Can a Fitness Questionnaire Predict Mortality Like a CPET?

A questionnaire can sort people into mortality-risk groups surprisingly well. That makes estimated cardiorespiratory fitness (eCRF) useful when exercise testing is unavailable. It does not make a questionnaire interchangeable with a direct measurement of fitness.

The distinction matters because eCRF formulas commonly combine age, body mass index (BMI), resting heart rate, and self-reported activity. Those variables predict mortality on their own. An eCRF score therefore answers a mixed question: how fit does this person likely seem, given their demographics, habits, and basic physiology? A maximal exercise test asks a narrower question: how much aerobic work can this person actually perform?

For screening a population or getting a rough baseline, the shortcut has value. For deciding whether your training raised VO2max, assessing an individual clinical threshold, or interpreting a score in an older adult or someone taking a beta-blocker, the shortcut can become a blurry map.

Fitness predicts mortality regardless of how researchers assess it

Cardiorespiratory fitness (CRF), often expressed in metabolic equivalents (METs), has a consistent observational relationship with survival. In an updated 2022 meta-analysis by Laukkanen et al., 37 cohorts included 2,258,029 participants and 108,613 all-cause deaths. CRF came from direct cardiopulmonary exercise testing or maximal exercise testing. Each 1-MET higher CRF was associated with an 11% lower risk of all-cause mortality (risk ratio 0.89; 95% confidence interval 0.86 to 0.92). Participants in the highest versus lowest CRF tertile had a risk ratio of 0.55 (95% CI 0.50 to 0.61).

Those are associations, not proof that raising fitness by one MET will reduce any one person's mortality risk by exactly 11%. Healthier people may differ in ways that exercise testing cannot fully capture. Still, the consistency and gradient make CRF one of the more informative markers available in preventive medicine.

A 2025 systematic review and meta-analysis by Singh et al. compared direct and estimated CRF methods across 42 studies, 35 cohorts, and 3.8 million observations. It included direct measurement, exercise-based estimates, and non-exercise estimates. Every tier showed statistically significant, graded associations with all-cause and cardiovascular mortality. Pairwise comparisons found measurable differences between tiers, yet non-exercise estimates retained meaningful population-level prognostic value.

That is the practical answer to the waiting-room question: an estimate is not worthless because it is an estimate. It may identify people who would benefit from a closer look at physical activity, symptoms, blood pressure, metabolic health, and formal exercise testing.

Why eCRF can look better than measured fitness on paper

A larger mortality association does not necessarily mean a better fitness measurement. In an 2021 meta-analysis by Qiu et al., eight cohorts with more than 170,000 participants linked each 1-MET higher non-exercise eCRF to lower cardiovascular mortality (hazard ratio 0.83; 95% CI 0.80 to 0.86) and lower all-cause mortality (hazard ratio 0.83; 95% CI 0.78 to 0.88). That corresponds to a 17% lower relative hazard per MET, numerically larger than the 11% association in the direct-testing meta-analysis.

The sensible explanation is confounding by construction. Age, BMI, resting heart rate, and physical activity contribute both to many eCRF equations and to mortality risk. The score can therefore predict mortality well partly because its ingredients already contain mortality information. It is a useful risk score, but it is not a pure physiological readout of aerobic capacity.

Qiu et al. found that eCRF discriminated mortality better than any single component such as activity, resting heart rate, or BMI. The authors described its discrimination as slightly lower than exercise-tested CRF, but did not report specific AUC or C-statistic values in the available summary. That missing number prevents a clean statement of how much individual-level discrimination is lost.

One cohort shows the gap between group prediction and personal precision

A 2022 study by Myers et al. tested this issue within the same clinically referred cohort. The retrospective single-center study followed 1,545 people undergoing exercise testing; their mean age was 60 ± 13 years. Researchers compared treadmill-measured CRF with an estimate based on the Veterans Specific Activity Questionnaire (VSAQ), symptoms, and clinical variables.

At the group level, the results looked remarkably similar. Mean estimated and measured CRF were 8.7 ± 4.7 and 8.6 ± 5.0 METs, respectively (p=0.27). The highest versus lowest fitness groups had 87% lower mortality risk using estimated CRF and 88% lower risk using measured CRF. Measured CRF was associated with about a 19% lower mortality hazard per MET (hazard ratio about 0.81), while VSAQ-estimated CRF was associated with about a 26% lower hazard per MET (hazard ratio about 0.74).

Yet the individual-level match was imperfect: VSAQ and treadmill CRF had a multiple correlation of R=0.67 (p<0.001). That explains about 45% of the variation in measured CRF, leaving substantial room for an individual estimate to miss the mark. The paper did not report a standard error of estimate, so readers should not treat a questionnaire score as a laboratory value with laboratory-level precision.

This is why a score can stratify groups well while still misclassifying particular people near a decision boundary. A risk tool can be useful without being a personal oracle. Most health technology eventually learns this lesson; spreadsheets tend to learn it last.

The low-fitness end may matter most

Directly measured CRF has a curvilinear relationship with mortality: the largest risk differences appear at the lower end of the fitness range, then tend to flatten as CRF rises. Singh et al. and the broader direct-measurement evidence describe this pattern, although the available reports do not specify a numerical MET threshold where the curve changes shape.

Qiu et al. reported a linear dose-response between non-exercise eCRF and mortality. The two patterns may differ because eCRF compresses several predictors into one score and smooths over the lowest-fitness range. That creates a practical concern: an estimate that shifts a truly deconditioned person into a middling category could obscure the part of the curve where identifying low fitness matters most. The evidence does not establish how often that happens or identify a precise cutoff, so this remains a reason for caution rather than a quantified clinical rule.

Who should be especially cautious with non-exercise estimates?

Age and clinical context limit how confidently an eCRF equation travels. A 2024 study by Schumacher et al. developed machine-learning, non-exercise VO2max prediction models for older adults because virtually no existing non-exercise equation had been designed or validated for that population. Adults over 60 or 65 carry the highest absolute mortality risk, yet they sit in a major validation gap for common equations.

Medication can also distort an estimate. Many non-exercise formulas use resting heart rate; beta-blockers lower it independently of fitness. Available studies did not quantify the resulting error, but the direction of bias is a plain pharmacologic concern: a lower drug-induced heart rate can make an eCRF score look better than true exercise capacity.

The large Singh synthesis also excluded people with prevalent chronic conditions at baseline. Its broad comparison therefore does not automatically settle performance in people with hypertension, type 2 diabetes, or cardiovascular disease. A 2023 study by Peterman et al. examined multiple CRF estimation methods in patients with cardiovascular disease, but the available published information does not provide enough full results to quantify its four-method comparison here.

Choose the tool for the decision

• Use eCRF for baseline awareness and broad screening. It predicts mortality better than any one input variable and can flag a need for more attention to fitness and cardiometabolic risk.
• Use direct exercise testing when the individual number will drive a high-stakes decision. Examples include pre-surgical assessment, cardiac rehabilitation thresholds, unexplained exercise intolerance, or a clinical decision that turns on a narrow fitness category.
• Use direct measurement to track training when precision matters. A 2023 JACC study by Kokkinos et al. found that changes in directly measured CRF over time predicted survival independently of baseline fitness in people with and without cardiovascular disease. A questionnaire-based estimate adds shifting self-reported activity and resting-heart-rate inputs, making small personal changes difficult to interpret.
• Apply extra skepticism after age 60 and when taking heart-rate-altering drugs. Ask whether the equation was actually validated in people like you before treating the result as a measurement.

Use an estimated CRF score as a useful screen for risk, not as proof that a change in the score came from a change in aerobic capacity.

The broad evidence supports knowing your CRF by whatever credible method you can access. Direct testing gives the clearest physiological answer; eCRF gives a cheaper, more accessible risk estimate with demographic and behavioral information baked in. The right choice depends on whether you need a population-level signal or a precise answer about your own fitness.