Fig. 2: Development, robustness and comparators of EMRAge.
From: OMICmAge quantifies biological age by integrating multi-omics with electronic medical records
a, Pairwise correlation (Corr) between four different estimates of EMRAge at the following time points: 1 January 2008, 2010, 2012 and 2014. ***P < 0.001. b, HRs and confidence intervals (CIs) of a 1-s.d. change to onset of aging-related diseases. These values were estimated in a subset of the testing dataset from the MGB Cohort, in which PhenoAge is available (N = 5,171). The models were adjusted for chronological age (if EMRAge or PhenoAge), sex, race, body mass index (BMI), smoking status and alcohol consumption. c, Kaplan–Meier plot of EMRAge versus PhenoAge versus chronological age in the testing dataset. Error bands show the 95% CIs for the estimated survival probabilities in each age group. d, Forest plot of HRs and ORs per s.d. change of EMRAge or PhenoAge for aging-related health outcomes when both variables are included as predictors in a single model, without additional adjustment for covariates. Error bars show the 95% CIs of the estimated HRs/ORs. The number of incident cases (n) and sample sizes (N) for each phenotype are as follows. Incident cases: all-cause mortality, N = 5,171 (n = 496); stroke, N = 4,794 (n = 443); type 2 diabetes, N = 4,072 (n = 339); COPD, N = 4,642 (n = 217); depression, N = 3,549 (n = 523); CVD (excluding stroke), N = 3,040 (n = 699); cancer, N = 4,172 (n = 385). Prevalent cases: The total sample size for prevalent diseases is 5,171, with the following case numbers; stroke, n = 377; type 2 diabetes, n = 1,099; COPD, n = 529; depression, n = 1,622; CVD (excluding stroke), n = 2,131; cancer, n = 999.
