QRISK2 validation by ethnic group

Julia Hippisley-Cox, Professor of General practice & Clinical Epidemiology,

Other Contributors:

November 13, 2013

Tillin et al recently reported a cohort study[1] comparing the performance of QRISK2 and Framingham in the Southall and Brent cohort in London. We have a number of comments on the study and the interpretation of results.

1. Number of events : The main problem with the paper, is that numbers are very small and given the resulting wide confidence intervals the authors have overstated their findings. There is no information on how many patients had the outcome in the different ethnic groups but there were only 78 events in women overall, so the sub-group numbers are likely to be extremely small. The fact that the authors reported little difference between the two scores is likely to be due to the very wide confidence intervals and small numbers.

2. Definition of outcome: We suspect the under-prediction of QRISK2 is due to their broader definition of the cardiovascular disease outcome which includes HES data and patient reported outcomes. They have also included revascularization procedures. They do not report how many events were from different sources or an assessment of how they validated the retrospective self-reported events. From 2014 - QRISK2 annual updates include linked hospital episode data which increases ascertainment of events by approximately 10%.

3. Characteristics of the cohort: The SABRE cohort largely comprised first generation migrants with baseline measurements made over 20 years ago. The QResearch database, which is used to develop QRISK2, includes a nationally representative group of diverse ethnic groups many of whom will be second or third generation at lower risk than first generation migrants.

4. Calculation of the QRISK2 score: A number of the data items needed to calculate a QRISK2 score were missing in their dataset. There was no baseline information on rheumatoid arthritis, family history of premature IHD, chronic kidney disease or atrial fibrillation. Absence of this information is likely to have lowered discrimination. Similarly, the cohort consisted of patients aged 40-69 which will have lowered discrimination compared with evaluation of QRISK2 across the wider age range of 30-84 years.

5. Reporting of performance by ethnic group. The authors mention that the QRISK2 score has been validated in internal and external datasets but that the performance of QRISK2 has not been reported separately by ethnic group. We would like to draw readers attention to the QRISK-2013 update information which was published online in April 2013 and which includes validation statistics separately for each ethnic group[2]. The tables at http://www.qrisk.org/QRISK2-2013-Annual-Update-Information.pdf demonstrates good performance in all ethnic groups.

References 1. Tillin T, Hughes AD, Whincup P, et al. Ethnicity and prediction of cardiovascular disease: performance of QRISK2 and Framingham scores in a UK tri-ethnic prospective cohort study (SABRE--Southall And Brent REvisited). Heart 2013 doi: 10.1136/heartjnl-2013-304474[published Online First: Epub Date]|. 2. Hippisley-Cox J, Coupland C. QRISK2-2013 Annual Update Information, 2013:5. http://www.qrisk.org/QRISK2-2013-Annual-Update-Information.pdf

Conflict of Interest:

JR chaired, and PB was a member of, the NICE guideline development group on cardiovascular risk assessment"The modification of blood lipids for the primary and secondary prevention of cardiovascular disease."JHC is codirector of QRESEARCH, a not for profit organisation that is a joint partnership between the University of Nottingham and EMIS (leading supplier of information technology for 60% of UK general practices).JHC is also director of clinrisk which supplies open and closed source software to implement QRISK2.

Conflict of Interest

None declared