Explaining annual updates
We release a new version of QRISK®2 every April.
Annual updates are required because of:
These factors require us to remodel the QRISK®2 algorithm to the latest version of the QResearch® database each year to ensure the algorithm keeps up to date. If the algorithm is not re-calculated, then its performance would gradually decay and its clinical value would diminish as a result. We have therefore re-fitted the algorithm using a three quarter random sample of the QResearch® practices and incorporated this into the annual update of the software.
- Changes in population characteristics - for example, incidence of cardiovascular disease (CVD) is falling; obesity is rising; smoking rates are falling;
- Changes in requirements for how the risk prediction scores can be used, e.g. changes in age ranges.
- Improvements in data quality - for example the recording of exposures and also clinical outcomes becomes more complete over time. This is especially true for recording of ethnicity data which is becoming more complete. All 1243 practices in England currently contributing to the QResearch® database (version 40) now have data linked at patient level to the Hospital Episode Statistics (HES) database. This means it is possible to identify patients who have been admitted to hospital for cardiovascular disease where this information is not already recorded on the GP record or linked mortality record from the Office of National Statistics (ONS). Both HES and ONS mortality data linkages extend back to 1997.
The 2016 update
We updated the coefficients for the QRISK®2-2016 algorithm using the latest version of the QResearch® database available at the time.
The annual update documentation linked from the 2016 part of the site shows details of how risk factor recording and CVD incidence rates have changed over time.
In summary, CVD rates are declining and there has been a marked improvement in recording of self-assigned ethnicity over time.
The improvement in recording of ethnicity has resulted in more accurate estimation of risk for the different ethnic groups as it's based on a much larger sample size.
The practical result is an increase in the weighting applied to the non-white ethnic groups which will result in a small increase to the corresponding risk scores.
As with any re-calibration, there are other changes to the relative risks given to the various input parameters. This is just what happens when a model is fitted to new data.
Still using QRISK®2-2015
That's just fine. QRISK®2-2015 works very well. We recommend that those who integrate QRISK®2 into their own systems move to the 2016 score by the autumn -- this means that they have the freedom to schedule the work needed into a convenient point of their release cycle.
Where are the calculators?
and here is