The arithmetic is simple; the interpretation is not
If a model expressed in age-equivalent units estimates 48 years for a 52-year-old person, that is −4 Δ-years, interpreted as a 4-year age-equivalent reward. If it estimates 56 years, that is +4 Δ-years, interpreted as a 4-year age-equivalent penalty.
The important point is that the “years” belong to the model’s reference scale. They do not automatically represent years of life gained, lost, accelerated, or reversed.
The meaning depends on the model
In KDM, the age-like output emerges from relationships between biomarkers and chronological age. In Levine PhenoAge (clinical), the age scale is tied to estimated mortality risk. In a longitudinal digital model, the scale may be used to make a multi-signal deviation easier to understand.
Those are related but distinct interpretations.
Why short-term Δ-years can move quickly
If the underlying model includes daily physiological or behavioral signals, its Δ-year output may change when those inputs change. That is expected model behavior, not proof of rapid biological rejuvenation or deterioration.
A responsible interface should therefore show contributors, trends, and data quality rather than presenting the Δ-year value as a self-explanatory truth.
BioMIR’s use of Δ-years
BioMIR uses Δ-years as a common age-equivalent scale for Adaptive BioAge and its daily domain contributions. Negative values are interpreted as age-equivalent reward and positive values as age-equivalent penalty. The result is intended for longitudinal interpretation alongside the underlying data, Data Confidence, and CMA Freshness.