n  Information for professionals and suppliers

What is the difference between QRISK1 and QRISK2

QRISK2 is a revised version of the original QRISK algorithm and takes into account additional factors which are known to affect an individual risk of developing heart disease or having a stroke/TIA attack.

In particular, it now includes self assigned ethnicity and diabetes as well as additional factors listed in the below. By including additional information it is likely that we can more accurately estimate risk for an individual patient.

The new QRISK2 analysis has been based on data collected between 1993 and 2008 (rather than up to 2007 as in QRISK1). The new analysis also incorporates cause of death using ONS death certificate data linked to individual patients who died during the 15 year study period. We used a more complex modelling technique to better account for how risk factors change with age which has tended to increase the risk estimates among younger people.

What is the difference between QRISK2 and the traditional Framingham score?

The QRISK2 CVD score contains many of the traditional risk factors included in Framingham (such as age, sex, cholesterol/HDL ratio blood pressure, diabetes and smoking status) but also contains important additional risk factors:

n  Self assigned ethnicity

n  Family History of premature coronary heart disease in a first degree relative under the age of 60

n  Deprivation (measured using the Townsend deprivation score)

n  Blood Pressure Treatment

n  Body Mass Index

n  Rheumatoid Arthritis

n  Chronic Kidney Disease

n  Atrial Fibrillation

Why was a new CVD risk score needed for the UK when we already have Framingham?

n  Estimates of CVD risk derived from equations are not an exact science but are better than clinical judgment alone for the estimation of CVD risk.

n  A number of risk assessment equations are available that estimate cardiovascular risk in individual patients. They have been derived from studies of individuals who have been followed up often for substantial lengths of time.

n  Risk assessment equations predict risk best in the type of population from which they were derived. The Framingham equations were derived from North American populations from the 1960s to the 1980s when coronary heart disease (CHD) was at its peak and they overestimate risk in contemporary European populations by around 100% in Southern European populations and by 50% or more in Northern European populations including the UK.

n  Conversely, such equations may underestimate risk in populations such as people with diabetes, South Asian men or the most socially deprived who are at higher than average risk. Overall the Framingham risk equation is likely to overestimate risk in the current UK population.

n  They may also underestimate risk in people with extreme risk factor levels or other clinical risks not included in the model.

Why was a measure of socio-economic deprivation included in the QRisk equation?

n  Cardiovascular risk is closely associated with socio-economic status such that people from deprived areas have higher risks.

n  Framingham equations do not include socio-economic status and underestimate risk in people who are relatively socially deprived.

n  The use of equations that do not include a measure of socio-economic status may exacerbate inequalities in CVD ie the difference between rich and poor.

What is the Townsend score, what does it measure and why was it used?

n  The Townsend Score is a measure of material deprivation based on where a person lives and obtained using their postcode and includes four variables obtained from census data: unemployment (lack of material resources and insecurity), overcrowding (material living conditions), lack of owner occupied accommodation (a proxy indicator of wealth) and lack of car ownership (a proxy indicator of income).

n  This score is considered the best indicator of material deprivation currently available and has been widely used in medical research including a range of studies conducted on the QResearch database.

How has the QRisk score been validated?

n  The original QRISK1 score and the QRISK2 score were both initially validated in a one third sample of the QResearch database by comparing its performance against the traditional   Framingham score which is currently in use in the UK. This research was published in the British Medical Journal in July 2007.

n  A second validation study was performed using the THIN database (which is a similar UK research database consisting of the electronic records of patients using a different computer system). This was published in the Heart journal in January 2008.

n  An editorial accompanying this study explains that the two main measures by which a risk prediction tool should be judged are calibration and discrimination. Calibration relates to how close the predicted risk is to the observed risk. Discrimination is the ability of the tool to differentiate between people who will have an event and those who will not, over a defined period of time (often five to ten years).

n  On both measures and in both studies, the QRISK2 here score outperformed Framingham indicating that it is likely to be more accurate than Framingham at estimating cardiovascular risk.

Has QRISK been compared against any other cardiovascular risk scores apart from Framingham and what were the results?

n  Yes, QRISK1 was compared against the Scottish ASSIGN score and the results published in our original BMJ paper. The results showed that QRISK1 was better calibrated to the UK population than ASSIGN and that QRISK1 had better discrimination. We have found similar results when comparing QRISK2 against the ASSIGN score in the QResearch database as shown in the table (higher scores represent better performance).

What further validation of QRISK2 is planned?

n  We are working with colleagues in other Universities who will be able to validate QRISK2 in independent populations including those from other countries. Our aim is to derive a CVD score which has international application. Colleagues interested in offering time and access to suitable data to undertake validation studies should contact Julia.hippisley-cox@nottingham.ac.uk in the first instance. There is no funding available, however, to support this work.  

Where patients are on antihypertensive treatment, should a pre-treatment blood pressure be used when calculating their risk?

n  No. QRISK2 has been designed such that if a patient is taking antihypertensive medication then their current blood pressure on treatment can be used rather than a pre-treatment value.

What is the difference between an ‘estimated’ QRISK2 CVD score and an ‘actual’ QRISK2 CVD score?

n  In EMIS the actual QRISK2 CVD score can be calculated where all values needed to calculate the score are available in the patient’s electronic health record.

n  In EMIS the estimated QRISK2 score is a CVD score which has been calculated using the data recorded in the patient’s electronic health record but also using reference values based on the patient’s age and sex where some data are missing.

n  For example, if a patient is a 55 year old male and has all the data for calculating a QRISK2 score in his ‘e’ health record except a systolic blood pressure, then the system will select a default value from a reference table for males aged 55 years.

n  An actual QRISK2 score is one which has been calculated using the patient’s actual recorded data with all the risk factor values available within a clinical consultation with the patient present. It doesn’t have to be done within a consultation.  An automatically calculated score with all the risk factor information would still be the ‘actual’ QRISK2 score. It will not be based on estimated default values.

I am a clinician and have a patient with a QRISK2  score of just under 20% - what should I do?

n  If the risk estimate is marginally below the threshold, clinical judgement should be used to determine whether further treatment of risk factors should be offered (for example, South Asian males).

Why is it important to measure HDL cholesterol in order to get the best estimate of CVD risk?

n  It is the ratio of total cholesterol to HDL cholesterol that is the best predictor of risk, better than either total cholesterol or HDL cholesterol alone.

n  Someone with a total cholesterol of 7.5mmols/l and an HDL cholesterol of 1.8mmols/l has a ratio of 4.2 which is associated with a considerably lower risk than someone with a total cholesterol of 5.8mmols/l and an HDL of 0.8 (ratio 7.2)

What about other factors which may increase CVD risk but are either not included in the 'score' or fully accounted for?

n  There are other factors which can increase CVD risk which are not directly accounted for in the QRISK2 algorithm.

n  These include, for example, alcohol excess, very heavy smoking, poor diet, lack of exercise and extreme obesity. These factors will all tend to increase risk of cardiovascular disease.

n  There is a strong relationship between deprivation and CVD risk, and so the inclusion of deprivation in the QRISK2 algorithm will take account of this to some extent.

Has NICE now recommended QRISK2?

n  QRISK was provisionally recommended by NICE following the commissioned expert reviews of QRISK. Finally however, NICE chose not to take the advice of its three specially commissioned international experts and decided to develop a new version of the Framingham equation by applying ‘fudge factors’ for family history and ethnicity.

Will the QRISK2 CVD algorithm be updated and change over time?

n  QRISK2 is a dynamic risk factor score developed from live anonymised electronic health records recorded by thousands of family doctors in the UK. We know that the characteristics of the population change over time and this will affect the equation itself. For example, the incidence of heart disease itself has fallen over the last 30 years, blood pressure has fallen, obesity is rising, smoking patterns have changed.

n  QRISK2 will therefore be updated periodically in order to reflect these population changes and also to take advantage of the continual improvements in the quality of electronic health records and the latest evidence regarding new or additional risk factors.

When will QRISK next be updated?

n  We expect that the next version of the QRISK algorithm will be available in 2009/2010.

I am a GP and use EMIS clinical computer system. Will QRisk be available in the clinical system?

n  EMIS are developing software which will enable all practices to generate a list of high risk patients for primary prevention in ‘population manager’ based on the new QRISK2 equation.

n  This software will flag patients according to what assessments and interventions are needed so that the appropriate patients can be recalled.

I am a GP who uses other clinical computer systems apart from EMIS  - will QRisk be available within my clinical system?

n  QRISK2 will be available under license to the system suppliers for all clinical computer systems in England and Wales.

n  It will be up to the system supplier to decide on whether to implement QRisk within the clinical system or not.

I am the CHD lead for a PCT and would like to use QRisk in all the practices in my PCT as we have a Locally Enhanced Service.

How can I achieve this?

What data is available to the practices themselves and to me as a PCT.

What training is available?

n  Practices using EMIS will have QRISK2 incorporated into the clinical system so that they can generate a CVD recall list and establish a CVD primary prevention register. Other computer suppliers may also provide this for their practices.

n  In addition, there are third party software companies which have CVD prevention software. A list of approved suppliers which hold a QRisk license will be available.

n  EMIS\QResearch may in future provide a PCT level extraction service for EMIS practices subject to governance arrangements, practice consent and costs being met.

n  EMIS field staff will be providing training for EMIS practices.

n  We expect PCTs and health communities to provide training locally. No other training is being provided by QResearch or EMIS.

Will QRisk be available free for research and non commercial use?

n  Yes – QRISK2 will be available on CD from EMIS for njon commerical academic and personal use. The CD will have a individual windows based patient calculator which can be installed on a PC, a 'batch processor' which will apply the QRISK2 score to mutliple records. The integrated package will include the postcode and clinical reference tables. This will be £50 per annum and will cover postage, package and administration costs for quarterly updates to the reference tables and annual updates to the CVD algorithm where appropriate.

n  There will be an annual license fee for commercial use.

I am interested in using the QRisk CVD algorithm within my own software, where can I find out more information?

n  We have developed a QRISK2 software development kit designed to be used by commercial companies who wish to embed QRISK2 in their applications.

n  For further details contact information@qrisk.org

I am an academic who is currently developing some software with a commercial company which we plan to sell to third parties. I am being paid for my expertise. Can I have a free academic license?

n  No. The arrangements are the same as for a commercial supplier as it is the intended use which is important.

What is the definition of commercial use?

n  Use for research purposes means use within a specific research project where the output is a research paper intended for publication in a peer reviewed journal.

n  Research expressly excludes service delivery and Commercial Use and the benefits derived from such use must not constitute a financial gain.  “Commercial Use” means any purposes which seek to exploit the data for financial gain or any purpose which is likely to place the use of the data in competition with a third party who is seeking to exploit the licensed data.  “Financial Gain” means a benefit accruing where the licensee or any third party used by, or connected to, the licensee receives any revenue or financial credit when using the data. 

n  To avoid any doubt, commercial use (or any other use which falls outside the definition of research) will require further direct discussion and a further licence to be obtained.

As clinical director of a primary care clinical computer system, I have been following the discussions around QRisk. I am keen to review how this scoring system could be integrated within our clinical system. Can you tell me what arrangements are in place for suppliers to have access to the scoring algorithm and supporting data?

We have published the technical documentation which describes the input variables/search definition and also how to generate the score with the coefficients using the QRISK batch processor utility. In addition, we have also prepared a mini software development kit which has a set of DLL class libraries (and equivalent for Linux) which will be licensed and will:

n  'Fill in' missing values for blood pressure, BMI, cholesterol ratio using age-sex reference data where these values are missing.

n  Generate an 'actual' QRISK2 score given the right input parameters for patients with complete data or an 'estimated' QRISK2 score where one or more data items are missing.

n  The software can then be used in single use mode or batch mode (eg to run against the practice database to generate a recall list).

The package will be available under annual license fee. In addition to the above, it include the following components:

n  The read code and search definition needed to extract the patient level data which needs to be fed into the algorithm.

n  A postcode to deprivation mapping table to generate the deprivation score.

n  The age-sex reference data for the blood pressure, cholesterol, body mass index.

n  Test harness/test set of data.

The idea behind this approach is to ensure 

n  The QRISK2 score is implemented as simply and correctly by suppliers as possible as it is fiddly to calculate.

n  Also, it is likely that we will need to update the score every couple of years (as per NICE recommendation) so will want to be able to issue updates simply and reliably.

I have further questions on QRisk which haven’t been answered here – who can I contact?

Please email information@qrisk.org  and put QRISK2 in the title of the email.

The prevalence of diabetes in the QRISK2 paper seems lower than I would have expected compared with QOF data. Can you explain this please?

The QRISK2 study started in 1993 and ran over a 15 year period during which the prevalence of diabetes has more than doubled as you can see from chart 11 in a previous analyses of the Qresearch database. Also, at baseline, we also excluded patients with cardiovascular disease, some of whom would also have a diagnosis of diabetes. This would tend to have the effect of reducing the prevalence of diabetes in the remaining population. Another study using the QResearch database reported a prevalence of 3.54% which is very similar to the national QOF data from 2006.

Ethnicity recording is likely to be increasing now it has been included in the GP Quality and Outcomes Framework. What percentage of patients have ethnicity recorded on the QResearch database on 01 April 2008?

The recording of ethnicity is increasing within general practice and this is reflected on the QResearch database. Overall one third of patients registered with practices contributing to QResearch have self assigned ethnicity recorded in their medical records. The increase is likely to continue and we will therefore keep updating QRISK to take advantage of these data in order to get the estimates for people from different groups as precise as possible.