Sleep Duration
Evidence BWhat the evidence actually shows
Cappuccio and colleagues meta-analysed prospective cohorts covering more than a million people and found the now-familiar U shape: relative to about seven hours, short sleepers carried a modest excess mortality (hazard ratio around 1.12) and long sleepers a larger one (around 1.30). The bottom of the U sits somewhere in the seven-hour region, which is reassuringly close to what most people actually need and unhelpfully far from the "just sleep more" and "sleep is for the weak" camps alike.
The short-sleep arm has a plausible mechanistic story — sympathetic activation, blood pressure, glucose handling, appetite dysregulation — and it lines up with what controlled sleep-restriction experiments do to intermediate markers over days to weeks. So the short end is the part of this curve we take reasonably seriously as potentially causal.
How big the effect really is, in years
Converting these hazard ratios to life expectancy gives something on the order of one to two years for habitual short or long sleep versus the optimum. We deliberately keep this a grade B effect with a wide hedge, because self-reported sleep duration is a famously unreliable measurement and because the U-curve's two arms almost certainly have different amounts of real causation behind them.
The blunt version: a chronic five-hour-a-night pattern is worth worrying about at the scale of maybe a year or so; a one-off bad week is worth nothing at the actuarial level. This is not a lever that produces dramatic year-swings, and anyone selling sleep optimisation as a decade of added life has confused a real, modest effect with a marketing narrative.
The catch: the long-sleep arm is mostly sickness
The long-sleep half of the U is where reverse causation runs riot. People sleeping nine or ten hours are disproportionately those with depression, undiagnosed cancer, heart failure, chronic inflammation, or the general deconditioning that precedes death. Long sleep is a symptom of being unwell far more often than it is a cause of becoming unwell. So the hazard ratio of 1.30 at the long end substantially overstates any causal effect of the sleeping itself.
This matters for interpretation: if you naturally need eight and a half hours and feel fine, the data does not credibly say you are shortening your life. It says that the population of people sleeping that long contains a lot of sick people. Do not confuse the average of a mixed group for your own risk. The short arm is likelier to be genuinely causal; the long arm is likelier to be a marker.
How this feeds your actuarial age
The sleep modifier applies a gentle U-shaped adjustment centred near seven hours, and we deliberately make the long-sleep penalty milder than the raw hazard ratio implies, precisely because so much of that arm is reverse causation rather than a real cost of sleeping. We are pricing your risk, not the risk of the sick people who happen to share your reported sleep duration.
Because the whole effect is small and the measurement is noisy, this modifier moves your actuarial age only modestly, and it should. If it moved a lot on the strength of a single number you estimated about your own bedtime, we would be over-fitting to a variable that barely deserves the trust. Treat it as a nudge, not a verdict, and if you suspect a genuine sleep disorder like apnoea, that is a clinician conversation, not a slider.
This factor feeds directly into your actuarial age. Run the calculator →
Sources
Reviewed 2026-07-06 by Dmytro Dubina, Actuary · MSc Probability & Statistics · 20+ years in insurance. Population statistics, not medical advice.