Objective This systematic review assessed the completeness of accelerometer reporting in physical activity (PA) intervention studies and assessed factors related to accelerometer reporting.
Design The PubMed database was used to identify manuscripts for inclusion. Included studies were PA interventions that used accelerometers, were written in English and were conducted between 1 January 1998 and 31 July 2014. 195 manuscripts from PA interventions that used accelerometers to measure PA were included. Manuscript completeness was scored using 12 questions focused on 3 accelerometer reporting areas: accelerometer information, data processing and interpretation and protocol non-compliance. Variables, including publication year, journal focus and impact factor, and population studied were evaluated to assess trends in reporting completeness.
Results The number of manuscripts using accelerometers to assess PA in interventions increased from 1 in 2002 to 29 in the first 7 months of 2014. Accelerometer reporting completeness correlated weakly with publication year (r=0.24, p<0.001). Correlations were greater when we assessed improvements over time in reporting data processing in manuscripts published in PA-focused journals (r=0.43, p=0.002) compared to manuscripts published in non-PA-focused journals (r=0.19, p=0.021). Only 7 of 195 (4%) manuscripts reported all components of accelerometer use, and only 132 (68%) reported more than half of the components.
Conclusions Accelerometer reporting of PA in intervention studies has been poor and improved only minimally over time. We provide recommendations to improve accelerometer reporting and include a template to standardise reports.
- Physical activity
- Intervention effectiveness
Statistics from Altmetric.com
Physical activity (PA) is related to positive health outcomes in adults1 and youth.1 ,2 However, <5% of American adults and <50% of American youth meet PA guidelines when assessed via objective monitoring,3 prompting researchers and practitioners to devise interventions aimed at increasing PA. In order to determine if a PA intervention has been successful, researchers must be able to accurately assess PA. Many methods exist for PA assessment, and the best measure to use in any given study depends on the research question(s) being addressed.
Accelerometry is often employed for use in PA intervention studies and has become a preferred tool for assessing PA in adults and youth on the basis of its objective nature, relative ease of use for the researcher and the study participant and high sensitivity for detecting changes in PA.4 ,5 However, it is not yet as well established as some other methods, such as surveys or direct observation. In 2004, a meeting titled ‘Objective measurement of physical activity: closing the gaps in the science of accelerometry’ took place at the University of North Carolina at Chapel Hill. The purpose of the meeting was to examine the use of accelerometry, which was an emerging technology, and potentially come to consensus regarding best practices for many aspects of its use. Although the meeting participants were not able to reach consensus regarding all issues, most of the important issues regarding field-based implementation of accelerometry were highlighted in a publication by Trost et al.6 Available research findings and considerations for accelerometer type, placement on the body, selection of epoch length, number of days of monitoring and minutes per measurement days required and methods for distributing and collecting monitors were among the important topics noted in Trost et al's manuscript. The manuscript did not provide hard recommendations on accelerometer use but rather presented considerations for their use and represented an important contribution to the PA assessment literature.
As a follow-up to the 2004 meeting, the National Institutes of Health hosted a workshop in 2009 titled ‘Objective measurement of physical activity: best practices and future directions’. One of the purposes of the workshop was to update recommendations for best practice in use of accelerometry. Matthews et al 7 expanded on recommendations from Trost et al's manuscript. The authors suggested that, when disseminating study findings, researchers also need to note the brand and location of accelerometer placement, a priori goals for sampling periods, methods for determining wear time, quality control checks, compliance criteria for a valid day and methods used to create key outcome variables. In fact, Matthews et al provided a checklist for researchers to use when reporting use of objective monitoring, and a recent methodological manuscript by Tudor-Locke et al 8 presents a very similar model for presenting accelerometer use and data. Thus, experts in the field have provided frameworks to follow while designing, implementing and reporting studies using objective PA assessment for all types of study designs that involve population-based research, including intervention studies.
Despite the existence of published best practices for use and reporting of accelerometry, there is no requirement for researchers to follow and report application of expert guidance in journal publications. Given the current interest conducting PA intervention studies in adults9 and youth,10 it is imperative that researchers report important aspects of data collection and analysis relative to PA as their primary outcome measure, which is ever more frequently assessed by the accelerometer. Otherwise, studies are not replicable, and it is difficult to discern if intervention effects (or lack thereof) are related to implementation issues, measurement issues, biological factors or an ineffective intervention protocol. The purpose of this systematic review was to examine reporting of accelerometry best practice elements in PA intervention studies in adults and youth over the past 17 years, encompassing studies published before and after formal accelerometer use guidelines were published.
Search methods and selection criteria
A comprehensive search of the existing accelerometry literature was conducted between July and August 2014 using the PubMed database. A date range from 1 January 1998 to 31 July 2014 was applied to all literature searches, and only manuscripts published within this time frame were included. Dates before the year 1998 were not included due to the lack of widely available methods for processing accelerometer data (eg, cut-points) prior to Freedson et al's11 study published in 1998. Manuscripts published online ahead of official journal publication date were also included if the online publication was available before 31 July 2014. Search terms consisted of (1) ‘acceleromet* AND behavior AND intervention’, (2) ‘acceleromet* AND activity AND intervention’ and (3) ‘acceleromet* OR MTI OR CSA OR ActiGraph OR Actical OR RT3 OR Tritrac OR Armband OR activPAL OR Biotrainer OR Dynaport OR Caltrac OR Actiheart OR Tracmoor OR GENEA OR Minilogger OR Actiwatch OR Actitrac AND intervention AND activity NOT pedometer AND objective’. Published review manuscripts found in this search were also examined to look for any relevant manuscripts that had not appeared in the original PubMed search. Only original research manuscripts that were published between January 1998 and July 2014 (inclusive), were written in English, cited PA as a primary study outcome, used accelerometers to assess PA and described intervention studies (not observational or validation studies) were included. Interventions in all populations and all ages (children, adults) were eligible for inclusion provided that they met the above criteria. Interventions focusing only on sedentary behaviour were not included in this review. This decision was made for three reasons: (1) use of accelerometer in PA research has explicit guidelines for reporting, which is not yet the case for accelerometer use in sedentary behaviour research; (2) to limit the number of eligible manuscripts for inclusion and (3) sedentary behaviour interventions are newer than PA interventions, and inclusion of sedentary behaviour interventions would inflate the number of recent publications and could confound potential differences in reporting completeness over time.
Manuscript scoring and data analysis
Manuscripts in this review were evaluated based solely on completeness of accelerometer reporting, not the correctness of decisions the study authors made towards accelerometer use. This review assessed the frequency with which manuscripts reported the type/brand, model and epoch length of the accelerometer and information on device placement, number of days of data collection, minimum minutes per day or days required for data to be considered valid, procedure for identifying and handling non-compliant participants and data interpretation. Twelve questions were formulated to address these issues (table 1), and questions fell into one of three categories: (1) accelerometer information (seven questions), (2) accelerometer data processing and interpretation (four questions) and (3) protocol non-compliance (one question). Manuscripts were given a ‘+’ score for each question where the manuscript adequately reported the item and a ‘−’ score for each question not adequately reported, most often when an item was not mentioned in the manuscript (eg, cut-points used were not mentioned or referenced). Examples of adequate reporting are shown in table 1. Occasionally, a ‘0’ score was given if the item was not relevant to the manuscript; for example, a ‘0’ was given when a monitor brand had only one model (ie, the early Actical monitor). In order to score reporting completeness, the number of ‘−’ scores was summed and counted as a negative score (eg, a manuscript given 5 ‘−’ scores had a completeness score of −5). Each manuscript was, therefore, given a reporting completeness score of 0 to −12, with scores closer to 0 indicating more complete reporting and scores closer to −12 indicating less complete accelerometer reporting. Although not a primary aim of this review, we collected examples of incorrect accelerometer use as they were uncovered to gain understanding of errors made when using accelerometers.
Two authors (AHKM and RWM) were responsible for reviewing and scoring manuscripts. Approximately 20% (41 manuscripts) were scored independently by both to determine inter-rater agreement; completeness scores were within 1 point of each other for 38 of the 41 manuscripts. Each reviewer then scored 80 manuscripts independently. Any manuscripts with questionable reporting were scored by the other reviewer to arrive at a scoring consensus; this occurred in four instances. Examples of acceptable reporting are given for each question in table 1, with underlining signifying the critical information needed to address each aspect of reporting; these examples illustrate how acceptable accelerometer reporting can be accomplished without adding considerable length to a manuscript.
The overall completeness of accelerometer reporting for each manuscript was first evaluated by journal of publication. In separate analyses using independent-sample t-tests, manuscripts were stratified into categories based on publication in high versus low impact factor journals (median split), publication in PA-focused versus non-PA-focused journals (PA-focused journals: journals with PA mentioned in journal aims or scope; non-PA-focused journals: all other journals) and sample being children/adolescents versus adults/older adults. Additionally, Spearman rank-order correlations were calculated to assess the relationships of publication year and completeness score as well as journal impact factor and completeness score. Current impact factors for journals were obtained from the journal website or Researchgate (ResearchGate, Berlin, Germany). Statistical significance was set at a value of p<0.05 for all analyses.
A total of 372 research manuscripts were originally identified; 201 met initial search criteria and 195 met criteria for analysis. Criteria for selection of the final 195 manuscripts can be found in figure 1. A online supplementary table is available on the journal website with scores and references for all 195 manuscripts included in this analysis.
References and raw scores from each manuscript included in analysis.
Ninety journals were used to publish the 195 manuscripts. The journals with the most published manuscripts were Preventive Medicine (n=17), American Journal of Preventive Medicine (n=11), International Journal of Behavioral Nutrition and Physical Activity (n=12) and Medicine and Science in Sports and Exercise (n=9). Four journals did not have a current impact factor, and manuscripts published in those journals were removed from the analysis examining differences in accelerometer reporting between high and low impact factor journals (n=4 manuscripts excluded). The median impact factor from the remaining 86 journals was 2.94. Each journal, based on current impact factor, was classified as either a low impact factor journal (impact factor below median) or a high impact factor journal (impact factor at or above median), and the mean (±SD) completeness score was calculated for each group. High impact factor journals had an overall completeness score of −4.2±2.0 while low impact factor journals had an overall score of −4.4±1.8; these were not statistically significantly different (p=0.54). Additionally, the impact factor was not associated with the accelerometer reporting score (r=0.04, p=0.96).
We also evaluated accelerometer reporting in manuscripts published in PA-focused journals versus non-PA-focused journals. Nine of the 90 journals with manuscripts included in this analysis were PA-focused journals, and the remaining 81 were non-PA-focused journals. PA-focused journals published 46 of the 195 manuscripts included in this analysis. The PA-focused journals had completeness scores significantly closer to 0 (better) than completeness scores in the non-PA-focused journals (−3.6±2.0 and −4.5±1.9, respectively; p=0.005).
Table 2 displays the number of manuscripts with each reporting score. Only seven manuscripts (3.6%) reported all aspects of accelerometer use, and only 132 (67.7%) reported more than half of these items. Further evaluations of accelerometer reporting are noted in table 1. Of the 195 manuscripts, all but five reported the type/brand of accelerometer used. At 69.2%, the majority of studies used the ActiGraph accelerometer brand, and the Actical was the next most used (8.2% of studies). A total of 135 manuscripts reported accelerometer placement, but only 95 included the side of the body on which the accelerometer was worn. For data management, just under half of the studies (n=96) reported wear-time requirements to be considered a valid day; similarly, just under half (n=92) reported the number of days needed for inclusion in analyses. Only 77 (39.5%) of the manuscripts reported both, and 80 (41.0%) reported neither wear-time requirement.
Of the 162 manuscripts that adequately reported how accelerometer data were interpreted into measures of PA, 116 (71.6%) used cut-points for deriving time spent in PA intensities. Of the children/adolescent manuscripts using cut-points (n=71), 15 used the Freedson cut-points, followed by the Evenson (n=10), Puyau (n=9; 6 with ActiGraph monitor, 3 with Actical monitor), Sirard (n=8) and Treuth (n=7) cut-points. Fourteen previously published sets of cut-points were used, and three studies used proprietary or individually created cut-points. For the adult/older adult manuscripts using cut-points (n=45), the majority used the Freedson 1998 cut-points (n=24), followed by the Matthews, Swartz and Copeland/Esliger cut-points (n=2, each). Twelve previously validated sets of cut-points were used, and eight manuscripts used proprietary or individually created cut-points. Alternatives to cut-points included using average counts/min, total counts/day or continuous METs/energy expenditure prediction. Notably, none of the manuscripts evaluated in this review used machine learning for analysing accelerometer data. For protocol non-compliance reporting, only 64 (33.5%) of the manuscripts had adequate reporting. Monitor malfunction, while not included in scoring, was examined and was reported infrequently, with only 27 (13.8%) manuscripts adequately reporting if and when accelerometer malfunction occurred. Furthermore, only 19 (9.7%) of manuscripts adequately reported protocol non-compliance and monitor malfunction.
Trends in publication numbers and completeness of published manuscripts
Figure 2 shows trends in the number of manuscripts published per year, in total numbers as well as split for PA interventions in children/adolescents and adults/older adults separately. Large increases were seen from 2006 to 2007, 2010 to 2011 and 2012 to 2013 for total numbers, driven by large increases from 2006 to 2007 and 2010 to 2011 in adults/older adults manuscripts and large increases from 2010 to 2011 and 2012 to 2013 for children/adolescents manuscripts.
Figure 3 provides a graphical depiction of the change in accelerometer reporting as the field has matured, overall and separated into manuscripts using adults/older adults and children/adolescents. Completeness of accelerometer reporting was weakly associated with publication year (r=0.24, p<0.001). When splitting by age category, the reporting score was also weakly associated with publication year for manuscripts with adults/older adults (r=0.32, p=0.003) and trended towards a significant correlation for children/adolescents (r=0.18, p=0.064).
Starting in 2005–2006, accelerometer reporting appeared to improve with all manuscripts and adults/older adults manuscripts only (figure 3A,B), and this same trend was present in the manuscripts for children/adolescents starting in 2003. There was a nonsignificant trend towards poorer completeness of reporting in manuscripts evaluating adults/older adults versus children/adolescents (−4.6±2.2 vs −4.1±2.1, respectively, p=0.098).
As noted in table 1, characteristics of accelerometer reporting were divided into accelerometer information, data processing and interpretation and protocol non-compliance. We examined changes in accelerometer reporting in these categories by year (figure 4); a figure was not created for reporting protocol non-compliance since the category contained only one question. Figure 4A,B shows trends towards better reporting of accelerometer information and data processing starting around 2006, with more consistent improvements seen for data processing (figure 4B).
Since PA-focused journals had better accelerometer reporting, we evaluated accelerometer reporting in manuscripts in PA-focused and non-PA-focused journals by publication year in figure 5. With the exception of 2008 and 2010 (two publications each year), accelerometer reporting improved over time for manuscripts published in PA-focused journals (figure 5A). Conversely, manuscripts published in non-PA-focused journals (figure 5B) showed little change. Correlation analyses for journal type and completeness of accelerometer reporting can be found in table 3. Correlations between completeness of accelerometer reporting for all questions and for accelerometer information questions were similar between journal types. However, correlations were significantly higher for PA-focused journals than non-PA-focused journals for completeness of reporting data processing questions.
Incorrect accelerometer use examples
Although it was not our aim to identify incorrect accelerometer use, there were several instances of incorrect use noted that we present as examples of errors when using accelerometers to measure PA. One error (five manuscripts) was citing of studies using a certain set of cut-points as justification for using those cut-points, rather than citing the original validation manuscript for the cut-points; four of these instances occurred in PA-focused journals. Another error was application of cut-points in a population considerably different from the population in which the cut-points were validated, that is, use of cut-points developed for adults to measure PA in children (one manuscript) and use of cut-points for preschool children in grade-school children (one manuscript); both instances occurred in PA-focused journals.
Another error was application of cut-points developed for vertical axis accelerometer counts to vector magnitude counts (two manuscripts, both in non-PA-focused journals). We also found 12 manuscripts that reported collecting data with triaxial accelerometers and using uniaxial cut-points for analysis without explicitly describing if one or three axes of collected data from the accelerometer were used. As it is incorrect to apply uniaxial cut-points to triaxial data, we included these as examples of potential error. Of these 12 instances, 10 occurred in manuscripts published in non-PA-focused journals.
This systematic review evaluated the completeness of accelerometer reporting in PA intervention studies using 12 questions derived from reporting recommendations in seminal publications by Trost et al 6 in 2005 and Matthews et al 7 in 2012. Characteristics of the published manuscripts (children/adolescent vs adult/older adult sample, year published) and journals in which they were published (high vs low impact factor, PA vs non-PA focus) were examined to hypothesise factors potentially affecting differences in completeness of accelerometer reporting.
Completeness of accelerometer reporting was poor. Only 7 of the 195 manuscripts evaluated adequately reported all necessary aspects, and <68% reported over half of the items. Given that these 12 items are required for proper understanding and replicating PA assessment using accelerometers, it is concerning that so many of the manuscripts did not report so many necessary elements. It is important to note that manuscripts published prior to 2005 were less likely to meet criteria since the standards/checklists were not yet published or available. This fact likely contributed to the small reporting improvement seen in more recent publications. However, improvements in accelerometer reporting improved minimally over time, even after 2005 (figure 3). Given the increased use of accelerometers in intervention studies (figure 2), the growth of scientific meetings focused on PA assessment and publication of pivotal recommendations for accelerometer use and reporting, such small improvements are disappointing.
Evidence that familiarity with accelerometers was related to more complete reporting was found when comparing accelerometer use reporting in PA-focused versus non-PA-focused journals, with average reporting scores of manuscripts published in PA-focused journals significantly higher than manuscripts published in non-PA-focused journals. Additionally, correlation analyses showed greater improvement in reporting scores over time in PA-focused journals than non-PA-focused journals (table 3), suggesting that reviewers and editors with ties to PA-focused journals may require more complete accelerometer reporting. Our results suggest that the type of the journal may be more informative regarding reporting completeness than the impact factor of the journal.
Poor reporting of protocol non-compliance and accelerometer malfunction was a notable issue uncovered. Only 34% of studies evaluated reported the portion of the sample non-compliant to the accelerometer wear requirements, and <15% reported if/when accelerometer malfunction occurred. When assessing the effectiveness of PA interventions, it is important to understand compliance to the intervention and PA assessment. An intervention shown to improve PA in a population may be less meaningful if intervention compliance is low. Similarly, if only a small portion of individuals included in an intervention met wear-time inclusion criteria, findings of the study may not be generalisable to the entire sample or to a different population. Monitor malfunction is also an important consideration when choosing a brand/type of the accelerometer, although recent technological advancements have made accelerometers more robust and reliable. Additionally, the distribution method of accelerometers (in person or mail) is important for evaluating compliance and confidence in data quality. Mailed accelerometers must be programmed to start recording days before they are to be worn to ensure that they capture the entire period they are worn.12 Methods to differentiate non-wear from in-mail transit time exist, but they are not the same as identifying non-wear during a wear period.13 Moreover, supervised attachment of accelerometers by trained personnel can enhance confidence in correct accelerometer wear. Therefore, accelerometer distribution methods need to be reported since they can affect trustworthiness of the data.
Potential changes in accelerometer use in future studies
Wrist accelerometer wear and raw data
This review uncovered several important aspects of accelerometer use that will likely be changing in the near future. First, this review found that very few studies reported using wrist-worn accelerometers (n=1) or reported collecting raw accelerometer data (n=2). While conventional data collection and processing methods worked poorly for analysing data from wrist-worn accelerometers,14 the transition to raw data collection and more advance modelling methods (eg, machine learning) have allowed for high measurement accuracy for energy expenditure, sedentary behaviour, sleep and activity type assessment using wrist-worn accelerometers.15–22 Additionally, recent studies show greatly improved compliance with wrist-worn accelerometers compared to hip-worn accelerometers in children and adults.23 ,24 Given these advantages and the adoption of wrist-worn accelerometers by large studies including NHANES and the UK Biobank, we expect to see an increase in the use of wrist-worn accelerometers in future interventions. However, as this occurs, similar questions of how to evaluate compliance, the best methods for collecting and analysing the accelerometer data, choice of wrist placement (left vs right, dominant vs non-dominant) will need to be considered and reported in these studies.
The vast majority of manuscripts evaluated in this review used ActiGraph accelerometers, likely due in part to the importance of having comparable data to other studies. With older accelerometer models and when using activity counts, research demonstrated that activity counts collected by different accelerometer brands were not comparable.25–27 These differences among different monitor brands necessitated the development of brand-specific cut-points and resulted in issues comparing PA estimates across studies using different accelerometer brands. As recent technological improvements have allowed for collection of raw accelerometer data, there is hope that raw data will be comparable across accelerometer brands and allow for better data processing methods and study comparability. Recent studies have conflicting findings regarding raw data comparability between several accelerometer brands,28 ,29 but it is likely that improved hardware and data filtering will improve raw data comparability among brands. As accelerometer brand comparability improves, it will be less critical to use the same accelerometer brand to be able to compare across studies. Accordingly, future studies may see more diversity in accelerometer brands used depending on desired battery life, design, features (eg, feedback or not) and accelerometer placement.
Advanced data modelling techniques
A notable study finding was that none of the manuscripts reviewed used machine learning or other pattern recognition techniques for processing and interpreting accelerometer data. This finding is informative, given that machine learning approaches have been recognised for their potential in improving the measurement of physical behaviours since the mid-2000s,30 ,31 with many recent validation studies showing improvement of energy expenditure prediction compared to traditional methods and the ability to identify PA intensities and types and for activity count and raw accelerometer data.17 ,20 ,32–36 However, machine learning is more complex than currently used data analysis techniques, and automated processes are not yet available for machine learning. The interventionists using accelerometers to measure PA are often not the same people as the researchers developing the modelling techniques; until machine learning models are automated and made into a more user-friendly, ‘plug-and-play’ format, there will be significant barriers to their adoption by those who are not measurement specialists. With the high interest in these techniques, we anticipate that automation of machine learning will soon be available for use by interventionists. As use of machine learning increases, additional care must be taken in accelerometer reporting, examples of which we present in table 4.
Consumer-based PA monitors
Only two manuscripts in this review used ‘consumer-based PA monitors’ (eg, Fitbit). Consumer-based PA monitors use accelerometer technology but are designed for use by individuals tracking personal PA and other lifestyle behaviours. They report PA variables such as steps, calories, distance, active time and flights of stairs climbed. Although most companies selling consumer-based PA monitors did not exist even a decade ago, their products have shown similar accuracy to ‘research-grade’ accelerometers.37 ,38 Additionally, many devices allow for competitions with other users and have goal-setting and real-time feedback (eg, vibrations, lights), and these devices have potential for encouraging behaviour change.39 As more interventions use consumer-based PA monitors, for assessment of PA and/or to elicit behaviour change, adequate reporting of their use remains critical for proper study evaluation.
The 2004 and 2009 workshops focused on using objective PA measures did not create consensus for the optimal use of accelerometers for PA assessment, in part to continue to encourage creativity and ingenuity in developing better ways to analyse accelerometer data. Accordingly, one difficultly with accelerometer use is the lack of standardised methods for accelerometer brand, placement, wear-time, data processing, interpretation, etc. This review highlights the reality that accelerometer reporting in PA intervention studies has significant room for improvement. Familiarity of reviewers with PA and accelerometer use and reporting is important, as evident by better accelerometer reporting in PA-focused journals compared to non-PA-focused journals. Those unfamiliar with accelerometer use to assess PA cannot be expected to know and critique accelerometer reporting or the quality of those choices.
We have created a table (table 5) that we recommend as a checklist of necessary elements of accelerometer reporting. Use of this table would help to standardise accelerometer reporting, providing an easy and effective way for researchers to report, reviewers to evaluate and readers to identify accelerometer use in PA interventions as well as sedentary behaviour interventions or other population-based studies using accelerometers. We also give the following recommendations:
All PA intervention studies using accelerometers should fully report accelerometer use decisions. If word and table/figure limits allow, studies should consider using a template such as table 5, which provides an easy, effective way for researchers to report accelerometer use in PA interventions and other population-based studies using accelerometers. This will allow reviewers and readers to have a concise checklist with which to evaluate accelerometer use.
Journal editors should continue to make concerted efforts to select at least one reviewer with accelerometer knowledge when evaluating submitted manuscripts using accelerometers for PA assessment. We realise the difficulty editors and journals face in recruiting well-qualified reviewers but stress the importance of needing correct accelerometer use and reporting in order to enhance quality of readers' understanding of the effectiveness of PA interventions.
Grant reviewers must similarly make efforts to critically evaluate the quality of accelerometer use plans described in applications. Grants supporting studies with thoughtfully developed and well-supported PA assessment plans will likely enhance the quality of accelerometer use and reporting in manuscripts published from the funded studies.
The research community needs to better use current recommendations7 for accelerometer reporting. There is no consensus on optimal methods for accelerometer use, so decisions on accelerometer use must be adequately reported.
Strengths and limitations
This review has notable strengths. First, our team was able to comprehensively evaluate accelerometer reporting using 12 questions taken from accelerometer use recommendations created by experts in the field of device-based PA measurement.6 ,7 We were also able to assess the role of factors such as publication year, journal type and impact factor and age of sample.
There were limitations that must also be considered. Our review necessarily had a narrow scope, examining only PA interventions using accelerometers. Our review did not examine the use of other measurement instruments such as pedometers or questionnaires, nor did it include other types of studies or sedentary behaviour interventions. Given the poor accelerometer reporting found in this systematic review, further evaluation of reporting in other study designs is warranted, as is examining other types of PA assessment tools and interventions developed to reduce sedentary behaviour. Additionally, we did not assess reporting of other potential variables of interest such as validity or reliability of the accelerometer and/or data interpretation techniques used. Finally, we did not take into account the experience of the authors on the manuscripts, which could play a role in accelerometer reporting.
Our systematic review found poor accelerometer reporting in PA intervention studies. Small improvements in reporting completeness were seen over time, although better completeness and greatest improvement were noted in manuscripts published in journals with a PA focus. This review highlights important considerations in accelerometer reporting and offers a table template that can be inserted into manuscripts for ease of reporting by researchers and reviewing by reviewers, editors and readers. Accelerometer reporting is vitally important for evaluation and replication of intervention studies, and the importance of clear reporting will become even more important as the field moves to more complex methods for analysing accelerometer data.
What are the findings?
There are deficiencies in accelerometer reporting in physical activity (PA) intervention studies. Variables such as journal of publication and year published are related to completeness of accelerometer reporting.
This study developed an evidence-based template for reporting all necessary aspects of accelerometer use and for standardising reporting in studies using accelerometers to measure PA.
Accelerometer reporting will become even more important as more advanced analytic techniques (eg, machine learning) gain wider use for analysing accelerometer data. This study offered a preliminary list for accelerometer reporting as these advanced techniques are adopted by the research community.
Twitter Follow Alexander Montoye at @alexmontoye
Contributors AHKM led data analysis, collected data and prepared final version of the manuscript. RWM collected data and conducted data analyses. HRB and KAP planned the review and oversaw data collection. RK led collection of manuscripts for inclusion in study. All authors contributed to writing of the manuscript.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.