Study analysis · SSM - Population Health · 2026
Your step counter says 10,000 steps—but this 21,000-person study found healthcare spending flattens around 11,000, and the '20,000-step sweet spot' is probably a math artifact.
In long-term app users aged 40+, predicted healthcare spending dropped as daily steps rose, then leveled off around 11,000 steps/day, but the study can’t prove walking caused the lower costs.
Overview
What the study found
The study in plain English — the bottom line, every takeaway we extracted, and what to do with them.
In simple terms
This study watched people over time to see whether those who walked more spent less on healthcare. It can show a link or pattern, but it cannot prove that walking more causes lower spending, because healthier people might walk more for other reasons. It is a clue, not proof.
What’s the bottom line?
Researchers studied how many steps people took each day and how much they spent on healthcare in the next 6 months. They found that people who walked more tended to spend less, but the benefit leveled off around 10,000–12,000 steps per day. Walking more than that didn't clearly help further. This was an observational study, so it can't prove that walking more causes lower costs.
How strong is this study?
The study used a lot of people and phone step counts, and it tried to adjust for some differences. But it only included people who chose to use a health app in one area of Japan, who may be healthier and more tech-savvy than average, and some important information was missing. That means the results are useful for ideas, but they might not apply to everyone.
40 / 100
- COI disclosure+40/40
- Data availabilitydata not shared
- Code availabilitycode not shared
38 / 100
- Randomizationnot randomized
- Blindingnot blinded
- Control groupno control group
- Sample size (n=20950)+20/20
- Follow-up+10/10
100 / 100
77 / 100
- P-values+15/15
- Effect size+20/20
- Confidence intervals+15/15
- Pre-registrationnot pre-registered
Each component is scored out of 100 and then capped by the study design — a case series cannot reach the ceiling a randomised trial can, however well it is reported.
Where it sits
RCT reviewsReviews of RCTs (Meta-analyses)
Max 100Randomized TrialsRandomized Trials
Max 90Reviews of Cohort StudiesReviews of Cohort Studies
Max 85Cohort StudiesCohort Studies
Max 72Reviews of Case-Control StudiesReviews of Case-Control Studies
Max 63Case-Control StudiesCase-Control Studies
Max 58Cross-Sectional & Case SeriesCross-Sectional & Case Series
Max 50Expert OpinionExpert Opinion
Max 560 / 100
Probability of being correct
Groups of people are followed over time to see who develops an outcome. Strong for identifying risk factors and associations, but cannot prove causation as firmly as RCTs.
This design cannot establish causation — the findings describe an association, not a cause. Retrospective observational cohort without randomization or a control group. It cannot rule out confounding, reverse causation, selection bias, or unmeasured factors. The authors explicitly state that the associations cannot establish that increasing walking causally lowers healthcare expenditures.
No Conflicts
No conflicts of interest identified
No conflicts of interest or funding statement was present in the provided text; the study uses a government-run mHealth app and administrative data, with no apparent industry ties.
The provided excerpt lacks a COI or funding statement. The study relies on Osaka Prefecture's official A-Smile app and administrative claims data, which may imply a non-financial government affiliation, but no author affiliations or funding roles are specified in the text. Absence of disclosure means conflicts cannot be fully ruled out.
Key takeaways
- 01
In about 21,000 long-term app users aged 40+ in Japan, predicted 6-month healthcare spending was lowest at around 11,000 steps/day (predicted geometric mean 17,170 yen).
- 02
For older adults (65+), the lowest predicted spending was 48,407 yen at ~10,600 steps/day.
- 03
For middle-aged adults, it was 1,514 yen at ~11,300 steps/day.
- 04
Spending was higher for people who walked less.
- 05
A minimum at ~20,000 steps/day was not reliable.
- 06
The study found that predicted healthcare spending was lower for people who walked more, up to about 11,000 steps/day, after which it flattened.
- 07
For older adults, the predicted 6-month spending at the low point was 48,407 yen; for middle-aged adults, it was 1,514 yen.
- 08
However, these are predicted averages from a selected group of app users, not a random sample, and the study cannot prove that walking more directly reduces costs.
- 09
The absolute difference in spending between people with different step counts was not reported as a simple comparison; the curve shows a steep decline from low steps.
Surprising findings
- Older adults had slightly higher mean daily steps than middle-aged adults: 5,363 vs 5,092 steps/day, a 271-step difference.Physical activity usually declines with age. The authors suggest this is likely due to stronger health and digital selection among older app users, not a true age effect.
- Predicted expenditures turned downward below about 2,500 steps/day, suggesting lower costs at very low activity.This contradicts the idea that being completely sedentary is associated with the highest costs. The authors say it is likely a measurement artifact from incomplete device carriage or app malfunction.
- The non-linear step term significantly improved model fit but added almost no predictive value.A statistically significant curve can still be practically unimportant for prediction. Here, baseline medical costs and age did almost all the work.
- The apparent 20,000-step global minimum was not statistically distinguishable from the 11,000-step local nadir.Many headlines might claim '20,000 steps is optimal,' but the confidence interval for the difference crossed zero.
Practical takeaways
If you’re already active, aim for the 10,000–12,000 steps/day range rather than chasing extreme step counts.
This is an observational association in a selected cohort of long-term app users, not proof that hitting 11,000 steps will lower your personal healthcare costs.
medium confidenceOlder adults may get more financial-health relevance from maintaining ~10,000–11,000 steps/day than middle-aged adults.
The absolute predicted expenditures were much higher in older adults, but the study cannot show that increasing steps causes the lower costs.
medium confidenceDon’t reorganize your life around a 20,000-step target based on this study.
The 20,000-step minimum was unstable, data-sparse, and shifted with modeling choices.
high confidenceUse step counts as a general activity nudge, not as a personal medical-cost prediction tool.
Daily steps added only ~0.03% to explained deviance beyond baseline health status.
high confidenceWhy this study matters
The 11,000-step plateau
Predicted 6-month healthcare expenditures decreased steeply from low activity and plateaued at about 10,000–12,000 steps/day. The local nadir was ~11,000 steps/day, with a predicted geometric mean expenditure of 17,170 JPY (95% CI 16,174–18,229). The study explicitly says this is observational and not causal.
It challenges both the classic 10,000-step mantra and the idea that more steps always mean lower costs. The curve flattens, so extra steps beyond ~11k may not buy much in this cohort.
Older adults see much bigger absolute spending differences
For older adults (≥65), the local nadir was 10,612 steps/day with predicted expenditure 48,407 JPY (95% CI 45,568–51,423). For middle-aged adults (40–64), it was 11,317 steps/day with 1,514 JPY (95% CI 1,360–1,685). The step benchmark is similar, but the absolute money at stake is about 32 times higher in older adults.
It suggests the same step target may matter far more financially after 65, even though the optimal step range looks similar.
The 20,000-step 'minimum' is an extrapolation artifact
A global minimum appeared at ~20,000 steps/day, but it was not statistically distinguishable from the ~11,000-step local nadir (difference 1,048 JPY; 95% CI −4,892 to 7,006). Its location shifted between 16,000 and 22,000 steps/day depending on spline flexibility.
Viral claims about an 'optimal' 20,000 steps are likely overinterpreting sparse, unstable data. The robust finding is a plateau, not a high-step optimum.
Steps add almost nothing to predicting individual costs
Adding a non-linear step term significantly improved model fit (ΔAIC = 167.4), but it explained only about 0.0003 extra pseudo-R². The full model explained 32.6% of deviance, mostly from baseline medical expenditure and age.
Even though the curve is real at the population level, your daily step count is a poor tool for forecasting your personal healthcare costs.
Only 15.1% of registered users made the final sample
The analytic cohort was 20,950 users out of 138,344 registered A-Smile users. Inclusion required sustained app use, successful claims linkage, and enrollment in specific insurance schemes. The sample skewed healthier, more digitally literate, and higher socioeconomic status.
The findings may not generalize to all middle-aged and older adults. The people studied are a selected, health-conscious group.
Want the whole report?
Detailed mode opens the full scientific breakdown — every score component, the methodology, conflicts of interest, the evidence analysis behind each claim, and the raw study data.
Overview
What the study found
The study in plain English — the bottom line, every takeaway we extracted, and what to do with them.
Not medical advice. For informational purposes only. Always consult a healthcare professional. Terms
Researchers studied how many steps people took each day and how much they spent on healthcare in the next 6 months. They found that people who walked more tended to spend less, but the benefit leveled off around 10,000–12,000 steps per day. Walking more than that didn't clearly help further. This was an observational study, so it can't prove that walking more causes lower costs.
Research results
In about 21,000 long-term app users aged 40+ in Japan, predicted 6-month healthcare spending was lowest at around 11,000 steps/day (predicted geometric mean 17,170 yen). For older adults (65+), the lowest predicted spending was 48,407 yen at ~10,600 steps/day. For middle-aged adults, it was 1,514 yen at ~11,300 steps/day. Spending was higher for people who walked less. A minimum at ~20,000 steps/day was not reliable.
What this means - more context
The study found that predicted healthcare spending was lower for people who walked more, up to about 11,000 steps/day, after which it flattened. For older adults, the predicted 6-month spending at the low point was 48,407 yen; for middle-aged adults, it was 1,514 yen. However, these are predicted averages from a selected group of app users, not a random sample, and the study cannot prove that walking more directly reduces costs. The absolute difference in spending between people with different step counts was not reported as a simple comparison; the curve shows a steep decline from low steps.
To elucidate the non-linear association between mHealth-measured daily steps and subsequent healthcare expenditures among middle-aged and older adults, using a large linked dataset of long-term app users in Osaka, Japan.
This retrospective cohort study analyzed 20,950 long-term mHealth app users aged ≥40 in Osaka, Japan, linking objective step counts to administrative claims data. Using generalized additive models, it found that predicted 6-month healthcare expenditures decreased steeply from low step counts and plateaued at approximately 10,000–12,000 steps/day, with a local nadir at ~11,000 steps/day (predicted geometric mean 17,170 JPY, 95% CI 16,174–18,229). The association was non-linear and non-monotonic, but observational and not causal. A high-step global minimum at ~20,000 steps/day was unstable and likely an extrapolation artifact. The non-linear step term significantly improved model fit (ΔAIC=167.4) but added little explained deviance (Δ pseudo-R²≈0.0003).
Methods Used
Retrospective longitudinal cohort using linked mHealth step counts and administrative claims data from 20,950 A-Smile app users aged ≥40 in Osaka Prefecture, Japan (2020–2023). Exposure was monthly mean daily steps; outcome was total healthcare expenditures in the subsequent 6 months. Generalized additive models with penalized splines adjusted for age, gender, BMI, baseline medical expenditures, health checkup attendance, and COVID-19 emergency periods. Person-month analysis with cluster bootstrap for uncertainty intervals.
Main Finding
Predicted 6-month healthcare expenditures showed a non-linear, non-monotonic association with daily steps, decreasing steeply from low activity and plateauing at approximately 10,000–12,000 steps/day. The local nadir was at ~11,000 steps/day (predicted geometric mean 17,170 JPY, 95% CI 16,174–18,229). For older adults (≥65), the nadir was at 10,612 steps/day with predicted expenditure of 48,407 JPY (95% CI 45,568–51,423); for middle-aged adults (40–64), it was 11,317 steps/day with 1,514 JPY (95% CI 1,360–1,685). The global minimum at ~20,000 steps/day was not statistically distinguishable from the local nadir (difference 1,048 JPY, 95% CI −4,892 to 7,006) and is considered an extrapolation artifact. The association is observational and cannot establish causation.
Confidence Level
Moderate. The study benefits from a large sample, objective step measurement, and advanced modeling, but is limited by strong selection bias (only 15.1% of registered users included), residual confounding, and minimal incremental predictive value of daily steps beyond baseline health status. Generalizability is restricted to a selected cohort of long-term app users.
Study Flags
Red Flags
- •Strong selection bias: only 15.1% of registered users included; sample skews toward healthier, digitally literate, higher socioeconomic status individuals
- •Residual confounding from unmeasured factors such as socioeconomic status, lifestyle habits, and subclinical conditions
- •Limited practical predictive value: daily steps added only ~0.03% to explained deviance beyond baseline health status
No biological mechanisms were identified in this study. This may be an epidemiological, observational, or survey-based study that reports associations rather than proposing causal biological pathways.
Surprising Findings
Older adults had slightly higher mean daily steps than middle-aged adults: 5,363 vs 5,092 steps/day, a 271-step difference.
Physical activity usually declines with age. The authors suggest this is likely due to stronger health and digital selection among older app users, not a true age effect.
Practical Takeaways
If you’re already active, aim for the 10,000–12,000 steps/day range rather than chasing extreme step counts.
RCT reviewsReviews of RCTs (Meta-analyses)
Max 100Randomized TrialsRandomized Trials
Max 90Reviews of Cohort StudiesReviews of Cohort Studies
Max 85Cohort StudiesCohort Studies
Max 72Reviews of Case-Control StudiesReviews of Case-Control Studies
Max 63Case-Control StudiesCase-Control Studies
Max 58Cross-Sectional & Case SeriesCross-Sectional & Case Series
Max 50Expert OpinionExpert Opinion
Max 560 / 100
Probability of being correct
Groups of people are followed over time to see who develops an outcome. Strong for identifying risk factors and associations, but cannot prove causation as firmly as RCTs.
Human Cohort Study
Subject
Moderate probability
on the GRADE evidence scale
This study watched people over time to see whether those who walked more spent less on healthcare. It can show a link or pattern, but it cannot prove that walking more causes lower spending, because healthier people might walk more for other reasons. It is a clue, not proof.
No conflicts of interest were detected in this study. No score impact.
Strengths
- Large linked dataset of 20,950 individuals and 338,706 person-months.
- Objective mHealth-measured step counts rather than self-reported physical activity.
- Administrative claims data for healthcare expenditures, reducing recall bias for the outcome.
Weaknesses
- Observational retrospective cohort design without randomization or control group.
- High risk of selection bias and healthy-user effect from voluntary mHealth app participation.
- Potential unmeasured confounding by socioeconomic status, lifestyle, diet, smoking, alcohol, and genetics.
Methodology
Evidence Keywords
Statistical Reporting
Not medical advice. For informational purposes only. Always consult a healthcare professional. Terms
Researchers studied how many steps people took each day and how much they spent on healthcare in the next 6 months. They found that people who walked more tended to spend less, but the benefit leveled off around 10,000–12,000 steps per day. Walking more than that didn't clearly help further. This was an observational study, so it can't prove that walking more causes lower costs.
Research results
In about 21,000 long-term app users aged 40+ in Japan, predicted 6-month healthcare spending was lowest at around 11,000 steps/day (predicted geometric mean 17,170 yen). For older adults (65+), the lowest predicted spending was 48,407 yen at ~10,600 steps/day. For middle-aged adults, it was 1,514 yen at ~11,300 steps/day. Spending was higher for people who walked less. A minimum at ~20,000 steps/day was not reliable.
What this means - more context
The study found that predicted healthcare spending was lower for people who walked more, up to about 11,000 steps/day, after which it flattened. For older adults, the predicted 6-month spending at the low point was 48,407 yen; for middle-aged adults, it was 1,514 yen. However, these are predicted averages from a selected group of app users, not a random sample, and the study cannot prove that walking more directly reduces costs. The absolute difference in spending between people with different step counts was not reported as a simple comparison; the curve shows a steep decline from low steps.
To elucidate the non-linear association between mHealth-measured daily steps and subsequent healthcare expenditures among middle-aged and older adults, using a large linked dataset of long-term app users in Osaka, Japan.
This retrospective cohort study analyzed 20,950 long-term mHealth app users aged ≥40 in Osaka, Japan, linking objective step counts to administrative claims data. Using generalized additive models, it found that predicted 6-month healthcare expenditures decreased steeply from low step counts and plateaued at approximately 10,000–12,000 steps/day, with a local nadir at ~11,000 steps/day (predicted geometric mean 17,170 JPY, 95% CI 16,174–18,229). The association was non-linear and non-monotonic, but observational and not causal. A high-step global minimum at ~20,000 steps/day was unstable and likely an extrapolation artifact. The non-linear step term significantly improved model fit (ΔAIC=167.4) but added little explained deviance (Δ pseudo-R²≈0.0003).
Methods Used
Retrospective longitudinal cohort using linked mHealth step counts and administrative claims data from 20,950 A-Smile app users aged ≥40 in Osaka Prefecture, Japan (2020–2023). Exposure was monthly mean daily steps; outcome was total healthcare expenditures in the subsequent 6 months. Generalized additive models with penalized splines adjusted for age, gender, BMI, baseline medical expenditures, health checkup attendance, and COVID-19 emergency periods. Person-month analysis with cluster bootstrap for uncertainty intervals.
Main Finding
Predicted 6-month healthcare expenditures showed a non-linear, non-monotonic association with daily steps, decreasing steeply from low activity and plateauing at approximately 10,000–12,000 steps/day. The local nadir was at ~11,000 steps/day (predicted geometric mean 17,170 JPY, 95% CI 16,174–18,229). For older adults (≥65), the nadir was at 10,612 steps/day with predicted expenditure of 48,407 JPY (95% CI 45,568–51,423); for middle-aged adults (40–64), it was 11,317 steps/day with 1,514 JPY (95% CI 1,360–1,685). The global minimum at ~20,000 steps/day was not statistically distinguishable from the local nadir (difference 1,048 JPY, 95% CI −4,892 to 7,006) and is considered an extrapolation artifact. The association is observational and cannot establish causation.
Confidence Level
Moderate. The study benefits from a large sample, objective step measurement, and advanced modeling, but is limited by strong selection bias (only 15.1% of registered users included), residual confounding, and minimal incremental predictive value of daily steps beyond baseline health status. Generalizability is restricted to a selected cohort of long-term app users.
Study Flags
Red Flags
- •Strong selection bias: only 15.1% of registered users included; sample skews toward healthier, digitally literate, higher socioeconomic status individuals
- •Residual confounding from unmeasured factors such as socioeconomic status, lifestyle habits, and subclinical conditions
- •Limited practical predictive value: daily steps added only ~0.03% to explained deviance beyond baseline health status
No biological mechanisms were identified in this study. This may be an epidemiological, observational, or survey-based study that reports associations rather than proposing causal biological pathways.
Surprising Findings
Older adults had slightly higher mean daily steps than middle-aged adults: 5,363 vs 5,092 steps/day, a 271-step difference.
Physical activity usually declines with age. The authors suggest this is likely due to stronger health and digital selection among older app users, not a true age effect.
Practical Takeaways
If you’re already active, aim for the 10,000–12,000 steps/day range rather than chasing extreme step counts.
RCT reviewsReviews of RCTs (Meta-analyses)
Max 100Randomized TrialsRandomized Trials
Max 90Reviews of Cohort StudiesReviews of Cohort Studies
Max 85Cohort StudiesCohort Studies
Max 72Reviews of Case-Control StudiesReviews of Case-Control Studies
Max 63Case-Control StudiesCase-Control Studies
Max 58Cross-Sectional & Case SeriesCross-Sectional & Case Series
Max 50Expert OpinionExpert Opinion
Max 560 / 100
Probability of being correct
Groups of people are followed over time to see who develops an outcome. Strong for identifying risk factors and associations, but cannot prove causation as firmly as RCTs.
Human Cohort Study
Subject
Moderate probability
on the GRADE evidence scale
This study watched people over time to see whether those who walked more spent less on healthcare. It can show a link or pattern, but it cannot prove that walking more causes lower spending, because healthier people might walk more for other reasons. It is a clue, not proof.
No conflicts of interest were detected in this study. No score impact.
Strengths
- Large linked dataset of 20,950 individuals and 338,706 person-months.
- Objective mHealth-measured step counts rather than self-reported physical activity.
- Administrative claims data for healthcare expenditures, reducing recall bias for the outcome.
Weaknesses
- Observational retrospective cohort design without randomization or control group.
- High risk of selection bias and healthy-user effect from voluntary mHealth app participation.
- Potential unmeasured confounding by socioeconomic status, lifestyle, diet, smoking, alcohol, and genetics.
Methodology
Evidence Keywords
Statistical Reporting
Scoring
How strong is this study?
The study used a lot of people and phone step counts, and it tried to adjust for some differences. But it only included people who chose to use a health app in one area of Japan, who may be healthier and more tech-savvy than average, and some important information was missing. That means the results are useful for ideas, but they might not apply to everyone.
40 / 100
- COI disclosure+40/40
- Data availabilitydata not shared
- Code availabilitycode not shared
38 / 100
- Randomizationnot randomized
- Blindingnot blinded
- Control groupno control group
- Sample size (n=20950)+20/20
- Follow-up+10/10
100 / 100
77 / 100
- P-values+15/15
- Effect size+20/20
- Confidence intervals+15/15
- Pre-registrationnot pre-registered
Each component is scored out of 100 and then capped by the study design — a case series cannot reach the ceiling a randomised trial can, however well it is reported.
Where it sits
RCT reviewsReviews of RCTs (Meta-analyses)
Max 100Randomized TrialsRandomized Trials
Max 90Reviews of Cohort StudiesReviews of Cohort Studies
Max 85Cohort StudiesCohort Studies
Max 72Reviews of Case-Control StudiesReviews of Case-Control Studies
Max 63Case-Control StudiesCase-Control Studies
Max 58Cross-Sectional & Case SeriesCross-Sectional & Case Series
Max 50Expert OpinionExpert Opinion
Max 560 / 100
Probability of being correct
Groups of people are followed over time to see who develops an outcome. Strong for identifying risk factors and associations, but cannot prove causation as firmly as RCTs.
This design cannot establish causation — the findings describe an association, not a cause. Retrospective observational cohort without randomization or a control group. It cannot rule out confounding, reverse causation, selection bias, or unmeasured factors. The authors explicitly state that the associations cannot establish that increasing walking causally lowers healthcare expenditures.
No Conflicts
No conflicts of interest identified
No conflicts of interest or funding statement was present in the provided text; the study uses a government-run mHealth app and administrative data, with no apparent industry ties.
The provided excerpt lacks a COI or funding statement. The study relies on Osaka Prefecture's official A-Smile app and administrative claims data, which may imply a non-financial government affiliation, but no author affiliations or funding roles are specified in the text. Absence of disclosure means conflicts cannot be fully ruled out.
Standing
Who’s using this study?
The videos and claims on this site that lean on this study, and the researchers who wrote it.
1 video from Menno Henselmans cite this study, drawing 1 claim from it.
- Conflicting evidence
Evidence points in both directions — no clear conclusion yet.
Evidence
Authored by
1 researcherIf this is your work, this is how we attribute it on Fit Body Science. Haruka Kato is listed as the lead author.