct symmetric heterogeneous networks of humans

and digital objects, all treated alike

human and non-human are treated as actors

all actors are assigned equal amounts of agency

for each actor, a specific (heterogeneous) network is

formed



heterogeneous networks & ANT

symmetry: human and non-

human actors are capable of

exerting force,

changing and being changed

by one another

allows for different views of the system (e.g. patient, doctor, disease, risk factor, treatment, …)


objects



ANT concepts

translation: actors coming together, changing one another to

form links

e.g. enhance weak relationships (e.g. may_cause) via

multiple connections based on evidence based medicine

findings

black boxes: stabilized parts of the network (which can

always be re-opened)

a chronic condition with all its inflicting factors,

monitoring, treatments etc. (re-opened when new

evidence is available)

15



ANT concepts

immutable mobiles: can be silent, ignored of overriden by

other actors, but exhibit enough solidity to move about and

still hold their relations

e.g. health & lifestyle behaviours, habits, proven risk

factors, etc. allow for pattern extraction

obligatory points of passage: central groupings through

which all relations must flow at some time

crucial for lifestyle and disease management, and

clinical protocol alignment in comorbidities

16



building agency into non-human agents

agency: the ability to produce an effect,

change relationships among actors

no effect implies no agency

building agency into non-human agents

via semantic connections and enrichment of metadata


sources of personal information

life style related data

basic demographic

data

healthcare related data

monitoring data

genetic/familial history Linked

Data Cloud

the Web

manually adjusted

web service

manually adjusted or automatically

harvested from related system

patient (meta)data

a health related profile for healthy or at risk individuals & patients

sources of formal information


medical evidence

monitoring data

patient info pages

Linked Data Cloud

the Web

building agency for conditions and

interventions

automatically harvested

from related sources

health/medical interventions

health/medical conditions

condition, risk factor, intervention


metadata interlinking and enrichment

semantic enrichment of resource profiles by harvesting

generic and domain specific controlled vocabularies/ontologies

SNOMED-CT (clinical terms), ICD9/10 (diseases), LOINC (lab

exams), MeSH (medical literature), … FOAF, SIOC, …

UMLS (Unified Medical Language System)

a) semantic network of concepts (135 types) via sets of

semantic relationships (54 relationships)

b) mappings between more than

100 medical controlled vocabularies

further linking of enriched resources via Linked Data Cloud

e.g. scientific publications, medical evidence, educational

material, personal social status information, etc



21


Risk factor characteristics/attributes:

name: risk factor name (e.g. weight)

strength: value, which can be literal (e.g. 60 Kg), binary (exists or not, e.g. for a gene), or qualitative

sourceType: environmental, genetic, behavioral, biomedical, and demographic

correlate: type depending the ability of the risk factor to change its value, i.e., “fixed marker”, “variable marker”, “variable risk factor”, “causal risk factor”

temporalType: “continuous” or “intermittent”

duration: duration of exposure to the risk factor. This is a free text string

associatedCondition: the resultant condition attached to the risk factor

impact: the evidence-based percentage of risk factor’s impact on the associated condition

evidenceSource: authoritative source of evidence for risk factor impact

lifecycle: expected duration/upgrade of evidence

UMLS Semantic Network, associations between a risk factor and the associated condition include:

issue_in: the risk factor is a point of discussion for a condition

affects: the risk factor produces a direct effect on the condition

causes: the risk factor brings about the condition

complicates: the risk factor causes another (risk) factor to become more complex (recursive).


enriching educational resources

semantic enrichment of resource profiles by harvesting

domain specific controlled vocabularies

further linking of enriched resources via the

Linked Data Cloud

e.g. scientific publications, formal curricula, ECTS, etc

exposing enriched resource profiles to the

Linked Data cloud to be part of the semantic web



the web

Linked Data

semantic linking of data and services

metadata enrichment

the web

Linked Data metadata enrichment

NCBO Bioportal

http:// bioportal.bioontology.org

an open repository of biomedical ontologies

258 ontologies 5 million standardized terms

accessed via APIs

http://bioportal.bioontology.org/


educational resource description

before metadata enrichment


educational resource description after metadata enrichment

the Web

mesh and enrich

partial

information to

determine

current status

of individual

sources of ground medical

knowledge

determine state-

of-the-art for

disease &

comorbidities,

co-dependencies,

risk factors,

treatments, etc.

data interlinking and clustering for personalized cardio-renal disease and comorbidities model

sources of personal

information

personal monitoring data

attention metadata from social media presence

personal medical data

Linked Data Cloud

clinical/medical evidence

medical scientific literature

patient educational material

work in progress


work in progress

personalized empowerment and shared decision support services

1) generic model for comorbidities description and management

2) integration of heterogeneous sources, including

medical ground knowledge (ontologies, vocabularies, etc.)

medical evidence (literature and authoritative data sets)

personal data, including:

sensor data, medical data, and personal disposition/lifestyle

educational data (patient pages)

3) cardiorenal comorbidities model, updated with medical evidence

and personalized to the patient

30



work in progress

personalized empowerment and shared decision support services

4) visualization of the personalized model

disease progression pathways

comorbidity trajectories

risk calculations

comparison of personal state with current medical evidence & statistical views of ‘similar’ patients

5) data/model driven decision support system

shared decision support services for the patient and the medical professional

31



acknowledgements

this work has been partly funded by the project:

HELCOHOP

Development of Electronic Clinical Protocols

Thalis MIS 375876

EU and Greek National co-funding

2012-2015


http://www.google.com/imgres?imgurl=http://www.topnews.in/files/EU-Logo.jpg&imgrefurl=http://www.topnews.in/regions/brussels?page=7&$Version=0&$Path=/&$Domain=.topnews.in&h=474&w=726&sz=31&tbnid=EqNHM11FCeAANM::&tbnh=92&tbnw=141&prev=/images?q=EU+LOGO&usg=__qQZEiI1d5U2lKDJ31eLJ2jQKRXc=&ei=pu2zSbrlGZK60AXy1cF-&sa=X&oi=image_result&resnum=1&ct=image&cd=1


work in progress will be funded by the project

CARRE: Personalized patient empowerment and shared

decision support for cardiorenal disease and comorbidities

FP7-ICT-2013-10, No. 611140

2013-2016

(starting Nov. 2013)

acknowledgements


http://www.beds.ac.uk/

thank you!


cite this presentation as

E. Kaldoudi, N. Dovrolis,

“Semantic Social Networks for Integrated Healthcare”,

in L.M. Camarinha-Matos, R.J. Scherer (Eds.) Collaborative

Systems for Reindustrialization. 14th IFIP WG 5.5 Working

Conference on Virtual Enterprises PRO-VE 2013, Dresden.

Springer, Heidelberg, 2013 (pp. 729-736)

35


semantic social networks for integrated healthcare · cardiorenal disease and comorbidities ......

Documents