The STRATIFY-CVD risk prediction model demonstrated excellent discrimination (C-statistic 0.832; 95% CI 0.831-0.833) and better calibration than the QRisk2 model for predicting cardiovascular events.
Observational (n=5,578,002)
Yes
Does the STRATIFY-CVD model improve cardiovascular risk prediction compared to the QRisk2 model in patients aged 40 and older with elevated blood pressure?
The STRATIFY-CVD model provides better calibration and discrimination than QRisk2 for predicting 10-year cardiovascular risk in a broad population including older adults, those with prior CVD, and statin users.
Effect estimate: C-statistic 0.832 (95% CI 0.831-0.833)
Objective: Antihypertensive treatment has long been prescribed with consideration of an individual's overall cardiovascular risk, using risk prediction models such as QRisk2 in the UK. However, QRisk2 was not developed for use in individuals with previous CVD, or existing statin treatment users, or those aged 85 years or older, and did not account for competing risk of death. The present study aimed to develop and validate a comparable cardiovascular risk prediction model for use in a broader population. Design and method: Participants aged 40 years and above, registered to an English primary care practice within the Clinical Practice Research Datalink (CPRD), with at least a single blood pressure reading between 130-179 mmHg were included the study. The outcome investigated was first non-fatal myocardial infarction or stroke leading to hospital admission or any cardiovascular death within 10 years of the index date. Model predictors were pre-specified based on the QRisk2 model, with the addition of CVD history and statin prescription. Performance of the model was assessed in both CPRD GOLD (apparent validation) and an independent dataset, CPRD Aurum (external validation). A Fine-Gray competing risks approach was used to account for death from other causes. Missing data was handled using multiple imputation. Results: A total of 5,578,002 patients (mean age of 59 years; 52% female) were included in the study. The cardiovascular risk model included 18 predictors and showed excellent overall calibration (Observed/Expected ratio 1.000, 95%CI 0.996-1.004) after recalibration and discrimination (C-statistic 0.832, 95%CI 0.831-0.833) upon external validation. The STRATIFY-CVD model typically predicted lower risks than the QRisk2 model. Overall, the calibration of the STRATIFY-CVD model was better than the QRisk2 model, which typically overestimated the risk of CVD across all age groups. Conclusions: In this study, a clinical prediction model for CVD events was developed and validated for use in older patients with elevated blood pressure and with a history of CVD and existing statin treatment. The model showed good discrimination and calibration upon external validation. Using this model, could improve informed decision making compared with existing risk prediction models such as QRisk2.
Wang et al. (2026) conducted an observational in Elevated blood pressure (n=5,578,002). STRATIFY-CVD risk prediction model vs. QRisk2 model was evaluated on First non-fatal myocardial infarction or stroke leading to hospital admission or any cardiovascular death within 10 years (C-statistic 0.832, 95% CI 0.831-0.833). The STRATIFY-CVD risk prediction model demonstrated excellent discrimination (C-statistic 0.832; 95% CI 0.831-0.833) and better calibration than the QRisk2 model for predicting cardiovascular events.