Prognostic models with competing risks: methods and application to coronary risk prediction

• Published in: Epidemiology (Cambridge, Mass.) Vol. 20; no. 4; p. 555
• Main Authors: Wolbers, Marcel, Koller, Michael T, Witteman, Jacqueline C M, Steyerberg, Ewout W
• Format: Journal Article
• Language: English
• Published: United States 01.07.2009
• Subjects: Aged; Aged, 80 and over; Coronary Artery Disease - epidemiology; Female; Humans; Middle Aged; Netherlands - epidemiology; Prognosis; Proportional Hazards Models; Risk Assessment - statistics & numerical data; Netherlands
• DOI: 10.1097/EDE.0b013e3181a39056

Clinical decision-making often relies on a subject's absolute risk of a disease event of interest. However, in a frail population, competing risk events may preclude the occurrence of the event of interest. We review competing-risk regression models with a view toward predictive modeling. We show how measures of prognostic performance (such as calibration and discrimination) can be adapted to the competing-risks setting. An example of coronary heart disease (CHD) prediction in women aged 55-90 years in the Rotterdam study is used to illustrate the proposed methods, and to compare the Fine and Gray regression model to 2 alternative approaches: (1) a standard Cox survival model, which ignores the competing risk of non-CHD death, and (2) a cause-specific hazards model, which combines proportional hazards models for the event of interest and the competing event. The Fine and Gray model and the cause-specific hazards model perform similarly. However, the standard Cox model substantially overestimates 10-year risk of CHD; it classifies 18% of the individuals as high risk (>20%), compared with only 8% according to the Fine and Gray model. We conclude that competing risks have to be considered explicitly in frail populations such as the elderly.