QResearch - qriskgeneralinformation

Technical Information about QRISK1 and QRISK2

This section is for those interested in understanding more of the technical details about QRISK®2 rather than the general public.
There is a search definition which describe the data items which are needed to calculate the QRISK®2 score (note this uses UK clinical coding schemes - an international version of QRISK®2 is in development).
There is a detailed description and discussion of the multiple imputation methods used in QRISK®1 (similar methods were used in QRISK®2). There are other statistical and technical notes in which can be found within the download section of this website.
Here is a pdf of the QRISK®2 paper as published in the British Medical Journal in June 2008
For a powerpoint presentation about advances in cardiovascular risk assessment by Dr Peter Brindle which was the plenary presentation at the recent HEART-UK conference (June 2008).
If you are an academic and are interested in validating QRISK®2 on another dataset or using it for a research project, then please contact us. There is some simple software available which will enable you to calculate QRISK2 scores reliably for your sample so you can be sure you are validating the right thing. We are happy comment on a protocol as appropriate and will be interested in seeing the results in a peer reviewed journal.

Glossary of terms

QRISK®1: This was the original CVD risk prediction algorithm publihsed in the BMJ in 2007. As well as traditional risk factors such as age, sex, systolic blood pressure, cholesterol/HDL ratio, it also included deprivation, family history of premature CHD, body mass index and use of antihypertensive medication.
QRISK®2: This is version two of the QRISK® CVD risk predition algorithm. QRISK2 was published the BMJ in 2008. In addition to the variables included in QRISK®1, QRISK®2 also included self assigned ethnicity, rheumatoid arthritis, chronic renal disease, diabetes and atrial fibrillation and age interaction terms. This is the first clinical release version which will be used in the UK.
QResearch: QResearch is the name of the database which contains electronic health records from primary care and which has been used to derive the algorithm.
Risk prediction algorithm: This is a mathematical formula which uses risk factors to generate a percentage risk of developing a particular outcome (eg Cardiovascular Disease)
Outcome: The outcome refers to risk of cardiovascular disease.
Cardiovascular disease (CVD): This refers to coronary heart disease, stroke or transient ischaemic attack.
Survivor function. The survivor function is a term which is the percentage of patients who haven’t had the outcome at X years. In the algorithm, this term has been adjusted for other variables in the regression analysis.
Years over which risk is predicted: Patients may wish to know their risk over 1, 2, 5 etc years so a QRISK2 can be calculated for different time periods or for different ages.
Risk Factors: Risk factors are variables which are used to predict an individual’s risk of developing CVD. Broadly speaking, they can be divided into ‘modifiable’ risk factors (which the patient can change, such as body mass index) and ‘not modifiable’ (eg age).
Missing data: Missing data refers to when a particular variable (eg body mass index, systolic blood pressure, serum cholesterol/HDL ratio) is not recorded in the patient’s electronic record.
Multiple Imputation: This is a powerful statistical technique which takes accounts of missing data in the datasets used for the modelling and tends to result in more powerful and less biased results.
Complete data. Complete data refers to the situation when all the necessary variables to generate the score are recorded in the patient’s electronic record.
Fractional Polynomials: These are terms which used to model non-linear functions.
Non-Linear Function: This is a function (ie relatonship between two continuous variables) which is not linear ie there is not a straight line relationship between the variables but a curve.
Validation: Validation can mean various things according to the context. Validation of a prognostic score usually means how well does ths score 'work' when it is applied to another dataset. A validation is said to be independent when the score has been applied to a dataset other than the one which was used to develop the model.
High risk: For the purposes of the guidelines, high risk is a CVD risk of >=20% over ten years.
Calculated risk: The calculated risk is the patient’s CVD risk score based on data which is already recorded in their electronic health records or entered directly into the individual use calculator.
Estimated risk: The mean estimated risk is the patient’s CVD risk score which has been calculated on the bases of recorded data but where missing data for either BMI, systolic blood pressure or cholesterol/HDL ratio have been replaced using predicted mean values based on the patients age, sex, family history and smoking status. Zero will be substituted if townsend scores are missing as this is the mean value nationally.
Townsend score: This is a proxy mesure for material deprivation evaluated at output area level (approximately 125 households). The Townsend score is based on four measures from the 2001 census as defined by Townsend. These are: Unemployed - percentage of economically active people who are unemployed: Overcrowding - households with occupancy rating of -1 or less as a percentage of all households: Non car ownership - households with no car or van as a percentage of all households; Non home ownership - households renting as a percentage of all households. We have a verison of the QRISK®2 algorithm which will work with other deprivation scores apart from Townsend score. Please contact us if this is of interest.

Quick Launch