Big data and connected challenges in cardiovascular health

advertisement
Big data and connected challenges
in cardiovascular health
Harry Hemingway FFPH, FRCP
Professor of Clinical Epidemiology
Director Institute of Health Informatics, UCL
IUCL
n Farr Institute, London
Director
UCL Institute of Digital Health
22 June 2015
Farr Institute
Sanger Institute
Crick Institute
Turing Institute
The Farr vision: Big data target
Data
Precision Medicine
Citizen Health
Capacity
People
Programmes
Ethical, social
Tool
chain
Public Health
Data
Analytics
Drug Discovery
E-Infrastucture
Partnerships
Platforms
Public
Learning Health Care
The big data tapestry
Weber G, Mandl K, Kohane I, JAMA 2014
The big data tapestry
Weber G, Mandl K, Kohane I, JAMA 2014
William Farr (1807-1883) the original Mr Big Data
‘Compiler of Scientific Abstracts’ at General Register Office
•
• Photo, quote
• And a
•Computing and visualisation
•Hospital outcomes
•Public health statistics and intervention
•The first International Classification of Disease (ICD)
CALIBER 2 million adults, 5 billion data
points from linking 4 sources of EHR
Healthy, GP
registration
Myocardial infarction
hospitalization
See GP for
follow-up
Death
Time
Patient’s
experience
Primary
care
Pneumonia
hospitalization
Stable
angina
New patient
check: blood
pressure,
smoking status,
alcohol use etc.
Hospitalization
(HES)
Diagnosis of
stable angina.
Blood tests (e.g.
cholesterol).
Prescription of
aspirin, nitrates
etc.
Admit / discharge dates.
Primary diagnosis: Viral
pneumonia, not
elsewhere classified
Disease
Registry (MINAP)
Death Census
(ONS)
Denaxas et al., CALIBER, Intl J Epidemiology 2012
Herrett et al., CALIBER, BMJ 2013
Diagnosis of
myocardial
infarction
Blood tests,
blood pressure.
Prescriptions of
beta blocker,
statin, ACEi etc.
Sudden
death
Admit / discharge dates.
Primary diagnosis: Acute
myocardial infarction
Procedure: Percutaneous
coronary intervention
ECG, cardiac markers.
Diagnosis: STEMI
Date of death. Cause:
1) Rupture of abdominal
aortic aneurysm
2) Old myocardial
infarction
What does linked record data look
like?
The Farr vision: Big data target
Data
Precision Medicine
Citizen Health
Capacity
People
Programmes
Ethical, social
Tool
chain
Public Health
Data
Analytics
Drug Discovery
E-Infrastucture
Partnerships
Platforms
Public
Learning Health Care
Outcomes assessment: importance of linking
multiple record sources
Herrett et al, CALIBER, BMJ 2013;346:f2350
Cardiovascular diseases: lumpers or
splitters?
How does CVD first present?
In the real world, today
Intracerebral
haemorrhage 2%
Subarachnoid
haemorrhage 1%
Abdominal aortic
aneurysm 2%
Ventricular
arrhythmia/sudden
cardiac death 3%
MI/Fatal CHD 18%
Unstable angina 5%
CHD 10%
Ischaemic stroke
13%
Peripheral arterial
disease 11%
Stable angina 12%
Transient ischaemic
attack 11%
Heart failure 12%
George et al CALIBER 2015, in revision
N=1.93 million patients
>110K CVD events
5 year median follow-up
The Farr vision: Big data target
Data
Precision Medicine
Citizen Health
Capacity
People
Programmes
Ethical, social
Tool
chain
Public Health
Data
Analytics
Drug Discovery
E-Infrastucture
Partnerships
Platforms
Public
Learning Health Care
‘Risk factors’ have heterogeneous
associations with specific CVDs
Example: blood pressure
Example Abdominal aortic
aneurysm and discordance of
systolic and diastolic BP associations
Rapsomaniki et al
CALIBER Lancet 2014
Higher resolution epidemiology: type 2
diabetes and 12 CVDs
Shah et al CALIBER
The Lancet D&E 2014
EHR added to omic/imaging
biobanked cohorts
• EHR record linkages for CVD and other
disease outcomes
• EHR Phenotype algorithms
EHR added to genomic sequence
100, 000 Whole genome sequences
patients with cancer, infections, rare diseases (including
cardiac)
NHS Transformation Project
The Farr vision: Big data target
Data
Precision Medicine
Citizen Health
Capacity
People
Programmes
Ethical, social
Tool
chain
Public Health
Data
Analytics
Drug Discovery
E-Infrastucture
Partnerships
Platforms
Public
Learning Health Care
Personalised trajectories…..for risk
prediction
Crowther et al., CALIBER, 2015
Using all the data for prediction
1000s of events in one patient cosmos all clinical data over time:
Diagnoses, drugs, blood tests, consultations
Events
Luscombe Lab, CALIBER, 2015
The Farr vision: Big data target
Data
Precision Medicine
Citizen Health
Capacity
People
Programmes
Ethical, social
Tool
chain
Public Health
Data
Analytics
Drug Discovery
E-Infrastucture
Partnerships
Platforms
Public
Learning Health Care
Global reality: data-free medicine
The Learning Health System series
Institute of Medicine of the National Academies, 2012
Better outcomes
Sure
But better than what?
Hospitals: NIHR Health Informatics
Collaboration
A revolution in science ?
• Nearly all that we know about how to maintain and improve
our health in general and cardiovascular health in particular
comes from an era
• where data, information and knowledge were
–
–
–
–
–
Small
Simple
Sequestered (not shared)
Expensive
And interrogated in ‘mono’ disciplines
Farr London Investigators
CARDIOVASCULAR
• Mike Barnes, Director of Bioinformatics
• James Carpenter, Professor of Medical Statistics
• John Deanfield, Professor of Paediatric Cardiology
• Mark Caulfield, Professor Clinical Pharmacology
• Spiros Denaxas, Health Informatics Senior Research
Associate
• Nicholas Freemantle, Professor of Clinical Epidemiol
and Biostatistics
• Harry Hemingway, Professor of Clinical Epidemiology
• Aroon Hingorani, Professor of Genetic Epidemiology
• Steffen Petersen, Reader in Advanced Cardiovascular
Imaging
• John Robson, GP, Clinical lead for the Clinical
Effectiveness Group
• Liam Smeeth, Professor of Epidemiology
• Adam Timmis, Professor of Clinical Cardiology
INFORMATICS
• Anne Blandford, Professor of Human–Computer
Interaction
•
•
•
•
Peter Coveney, Professor of Physical Chemistry
James Freed, Head of Health Intelligence and Standards
Dipak Kalra, Professor of Health Informatics
John Shawe-Taylor, Professor of Computing
• Paul Taylor, Reader in Health Informatics
• Alan Wilson, Professor of Urban Regional Systems
MOTHER & CHILD
•
•
•
•
•
Peter Brocklehurst, Professor of Women's Health
Tito Castillo, Chief Operating Officer, LIFE Study
Carol Dezateux, Professor of Paediatric Epidemiology
Ruth Gilbert, Professor of Clinical Epidemiology
Irene Petersen, Senior Lecturer Epidemiology and Medical
Statistics
• Judith Stephenson, Professor of Reproductive and Sexual
Health
• Phil Koczan, Chief Clinical Information Officer
• Irwin Nazareth, Professor of Primary Care and Population
Science
• Max Parmar, Director of MRC Clinical Trials Unit
INFECTION
• Mike Catchpole, Head of Epidemiology and Surveillance
• Andrew Hayward, Senior Clinical Lecturer in Infection
• Richard Pebody, Head of the Seroepidemiology Programme
• Deenan Pillay, Professor of Virology
PHASE 2 CLINICAL WORKSTREAMS
• Andy Goldberg, Senior Lecturer in Trauma and Orthopaedics
• Anthony Moore, Professor of Ophthalmology
• Kathy Pritchard-Jones, Professor of Paediatric Oncology
• Martin Rossor, Professor of Neurology & Director of DeNDRON
Farr Institute’s First
International Conference,
St Andrews, Scotland,
26-28 Aug 2015
Thank you !
Download