I found this paper interesting and particularly the bit which identified key biomarkers that were significant (see after link)
Authors: Wei Qiu, MSc * Hugh Chen, PhD , Prof Matt Kaeberlein, PhD , Prof Su-In Lee, PhD
https://www.thelancet.com/journals/lanhl/article/PIIS2666-7568(23)00189-7/fulltext
Overall, for all-cause mortality ENABL Age, we identified nine features that appear in the top 20 most important risk factors for more than half of the randomly selected samples (n=30 000) from the test set. These features, and their respective percentages of individuals in which they appear among the top 20 most important risk factors, are as follows: long-standing illness, disability, or infirmity (29 376 [97·92%] of 30 000); cystatin C (23 487 [78·29%] of 30 000); overall health rating (21 372 [71·24%] of 30 000); average total household income before tax (21 024 [70·08%] of 30 000); red-blood-cell distribution width (20 370 [67·90%] of 30 000); pack years of smoking (18 072 [60·24%] of 30 000); past tobacco smoking (17 646 [58·82%] of 30 000); sex-hormone-binding globulin (15 534 [51·78%] of 30 000); and lymphocyte percentage (15 462 [51·54%] of 30 000). For neoplasm-caused mortality ENABL Age, we identified seven features that appeared in the top 20 most important risk factors for more than half of the same random sample: past tobacco smoking (27 279 [90·93%] of 30 000), pack years of smoking (24 384 [81·28%] of 30 000), cystatin C (22 929 [76·43%] of 30 000), red-blood-cell distribution width (21 357 [71·19%] of 30 000), C-reactive protein (21 261 [70·87%] of 30 000), age at last cancer diagnosis (20 793 [69·31%] of 30 000), and platelet distribution width (19 239 [64·13%] of 30 000).