Your result
0%
What's driving it
Next best move
What this means in practice
In plain language
Dr. Mithil Patel DO, DABOM, CPT
How much your current movement lowers your risk of dying from any cause, and which single change would lower it most from where you are now.
Your answers stay on this device. Nothing is sent to Elevé or saved, and it all clears when you close or refresh the page.
Dr. Mithil Patel DO, DABOM, CPT
Guideline-based tools compare you to a fixed target (150 to 300 minutes a week) and report pass or fail. This one places each input on its own published dose-response curve, combines the curves in log-hazard space the way a Cox model would, and then asks which single change moves the combined estimate the most. Four pieces of it have no counterpart in any published calculator we could find in October 2026: a wearable step channel with the age-specific plateau from Paluch (2022) reconciled against questionnaire minutes rather than double counted; the joint aerobic and resistance structure from Zhang, Lee and Giovannucci (BJSM, June 2026), including the 0.53 joint floor; a sitting penalty that decays continuously with activity volume instead of a yes/no "offset"; an optional measured-fitness channel (wearable or lab VO2 max) weighted against reported activity; and the ranked marginal-benefit output.
The output is a relative risk of all-cause mortality compared with an adult who does no structured exercise, averages about 3,500 steps a day and sits under four hours. It is the product of four terms. Each term is a piecewise-linear interpolation in log(RR) through published anchor points, flat beyond the last anchor. The interpolation is ours; every anchor is from a paper listed below.
RR = exp( max(ln RR_aerobic + ln RR_resist, ln 0.53) + ln HR_sitting )
Aerobic volume is converted to marginal MET-hours per week (mMET-h) with Garcia's values: moderate minutes at 3.5, vigorous minutes at 7.0. Short vigorous bursts in daily life (VILPA) are added to the vigorous minutes, following Stamatakis (2022), who found them to behave like vigorous exercise in people who don't otherwise train. Pattern across the week is deliberately ignored: dos Santos (2022) and Kany (2024) found no difference between weekend-warrior and spread-out patterns at the same volume.
| Term | Input | RR or HR | Source |
|---|---|---|---|
| Aerobic, questionnaire channel | 0 mMET-h/wk | 1.00 (ref) | Garcia 2023, Table 1, all-cause mortality |
| 4.375 (≈75 min moderate) | 0.77 | ||
| 8.75 (≈150 min moderate) | 0.69 | ||
| 17.5 (≈300 min moderate) | 0.66 | ||
| 35 and above | 0.63 | Lee 2022: 2 to 4 times the guideline adds 3 to 13% over guideline-level moderate activity; nothing clearly beyond that | |
| Steps, wearable channel | 3,553 steps/day (Q1 median) | 1.00 (ref) | Paluch 2022, 15 cohorts, quartile medians. Plateau 8,000 to 10,000 under 60, 6,000 to 8,000 at 60 and over |
| 5,801 | 0.60 | ||
| 7,842 (plateau at 60+) | 0.55 | ||
| 10,000 (plateau under 60) | 0.47 | ||
| Resistance, adjusted for aerobic | 30 min/wk | 0.91 | Momma 2022, pooled 10 to 17% at low volumes |
| 60 min/wk | 0.89 | Coleman 2022, one session a week | |
| 90 to 119 min/wk | 0.87 | Zhang 2026, 147,374 adults, 30 years; no added benefit above 120 | |
| 120 and above | 0.87 | ||
| Joint floor | Aerobic × resistance | 0.53 | Zhang 2026, lowest joint HR observed (≥45 MET-h aerobic, any resistance) |
| Fitness (optional) | Per 1 MET above 7.9 METs | 0.87 | Kodama 2009, 33 cohorts, 102,980 adults |
| Low (<7.9) vs high (≥10.9 METs) | 1.70 | ||
| Upper bound | 0.50 | Our cap; Mandsager 2018 found no upper limit, elite vs low HR 0.20 | |
| Sitting | >8 h/day vs <4, lowest activity | 1.25 | Ekelund 2016, 1.59 ÷ 1.27; at >35.5 MET-h/wk the excess falls to 1.04 |
| Same, most active quartile | 1.04 |
Self-reported exercise minutes are unreliable; the questionnaire-based cohorts behind these curves accept that and still find the dose-response because the error averages out over tens of thousands of people. For one patient it doesn't average out, and a typed "90 minutes" carries a precision nobody has. So each activity input is a band with a worked example of what that band looks like in a real week, and the band is scored at a representative value: the midpoint, or the published anchor when the band was drawn around one (the 60 to 120 minute resistance band scores at 90, where Zhang found the lowest risk; the step bands are centred on Paluch's quartile medians). The ranges are wide enough that someone who misjudges by a third usually lands in the same band. Patients who track with a device can pick the band their 30-day average falls in. Age and VO2 max stay numeric because they're read off a screen, not recalled.
A step count and a questionnaire both measure aerobic volume, with different error. Device measures produce stronger associations than self-report (Garcia notes this directly), partly because they catch light and incidental movement that nobody remembers to report. Adding the two curves together would count a brisk walk twice. So the aerobic term is a weighted average of the two log-risks, default weight 0.5 on each; the clinician panel exposes the weight. When no step count is entered, the questionnaire curve is used alone. Steps below 3,553 are treated as the reference rather than extrapolated upward, because the meta-analysis didn't model that range.
Cardiorespiratory fitness is the strongest single predictor in this literature, so when a VO2 max is entered it takes half the aerobic term by default, with the questionnaire and step channels sharing the other half. The curve is Kodama's: relative risk 0.87 for each MET of maximal aerobic capacity above the 7.9-MET low-fitness cutoff (one MET is 3.5 mL/kg/min), bounded at 1.70 on the unfit side (Kodama's low versus high) and 0.50 on the fit side. Mandsager found no upper limit of benefit out to the 97th percentile, but a watch estimate doesn't justify extrapolating past that bound. The reference of 7.9 METs is not age- or sex-normalised, which means a 70-year-old at 28 mL/kg/min is scored as average rather than as the well-above-average person they are; the age-and-sex-matched version is a planned refinement that needs normative tables. Wearable estimates run several mL/kg/min off a lab measurement in either direction, which is why the weight is exposed to the clinician.
Ekelund's 4×4 matrix gives a 25% excess for sitting more than 8 hours at the lowest activity quartile, falling to 4% (not significant) at the highest. The calculator turns that into a continuous penalty: it ramps from zero at 4 hours of sitting to full at 8, and it decays linearly with activity volume to 18% of its size at 25 mMET-h/wk, roughly the 60 to 75 minutes a day of moderate activity Ekelund identified. Volume for this term is the larger of reported mMET-h and an estimate from steps above the reference (100 steps per minute at 3.5 mMET).
The relative risk is the honest output; years of life is what people ask for. The practical section converts one to the other with a single constant calibrated to Moore (2012), who computed life expectancy after age 40 across 654,827 adults in six cohorts: 1.8 years gained for the lowest active category (up to about 75 minutes a week of brisk walking, hazard ratio about 0.80) and 4.5 years at the top (22.5 or more MET-hours a week, hazard ratio about 0.61). Both points fit years ≈ 9 × (−ln RR) to within a few months, so that is the formula, applied to the composite RR and to the change each next move would make. It is a population average for someone around 40 to 50; the gain is smaller for a 75-year-old (fewer years left to protect) and the formula says nothing about any one person. Shown as a range of ±25% to keep it from looking more precise than it is.
The function statements draw on the LIFE trial (Pahor 2014): 1,635 sedentary adults aged 70 to 89 with physical limitations, randomised to a walking, strength and balance programme or to health education. Over 2.6 years, 30% of the exercise group lost the ability to walk 400 metres compared with 36% of controls (HR 0.82), and persistent disability fell by 28%. That is the trial evidence that the same activity which lowers mortality also delays the loss of independence, in people who started late and started frail.
Everything here is observational. The curves are population associations adjusted for the usual confounders, and they are not a prediction for one person. The reference group is defined by the studies, so a result above 100% (possible for someone inactive who sits a lot) means higher risk than an inactive adult who sits little, not a prediction of death. The resistance curve is flat rather than J-shaped above 120 minutes because Zhang (2026) found no added benefit and no harm there; Momma's earlier meta-analysis saw possible attenuation above about 130 minutes, so we apply no credit and no penalty. The age split applies only to steps; Coleman found only small aerobic gains beyond three hours a week in older adults, which the flat tail already reflects. The fitness term is optional and, when present, uses a fixed reference rather than an age-matched one; a wearable VO2 max is an estimate, and a lab measurement is better.
Citations retrieved through PubMed, October 2026.