Table 1.
Demographics in RealAge California Full Identifier Sample.
Table 2.
Causes of Death in RealAge California Full Identifier Sample.
Table 3.
Summary Statistics in RealAge California Full Identifier Sample, ages 30 to 79, no diabetes or pre-existing circulatory conditions.
Table 4.
Comparison of Age and RealAge.
Table 5.
Framingham Hard Coronary Heart Disease 10-Year Risk Model results.
Figure 1.
Comparison of RealAge, Age, and Framingham ATP-III models. (Time-dependent ROC curve at 5 years).
This figure shows the discriminative ability of Cox proportional hazards models for age, age and RealAge, and the Framingham ATP-III scores. Models with ROC curves higher and to the left show better discrimination between unhealthy and healthy respondents–here meaning that, between a pair of randomly chosen respondents, the deceased user was assigned a higher predicted mortality risk. Calibration plots are included in Appendix S1 (Figures S7 through S11), and show that models with unscaled age are well-calibrated. Results in the text (the net reclassification improvement and the integrated discrimination improvement) show that the RealAge score provided improved case-by-case risk predictions compared to the Framingham models.
Figure 2.
Comparison of RealAge score distributions for all deceased/surviving users and deceased/surviving “healthy” users - heart disease.
The red lines indicate that the users are deceased and black lines indicate that the users were not identified as deceased. The dotted lines indicate that the group excluded users with a history of heart attack, stroke (or interest in either of heart attack or stroke topics), or smoking.