why is ehealth so hard

Hildegard Franke Dr Ian McNicoll

Why is eHealth so hard?

Is it so hard?

The banks can do it ….

why not healthcare ?

but in the real-world …Large (sometimes whole) populations involved

large amounts of data

imaging, genomics, life-long records

Segmented populations adult, paediatric care, obstetric care, mental health

Safety critical industry very distractible environment

very mobile workforce

Generally poor IT infrastructure

Connectivity patchy, limited devices

.. and ..Healthcare is hugely expensive

highly politicised

“all change” at the next election

Highly culturally nuanced “That’s not how we do things here”

Ageing Population timebomb

… not forgetting …Patients

Increasingly mobile, national and international

Access to records ? ownership

mHealth and the ‘Quantified self’

Information governance Data security

Patient privacy

Data quality

Data retention / audit trails

Jargon Jungle

Jargon Jungle

Jargon Jungle

Jargon Jungle

Jargon Jungle

ADL

Jargon Jungle

Jargon Jungle

CKM

Jargon Jungle

Jargon Jungle

REST

Jargon Jungle

Jargon Jungle

Healthcare is powered by information To facilitate immediate care

Aide-memoire for clinician

Automate business processes

Trigger decision support

To facilitate review of care quality Quality registers / audit

To inform healthcare management

To inform research ‘Big data’, analytics

Clinical trials

Sharing health information

Most health information is captured on paper

Even when electronic it is largely text

Even when computable it is generally non-standardised

Even when computable it is often of variable quality

The challenge of ‘semantic interoperability’Complexity of health systems

makes interoperability an issue

even within single vendor applications

Between applications globally largely absent in spite of huge investment

Some success Lab requesting, reporting

GP2GP project

eReferral, EPS

Why is health information so hard?Very different kinds of information

Biological concepts

Documentation of care

Compressed, summarised information Reporting

Making sense of complex care

Large number of complex datatypes not just string, numeric, url, date

Health information: IBiological ‘real-world’ concepts

Symptoms, examination signs

Lab tests, imaging results

Illnesses, procedures

Asthma

Depression

Appendicectomy

Medications, devices

Penicillin 250mg tablets

Hip arthroplasty component

Health information: IIDocumentation of care

Diagnosis

Name = Diabetes (biological concepts)

Date of onset

Date diagnosed

Diagnosed by

Grade or stage

Health information: IIIDocumentation of care

Medicolegal context

Who, when, why, what, how?

Patient, clinical author, committer, committal date, organisation, care setting, episode of care identifier

Lab test

Name = Oral glucose tolerance test

Result = Abnormal

Diagnosis

Name = Diabetes (biological concepts)

Date of onset = 2015

Health information: IVCompression and processing of health information

Registry and research

Diagnosis = ‘Diabetes’ (CodedText)

Diabetes Y/N (boolean)

Handling ‘health status’

‘Curated’ list of patient’s current problems which change and morph over time

Remove ‘inactive’ or trivial problems

Add new problems ? all problems

From whose perspective?

Complex datatypes IPlain text / marked-up text

Coded terms external ‘ reference’ terminologies

local codes and mappings

Dates and times Partial dates “July 2014”

Date intervals

Complex datatypes IIDurations

length of symptoms

“take for 7 days’

ages

‘Ordinals’ Coded terms with associated numerics

Multimedia audio, video, images

Complex datatypes IIIProportions (70% oxygen, 0.7)

Quantities with units SI units (3mg, 4 mmol/l, cm, inches)

Dose units (2 tablets, 4 drops)

Normal ranges

Approximations (approximately 7mg/l)

Out of range statements (> 7mg/l)

Interpretations very high, high, normal, low, very low

Medication ‘dose syntax’“Co-codamol 8mg/500mg/5ml oral suspension 5-10mls 4-6hourly for 7 days for pain, maximum 40mls daily”

Medication ‘dose syntax’

Medication ‘dose syntax’

How do we cope with product versus dose prescribing?

Task / workflow management

Ordering “Request referral”

Track progress of referral

Referral requested

Referral scheduled

Referral performed

Referral cancelled

Information model is critical

Information model is critical

Health terminologies

Needed to represent biological and scientific concepts

500,000+ SNOMED CT concepts

even more relationships which support ‘inferencing’

e.g. Diabetes Mellitus Type 1 is_a metabolic disease AND is_a disease of the pancreas

A mixed economyIn reality we need

‘Clinical content’ modelsunderpin persistence

underpin API / messaging

+ Terminologies

label/ classify biological concepts

power inferencing

Blended approachno ‘one true way’

Health information is “Nasty”

Health information is Complexity is a challenge for any e-Health developer

building APIs, messages

building apps

Need good requirements from clinicians

How do we know they are widely applicable?

How do we make sure that we capture and share that clinical knowledge?

“Nasty”

Health information is Complexity is a challenge for any e-Health developer

building APIs, messages

building apps

Need good requirements from clinicians

How do we know they are widely applicable?

How do we make sure that we capture and share that clinical knowledge?

Health information is Complexity is a challenge for any e-Health developer

building APIs, messages

building apps

Need good requirements from clinicians

How do we know they are widely applicable?

How do we make sure that we capture and share that clinical knowledge?

“Superknifflig”