How We Report the 24-Hour Day: Landmark Review Finds Inconsistent Standards in Movement Behavior Research

Every second of your day belongs to one of three buckets: sleep, sedentary behavior, or physical activity. Spend more time in one and you necessarily have less for the others. This simple arithmetic truth has spawned a statistical revolution in health research over the past decade, one built on a technique called compositional data analysis, or CoDA. Rather than treating sleep, sitting, and movement as independent variables, CoDA models them as relative parts of a finite 24-hour whole, converting absolute minutes into proportions and then into log-ratios that respect their co-dependency. The approach has exploded in popularity since it was first applied to movement behaviors in 2015. But a sweeping new systematic review now reveals an uncomfortable secret beneath that growth: the field’s reporting practices are a patchwork, so inconsistent that the scientific community may be unable to pool its own findings.

The review, published in the Journal of Activity, Sedentary and Sleep Behaviors, was led by Denver M. Y. Brown of Kansas State University together with colleagues from institutions across North America. The team searched eight databases from their inception through late 2023, supplemented by forward and backward citation searches, a manual scan of the International Network of Time-Use Epidemiologists’ publication list, and direct consultation with experts. From 7,212 identified records, 102 observational studies met the strict inclusion criteria: each had to use CoDA with log-ratio transformations, capture all three movement behaviors across the full 24-hour day, and test associations with indicators of physical, cognitive, or mental health. The studies, published between 2016 and 2024, drew on samples from 29 countries, with analytic samples ranging from just 28 participants to more than 130,000, and participant ages spanning from infancy to late adulthood.

To make sense of such a heterogeneous literature, the researchers classified reporting practices into seven domains: methodological justification, behavioral measurement and data handling, composition construction, analytic plan, composition-specific descriptive statistics, model results, and auxiliary information. The extraction framework was reviewed by eleven international experts and drew on existing reporting checklists such as STROBE, designed for observational epidemiology, and CREMAS, developed for ecological momentary assessment studies. Each study was also scored with an adaptation of the National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-sectional Studies, yielding a sobering baseline: 81 of the 102 studies were rated fair quality, 19 good, and two poor.

The brightest spot was methodological justification. Nearly every study, 100 of 102, mentioned CoDA in the title or abstract, and 94 explained why a compositional approach was appropriate for 24-hour time-use data. This is more than academic housekeeping. Because the 24-hour movement paradigm remains unfamiliar to many journal audiences, clear justification helps readers grasp why the co-dependency of behaviors demands special statistical treatment. It also aids discoverability: when authors flag CoDA prominently, systematic searches can actually find these studies, a prerequisite for any future synthesis of the evidence.

Beyond that strong opening, the picture fragments. In the measurement domain, most studies did describe how accelerometer data were scored and processed for each behavior, where devices were worn, how many minutes constituted a valid day, and how many valid days were required. Yet only 44 of the 90 device-based studies clearly stated which valid days entered the analysis, whether averages across all days, random selections, or weighted combinations of weekdays and weekends. That omission matters: recent work has shown that different data-handling strategies can materially change conclusions about how movement behaviors relate to obesity. Sleep fared worst of all. Only about half the studies clearly conceptualized what sleep they were measuring, whether nocturnal sleep, 24-hour sleep, or daytime naps, and fewer than 40 clearly defined the sleep variable itself. Only 17 studies explicitly included naps in the composition, and in 44 studies it was simply unclear. Without sleep logs, daytime sleep captured by accelerometers risks being misclassified as sedentary time, quietly corrupting the very compositions the analyses depend on.

Composition construction revealed similarly troubling gaps. All 102 studies reported how many parts the day was partitioned into, most commonly four: sleep, sedentary behavior, light physical activity, and moderate-to-vigorous physical activity, with compositions ranging from three parts up to nine. And 101 of 102 reported the log-ratio transformation used. But only six studies specified how a day was defined, midnight-to-midnight versus wake-to-wake, a detail that shapes what the composition actually represents. Most striking, only 27 studies addressed zeros in the behavioral data. Zeros are poison for log-ratio mathematics, since the logarithm of zero is undefined, and how researchers replace them can distort the entire analysis. Encouragingly, of the 19 studies that observed zeros, most used principled methods such as log-ratio expectation maximization imputation or multiplicative replacement, approaches recent methodological work has endorsed over crude substitution.

The analytic plan and results sections showed the deepest inconsistencies, and these are the ones with the greatest consequences for cumulative science. Nearly all studies reported their post-transformation analytic technique and whether covariates were adjusted, but fewer than a third compared their analytic sample to the full sample, and only 15 provided a power analysis, a gap the authors attribute largely to the field’s reliance on secondary data analysis. In the results, only 64 studies reported whether the overall 24-hour composition was significantly associated with the health outcome at all, and just 25 included a standardized effect size such as R-squared. From there, practices diverged into what the authors describe as selective reporting: some studies examined individual behaviors or ran isotemporal substitution models only when the overall model was significant, others ran them regardless, and some never stated whether the overall test was significant. Several studies reported associations for only some components of the composition, omitting exactly the kind of complete, per-behavior estimates that meta-analysis requires.

Isotemporal substitution, the technique that estimates what happens to a health outcome when minutes are swapped between behaviors, was the single most popular analytic approach, appearing in 84 studies. Here, reporting was somewhat better: 83 of the 84 studies clearly stated whether substitutions were one-to-one, such as ten minutes of sitting replaced by ten minutes of brisk walking, or proportional, where reallocated time is spread across all other behaviors. Most reported model statistics with confidence intervals or standard errors. Yet even here, roughly half only examined reallocations for outcomes already found significant, and 15 reported substitutions for only a subset of behaviors. Auxiliary reporting told a similar story of selective diligence: funding sources and conflicts of interest were almost universally disclosed, likely driven by journal requirements, but only 16 studies acknowledged the limitations of CoDA itself, and only 22 discussed whether their effect sizes were clinically meaningful, a crucial question when translating compositional proportions back into actionable minutes of activity.

The stakes of this inconsistency extend well beyond academic tidiness. The 24-hour movement paradigm underpins integrated guidelines now adopted across multiple countries for children, adults, and older adults, and it increasingly informs public health policy about how we should structure our days. If individual studies cannot be compared, and if selective reporting blocks the extraction of complete effect estimates, the field cannot perform the meta-analyses needed to establish robust, population-level conclusions about the optimal distribution of sleep, sitting, and movement. The review’s authors point to a proven path forward: the CREMAS checklist for ecological momentary assessment was born from exactly this kind of systematic characterization of existing practices, and they propose a similar route for CoDA, using Delphi methodology, an iterative consensus process among experts and end-users, to develop a formal reporting checklist for observational compositional studies. They also caution that their conservative approach, counting unmentioned items as unreported, may underestimate actual practice, and that emerging techniques will require the standards to evolve. Still, the message is unambiguous. A field that has grown from two studies in a 2020 systematic review to more than a hundred in under a decade now needs to grow up statistically: consistent, transparent, and complete reporting is the price of turning a promising method into trustworthy evidence about how the whole 24-hour day shapes human health.

Subject of Research: Reporting practices in observational studies using compositional data analysis to examine associations between 24-hour movement behaviors and health indicators

Article Title: A systematic review of research reporting practices in observational studies examining associations between 24-h movement behaviors and indicators of health using compositional data analysis

Article References: Brown, D. M. Y., Burkart, S., Groves, C. I., Balbim, G. M., Pfledderer, C. D., Porter, C. D., Laurent, C. S., Johnson, E. K., & Kracht, C. L. (2024). A systematic review of research reporting practices in observational studies examining associations between 24-h movement behaviors and indicators of health using compositional data analysis. Journal of Activity, Sedentary and Sleep Behaviors, 3(1), Article 23. https://doi.org/10.1186/s44167-024-00062-8

Image Credits: AI Generated

DOI: 10.1186/s44167-024-00062-8

Keywords: compositional data analysis, 24-hour movement behaviors, physical activity, sedentary behavior, sleep, systematic review, reporting practices, time-use epidemiology, isotemporal substitution, accelerometry, research methods, public health