Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Machine-learning fitness programs added steps in review of 10 trials with 2,269 adults

Study title: The impact of machine learning on physical activity–related health outcomes: A systematic review and meta‐analysis

Evidence type
Systematic review or meta-analysis
Sample
n = 2,269
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2742cite or check it

Study title: The impact of machine learning on physical activity–related health outcomes: A systematic review and meta‐analysis

A systematic review and meta-analysis published in International Nursing Review in May 2025 pooled randomized trials from 2013 to 2024 that tested machine-learning-based programs for getting adults to move more. Ten trials with 2,269 participants (1,349 in intervention groups, 920 in control groups) were included. Five of them fed the meta-analysis of daily steps, which found a small-to-moderate gain for the machine-learning groups (Hedge's g = 0.402, 95% CI 0.231-0.573).

Read the study abstract

AIM: To analyze randomized controlled trials evaluating the effectiveness of machine learning (ML)-based interventions in promoting physical activity. BACKGROUND: Evidence on the effectiveness of ML-based interventions to increase physical activity from randomized controlled trials is limited. Synthesizing existing evidence is crucial for nurses to integrate such advancements into their care and implement health-promoting interventions. METHODS: Randomized controlled trials from 2013 to 2024 have been accessed b…

The gain

A meta-analysis of five randomized trials found that machine-learning-based physical activity interventions raised adults' daily step counts more than non-ML comparison programs (Hedge's g = 0.402, 95% CI 0.231-0.573).

In brief

The trials used machine learning in different ways: six to deliver the intervention, two for both delivery and data collection, one for data collection and one to predict frailty after an exercise program. Most delivered the programs through mobile apps. Results on moderate-to-vigorous activity came from three trials and were not pooled.

The review compared ML-based programs with standard or non-ML ones, and the gain was found for steps rather than for broader health outcomes. One of the six step-count trials, from Brazil, found no significant difference between groups, and the review reports little work led by nurses.

Main points

  1. The search found 2,761 records; 163 were assessed in full, ten randomized trials were included and five went into the step-count meta-analysis.
  2. Seven trials were in the United States, and one each in China, Australia and Brazil; participants were adults, including one trial of people over 65 and one of college students.
  3. The five trials pooled for steps showed no heterogeneity (I2 = 0.000%), and Egger's and Kendall's tests showed no evidence of publication bias.
  4. One trial found that algorithm-driven personalization raised step count more than random text messages or no intervention.
  5. A Brazilian trial found no significant difference in average daily steps between intervention and control groups.
  6. In one trial, the intervention group's moderate-to-vigorous activity minutes per day rose while the control group's fell.
  7. In one trial of older adults, a stacking model outperformed other algorithms at predicting whether people would reverse frailty after an exercise program.

What the review found when it combined the studies

These are the review’s own pooled results, copied from its abstract. Truvace did not recalculate them.

  • Daily step count in people receiving machine learning-based physical activity interventions compared with controls

    Hedges' g 0.402 (95% CI 0.231 to 0.573)

    499 people

    The review’s own words
    Meta-analysis revealed that ML interventions significantly increased daily step count (Hedge's g = 0.402, 95% CI: 0.231-0.573, p<0.000).

Code checks every number above against the review’s text before it is shown. The short labels are an AI model’s reading of that text, so the review’s own sentence sits under each one. A pooled figure is only as strong as the studies behind it. How Truvace uses pooled results

The full story

The authors searched PubMed, EBSCO, Cochrane and Turkish national databases from October 2023 to December 2024 under the PRISMA statement, with a protocol registered on PROSPERO. Eligible studies were randomized trials in adults that compared an ML-based intervention with a standard or non-ML one and reported outcomes such as steps, moderate-to-vigorous activity or goal adherence.

Risk of bias was low in most studies, though some lacked allocation concealment or blinding of participants and outcome assessors. The authors rated the overall evidence quality high on the GRADE method. Six trials covering 499 participants reported on daily steps, and five of them were pooled.

Beyond steps, the trials reported mixed outcomes: moderate-to-vigorous activity rose in two trials, and in a third the findings concerned who met their activity goal. A website trial found similar usage and satisfaction for video-tailored and text-tailored versions. Self-reported activity rose in groups using the mPED app.

The authors note that some trials had small samples and limited diversity, so the effects apply to specific groups, and that access to these tools can be harder for older people, people with limited digital literacy and those in low-income areas.

Sources

1 on file
  1. International Nursing Reviewdoi:10.1111/inr.70019
    Peer-reviewed2025-05-12

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

Machine-learning fitness programs added steps in review of 10 trials with 2,269 adultshttps://truvace.com/item/the-impact-of-machine-learning-on-physical-activity-related-health-outcomes-a-sy

  1. Who gets this benefit, and who was left out of the study or story?

    The gain A meta-analysis of five randomized trials found that machine-learning-based physical activity interventions raised adults' daily step counts more than non-ML comparison programs (Hedge's g = 0.402, 95% CI 0.231-0.573).

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a systematic review or meta-analysis with a sample of 2,269, from International Nursing Review.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat Some included trials had small samples and limited diversity, few were led by nurses, and the review included only English and Turkish publications, so the effects apply to specific groups.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Health