QRISK a new CVD risk score development validation

QRISK: a new CVD risk score development & validation Julia Hippisley-Cox Calgary 23 Oct 2007 www. qrisk. org

Acknowledgments coauthors Carol Coupland Yana Vinogradova John Robson Margaret May Peter Brindle statistician GP www. qrisk. org

Acknowledgments • EMIS practices • David Stables & Andy Whitam (EMIS) • Dept Health www. qrisk. org

Goals of presentation • Background to CVD risk prediction • Overview of development of QRISK (new score) • Key results & validation • Web calculator www. qrisk. org

www. qrisk. org

What is Framingham? • Framingham small town near Boston • Half the size of Mansfield • Recruited a cohort of 5, 000 people in 1950 • Followed them for > 50 years • It has been hugely important www. qrisk. org

www. qrisk. org

Framingham Highlights 1959 – described the ‘silent MI’ 1960 – smoking, high BP & high chol bad 1967 – exercise good 1976 – menopause bad (for the heart!) 1978 – psychological factors can be bad 1988 – high HDL good 1991 – risk prediction equation used ALL AROUND THE WORLD www. qrisk. org

Why a new CVD risk score? – Small cohort 50 years ago from one American town – Almost entirely white – Developed during peak incidence CVD in US – Overpredicts CVD risk by up to 50% – Doesn’t include BMI, family history, blood pressure Rx, deprivation – But crucially it under estimates risk in patients from deprived areas www. qrisk. org

Policy context • NICE publication of lipid modification guidelines July 2007 • Statins recommended if CVD risk > 20% • Need to population screening tool to identify high risk patients • Dept Health considering ‘life check’ • Potential utility of routinely collected data • Interested in self assessment • Concern about health inequalities www. qrisk. org

Inverse equity hypothesis • This shows that when new interventions are introduced – Inequalities initially worsen (uptake quickest in the healthy and wealthy) – Eventually there is a ‘catch up’ but only when the wealthy reach a ‘ceiling’ • Need to be proactive to avoid new policies exacerbating health inequalities www. qrisk. org

AIM for QRISK • • • New CVD risk score Calibrated to UK population Better discrimination Use routinely collected GP data Include additional known risk factors (eg FH, deprivation, BMI, BPRx) www. qrisk. org

QRISK – general approach • QRISK is a new approach designed – to tailor management to the individual patient – to identify patients at high risk of disease – to identify those most likely to benefit or be harmed by treatment • Present risks and benefits back to patients at the point of care in an accessible way www. qrisk. org

QRESEARCH database – the largest GP database worldwide – 525 practices, 10 million patients ever – Good historical data > 12 years – Numerous validation studies www. qrisk. org

QRISK study cohort • Derivation cohort (2/3 rds practices) & validation cohort (1/3 rd) • All patients registered 1995 -2007 • Men and women aged 35 -74 • UK sample free from CVD & diabetes • Ethnically & socially diverse • 66, 000 Cardiovascular disease events • 8. 3 million person years www. qrisk. org

Cardiovascular disease outcomes • Computer recorded clinical diagnosis of – Coronary heart disease – TIA or Stroke • Outcome similar to that in JBS 2 • Validation against ONS certified cause of death 94% ascertainment www. qrisk. org

QRISK risk factors Traditional risk factors – Age, Sex, Smoking status – Systolic blood pressure – TSC/HDL ratio – (LVH – recorded prevalence too low) New risk factors – Deprivation (townsend score output area) – Family history premature CVD 1 st degree relative < 60 years – Body mass index – BP treatment www. qrisk. org

Validation Comparison against Framingham Independent one third of the database – Various statistics – Predicted vs observed CVD events – Clinical effect in terms of reclassification of patients into high/low risk www. qrisk. org

Validation statistics (note: higher scores are better) D statistic* (women) QRISK Framingham 1. 52 1. 39 R squared* 35. 5% (women) D statistic* 1. 42 (men) 31. 7% R squared* 32. 4% (men) 29. 1% 1. 31 www. qrisk. org

Degree of over prediction QRISK Framingham Women 2% 18% Men 0% 47% www. qrisk. org

www. qrisk. org

% of the UK population 35 -74 years at high risk >20% % All patients 3574 years Estimated numbers 2005 QRISK Framingham 8. 5% 12. 8% 3. 2 million 4. 7 million www. qrisk. org

Clinically important issue is degree of reclassification If we use QRISK rather than Framingham – Overall one in 10 reclassified – QRISK identifies different group of patients who are at higher risk – Incorrect classification affects patients from deprived areas (ie more high risk patients missed) www. qrisk. org

Strengths of QRISK – Better calibrated for UK than Framingham – Less likely than Framingham to over predict risk – Better at identifying patients likely to clinically benefit from treatment – Includes deprivation which makes it fairer – Can be implemented into GP computer systems – Can be periodically updated and refined www. qrisk. org

QRISK: Web calculator • Designed for patients to use • Also needs integration into clinical system Link http: //www. qrisk. org Username qrisk Password beta test www. qrisk. org