The review was registered on the International Prospective Register of Systematic Reviews (ID: CRD42020209672).

We considered experimental or quasi-experimental (pre-test post test) primary research, published in the English language. Eligibility according to the PICO (Population, Intervention, Comparator, Outcome) framework are described in detail below.

Searches, performed with the support of an information specialist, were conducted in seven bibliographic databases (MEDLINE [via Ovid], Embase [Via OVID], EMCARE [via Ovid], AMED [via Ovid], CINAHL [via EBSCO], Cochrane Library, Web of Science). Database selection was based on recommendations described by Bramer et al (26); and searches spanned the period from database inception to 16 November 2020. Reference list and citation searches of included articles and previous systematic reviews were conducted. Details of search strategies can be found in Supplementary Table 1. Table 1

List of definitions used to characterise exercise components

The web-based platform Rayyan (27) was used to screen titles and abstracts. Two reviewers screened results returned from searches, in a blinded manner. Full-texts were also reviewed blinded and in duplicate. Discrepancies at either phase were resolved by a third party. Multiple publications were collated to ensure we obtained the most detailed description of interventions (28).

The following data items were tabulated in an excel database:

Administrative: title, author, acronym, number of papers, year, country, design (main), design (sub-type), sample details (arm name, sample size), potential for meta-analysis.

Risk of bias is not relevant to this descriptive review and will be reported with the planned meta-analysis.

We grouped studies based on the number and type of exercise components they included on the supposition this would reveal the extent of heterogeneity within the category ‘multi-component’. To achieve standardisation, two blinded researchers reviewed all studies and coded the exercise intervention into six categories based on the description given in the paper or supporting texts. Exercise components were defined using WHO definitions (14) for aerobic physical activity, muscle strengthening exercises, flexibility exercise, balance exercises and functional exercises (Table 1). A further category for ‘respiratory exercises’ was used to include inspiratory muscle training. Exercises undertaken in different environments (e.g. hydrotherapy) were assessed and coded based on the components included.

In addition to characterising exercises included in interventions, we assessed the rationale behind intervention exercise content. Assessment was based on descriptions and references given within texts and categorised as either 1) based on pre-development work; 2) based on previous research studies; 3) based on national/international guidelines; 4) based on practitioner intuition/local practice. Studies that made no reference to either development work, previous research or guidelines were assumed to be based on ‘practitioner intuition/local practice’.

Seven databases were searched yielding a total of 51,001 papers, of these, 138 articles reporting unique interventions were included. Study references are provided in the supplementary material. Searches for associated texts to assess exercise types (e.g. reports of protocols, pilots and development work) resulted in a total of 259 articles.

This analysis found that multi-component exercise interventions targeting older adults with multi-morbidity include a great variety of combinations of exercise types (22 different combinations). The extent of this heterogeneity has significant implications for meta-analyses of effects. Grouping studies under the umbrella ‘multi-component’ could be problematic, as it is clear from this review that the interventions being grouped together can be manifestly different.

Many of the exercise types combined and the volume of exercise delivered by interventions, did not meet WHO recommendations for older adults with chronic conditions and/or mobility problems (14). Whilst we accept that many studies were not intentionally targeting either older adults nor multi-morbidity, they frequently targeted a chronic condition (64%) and all included both older adults with multi-morbidity. Lack of consideration of, and adaptation for, age-related and comorbid conditions within single-condition studies is concerning. Particularly as the types of chronic diseases or age related deficits which were the focus of many studies included in this review, are associated with multi-morbidity and guidance exists to support clinicians and researchers to do this (25, 34, 35).

A recent review of behavioural interventions reported that, from a sample of 600 studies, 68.3% excluded people with multiple chronic conditions (36). A larger study aimed at quantifying the exclusion of individuals with certain characteristics from 43,895 clinical trials listed in ClinicalTrials.gov, reported that 40.5% of adults ≥60 years and 91.1% of people with multi-morbidity were excluded (37). Within the present analysis, 67 studies were judged to be based on previous literature or guidelines that would have drawn upon research with similar stringent exclusion criteria. Therefore, it follows that researchers are extrapolating efficacy of interventions which may be effective in a single-condition, to people with that condition plus additional co-morbidity/ multi-morbidity. Within settings where there is a long history of implementation and a wealth of research, like cardiac and pulmonary rehabilitation, there is evidence demonstrating the modifying effects of aging and multi-morbidity on efficacy (11, 38, 39). This has led many to call for a move away from ‘disease focussed’ to ‘person-focussed’ care (11).

The inclusion of multiple types of exercises within a single session necessitates consideration of the therapeutic dose being delivered. Whilst there are targets for overall activity (14), guidance is less clear regarding the amount of time that should be devoted to each component, particularly within a multi-component context (16). Across studies, these data could not be extracted due to insufficient description (only total time). However, with the exception of 5-component studies, there was no clear pattern of increased exercise time to accommodate additional components. Training time has been found to be an important modifier of effect in strength and aerobic training in older adults (40, 41).

Only seven studies (5%) included respiratory training. This number was lower than anticipated given the proportion of studies in the review that focussed on either cardiovascular or respiratory conditions (n=38, 27.5%). Within the categories cardiovascular or respiratory disease, chronic obstructive pulmonary disease and heart failure were the target diseases 36% and 54% of time respectively. Given that both diseases frequently result in symptoms of breathlessness; and considering the mounting evidence supporting the efficacy of this type of training in a broad range of cardiovascular (40) and respiratory diseases (41); the relative absence of this type of exercise was surprising. Breathlessness has been identified as a significant barrier to exercise in qualitative studies of older adults with multi-morbidity (43). Recognition of the potential wider role of respiratory training in older adults, out with respiratory disease, was evidenced in only one study in this review, which targeted falls (42).

Seventy-eight percent of studies were wholly or partly centre based, however very few studies described providing transportation support. Accessibility, proximity, convenience and transportation costs have also been identified as barriers to exercising for people with multi-morbidity (43, 44). Very few studies detailed socialisation type activities, either within the exercise intervention itself or as an adjunct. Reducing social isolation has been identified as an important priority for people with multi-morbidity (45). Both characteristics suggest that priorities that are important to people with multi-morbidity are not being considered when designing exercise interventions.

Exploration of the rationale behind selection of the exercise type may provide insight as to why studies that are including older adults with multi-morbidity are not making appropriate provisions when designing their intervention. Our assessment suggests that insufficient attention is directed towards selecting appropriate exercise type, with researchers often defaulting to generic guidelines or simply replicating previous ‘successful’ programmes.

Whilst these findings are more relevant to research practice, they do hold some clinical utility. Clinicians working with patients who experience multi-morbidity should be aware of the weaknesses in the evidence base identified by this review. Notably, that lack of consideration of the degree of heterogeneity within ‘multi-component’ designs and limiting testing of interventions that are cognisant of the barriers and or priorities of those with multi-morbidity. As guidelines synthesise current evidence, it follows that these weaknesses will be reflected within current recommendations.

Previous reviews (2, 3, 18, 19, 46, 47) have retrieved a far narrower pool of studies (range: 10–38) which we attribute to more exclusive search strategies, definitions of multi-morbidity or inclusion criteria. Most of these reviews have focussed on efficacy, therefore they do not offer a relevant comparison with the findings here. In their review of community based rehabilitation programmes for older adults with chronic conditions, Mulligan et al. (46) reported that most programmes focussed on education about disease management with few including ‘active’ exercise components; moreover, the choice of components was unclear.

Whilst this review has many strengths, notably its inclusive sampling in comparison to other reviews, there are some important limitations. Firstly, there is no universal definition of multi-morbidity and inclusion of different conditions will invariably enrich or dilute multi-morbidity prevalence (48). The broad criteria used here may have resulted in inclusion of studies without definitively multi-morbid samples. In an effort to keep the size of the review manageable, we excluded studies without a control group. Although these studies may have been relevant to the objectives of this review, given the number of studies included (n=138) we believe that their inclusion would be unlikely to meaningfully change our findings.

Much of the analysis was based on the presence or absence of certain information within a report. Therefore, if a study did not describe a particular criterion, it was noted as absent. As a result, some studies may have inadvertently been judged less favourably, purely as a result of journal imposed word counts and reporting standards. Lastly, studies published after the search date would not have been included. A re-run of our search in January 2024 identified four studies that would have met our inclusion criteria. A re-run of our search conducted in January 2024 in one of the databases, identified four studies that would have met our inclusion criteria.