Designing and Disseminating Prognostic Models For Use In Routine Clinical Practice: Experience From The Framingham Heart Study

• Activity Number: 271
• Type: Topic Contributed
• Date/Time: Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Teaching of Statistics in the Health Sciences*
• Author(s): Lisa Sullivan*+ and Ralph D'Agostino, Sr.
• Affiliation(s): Boston University and Boston University
• Keywords: health risk appraisal ; clinical models ; prediction models

Abstract

The Framingham Heart Study has been a leader in the development of multivariable statistical models for estimating the risk of initial and subsequent stroke, coronary heart disease and cardiovascular disease. These models relate important risk factors, such as age, total, HDL and LDL cholesterol, blood pressure, diabetes, and smoking status to specific outcome events. While the statistical modeling has become more complex, the objective remains to develop and disseminate models that are useful to practitioners in everyday practice. Widely available computer software packages have facilitated the dissemination of sophisticated models, allowing practitioners to enter patients' risk factors to produce risk estimates. However, a requirement is access to a computer with the specific software. Score sheets have also been developed, which translate regression parameters into integer point systems, allowing clinicians to compute a patient's risk using simple paper and pencil computations. We will detail the methodology we use for producing these simple scoring systems. We will also discuss the appropriate interpretation of risk estimates, focusing specifically on precision.