D A Y S L E F T
Methodology
Statistical estimates, not point predictions.
In short
DaysLeft shows a biological-age estimate and a range. It does not predict an exact lifespan, diagnose disease, or replace a clinician. People with the same inputs can still have very different outcomes.
Two estimate layers
- Survey estimate: public life tables plus published epidemiology factors for behaviors such as smoking, activity, sleep, alcohol, and self-rated health.
- Precise estimate: the published clinical PhenoAge model from Liu / Levine 2018, using nine routine blood markers plus age.
PhenoAge inputs and weights
Albumin
Negative
Higher albumin lowers the PhenoAge mortality score.
Creatinine
Positive
Higher creatinine raises the score.
Fasting glucose
Positive
Higher glucose raises the score.
CRP
Positive
Higher log CRP raises the score.
Lymphocyte %
Negative
Higher lymphocyte percentage lowers the score.
MCV
Positive
Higher mean cell volume raises the score.
RDW
Positive
Higher red-cell distribution width raises the score.
Alkaline phosphatase
Positive
Higher ALP raises the score.
White blood cells
Positive
Higher WBC raises the score.
Age
Positive
Chronological age remains the strongest baseline signal.
Weight direction follows the PhenoAge coefficient signs implemented in the local algorithm. The UI summarizes contributions as an estimate for reflection, not as a medical interpretation of lab results.
Sources
- Liu Z. et al., PLOS Medicine 2018: clinical PhenoAge using NHANES data.
- Levine M. et al., Aging 2018: phenotypic age and mortality-risk framing.
- NHANES IV validation reported around AUC 0.88 for mortality discrimination.
- Internal research source: docs/research/09_validated_model_methodology.md.