Ten-year risk of all-cause mortality: assessment of a risk prediction algorithm in a French general population

Abstract

While assessment of global cardiovascular risk is uniformly recommended for risk factor management, prediction of all-cause death has seldom been considered in available charts. We established an updated algorithm to predict absolute 10-year risk of all-cause mortality in apparently healthy subjects living in France, a country with high life expectancy. Analyses were based on the Third French MONICA Survey on cardiovascular risk factors (1995–1996) carried out in 3,208 participants from the general population aged 35–64. Vital status was obtained 10 years after inclusion and assessment of determinants of mortality was based on multivariable Cox modelling. One-hundred-fifty-six deaths were recorded. Independent determinants of mortality were living area (Northern France), older age, male gender, no high-school completion, smoking, systolic blood pressure ≥ 160 mmHg, LDL-cholesterol ≥ 5.2 mmol/l, and diabetes. Score sheets were developed to easily estimate 10-year risk of death. For example, a non diabetic, heavy smoker, 46-year old man, living in South-Western France, who did not complete high-school, with LDL-cholesterol ≥ 5.2 mmol/l and systolic blood pressure < 160 mmHg, has a 17% probability of death in the ten coming years. The C-statistic of the prediction model was 0.76 [95% CI: 0.72–0.80] with a degree of overoptimism estimated at 0.0058 in a bootstrap sample. Calibration was satisfying: P value for Hosmer–Lemeshow χ² test was 0.483. This prediction algorithm is a simple tool for guiding practitioners towards a more or less aggressive management of risk factors in apparently healthy subjects.

Appendix

The following steps lead to the estimation of the 10-year absolute risk of all-cause mortality for a given subject:

Step 1

Equation A = 0.55 (for people living in Northern France) + 0.04 (for people living in North-Eastern France) + 0.00 (for people living in South-Western France) + 0.00 (for age ranging between 35 and 44) + 0.50 (for age ranging between 45 and 54) + 1.35 (for age ranging between 55 and 64) + 0.00 (for women) + 0.74 (for men) + 0.00 (if high school has been completed) + 0.45 (if high school has not been completed) + 0.00 (for non smokers) + 0.96 (for smokers < 15 pack-years) + 1.04 (for smokers ≥ 15 pack-years) + 0.00 (if LDL-cholesterol < 5.2 mmol/l (200 mg/dl)) + 0.49 (if LDL-cholesterol ≥ 5.2 mmol/l (200 mg/dl)) + 0.00 (for non diabetic subjects) + 0.50 (for diabetic subjects) + 0.00 (if systolic blood pressure < 160 mmHg) + 0.42 (if systolic blood pressure ≥ 160 mmHg).

Step 2

Equation A is calculated for the average level of each explanatory variable:

B = 0.55 × 0.33 + 0.04 × 0.31 + 0.50 × 0.35 + 1.35 × 0.30 + 0.74 × 0.50 + 0.45 × 0.66 + 0.96 × 0.06 + 1.04 × 0.15 + 0.49 × 0.08 + 0.50 × 0.12 + 0.42 × 0.08 = 1.7873

Step 3

C = A−B (where B = 1.7873) and D = e ^C

Step 4

The 10-year survival Kaplan–Meier value S(t) is exponentiated by D and subtracted from 1 to calculate the 10-year risk of all-cause mortality for a given subject, according to his baseline characteristics and risk factors.

P = 1−[S(t)]^D (where S(t) = 0.96)

Example

The 10-year absolute risk of all-cause mortality is 16% for a man living in South-Western France, aged 46, who did not complete high school, has smoked more than 15 pack-years, is non diabetic, and has LDL-cholesterol ≥ 5.2 mmol/l and blood pressure < 160 mmHg.

Step 1

A = 0.50 × 1 + 0.74 × 1 + 0.45 × 1 + 1.04 × 1 + 0.49 × 1 = 3.22

Step 2

B = 1.7873

Step 3

C = 3.22−1.7873 = 1.4327 and D = e ^1.4327 = 4.1899 and

Step 4

P = 1−0.96^4.1899 = 0.1572