Design: risk model derivation, regional recalibration and external validation
Population: 611,778 adults over 40 without prior cardiovascular disease for derivation; 1,336,824 for validation
Primary outcome: 10-year and 30-year incident heart failure
Effect: validation C-indices 0.827 (0.824-0.829), 0.839 (0.827-0.850), 0.874 (0.863-0.884)

SCORE2-HF was derived in 25 prospective cohorts across 14 countries - 611,778 people, 21,818 incident heart failure events - and predicts 10-year and 30-year risk of incident heart failure in adults over 40 with no previous cardiovascular disease. The model is sex-specific and adjusted for competing risk. Its inputs are the ones already in the notes: age, smoking, systolic blood pressure, antihypertensive treatment, BMI, eGFR, and type 2 diabetes including age at diagnosis and HbA1c.

It was recalibrated against contemporary incidence in four European risk regions using linked records on more than 36 million people, then validated in three further cohorts totalling 1,336,824 participants. C-indices in validation were 0.827, 0.839 and 0.874.

The illustrative figures show why regional calibration matters more than the model. A 70-year-old in a low-risk region with none of the four adverse factors has an average 10-year risk of 8% if male and 6% if female; with all four - smoking, type 2 diabetes, hypertension and BMI of 30 or more - that becomes 24% and 20%. In the very high-risk region the same four-factor profile averages 59% in either sex. India is not among the calibration regions, so the risk estimates cannot be read across; the ranking of who in your clinic is at higher risk survives the transfer better than the percentage does.

Use it to rank risk within your own population rather than to quote an absolute percentage to an Indian patient
Note that eGFR and HbA1c are inputs - the model needs bloods, not just a consultation
Age at diabetes diagnosis carries weight; record it rather than duration alone
A 30-year horizon is where the model speaks to younger patients with modifiable factors
Do not substitute this for an atherosclerotic risk score - it predicts heart failure, a different endpoint

Why it matters

Heart failure prevention has lacked the risk-estimation scaffolding that atherosclerotic disease has had for two decades.

The statistics, in plain English

A C-index of about 0.83 means that given one person who developed heart failure and one who did not, the model ranks them correctly around 83% of the time - good discrimination, and typical of the better risk models. Discrimination is not calibration: a model can rank people well and still be systematically wrong about absolute risk in a population it was not recalibrated for, which is exactly the caution for use outside the four European regions. The huge sample sizes make the confidence intervals narrow but do not correct that.