Validation of a new predictive risk model for mortality using Framingham Heart Study data

Abstract

BackgroundThe Institute for Health Metrics and Evaluation (IHME) developed a risk model to estimate 10-year mortality risk compared with a population with optimum risk factors in adults aged 30 years and older. IHME used data on exposure distributions of 12 behavioural and biometric risks in the US population, mortality rates by cause, and estimates of the proportional hazards of risk factor exposure from systematic reviews. The model has been validated using National Health and Nutrition Examination Survey (NHANES) data collected during 1988–94 and 1999–2004 (n=8331). MethodsWe will use data from participants attending the Framingham Heart Study (FHS) Offspring Examination 7 (1998–2001) and the Original Cohort Examination 20 (1986–90) to further validate the 10-year mortality risk model. These examinations were chosen since risk factors used in the IHME development mortality risk model were measured in these examinations. FHS participants aged 30 years and older at the time of examination will be included. Model performance will be assessed by (a) area under the curve to assess the ability of the model to discriminate between FHS participants who did and did not die within 10 years; and (b) Hosmer-Lemeshow goodness-of-fit tests comparing the FHS-observed mortality rate with the model-predicted mortality risk. FindingsThe FHS validation cohort consists of 3500 participants (mean age 62·4 years); 55% are women. Unlike NHANES, the FHS cohort sample is predominantly white. The FHS cohort is approximately 10·6 years (95% CI 10·1–11·1) older than NHANES, on average; FHS and NHANES have 55% and 53% females, respectively. The observed mortality rate is 14·3% (95% CI 13·2–15·5) in FHS and 8·4% (7·8–9·0) in NHANES cohort samples. We will present the findings of the validation on the FHS participants and compare with NHANES validation results. InterpretationAdequate performance of the risk model when applied to FHS, a sample with characteristics somewhat different from NHANES, will further support its validity and usefulness in the US population. FundingNational Heart, Lungs and Blood Institute at the National Institutes of Health (2 N01-HC-25195-06).