Cardiovascular Risk Scoring
The first risk scoring system was the Framingham Risk Score (FRS). It has been updated and improved several times, and other scoring systems have been developed. The approach involves statistical analysis to identify “risk factors” - identifiable and measurable physiological metrics - that predict, or at least suggest, the incidence of cardiovascular disease in the future. If a scoring system identifies an individual as being at high risk, the doctor is more likely to suggest a course of statins or other medicine. Indeed, guidelines issued by authorities explicitly use risk scores to recommended statin use.
To make the score useful and accurate is a challenge for the statistician. Statisticians use a process called Multivariate Regression to relate data about the patient to a risk number. You want the prediction model to be accurate (having a large number of inputs such as physiological metrics i desirable), but you also want the system to be user-friendly and that means asking for a low number of easily obtained metrics. Factors such as sex and age are easy to get, but the numbers from blood tests present a slightly higher barrier. Fortunately, routine blood tests yield LDL and TDL at low cost. However, the score architects still attempt to limit the number of factors they ask for.< The first Framingham coronary prediction method (1998) estimated the risk of coronary health disease (CHD) over the course of 10 years. Angina pectoris, myocardial infarction (heart attack), and death from coronary disease are considered CHD for the purposes of the scoring. Other heart and vascular diseases are not included.
The factors used to calculate the score include age, sex, total cholesterol level (mg/dl), HDL cholesterol level (mg/dl), systolic blood pressure (mm/Hg), and whether the person smokes. The answer is given as a percentage. Each person gets an individual score (Framingham Risk Score or FRS) that will change as the person ages and has changes in cholesterol and blood pressure levels.
After the publication of the FRS in 1998, other scoring systems were developed, and there were updates to FRS. All of these systems had limits (e.g. they claimed validity only for people in a given age range) and they had different input factors and predicted different types of cardiovascular diseases. The population of patients they used to develop the algorithms were different. Many systems have been validated – that is, independent testers applied the algorithms to patient pools and found results close to the predictions.
What is the purpose of these scores?
First, a high score can help change patient behavior by alerting them to risk. This "scared straight" method might be poor bedside manner, but it is effective for some patients.< More importantly, the score that results from the system helps doctors prescribe treatment. Indeed, the most current guidelines for US doctors explicitly call for the results of a score to be used when determining whether to administer a statin medication. You might think this is too simplistic, but use of a probabilistic method is partly a response to critics who objected to treating to target cholesterol numbers. The critics had said just focusing on cholesterol levels was wrong and that other factors should be considered. So that’s what the scoring systems do – they take into account multiple factors.
Inputs to scoring systems
Different systems require some or all of these:
Age – in years
Sec – male or female
Total Cholesterol (TC) – mg//dl
HDL – mg/dl
Systolic blood pressure (mmHg) – the higher number on the blood pressure reading
Blood pressure treatment (yes or no)
Current smoking (yes or no)
Family history of cardiovascular disease in first degree relative aged <60 years (yes or no)
C-reactive protein level in the blood – mg/dl
Some of the following:
Transient ischemic attack
Coronary insufficiency or angina
Fatal or nonfatal stroke
Transient ischemic attack
Validation and Simplicity
The major scoring systems recommended by public health authorities have been “validated” – meaning independent analysts have calculated scores for patients and looked at how those patients have fared and seen results match the predictions.
The great thing about the systems is that scores can be calculated from only a few measurements.
Joint British Societies Risk Calculator - http://www.jbs3risk.com/JBS3Risk.swf
ACC/AHA pooled cohort hard CVD risk calculator (2013) - http://tools.acc.org/ASCVD-Risk-Estimator/
Reynolds CVD risk score (both men and woman) - http://www.reynoldsriskscore.org/default.aspx
QRISK®2-2015 Web Calculator. - http://www.qrisk.org/
The Multi-Ethnic Study of Atherosclerosis - https://www.mesa-nhlbi.org/
Do the different scoring systems produce different estimates?
Of course they do. Otherwise there would be no point in developing a new estimation algorithm. They differ in inputs and outputs
One criticism is that a scoring system developed from data from one population may be inaccurate in predicting risk in another population. The FRS was developed from people living in Massachusetts and it has been criticized for both underestimating and overestimating risk in Europeans.
An article in the British Medical Journal found an "excess of models" for estimating cardiovascular risk and bemoaned the “methodological shortcomings, incomplete presentation, and lack of external validation and model impact studies.”
The website uptodate.com says anyone over age 40 “should undergo periodic CVD risk assessment every four to six years.” Who does this assessment? Presumably the person’s primary care doctor will. If people switch doctors the ball may be dropped, so the responsibility lies partly with the patient. This is a problem with our healthcare system, but perhaps people being encouraged to take more interest in their health and to maintain their own records will help mitigate this problem.