Background

Physical activity (PA) is a strongly recommended intervention for older adults, as randomized controlled trials (RCTs) have demonstrated positive effects on a variety of outcome domains (e.g., falls, cognition, mobility, and mood, among others) [1,2,3,4]. However, PA trialists lack appropriate guidance on which outcome domains to measure consistently, leading to considerable heterogeneity in outcome selection and reporting across PA RCTs for older adults. This heterogeneity contributes to bias and makes it difficult to compare, contrast, and combine results across trials [5,6,7]. In turn, lack of consistency in outcome selection and reporting makes it challenging to identify effective, ineffective, and unproven PA interventions in a timely manner, and thereby impedes future health research, health care decision making, and development of PA policy and public health programs to support aging.

For example, given strong associations with all-cause mortality [8] as well increased dependence in activities of daily living, hospitalization, and entry into nursing homes [9, 10], slow gait speed has been called a vital sign for older adults, and leading scholars and geriatricians have proposed that a gait speed of 0.6 m/s or slower should be considered as a diagnosis of dismobility [11]. However, gait speed is not measured consistently across PA RCTs for older adults, which impairs our ability to identify and implement the most effective interventions for increasing gait speed.

Defining a minimum and standard set of outcome domains to measure and report in all PA RCTs for older adults – called a core outcome set (COS) – would help to address the aforementioned problems [7]. Development and implementation of a COS leads to higher-quality evidence about interventions and less research waste [7, 12]. In turn, interventions that work may be available more quickly to those who need them, for example through scaling up of effective PA interventions to reach large populations [13,14,15].

The Core Outcome Measures in Effectiveness Trials (COMET) Initiative, launched in 2010, fosters the development, application, and promotion of COS in all health areas and supports collaboration among those developing COS [16]. Use of COS are endorsed by the Consolidated Standards of Reporting Trials 2010 statement [17] and the Standard Protocol Items: Recommendations for Interventional Trials 2013 statement [18, 19]. The field of COS was pioneered by the Outcome Measures in Rheumatology consensus initiative, which began in 1992 and has developed COS for many rheumatologic conditions, including osteoarthritis [20]. A COS for RCTs in the area of fall and injury prevention [21] has seen good adoption [22] and thereby enabled the production of high-quality systematic reviews and meta-analyses that summarize evidence in the field (e.g., [2]).

A critical first step in COS development is to review existing literature to determine which outcomes domains (and subdomains) have been measured and reported in past trials [23]. Accordingly, to support future development of a COS for RCTs of PA for older adults, the aim of this study was to conduct a rapid review to identify outcome domains and subdomains that have been used in RCTs of PA for older adults.

Methods

Protocol

We developed and followed a rapid review protocol using guidance from existing literature [24,25,26,27,28]. A checklist for reporting rapid reviews does not currently exist in the Enhancing the QUAlity and Transparency Of health Research library, but this study is reported according to the relevant quality elements of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension for scoping review (PRISMA-ScR) [29], to which the current study most closely aligns.

Inclusion criteria

Consistent with rapid review methodology, which streamlines components of the systematic review process, we sought to include recently published, English-language, high-quality RCTs that delivered a PA intervention to older adults with the goal of affecting one or more health-related outcomes. Operational definitions of each of these concepts are included in Table 1.

Table 1 Key concepts and operational definitions pertaining to the rapid review
Full size table

Search strategy

We searched MEDLINE [Ovid], since all journal titles of interest were indexed in MEDLINE. We searched first for journal titles of interest. Next, we used a combination of MESH terms and keywords to search the concepts of PA and older adult. We used the MEDLINE [Ovid] filter for RCTs. We limited by year of publication (2015 to 2021), and English language. An information scientist reviewed and provided feedback on the search strategy to optimize sensitivity and specificity. The search strategy is included in Additional file 1.

Screening process

We imported and managed articles in Covidence software for screening. Two reviewers (DCM, CLE) independently screened articles against eligibility criteria in two phases: citations (title and abstract screening); and full-text article screening. We resolved discrepancies by discussion.

Data abstraction process

We defined the data items for abstraction a priori, which included the following main categories: article information (e.g., first author, year of publication, journal title), study characteristics (e.g., country, continent, phase of trial, clinical trial registration), study design (e.g., study setting, sample size, % female participants, mean participant age, study arms, randomization unit), interventions (e.g., primary location, mode, type), adverse events, intervention adherence, and study outcomes.

We imported and managed articles in NVivo software for data abstraction. Data were abstracted from articles by highlighting the relevant content, assigning it to its corresponding node (data item), and exporting tables of highlighted content. In addition, categorical data items (e.g., was clinical trial registered: yes, no; participant sex: females only, males only, female and male sexes) were classified directly on a Google spreadsheet with pre-defined data validation.

Two reviewers (AC, AW) independently abstracted the study outcomes from all articles and resolved discrepancies through discussion. A third reviewer (DCM) verified the set of outcomes abstracted from all articles. All other data items were abstracted by a single reviewer (AC or AW). Reviewers completed their abstraction for each item from all articles before beginning abstraction of the next item, and they sought clarification and assistance when needed from each other and from DCM.

Synthesis

We classified study outcome domains according to the standard outcome classification taxonomy endorsed by the COMET initiative, which includes five core areas (death, physiological/clinical, life impact, resource use, and adverse events) and 38 outcome domains [23]. To increase descriptive power, we also defined custom outcome subdomains for those domains that appeared in at least 10 articles, as we felt that outcome representation across approximately 15% of included studies would give sufficient breath to understand ‘what’ may be critically important to measure in trials of PA in older adults. This is consistent with the objectives of this review to classify outcome domains and subdomains, which constitute ‘what’ to measure, rather than psychometric properties of outcome instruments, which constitute ‘how’ to measure. The core areas, outcome domains, and outcome subdomains we applied and used for results presentation are listed in Additional file 2. Applying the COMET taxonomy to some outcomes required discussion among the researchers; decisions about outcome classification are detailed in Additional file 3. Notably, some outcomes were classified under multiple domains, and specifically named adverse events (e.g., musculoskeletal injury) were categorized under the appropriate taxonomy domain and not under the generic adverse event domain [23].

Results

Literature search & screening

Our search yielded 548 citations (titles and abstracts). Citations were screened in four batches (n=25, n=25, n=150, n=348). Among the 548 citations screened, both reviewers (DCM, CLE) agreed on eligibility status for 514 (93.8%) and disagreed on 34 (6.2%), Cohen’s Kappa = 0.77. Reviewers readily resolved discrepancies by email or videoconference and made minor updates to the wording of eligibility criteria to improve clarity. In total, 82/548 (15.0%) articles passed citation screening and moved onto full text screening (Fig. 1. PRISMA flow diagram for rapid review) [34]. During full text screening of 82 articles, there were 13 discrepancies (15.9%); 12 disagreements about inclusion/exclusion (Cohen’s Kappa = 0.45), and one disagreement about reason for exclusion. These were readily resolved by email correspondence and brief discussion. Of the 67 articles that passed full text screening and were included in the review (Additional file 4), 30 were published in Geriatrics and Gerontology journals, 15 in Rehabilitation journals, 13 in Sports Science journals, and 9 in Medicine, General & Internal journals.

Fig. 1
figure 1

PRISMA flow diagram for rapid review

Full size image

Descriptive characteristics of randomized controlled trials

Of the 67 included articles, 35 (52%) were from trials in North America, 14 (21%) from Europe, eight (12%) from Asia, seven (10%) from Oceania (all in Australia), and three (5%) from South America (Table 2). In North America, 33 articles were from the United States and two from Canada. Clinical trial registration was reported in 41 (61%) articles; the most common registries were ClinicalTrials.gov (n=29) and the Australian New Zealand Clinical Trials Registry (n=8).

Table 2 Descriptive characteristics of included trials (n=67)
Full size table

Individual trial sample size ranged from 14 to 2157 participants, with median (IQR) of 94 (57-517); a total of 28,649 participants were included across all trials. A sample size calculation was included in 28 (41%) articles (Table 2). Trials were most commonly conducted in community settings (n=56, 84%), but four (6%) were conducted in nursing homes, one in assisted living, and six (9%) in other settings (e.g., research facility). Males and females were recruited in 57 (85%) trials; 10 (15%) trials recruited females only, and no trials recruited males only. Study populations were predominantly female (% female ranged from 40 to 100% with mean (SD) of 72 (15)).

Of the 67 included articles, 49 (73%) reported on main trial results, while 11 (16%) reported a secondary analysis, and five (8%) a subgroup analysis (Table 2). One article reported a trial protocol and one other article reported an ancillary study from a larger trial. Fourteen (21%) articles reported on data from either the Lifestyle Interventions and Independence for Elders (LIFE) Study [1] or the LIFE Pilot Study [35].

All 67 trials were superiority trials (designed to assess if one or more interventions was different from – better or worse – than control) [36] (Table 2). The majority were designed to test intervention efficacy/effectiveness (n=55, 82%), while 10 (15%) were proof-of-concept/feasibility/pilot trials, one was a large pragmatic trial, and one a cost-effectiveness trial. Sixty-three (94%) trials used a parallel study design, while four (6%) used a factorial design; no trials used cross-over, adaptive, or stepped wedge designs. Individual randomization was used in 61 (91%) trials and cluster randomization in six (9%).

Forty-two (63%) trials had two arms, 18 (27%) had three arms, and seven (10%) had 4+ arms. Active control groups were used in 36 (54%) trials, placebo/usual care control groups in 24 (36%), and waitlist control groups in two (3%) (Table 2); a control group was not explicitly identified in five (7%) trials. Most intervention arms used a combination of PA types, while 13 (19%) trials focused solely on strength/resistance activities, nine (13%) on aerobic/cardiorespiratory, and five (7%) on balance/proprioception. Many trials (n=27, 40%) also used a combination of group and individual modes of PA, while 19 (28%) were comprised solely of group-based activities, 13 (19%) were solely individual activities; mode was not reported in eight (12%) trials. Intervention adherence was reported to have been measured in 52 (78%) trials; however, measured intervention adherence was reported in only 48 (72%) trials. Adverse events were reported on in 45 (67%) trials.

Classification of study outcomes into domains and subdomains

We identified 21 of 38 possible outcome domains across the 67 articles, spanning 4 of 5 possible core areas: physiological/clinical, life impact, resource use, and adverse events (Table 3). No outcome domains were identified under the core area of death. Further, no outcome domains were reported in all RCTs.

Table 3 Outcome frequencies and examples (classified under core area; outcome domain; outcome subdomain) in n=67 articles
Full size table

The five most commonly reported outcome domains were physical functioning (included in n=51 articles), musculoskeletal and connective tissue outcomes (n=30), general outcomes (n=26), cognitive functioning (n=16), and emotional functioning/wellbeing (n=14) (Table 3). These five outcome domains were also the ones that met our criteria for defining subdomains, as they each appeared in at least 10 articles (Table 3). General outcomes referred to disorders and global measures and symptoms that affected the whole body and could not be attributed to a specific bodily system (see example outcomes in Table 3) [23].

We identified 10 unique outcome subdomains. For the physical functioning domain, we defined six subdomains (in order of frequency): mobility, fall-related, lifestyle, balance, quality of life, and other. For the musculoskeletal and connective tissue outcomes domain, we defined four subdomains: muscle performance, bone, body composition, and other. For the general outcomes domain, we defined three subdomains: fall-related, body composition, and other. For the cognitive functioning domain, we defined three subdomains: multiple cognitive functions, processing speed & executive function, and other. Finally, for the emotional functioning/wellbeing domain, we defined two subdomains: fall-related and quality of life. Notably, some subdomains were common to multiple domains, such as fall-related, body composition, and quality of life. We grouped outcomes found in n=2 articles or fewer into a category called “other.” Examples of outcomes classified under each domain, and subdomain (where applicable) are given in Table 3.

Discussion

In this rapid review, we observed considerable heterogeneity in outcome domains measured and reported in recently published, high-quality RCTs of PA for older adults, predominantly conducted in the community setting. Twenty-one unique outcome domains and 10 unique outcome subdomains were reported across 67 articles. This finding reflects the broad range of health benefits potentially derived from PA and also investigator interest to monitor a range of safety parameters that may be related to adverse events from PA. Notably, no outcome domains or subdomains were reported in all trials, which underscores the need to develop a COS to improve outcome reporting consistency and enhance evidence quality.

Physical functioning was the most dominant outcome domain reported in the RCTs included in this review, indicating that PA trialists have traditionally emphasized the importance of measuring the physical functioning benefits from PA. The physical functioning domain encompassed a number of distinct subdomains, the most prominent of which was mobility. The measures of mobility included in the articles we reviewed largely captured an individual’s capacity for mobility by assessing physical performance, such as gait speed, chair stand performance, etc. There were not many instances where measures of a person’s enacted mobility, such as the extent and frequency of movement away from one’s residence (which may further relate to role or social functioning), were used as outcomes. As a result, there is opportunity for more holistic and comprehensive measurement of mobility in future PA RCTs [37, 38].

Musculoskeletal and connective tissue was the second most dominant outcome domain in this review. Within, the most prominent subdomain was muscle performance, which included measures of muscle strength, power, torque, quality, and fatigue. This emphasis on muscle performance was likely due to the nature of interventions in the included trials, whereby about 20% exclusively targeted strength/resistance training and another 60% included combination training, some of which incorporated strength/resistance exercises. Strength/resistance training is heavily promoted for older adults for the prevention of falls and related injuries [2].

Fall-related outcomes also appeared prominently in the articles reviewed. The subdomain of fall-related outcomes was associated with three different outcome domains – physical functioning, general, and emotional – reflecting the wide range of impacts that falls have on the daily lives of older adults and the potential of PA to mitigate or prevent these impacts.

None of the trials included in this review measured the impacts of PA on sleep. As the importance of good sleep quality and quantity to overall health and wellbeing continue to be explored and documented [39, 40], we hypothesize that older adults and their health care providers will deem positive impacts on sleep to be an important reason for participation in PA. In addition, under the core area of resource use, the outcome domains of ‘hospital’, ‘need for further intervention’, and ‘society/carer burden’ were not reported in included articles. This may represent (1) a lack of input from health care professionals and/or older adult consumers in trial outcome selection, (2) a lack of suitable outcome measures (in terms of measurement feasibility, data accessibility, or psychometric performance), (3) limitations in trial resources (e.g., sample size required) and/or (4) a lack of expertise in health economics among teams of trial investigators. Future research will be necessary to determine the outcomes related to PA participation that older adults and their health care professionals view as critically important, and to overcome barriers to their quality measurement. In particular, despite the inherent methodological challenges involved in economic evaluation of health promotion programs [41], it is vital that cost-effectiveness data are provided in order to help justify future investment by policy makers in PA interventions for older adults.

We envision that the inventory of outcome domains and subdomains generated with this review will have multiple applications. First, we expect the review will serve as essential background information (an ‘informative brief’) toward the development of a COS for PA RCTs with older adults [28, 42]. Indeed, one third of published COS are preceded by a literature review to identify potentially relevant outcomes [23]. Thus, our review findings will be used as an evidence summary during future patient and health care provider engagement as well as consensus generation activities that lead to development of a COS. For instance, we will seek to compare patient and health care professional priorities for outcome domains with the inventory of outcome domains/subdomains generated by this review to identify both areas of overlap and difference. Second, as development work on a COS proceeds, the outputs of this review will be useful immediately to inform outcome selection for RCTs of PA for older adults and to guide reporting in systematic reviews and meta-analyses that synthesize RCTs of PA for older adults.

This study had certain limitations. First, our inclusion criteria were not designed to identify a complete and exhaustive set of older adult PA RCTs; rather, we sought to identify enough representative trials to achieve sufficient understanding of the concepts necessary to address the study objectives. We recognize, for example, that by including only RCTs published in journals with a top-10 ranking, we may have missed outcome domains and subdomains found in articles published in lower ranking journals. Second, in this rapid review, we did not formally assess study quality of the 67 included articles with a published checklist or tool, as would be required for a systematic review about intervention effectiveness. Formal quality assessment was not necessary to achieve the objectives of this review, as our purpose was not to assess effectiveness of PA interventions, but rather to create an inventory of outcome domains and associated subdomains that have been used in past PA RCTs. Moreover, we attempted to locate high-quality studies at the outset through searching exclusively for RCTs published in top-10 journals in the four most relevant disciplines. Third, we did not consider factors that may affect selection of outcome measurement instruments such as psychometric properties, feasibility, or mode of assessment, nor did we explore differences in outcome selection based on trial characteristics that may influence adoption of outcome measurement instruments. This was in keeping with the purpose of this review to generate an inventory of outcomes used in previous RCTs of PA for older adults to inform the selection of outcome domains (‘what’ to measure) for inclusion in a future COS. The selection of outcome instruments (‘how’ to measure) is distinct from, and typically comes after, the selection of outcome domains. Fourth, the results of this rapid review were influenced by outcome selection in the LIFE Study and the LIFE Pilot Study, as 21% of included articles reported results from these trials, reflecting the profound impact of these trials in the field of aging and PA. Nonetheless, a wide range of outcome domains were identified in these articles, including musculoskeletal and connective tissue, general, injury and poisoning, nervous system, psychiatric, vascular, physical functioning, social functioning, emotional functioning/wellbeing, cognitive functioning, and economic, diminishing the probability of overweighting certain domains from these two trials. Outcome selection in LIFE Study and the LIFE Pilot Study was governed predominantly by trial investigators [43, 44]; moreover, the inventory of outcome domains reported in this review most likely reflect the priorities of clinical trial investigators. Since outcomes selected for a COS must be meaningful to researchers, patients, and health care professionals, there is a distinct opportunity to advance the field by documenting and incorporating outcome preferences of older adults and their health care professionals in a COS [7, 45].

Conclusions

In conclusion this review provides an inventory of outcome domains and subdomains used in recent, high-quality RCTs of PA for older adults. There is strong potential to integrate the results of this review with the top research priorities of older adults and their health care professionals into a future COS for PA RCTs for older adults.