Industries / Media, arts and recreation / Fitness and gyms
Injury and safety
Injury and safety is the work of keeping people from being hurt or made ill by exercise: deciding who should see a doctor before starting, building machines that do not trap or throw the people near them, keeping shared rooms from spreading infection, and being ready when someone collapses.
7
Milestones achieved, from 1992 to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1992
A gentler screening form
The Physical Activity Readiness Questionnaire (PAR-Q) is a short yes-or-no form that tells anyone who answers yes to a question to see a doctor before becoming more active. Its authors' 1992 study at 40 accredited fitness testing centers across Canada found that the original wording screened out what seemed too many healthy older adults. New wording from an expert committee set up by Fitness Canada cut the number of the 399 people screened out from 68 to 48, mainly because fewer gave a wrong yes to the question on high blood pressure.
Lessona safety form that turns away too many healthy people becomes a barrier to the exercise it was meant to make safe.
Credit
S. Thomas, J. Reading and R. J. Shephard, for Fitness Canada
2010
Weight room injuries counted
Z. Y. Kerr, C. L. Collins and R. D. Comstock used the US Consumer Product Safety Commission's National Electronic Injury Surveillance System (NEISS), which records injuries seen in a sample of about 100 hospital emergency departments, to estimate weight training injuries from 1990 to 2007. The 25,335 cases in the sample stood for an estimated 970,801 injuries nationwide. Sprains and strains were the most common diagnosis (46.1%), weights dropping on the person were the most common cause (65.5%), and 90.4% of injuries involved free weights rather than machines.
Lessontwo thirds of these injuries came from handling and dropping the weight, not from the lift itself.
Credit
Z. Y. Kerr, C. L. Collins and R. D. Comstock
2013
Better odds in a gym
R. L. Page and colleagues reviewed every cardiac arrest in a public indoor place in Seattle and King County, Washington, from 1996 to 2008: 849 arrests, 52 of them at traditional fitness centers and 84 at other places people exercise, such as tennis courts, ice rinks and bowling alleys. Survival to leaving hospital was 56% at fitness centers, 45% at the other exercise sites and 34% at indoor places not used for exercise. Arrests per site were higher at indoor tennis, ice arenas and bowling alleys than at fitness centers. New York's General Business Law now requires every health club with 50 or more members to keep an automated external defibrillator (AED), a portable device that shocks a stopped heart back into rhythm, and to have someone trained in its use and in CPR on site during staffed hours.
Lessondefibrillators belong where arrests happen, and that includes rinks and bowling alleys as well as gyms.
Credit
R. L. Page and colleagues; New York State Legislature
2015
Screening by symptoms first
The American College of Sports Medicine (ACSM) dropped the counting of heart-disease risk factors from its screening before exercise, after studies suggested its old rules sent too many people to doctors and put them off exercising. The new model, from a roundtable held in June 2014, asks three things: how active the person already is, whether they have symptoms or known heart, metabolic or kidney disease, and how hard they want to exercise. In Canada, the PAR-Q+ that replaced the PAR-Q refers about 1% of people for further medical screening, by its authors' own account.
Credit
D. Riebe, B. A. Franklin, P. D. Thompson and colleagues, for the ACSM; Shannon Bredin and colleagues at the University of British Columbia, for the PAR-Q+
2020
Gyms reopened as a trial
While most gyms were shut for Covid-19, the TRAiN study group in Oslo, Norway, randomly gave 3,764 members of five fitness centers either access or no access from 22 May 2020, with 1 meter between people on the floor, 2 meters in hard classes and extra cleaning. Of 3,016 virus tests returned, one was positive, in the training group, and nobody in either group needed hospital care for Covid-19. In Chicago that August, by contrast, 55 of 81 people who attended in-person classes at one exercise facility caught Covid-19, and most attendees who answered said they seldom wore masks in class.
Lessonthe same kind of room was safe or not depending on how much virus was circulating outside, how far apart people stood and whether they wore masks; the Oslo authors limit their result to a time of low infection.
Credit
L. M. Helsingen, M. Bretthauer, M. Kalager and the TRAiN study group; US Centers for Disease Control and Prevention
2021
A recall and a fine
From December 2018 Peloton received reports of people, pets and objects pulled under the rear of its Tread+ treadmill. It began moving a warning label and studied adding a rear guard, but did not tell the US Consumer Product Safety Commission (CPSC) until 4 March 2021, the day after it learned a six-year-old child had died. The CPSC warned the public to stop using the Tread+ on 17 April, and about 125,000 were recalled on 5 May 2021. In January 2023 Peloton agreed to pay a civil penalty of $19,065,000: $16,025,000 for reporting late and $3,040,000 for delivering 38 recalled treadmills after the recall.
Lessonthe legal duty to report starts when incident reports point to a defect, not when someone dies.
Credit
US Consumer Product Safety Commission
2023
First spin class danger
Exertional rhabdomyolysis is the breakdown of muscle after very hard exercise; the released muscle protein can damage the kidneys. Doctors at Singapore General Hospital reviewed 93 people admitted with it after spin (indoor cycling) classes between March 2021 and March 2022: 87.1% were at their first spin class, 71% were women and one had kidney injury, which resolved the next day.
Lessonthe risk fell mostly on newcomers, so a first class needs to be run differently from a regular's class.
Credit
S. Y. W. Lim, C. J. Chong, Z. Liu and J. Y. L. Kan, Singapore General Hospital
Watch for
Camera form correction for safety
Peloton says the Movement Tracking Camera on its new Plus machines lets its Peloton IQ software give real-time form feedback and correction, count repetitions and suggest weights, to help members lift safely. The announcement gives no figures for how often the feedback is right.
Achieved whenan outside test publishes how often the camera flags known form faults, judged against certified coaches, across more than one exercise.
Claimed by
Peloton, October 2025
Open
01
A yearly count by machine
An open, repeatable count of US emergency room injuries from exercise equipment, broken down by machine (treadmill, exercise bike, elliptical, rower, bench, barbell, dumbbell and so on) and by what happened (fell off, weight dropped, caught in a belt, part broke). It reads each year's NEISS case narratives, sorts them with an AI model checked against people, and turns the counts into national estimates with error ranges using NEISS's own sample weights.
The briefCredit · unclaimed02
Measuring treadmill rear gaps
A public table of the openings at the back of motorized treadmills, where the belt wraps around the rear roller, measured on machines that are unplugged with the safety key removed. For each model it records the gap between the rear guard (if there is one) and the roller or belt, the gap between the bottom of the guard and the floor, the guard's width against the belt's, and photos, set against the limits the US standards committee drafted in 2022.
The briefCredit · unclaimed
Open 01 · the brief
A yearly count by machine
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, repeatable count of US emergency room injuries from exercise equipment, broken down by machine (treadmill, exercise bike, elliptical, rower, bench, barbell, dumbbell and so on) and by what happened (fell off, weight dropped, caught in a belt, part broke). It reads each year's NEISS case narratives, sorts them with an AI model checked against people, and turns the counts into national estimates with error ranges using NEISS's own sample weights.
Why it is still open
The counts that exist are one-off studies, each sorting the narratives its own way. Janessa Graves and colleagues at the University of Washington estimated 70,302 emergency room injuries from home exercise machines in 2007 to 2011, two thirds of them on treadmills, using text searches in Excel; they note that the type of machine is not routinely recorded and that NEISS does not code how an injury happened. In 2026 K. Coffey and colleagues sorted 15,348 resistance training cases from ten years of NEISS by injury and cause, and called for tracking that links each kind of lifting to its injuries. Machine learning has been used on NEISS narratives for other questions, such as what 91,325 young children fell from (E. Omaki and colleagues, 2023). A search found no study that applies an AI model to exercise equipment narratives, and no yearly series by machine.
Built with what exists
NEISS public case files, which the open-source R package neiss (github.com/hadley/neiss) already loads; a general model such as Claude, or a model built to answer fixed-choice questions in bulk such as Gemini Decision 1 (see the Note "A thousand questions, one pass"); the R survey package for weighted national estimates; a spreadsheet for two human coders.
First step
Load one year of NEISS cases under the exercise equipment product code (3277, the code Graves and colleagues used), hand-code 200 narratives, and score one model against them.
Done looks like
Every exercise equipment narrative from 2015 to 2025 coded; on a random 1,000 narratives coded separately by two people, the model matches their agreed answer on machine type at least 95% of the time and on what happened at least 90%; yearly national estimates with 95% intervals published with the code and labels, and updated within a month of each new NEISS year.
Open 02 · the brief
Measuring treadmill rear gaps
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of the openings at the back of motorized treadmills, where the belt wraps around the rear roller, measured on machines that are unplugged with the safety key removed. For each model it records the gap between the rear guard (if there is one) and the roller or belt, the gap between the bottom of the guard and the floor, the guard's width against the belt's, and photos, set against the limits the US standards committee drafted in 2022.
Why it is still open
CPSC staff asked the ASTM committee that writes the US treadmill standard, in April 2021, to deal with children pulled under the rear roller. At a May 2022 meeting, two prototype guards gave mixed results when balls were pushed against running belts, and members said that differences between treadmills "would make a simple performance test difficult". The draft moved instead to a guard with a gap under 9.5 mm to the roller and under 94 mm to the floor. Consumer Reports wrote in August 2023 that its preliminary tests suggest other treadmills could also pull objects under, "though there isn't yet one way to assess just how much of a hazard is posed by each machine". The public summary of the 2025 edition, ASTM F2115-25, does not say whether those limits were adopted. A search found no published measurements of rear gaps across models.
Built with what exists
Feeler gauges and a digital caliper for the small gap, a steel rule for the floor gap, a phone for dated photos, and a shared spreadsheet or static website for the table.
First step
With a store's permission, measure five unplugged floor models and publish the readings and photos.
Done looks like
50 current models from at least 10 makers measured, each by two people whose readings agree within 1 mm, with photos, the share of models that meet each draft limit, and the method published.