To compare the Framingham, PROCAM and modified Framingham risk scores for predicting cardiovascular events (CVE) in a cohort from a cardiovascular prevention program in the Caribbean Region of Colombia

Retrospective cohort study. A population of 128,265 patients enrolled in the “De todo corazón” program between 2013-2018 with hypertension (HTA), diabetes mellitus (DM), dyslipidemia or chronic kidney disease (CKD) was considered. Patients with 30 to 74 years of age, with at least 6 months of enrollment, without personal history of previous CVE or coronary revascularization were included. The PROCAM, Framingham and two modified versions of Framingham risk score were estimated (first: a 75% adjustment; second: Framingham>20, patient receiving statins, BMI> = 25, CKD, triglycerides> = 200mg/dL). Patients with insufficient information were excluded. A multivariate logistic regression model using clinical variables as covariates was used to predict first CVE and was compared with each risk scale. The Delong test and a Boostrap with 1,000 resamples were used to evaluate differences in the AUC-ROC statistic.

A sample of 85,798 (66.9%) patients with 2,664 (3.1%) CVE was obtained. The estimated AUC-ROC (95% CI) were: Framighan Score = 0.623 (0.603 - 0.643), Framighan Score 75% = 0.641 (0.621 - 0.660), Framighan Risk Factors = 0.687 ( 0.668 - 0.705), PROCAM Score = 0.599 ( 0.556-0.641), PROCAM score = 0.651 (0.625-0.677), PROCAM risk factors = 0.676 (0.656-0.695), High risk groups = 0.708 ( 0.688-0.727). Establishing the AUC-ROC Framighan Score as a basis, the differences with high-risk groups were significant (Delong test: p <0.001, Bootstrap test: p <0.001), Framighan risk factors (Delong test: p <0.001, Bootstrap test : p <0.001) and risk factors PROCAM (Delong test: p <0.001, Bootstrap test: p <0.001).

The high-risk group better predicts CVD. The use of this scale would improve the categorization of cardiovascular risk in this population.