A frequently asked question about HRAs has to do with their predictive validity. Nothing is considered for inclusion in the HRA unless it is based on scientific studies and data that have demonstrated validity. Within the HPN HRA Midlife version, there are multiple algorithms for each of 19 different causes of death (CODs), each of which carries its own specific validity.

For 23 of the 42 causes of death covered in the HPN HRA Midlife version, the research base was not adequate to identify with reasonable validity causal risk factors that can be used in a prediction algorithm yielding a risk level for a specific individual. For these 23 causes of death the best we can do is offer population average 10-year mortality estimates. The validity of these 23 lies in the fact that they are based on national mortality statistics averaged over three years and based on approximately 7.8 million death certificates. For practical purposes, this represents population data. A wide variety of data sources were employed for the other 19 models; heart attack and stroke are the best examples. For heart attack and stroke, the prediction models were derived from the Framingham Heart Study data(D'Agostino, Russell, Huse, Ellison, et al., 2000; D'Agostino, Belanger, Markson, Kelly-Hayes, & Wolf, 1995; Wolf, D'Agostino, Belanger, & Kannel, 1991.) This longitudinal study, covering more than a half century, is probably one of the best examples of predictive validity in the medical literature. In the midlife HRA, heart attack and stroke combined will account for some 31 percent of the mortality outcomes. During its early development a number of studies also addressed the issue of the validity of the HRA (Meeker, 1988; Kirscht, 1989.).

Subsequently, most of the literature has focused on its effectiveness as a tool in health promotion programs (Burton, Chen, Conti, Schultz, Pransky, & Edington, 2005; Yen, McDonald, Hirschland, & Edington, 2003; Burton, Chen, Conti, Schultz, & Edington, 2003; Wright, Beard, & Edington, 2002; Musich, Ignaczak, McDonald, Hirschland, & Edington, 2001; Edington , Karjalainen, Hirschland, & Edington, 2002; Yen, Edington, McDonald, Hirschland, & Edington, 2001; Braunstein, Li, Hirschland, McDonald, Edington, 2001; Musich, Napier, & Edington2001; Musich, Adams, DeWolf, & Edington, 2001; Burton, Conti, Chen, Schultz, & Edington; 1999; Edington, Yen, & Witting, 1997; Yen, Edington, & Witting, 1994; Yen, Edington, & Witting, 1991).