Jenny UreSocial Informatics Cluster

School of InformaticsUniv. of Edinburgh

Curating Complex Dynamic Data in Telehealth

OverviewOverview

• What data is collected in telehealth

• Why data curation is important for telehealth

• Why data curation in telehealth is particularly problematic (an engine not a camera)

• Why data collection and data curation of this complexity and on this scale requires a different approach

• What the future might look like – codex memex genex

Telemetry Supported Monitoring of Long-Term Conditions at Home -

Rapidly ageing populationExpected growth in elderly population

Increasingly unhealthy

Rise in Long Term Conditions

17.5 million UK adults are living with a chronic disease

By 2030 the incidence of chronic disease will double

Randomised Controlled Trials

HypertensionLargely asymptomatic conditions

Stroke/hypertension Older, frailer group with challenging targets

Diabetes/hypertension/obesityCo-morbid conditions

COPDSymptomatic, progressive, potentially unstable conditionCongestive Heart Failure

Models of telemetric supported self monitoring

Remote server

Record of readings and symptom scores

Internet or SMS

Urgent referral if required

Constant monitoring of potentially unstable conditions (e.g. COPD)

Call centre

Relatively stable conditions (e.g. diabetes, BP) checked intermittently (normally by practice nurse

GP practicePatient takes readings and enters symptom score

Home

Call to check situation

Reminders to self monitorAutomated feedback

Links to online information

Consultation arranged as appropriate

Telephone or videolink

Clinician view

Streaming Data

• Complete morbidity questions daily

Record any WORSENING of symptoms from usual daily level.

I am more breathless than usual

My sputum has increased in colour

My sputum has increased in amount

I have a cold (such as runny or blocked nose)

I have increased wheeze or chest tightness

I have a fever

I have an increased cough

I have a sore throat

<3 is OK

3-4 watch

5+ take action

The COPD telehealth system

The COPD telehealth system

• Complete morbidity questions daily

• Practice contacted according to an algorithm

• Call centre monitors

• Physiological measures (FEV1, SpO2)

Complex Dynamic DataComplex Dynamic Data

The novelty, scale, complexity and purpose of real-time mobile data present real challenges for every aspect of the data curation process.

Complex dynamic changeComplex dynamic change

The reconfiguration of the clinical, human and technical infrastructure invokes new dependencies, new axes of control, and emerging risks and uncertainties

A resource for diagnosis, analysis and intervention,not currently curated

You’d think you would find it easy to tell when you’re ill but it’s only afterwards that you know you are not well. But this technology is really brilliant

I don’t worry about him the same as I used tae. It’s all taken care of before it can get tae that level. That machine can tell [him] he’s ill even before he kens it himself

Initially a semi-automated protocol – led vision where technical artefacts may shape data

The social shaping of data to suit different purposes and contexts then became more apparent

Social ShapingSocial Shaping

Social shaping: the social life of dataSocial shaping: the social life of data

Awareness of how the data is Awareness of how the data is shaped in transit, and by contextshaped in transit, and by context

(It’s) hugely variable. I sent Dr X out to see them and he came out and said all their tests are clear but why are they scoring 9?. It just doesn’t make any sense. If you phone they’ll explain.

‘I was cleaning my windows before and it made me breathless (Nurse)

‘I knew that if I answered yes to that question I’d be carted off to hospital’

“I’ve had no problems whatsoever except when my son did it without me knowing and sent that figure through”

“Well when they changed the batteries the readings went back to normal”

The unexpected!

There’s one lady that says ‘No ‘(on the questionnaire) every day. But she’s probably the worst patient on it. Her chest is terrible. I know she’ll be struggling but she answers ‘No.’ She says ‘I don’t like to bother anyone’. (Nurse on COPD patient)

ScaleScale

There’s huge amounts of stuff when you think about it, there’s hundreds and hundreds of episodes of measurement and how do you summarise that data, and how do you look through it, and where do you store it and use it? These are the sort of questions that people don’t really have answers to at the moment.

(Primary Investigator)

I think what we need to make a decision with Telecare is what’s worth storing and in what format do you store that, in terms of do you store every reading that ever took place, or do you summarise those readings and store it in some sort of summary

(Database manager)

PurposePurpose

Since telemedicine records outlive the session in which they were created, clinical documents and other objective medical data need to be in a standard format to facilitate portability and accessibility…….. Finally, as telemedicine sessions are medical contacts, telemedicine records should be part of the patient’s life-long health record.

GP/Primary Investigator

Rights, roles, resources, Rights, roles, resources, responsibilities?responsibilities?

You know- what do we do with their data? Do we swipe it, or do we give it to them, how do we give it to them and how do we incorporate that in the GP’s record, would the GP want that incorporated in their record? These are the sort of questions that I think we’re beginning to ask really.

Different purposesDifferent purposes

Telehealth research to date has been dual purpose – with both a research driven agenda and a clinical one.

This impacts on what people regard as important, on the relationships they regard as salient and worth recording, and what they want to keep.

Ethical/Legal IssuesEthical/Legal Issues

I don’t know if we’ve got an ethics directive on that (data use of people who might withdraw from the study) and I’m not going to ask, because they might come up with a solution we might not like

The NHS would have to decide what it’s going to do with the data that’s already accrued but they don’t need the patient’s permission to keep that because the patient doesn’t own their record.

We’ll have the quantitative data which we’re using for the trial, and that will be a data set. And that probably could be anonymised and shared, and I don’t know how we’re going to store it yet. What I wrote on ethics is that hopefully by the time we’ve finished the University or the NHS will have a policy for us. (Prim. Investigator)

Ownership/AccessOwnership/Access

I mean, there is a question over who’s data it is, you know – that’s where the fluctuations are coming from because sometimes the data is regarded as belonging to the person and sometimes they’re regarded as belonging to the NHS, and there are different levels of belongingness depending on how identifiable it is to you

(Primary Care Specialist).

There is still an outstanding and increasingly urgent issue in relation to large data sets such as mobile monitoring data, where ethical and legal rights, IP, cost, storage, appraisal and resource implications for curation provide particular challenges and opportunities.

Challenges to investing Challenges to investing resources in curation?resources in curation?

.

Policy and ResourcesPolicy and Resources

…if people go for Research Council grants they are expected now to archive their data on the ESRC system. So presumably, yes, there has to be some protocol. It’s just that none of the funders we have at present have it as part of any policy.

(Primary Investigator)

MRC Data Discovery GatewayMRC Data Discovery GatewayThere is also potential scope for support from the emerging MRC Data Support Service, to be provided by the Science and Technology Facilities Council (STFC), in collaboration with Oxford University and University College London.

Stable, standardised, scalable computing infrastructures

Diverse, dynamic, distributed human infrastructures

Synergies are possible

Misalignments are more likely

Recurring problem: solution scenarios are rarely shared

Sharing

Curation

Access

Reuse

Preservation

SCARPcase studies

1.sampling

collecting coding cleaning linkage analysis use

the human process

the technical process

Barriers to Sharing DataBarriers to Sharing Data

Different environments

Different study designs

Different tools

Different populations

Different formats

Different study designs and procedures

Different Questionnaires

Coding? Format?

So what about data cleaning?: So what about data cleaning?:

A 46:36 waist/hip ratio reading – is it an input error or just a sample from West Lothian?

Harmonising kit and standardsHarmonising kit and standards

•Continua alliance• Benchmark testing• Algorithms

Disease Effect or Artefact?Disease Effect or Artefact?

Different equipment

Different populations

Different contexts

Different protocols

Different coding

Different metadata

Data curation requires an agreed framework for Data curation requires an agreed framework for data sharing or integrationdata sharing or integration

across sites (horizontal)

across scales (vertical …think Google Earth

Data curation requires ongoing support Data curation requires ongoing support for cross community working on for cross community working on

shared frames of reference

Shared standards

Shared naming conventions

Shared ethical and legal conventions

Shared costs and risks

Policy and governance

Recommendations includeRecommendations include

Build on existing shared infrastructure for clinical trials, and for ethics

Providing a forum for aligning the Providing a forum for aligning the efforts of different communitiesefforts of different communities

The vision of seamless data sharing is till more of a vision than a reality

HealthGrids, telehealth, biobanking are still working indiependently in the same disease domains

Using pilots as a laboratory Using pilots as a laboratory for strategyfor strategy

• The many telehealth pilots now operating in the same disease domain provide a laboratory for identifying emerging problems and opportunities on the ground and testing out approaches to managing data curation.

• Much of the challenge now derives from the need to provide opportunities and incentives for these communities to engage in the process of agreeing on data representation. curation and re-use in the same domain, and the process of renegotiating roles, risks, rights and ownership in the context of re-use.

Leverage Community Leverage Community Knowledge and AgencyKnowledge and Agency

The diabetes patients were bypassing the research convention and actually putting all of their own data, the diabetes patients, onto these websites, sharing it, and asking researcher to come and analyse it to answer their questions

( Primary Investigator/ Primary Care specialist)

If the community producing the data is responsible for data confidentiality and data quality there are huge benefits

(IT designer)

As web-based patient networks such as www.patientslikeme.com garner support, and patients increasingly have access to commercial storage facilities for electronic health records, (such as Microsoft HealthVault) they have begun to identify opportunities for user-led research that is curated by patient groups themselves.

These may provide unanticipated solutions to curational challenges relating to ownership, access and IP (Wyatt 2006)

Jenny.Ure@ed.ac.uk

The existence of a strong but disparate telehealth community makes it likely that a wiki –based approach (such as Neurowiki) might provide a lightweight first stage in building consensus around standards and metadata and there are already a number of projects exploring aspects of this at different levels as the full report indicates.

Use of different technical and mobile telecommunications standards is a particular issue here, and interoperability between tele-monitoring systems is limited, although initiatives such as the Continua Allianceare working with the wider community to agree on standards for interoperability that would facilitate the collection and federation of data from multiple sources and studies using different equipment.

In telehealth in particular, many context and patient-specific factors significantly impact on the viability of data for clinical or research use, and capturing this is a particular issue for distributed data from multiple studies. Telehealth, HealthGrid and Biobanking communities operate in the same disease domains, and share a number of comparable issues in managing the representation of data across sites, scales and over time, yet data harmonization efforts are still very community specific. Previous roadmapping workshops in this context highlighted very typical problems and useful strategies for addressing them, including (a) collaborative metadata and ontology development consortia sharing tools, content and standards, (b) road mapping opportunities to build on core measures of symptoms (Breton et al 2006) and (c) use of early prototypes and pilots as a vehicle for generating engagement and exploring data quality as well as problems in usability or opportunities that could be taken forward.

The increasingly distributed nature of technical and human agents that change and interact places new demands on both computational and curational strategies. This has driven the emergence of new models for mapping and managing change in complex, multi-scale health systems. Some of these concentrate on the implications for policy and practice (Bar Yam 2006; Plsek & Greenhalgh 2001), some on the potential for useful socio-technical alignment to make this achievable (Joslyn & Rocha 2003, Reddy et al 2009), and some on the potential of multi-agent-based models for representing or modelling change over time. In the context opf telehealth these include (Jami et al (2009); Masis et al (2006); De Toledo (2006), Latfi et al (2003) and D’Abiri et al (2003).

Recent work by Milner (2009) and others on bigraphs and reactive XML are being extended with the aim of better representation of dynamic change in such contexts.

“The world is increasingly populated with interactive agents distributed in space, real or abstract. These agents can be artificial, as in computing systems that manage and monitor traffic or health; or they can be natural, e.g. communicating humans, or biological cells. It is important to be able to model networks of agents in order to understand and optimise their behaviour.”

 

The many telehealth pilots now operating in the same disease domain provide a laboratory for identifying emerging problems and opportunities on the ground and testing out approaches to managing data curation. Much of the challenge now derives from the need to provide opportunities and incentives for these communities to engage in the process of agreeing on data representation. curation and re-use in the same domain, and the process of renegotiating roles, risks, rights and ownership in the context of re-use. As web-based patient networks such as www.patientslikeme.com garner support, and patients increasingly have access to commercial storage facilities for electronic health records, (such as Microsoft HealthVault) they have begun to identify opportunities for user-led research that is curated by patient groups themselves. These may provide unanticipated solutions to curational challenges relating to ownership, access and IP (Wyatt 2006).

References

www.continuaalliance.org

http://www.nesc.ac.uk/action/esi/contribution.cfm?Title=684

 

How does the process work on the ground?Changes in roles, risks, costs and benefits?Benefits and problems ?Barriers and enablers?Differences between practices/patients?We need more feedback from nurses!

Does monitoring change things?

Option 1 - Role Based AccessOption 1 - Role Based Access

The de facto standard

Persistent linked datasets more likely

Getting access is easier

Monitoring misuse is hard

Option 2 An additional layer?Option 2 An additional layer?

Checking for risks arising from linkage between particular datasets

Separating core and local processes

Option 3: Leveraging communitiesOption 3: Leveraging communities

Linkage assessment panel

Also combines ethical and quality roles

Existing roles and responsibilities support effective intervention to enhance security and quality