In 1948, the National Heart Institute embarked on an ambitious project in conjunction with Boston University, called the Framingham Heart Study. Its goal: to find which factors cause CVD. In 1948, 5,209 men and women were recruited from the town of Framingham, Massachusetts that had not yet been affected visibly by CVD (Framingham Heart Study, 2018a). Every 2 years, these individuals underwent a detailed medical history, physical examination, and laboratory testing. Over the years, new generations of patients were recruited and subjects have continued to return every 2 years for assessment, until today.
It was through this long-term, prospective study that the risk factors for CVD were first identified. The link between smoking and CVD was first reported in 1960 (Framingham Heart Study, 2018b). After that, cholesterol, high blood pressure, and diabetes were eventually linked to CVD as well. Starting in the 1990s, risk scores for developing specific types of CVD (for example, myocardial infarction, PVD, CHF) began being published. In 2008, the general Framingham Risk Score was published. The Framingham Risk Score assigns individuals a risk score for experiencing CVD events within a 10-year period based on five major CVD risk factors: age, high blood pressure, cholesterol levels, smoking status, and diabetes. The following is a summary of the general cardiovascular risk-scoring criteria for women (D'Agostino et al., 2008).
Men have similar criteria with slightly different point values:
| Points | Age (y) | HDL cholesterol | Total cholesterol | SBP not treated | SBP treated | Smoker | Diabetic |
| -3 | <120 | ||||||
| -2 | 60+ | ||||||
| -1 | 50-59 | <120 | |||||
| 0 | 30-34 | 45-49 | <160 | 120-129 | No | No | |
| 1 | 35-44 | 160-190 | 130-139 | ||||
| 2 | 35-39 | <35 | 140-149 | 120-129 | |||
| 3 | 200-239 | 130-139 | Yes | ||||
| 4 | 40-44 | 240-279 | 150-159 | Yes | |||
| 5 | 45-49 | 280+ | 160+ | 140-149 | |||
| 6 | 150-159 | ||||||
| 7 | 50-54 | 160+ | |||||
| 8 | 55-59 | ||||||
| 9 | 60-64 | ||||||
| 10 | 65-69 | ||||||
| 11 | 70-74 | ||||||
| 12 | 75+ |
Today, these same five risk factors (four of them being preventable) are continuously expounded by our doctors when we visit their offices.
To calculate the total score, the values for each of the six risk factors (age, HDL level, total cholesterol, SBP not treated, SBP treated, smoker, diabetic) are added. The following table shows how point values correspond to 10-year risk scores (D'Agostino et al., 2008):
| Points | Risk (%) | Points | Risk (%) | Points | Risk (%) |
| <-2 | <1 | 6 | 3.3 | 14 | 11.7 |
| -1 | 1.0 | 7 | 3.9 | 15 | 13.7 |
| 0 | 1.2 | 8 | 4.5 | 16 | 15.9 |
| 1 | 1.5 | 9 | 5.3 | 17 | 18.5 |
| 2 | 1.7 | 10 | 6.3 | 18 | 21.5 |
| 3 | 2.0 | 11 | 7.3 | 19 | 24.8 |
| 4 | 2.4 | 12 | 8.6 | 20 | 28.5 |
| 5 | 2.8 | 13 | 10.0 | 21 + | > 30 |
You may be wondering, "What is the method behind the madness?" To develop these scores, the study's authors used the Cox proportional hazards regression, which is similar to logistic regression, except that instead of determining how variables are associated with a binary outcome, it determines how variables are related to the quantity of time before an event occurs. They even calculated the C-statistic of their risk score (analogous to the area under the curve, discussed in Chapter 3, Machine Learning Foundations) which was 0.76 to 0.79. This is a very good score that can be obtained just from a patient history, physical examination, and some simple blood tests.