How it works

OpenLifeModel estimates life expectancy the way actuaries and epidemiologists do, and shows every step. Nothing is hidden in a server or a proprietary formula: the models are plain files and the engine is open source.

1. Start from a life table

A national life table gives the chance that a person of a given age and sex dies within the next year. The default baseline is the US Social Security Administration's 2023 period life table. On its own, it reproduces SSA's published life expectancies to within 0.02 years.

2. Adjust for personal factors

Studies report how much a factor such as smoking changes the risk of dying at any given age, as a hazard ratio: 2.0 means twice the yearly risk. A model multiplies the baseline risk by the hazard ratio for each of your answers. Because the life table already includes smokers and non-smokers, active and inactive people, the ratios are rescaled so that someone with population-average habits lands exactly on the national figure.

The default model, US adults: smoking, weight and exercise, takes its hazard ratios from three large studies: smoking and quitting from the US National Health Interview Survey (Jha et al., NEJM 2013), body mass index from a meta-analysis of 239 cohorts (Global BMI Mortality Collaboration, Lancet 2016) and exercise from a pooled analysis of 661,000 adults (Arem et al., JAMA Internal Medicine 2015). The population averages come from the US NHANES survey.

3. Build the survival curve

Applying the adjusted yearly risks age by age gives your survival curve: the chance of still being alive at each future age. Life expectancy is the area under that curve, and the median age at death is where it crosses 50%. The exact formulas are in the OLM specification.

What the results cannot tell you

• They are averages for people who share your answers, not a prediction for you. Two people with identical profiles can live very different lives.
• Period life tables assume today's death rates continue. If medicine keeps improving, real lifespans will be longer.
• Factors are treated as independent. Many are linked (weight, blood pressure and activity, for instance), so combining them can overstate their joint effect.
• The models do not know your medical history, family history or anything you did not enter.

This is not medical advice. Talk to a clinician about your own health.

Your privacy

The calculator runs entirely in your browser. Your answers are not sent to any server; they are remembered only in this browser so you do not have to retype them. There are no accounts, no analytics cookies and no tracking.

Open models, open format

Each model is an OLM file (.olm.yaml): a human-readable YAML document listing its baseline, factors, sources, assumptions and reference test cases. You can browse the models, edit one in the calculator, export it, and share it. Researchers can publish a model alongside a paper so that anyone can run and verify it.

The engine and calculator are open source under Apache-2.0, and the specification under CC-BY-4.0. Contributions and critique of the methodology are welcome on GitHub.