Study designs.

All reviews (n = 19) were descriptive with eight also including meta-analyses [31,32,34,35,37,38,42,43]. The included reviews reported on 545 studies in 604 papers. However, some studies were included in more than one review, thus the number of unique studies reported was 413 (an overlap of around 24% of studies). Notwithstanding this overlap, all identified reviews were included in the analysis due to their unique focus, methods of analyses and perspectives on the findings from the same studies [48]. Two particular reviews [37,42] originating from one research centre were based on a nearly identical set of identified studies, but both were included due to their different focus (effectiveness of PA promotion and characteristics of successful interventions–both of interest to this review). Additionally, despite a significant overlap in studies within two other reviews [32,33], both were included, as their aims differed and they were together providing complementary information.

The reviews generally included empirical studies, mainly randomised controlled trials (RCTs, 339 studies) and the remainder were quasi-experimental or pre-post studies. Reviews included between 6 and 79 studies, with a mean of 17.8 RCTs per review. One review, though not a review of reviews, also included some evidence coming from systematic reviews and meta-analyses [29] and two reviews did not specify the type of studies included [31,47].

Participants.

Total numbers of participants were reported in 18 reviews and ranged from 225 to nearly 60,000 per review. Seventeen reviews specifically concerned older people and two focused on adults in general [34,38] although most participants were older adults. The age of participants in the 17 reviews which included only older adults ranged from 40 to 91 years, with mean age ranging from 59.8 to 79 years (reported by 11 reviews).

Of the 14 reviews which specified gender of the participants, nine reported larger numbers of female participants (at least 60% women) [30,31,32,37,41,42,45,46,47], which is representative of a gender split in the older population. Seven reviews mentioned ethnicity [30,31,32,33,34,44,46] and reported that participants were primarily white with a small proportion of ethnic minorities. The majority of reviews either specifically focused on participants with no pre-existing medical condition or did not specify any medical conditions, with the exception of three reviews [32,33,43], which also included patient populations (Table 3).

Context/setting.

All included reviews focused on older adults living in the community. However, interventions were delivered in diverse settings and included: participants’ homes, general practice and occupational therapy service (Table 3). Most reviews included studies conducted in a range of settings. (In this review home-based interventions refer to instruction or prescription for participants to practice PA at or around home and not to personal supported training delivered by professionals at participants’ homes.)

Eleven reviews reported the country of research, indicating that a vast majority of studies were conducted in developed high-income countries (including USA, UK, Canada, Australia, Japan, China, New Zealand, Taiwan, The Netherlands, Belgium, Italy and Finland).

Aim of the reviews.

Fourteen reviews aimed to assess the effectiveness of interventions used to increase PA among older adults living in the community. Some of the reviews additionally applied a more specific focus on either the type of intervention (e.g.: remote feedback [36], walking in groups [38], non-face-to-face interventions [41], health coaching [43]); the length of the intervention (e.g. focusing on long-term effects [37,42]); or specific setting (e.g. interventions delivered through a GP service [45], within occupational therapy [29], or in rural or regional areas [40]). Three reviews were concerned with characteristics of interventions or determinants of change [35,40,44]. One review focused specifically on the time around retirement [30] and another on levels of participation in PA programmes [47].

Modes of delivery.

Due to heterogeneity and a lack of common approach to categorising interventions in the reviews, they are described here according to the prevailing modes of delivery (see also Table 4). Although interventions could broadly be categorised as face-to-face and non-face-to-face, these categories should be considered a guide only, as they were not always exclusive due to the complexity of the interventions (e.g. some of the remote interventions include an element of direct contact, while educational materials used in many of the face-to-face interventions may be considered remote modes of delivery). The interventions frequently incorporated lifestyle counselling and health education elements and typically took the form of a face-to-face group/individual counselling or training session followed by a scheduled remote contact to encourage further involvement in PA.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Interventions grouped according to face-to-face or remote modes of delivery.

https://doi.org/10.1371/journal.pone.0180902.t004

Face-to-face interventions only: Two reviews included interventions delivered through face-to-face contact only [38,40]; one review [38] focused on interventions to promote walking and another [40] on low- to moderate-intensity exercises and low-impact PA (including walking). Interventions in both reviews were predominantly group-based and consisted of multiple components, including PA itself, as well as education, discussion and counselling elements.

Remote interventions only: Four reviews focused entirely on non-face-to-face interventions [34,36,41,44]. Nearly all remote interventions were directed at individuals, instructing them to engage in self-selected PA or a guided exercise programme in their own homes or community. These interventions were primarily based on feedback provided with largely differing frequencies [36,44] and through a variety of modes, including telephone, internet and standard mail. Phone and print materials were the most frequently used modes of delivery in all four reviews, but internet interventions were also emerging [41].

Although remote approaches to PA promotion were the stated subject of the four reviews, three teams of authors [34,36,44] recognised that the primarily non-face-to-face interventions were likely to include an element of direct contact, commonly individual consultation, counselling or introductory session, usually at the beginning of the programme. For example, one review [34] included interventions which had a face-to-face element (the ‘primary dose of intervention’) and the principal remote ‘follow up dose’ (subsequent motivational phone calls or mailing of materials).

Remote interventions often included an educational element–either involving general PA information sent to participants or feedback-based tailored information to provide individualised reports/ leaflets/ exercise plans. A counselling element was also present, e.g. [44].

Combination of face-to-face and remote interventions: Thirteen reviews were concerned with either both face-to-face and remote modes of delivery or with complex multimodal interventions [29,30,31,32,33,35,37,39,42,43,45,46,47].

Interventions typically included multiple components and frequently combined education with lifestyle counselling and PA or exercise programme. Two reviews [46,47] categorised interventions as either home-based, group-based or education. The ‘home’ and ‘education’ modalities could be delivered either face-to-face or remotely.

Some authors reported on multimodal delivery format in most included studies [37], others considered PA promotion a part of larger health management and maintenance programmes [29], and some focused on tailored interventions, for which individualised referrals were made through the GP practice [45].

Five reviews focused primarily on BCTs and were not concerned with particular modes of delivery at the study selection stage [31,32,33,35,42] but included diverse multi-modal and multi-component interventions. Other reviews also frequently mentioned the use of BCTs as essential components of the interventions included.

Theoretical frameworks.

Ten reviews reported on theoretical frameworks purported to underpin the interventions [30,32,33,36,39,40,41,42,43,44]. The most common theoretical framework was Social Cognitive Theory (SCT), mentioned in nine reviews [30,32,33,36,39,41,42,43,44] and reported to have been used in at least 44 studies. The Transtheoretical Model (TTM) was the second most common framework, featuring in six reviews [32,33,41,42,44,43] and at least 21 studies.

Intervention components: BCTs.

Elements of BCTs were present in all of the discussed interventions and were particularly highlighted in five reviews with a specific focus on BCTs [31,32,33,35,42]. In two reviews the CALO-RE taxonomy [49] was used to code BCTs [35,42]. The five mentioned reviews reported effects of the BCT-based interventions on PA behaviour and one review additionally examined their impact on self-efficacy [35]. Of the five BCT-focused reviews, four performed meta-analyses [31,32,35,42] and their detailed outcomes as well as frequency with which the BCTs were used will be discussed in the outcomes section. One BCT-focused review [33], which did not perform meta-analyses, identified self-monitoring and health education as the most commonly used interventions, followed by goal setting, problem solving, feedback, reinforcement/contingencies, relapse prevention education, and modelling.

Seven other reviews, although not directly concerned with BCTs, reported on some of the techniques used in included interventions. The most commonly reported intervention components were: goal setting [36,37,39,40,41,43,44], self-monitoring [37,39,41,43,44], tailoring [36,41,44], feedback [36,37,39,43], relapse prevention training [39,44], strategies to increase motivation [36,41,43,44] and self-efficacy [36,43,44].

Types of physical activity.

Eleven reviews did not specify the types of PA promoted [29,30,31,32,34,37,41,42,43,44,45]. In the remaining eight, walking was the most frequently cited form of PA [33,35,36,39,40,46,47] and was the specific focus of one review [38]. In three reviews [33,35,36] at least one third of all included studies were concerned with walking as the main type of PA. Six reviews included other lifestyle PA (e.g. gardening [35], dance [39,46]) and, most commonly, low to moderate intensity activities, for example, a home-based exercise programme [36,39,46,47], tai chi [39,40], aquatic exercises [40] and yoga [46]. Although less vigorous activities were generally most common, reviews also mentioned aerobic-type exercise including jogging [35,36,39,46].

Some reviews made a distinction between interventions with a ‘prescribed’ or structured format and interventions in which participants were able to self-select their preferred PA type. In only one review [38] the type of PA was clearly defined for all studies and five reviews included both prescribed and self-select activities [32,34,36,40,45].

Intervention providers.

Just as interventions varied significantly, their providers included a range of professionals within the fields of: healthcare (GPs, nurses, physicians, occupational therapists, physiotherapists, health visitors), PA (exercise counsellors, PA coaches, health and fitness professionals, community exercise instructors, certified exercise trainers) and education (trained health educator, physical educator). Some interventions were reported to have been facilitated by a trained member of a research team [35,37,43] or a graduate student [33,46] and there was also mention of an intervention offered through an ‘automated computer’ [34]. As expected in interventions offered largely in the community and focusing on PA promotion rather than exercise classes, most reviews reported on participants taking an active role in the actual ‘delivery’ of the intervention. Elements of self-help and/or self-administered exercise under instruction [37,42] were present in many if not the majority of the studies. In some cases, PA was encouraged or sessions facilitated by peer mentors or lay leaders [30,33,35,38].

Intervention tailoring.

Six reviews provided information on intervention tailoring [29,33,37,41,44,45], with the majority of interventions individualised to some degree. One review [37], for example, stated that some interventions provided participants with tailored exercise prescriptions, while others (educational interventions) offered information specific to the individual, taking into account their health, environment, opportunities and goals. The remaining thirteen reviews did not provide details sufficient to assess the presence or degree of tailoring.

Dose and duration.

Reported interventions lasted between four weeks and three years, with durations of between three months and 12 months being most common. Shorter interventions of between four and 12 weeks were slightly less common, and duration of over 12 months was significantly less common. The longest reported intervention lasted for 90 months and was an ongoing PA programme [46], while the shortest intervention lasted for one day only and involved four different text messages sent to participants over the course of a single day [44].

The frequency of sessions varied largely between different types of interventions and within similar intervention type or component. In interventions which included face-to-face educational and lifestyle counselling component (e.g. lectures, motivational sessions), sessions were commonly offered on a weekly basis [29,33]. In interventions which actually monitored or at least recommended activity, participants were asked to engage in PA between one and seven times per week, with two to five times per week being the most common frequency of activity prescription [29,33,36,38,45,46]. Some reviews reported on interventions which incorporated daily activity [36,38], while the least frequent intervention reported activity once a month for a relatively longer period of time (90 to 120 minutes) than in the case of more frequent sessions [38]. In most interventions activity sessions lasted for between 10 and 90 minutes, with a range of 20 to 60 minutes being most common. Some interventions began with shorter sessions (e.g. 10–15 minutes) and eventually progressed to being active for longer periods of time (e.g. 45 or 60 minutes).

In case of remote feedback-based interventions (i.e. phone calls and text messages) there was a substantial variability across studies, with participants receiving messages and/or calls at either regular or varied intervals [44]. The frequency of contacts could be anything between three times per day (e.g. [44]) to once every two months (e.g. [45]). However, there seemed to be a common pattern of contacts being more frequent at the beginning of the intervention (often weekly for the first month) and maintained through less frequent calls/messages (biweekly/monthly) over the course of the intervention (e.g. [34]). Overall number of messages or phone calls was rarely reported in the reviews and ranged between three and 18 contacts [30,41].

Three reviews concerned with primarily with multimodal interventions [31,37,42] reported that number of sessions or contacts within an intervention ranged from one to 228 sessions and averaged between 15 and 37 sessions per intervention.

Reviews including meta-analyses: PA levels.

All of the reviews’ authors concluded that interventions aiming to increase PA were effective under differing conditions and with differing effect sizes observed (see Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Physical activity levels in reviews with meta-analyses, including sub-group analyses for short/long term outcomes.

https://doi.org/10.1371/journal.pone.0180902.t005

The review which focused on walking interventions [38] reported a medium-sized positive effect (d = 0.52) on PA and the largest pooled effect reported by any of the included reviews. This effect was also moderate for short term outcomes up to six months (d = 0.45) and even larger for outcomes measured beyond six months (d = 0.66). The difference between the two sub-groups was significant (p = 0.0004). In another review comparing short and longer term outcomes [32], multi-modal interventions were effective in both groups, but the effect size was moderate for outcomes up to 90 days (d = 0.42) and only small for outcomes over 180 days (d = 0.22), with a group difference significant at p < 0.05 level.

One review [37] reported positive small to moderate effects of multi-modal interventions on outcomes measured at 12 months (d = 0.19, p < 0.0001), but no significant effects at 24 months, while another review, focused on remote interventions [34], calculated comparably small effects for outcomes at 12 and 24 months (d = 0.20, for both conditions). In reviews which either did not specify times at measurement, took post-intervention measurement only or calculated combined effects [31,32,35,38,42,43], effects were positive and small to moderate (SMD varying between 0.14 to 0.52), suggesting that the multi-modal interventions examined were found to be effective at promoting PA behaviour among older adults.

Reviews including meta-analyses: Psychological outcomes.

One review [35] calculated the pooled effect of 24 studies with multi-modal interventions including BCTs on self-efficacy. A moderate positive effect (d = 0.37) was found (see Table 6). Another review [43] found no evidence of a differential effect of health coaching in improving quality of life (8 studies, d = 0.07) or mood (5 studies, d = 0.02) based on pooled calculations of measurements taken immediately after the intervention.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Psychological outcomes (self-efficacy levels) in a review with meta-analyses.

https://doi.org/10.1371/journal.pone.0180902.t006

Reviews including moderator analyses: Intervention characteristics.

Six reviews [31,32,34,38,42,43] performed moderator analyses related to the features of interventions examined and their impact on PA levels. These are categorised in the current review according to the characteristics of the participants, intervention providers, the setting and mode of intervention (see Table 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Moderator analyses: Intervention characteristics—participants, interventionist, setting, mode, other characteristics.

https://doi.org/10.1371/journal.pone.0180902.t007

Participants: In one review [31] interventions tested among healthy subjects were found to be more effective than those aimed at chronically ill participants. In contrast, another review [32] found significantly larger effect sizes for patient populations (cardiac problems, arthritis and diabetes) than for healthy participants and another review [43] found no differentiating effect of health conditions. In another paper [38] larger effect sizes were found for interventions directed at mixed gender groups than at women participants only and at adults aged over 60 years than at younger adults.

Intervention provider: In one review, interventions which featured higher levels of programme intensity (intense contacts with intervention providers) had higher effect sizes than those that featured relatively less intensive contact [32]. Three reviews calculated that interventions delivered by non-professionals had similar effect sizes to interventions facilitated by professionals, including health practitioners and exercise specialists [31,34,38].

Setting: One review found that studies testing centre-based activity had larger positive effect sizes than studies promoting home-based activity [32]. In another two reviews neither types of settings [31,42] nor their number [31] appeared to affect PA levels.

Mode of delivery: In two reviews the use of printed and mailed materials led to larger effect sizes compared to not using this delivery mechanism, but the effect was positive in one review [31] and negative in another [42]. Interventions that employed audio-visual media had a more positive effect on PA levels than those that did not use this form of delivery [31]. Using other modes, including written materials, discussion, lectures and phones led to similar effect sizes as not using these methods in three reviews [31,32,42]. Similarly, no difference was found in interventions which were and were not delivered face-to-face [31]. However, review that compared telephone and face to face delivery [43] found that, while both were effective, face to face interventions led to stronger positive effects. Of the two reviews that examined the effect of interventions offered to groups and individual people, one indicated that interventions directed at individuals resulted in smaller effect sizes than interventions delivered to groups [31], the other found no differences in effect [32].

Other intervention characteristics: One review which examined the effects of PA intensity [32] concluded that studies recommending moderate-intensity PA recorded significantly larger effect sizes than those recommending low-intensity PA. Additionally, interventions that made any recommendation on intensity levels had larger positive effects than interventions without an intensity recommendation. In the same review studies focusing on PA promotion specifically had higher effect sizes than studies designed to change multiple health behaviours [32]. Two reviews examined the impact of recommended PA activity type [32,34]: recommended walking was not found to lead to statistically different effect sizes compared to no specific recommendation [32] and there was no statistical difference between interventions that specified PA and those that allowed participants a choice [34]. Similarly, no difference in effects was reported for studies that generated their PA prescriptions by computers and by humans and for studies that used and did not use pedometers as part of intervention [34].

In addition to the described moderator analyses of dichotomous variables, one review [31] performed analyses of continuous variables, including: duration of intervention, length and number of sessions, participant baseline BMI and age and percentage of women. The calculations concluded that the interventions were similarly effective regardless of the different variables examined.

Reviews including moderator analyses: BCTs.

Four reviews [31,32,35,42] examined the impact of BCTs on PA levels. Two of these [35,42] used CALO-RE taxonomy [49], enabling direct comparisons. Other reviews did not specify the background for the adopted nomenclatures. Therefore, the current review attempted to categorise the BCTs included in moderator analyses of all four reviews for a clearer and more systematic report (see Table 8).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Moderator analyses: Behaviour change techniques–BCTs.

https://doi.org/10.1371/journal.pone.0180902.t008

Forty-four of the total 66 comparisons in the reviews resulted in non-significant effects. We identified 51 different BCTs. Fifteen of them were mentioned in more than one review. Effects of the 15 BCTs for which more than one comparison was available were predominantly inconsistent, with one case of a contradicting result (for BCT ‘feedback’) and only one case of significant effect confirmed by two reviews (for BCT ‘Provide information on consequences of behaviour to the individual’).

Theory and intervention type: Of the two reviews which looked at the use of theory [31,32], one reported that theory-based interventions had larger effect sizes than interventions without a stated theoretical basis [31], the other found no link between the use of a theoretical framework and effectiveness [32]. While in one review motivational-type interventions were reported to have been more effective than interventions without motivational element [31], another review found studies with and without motivational interviewing to result in similar effect sizes [35]. Two reviews examined the effects of behavioural- and cognitive-type interventions [31,32]: while single type of intervention was not associated with differences in effect sizes, interventions employing a combination of cognitive and behavioural strategies were found to be more effective [31].

Feedback: The impact of feedback was calculated in three reviews with largely differing results [31,35,42]. The two reviews which used CALO-RE taxonomy [35,42] found that providing feedback on performance led to significant differences in effect sizes, but the effect was positive in one review [42] and negative in another [35]. In the third review feedback was not associated with larger effect sizes [31]. One review did not perform calculations related to feedback, but, based on characteristics of included studies, indicated its positive impact on PA outcomes when combined with frequent and focused follow-up [34].

Self-monitoring: All four reviews examined the role of self-monitoring. Two of the reviews reported larger effect sizes of interventions that used self-monitoring than those that did not: in one review this effect was positive [32] and in the other the effect was negative [35].

Modelling: The presence or absence of social or role modelling was not associated with statistically significant differences in effect sizes in two reviews [31,32]. However, the BCTs ‘Model/demonstrate behaviour’ and ‘Provide normative information about others’ behaviour’ led to larger effect sizes in another review [35].

Social support: Social support was not associated with particular effects in two reviews [32,42], but was found to have a negative effect on PA levels in another review [35]. One review did not perform relevant calculations, but highlighted the observation that social support might be an effective component of successful walking interventions [38].

Goal setting: Three reviews [31,35,42] assessed the impact of goal setting and found no statistical difference in effect sizes in interventions which employed these strategies.

Problem solving: Problem solving techniques and barriers management were found to be effective in changing PA behaviour in two reviews [31,35] and no effect was recorded in another [42].

Education: All four reviews assessed the impact of specific educational practices on PA behaviour change and aspects of health education were found to be negatively associated with PA levels in three reviews [32,35,42]. Specifically, using BCTs ‘Provide information on consequences of behaviour to the individual’ and ‘Provide information on where and when to perform the behaviour’ led to larger negative effect sizes in two reviews [35,42].

Prompting: Of three reviews that examined the use of prompts in PA promotion, two did not find significant effects [31,42] and one reported negative effects of prompting on PA levels [35].

Counselling: Stress management techniques and counselling were not associated with larger effect sizes in two reviews that examined these BCTs [31,35].

Other BCTs: In one review [35] two other BCTs were associated with larger negative effect sizes: ‘Action planning’ and ‘Relapse prevention/coping planning’, and one BCT led to positively larger effect sizes: ‘Provide rewards contingent on successful behaviour’. Some other BCTs which were not associated with larger effect sizes are listed in Table 8.

Narrative reviews: PA outcomes.

All narrative reviews reported some kind of PA outcomes (see Table 9). Most often these were self-reported PA levels, assessed through questionnaires (e.g. CHAMPS, PASE, 7-day PA Recall) and/or activity logs. Three reviews [29,40,44] reported changes in PA levels measured objectively through pedometer or accelerometer and in two reviews it was unclear how the PA levels were measured [46,47]. At least 127 studies, examining interventions of duration between one and 36 months, reported positive outcomes and no change was recorded in at least 26 studies. Three reviews [33,36,40] reported on changes in walking time, speed or distance specifically, based on walking logs (13 studies reported increase of PA and no change was observed in one study).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Outcomes of interest and timings in the 11 narrative reviews, for which information was available.

https://doi.org/10.1371/journal.pone.0180902.t009

Narrative reviews: Psychological outcomes.

Seven reviews [29,30,34,39,40,44,45] reported on psychological outcomes pertaining to PA, including: self-efficacy, quality of life, and depressive symptoms.

Self-efficacy was reported in six reviews [29,30,39,40,44,45]. Improvements in self-efficacy were noted in eleven studies measuring effects at 4 months to 2 years [29,30,39,40,45], however this effect was not sustained in two studies beyond 12 months and 3 years [29]. One review reported decrease in self-efficacy at 12 weeks in one study [44].

Quality of life was equally often reported outcome [29,30,34,39,40,45], usually measured with the use of SF-36 questionnaire, consisting of sub-scales primarily relating to health and functional outcomes. Improvements in quality of life, based on the SF-36 scale, were noted in nine studies at times varying between 11 weeks and 12 months [29,30,34,39,40,45]. No difference was observed in three studies [29,34,45] and one study reported a decrease in quality of life at 12 months [45]. One review [39] reported improvements in all SF-36 sub-scales and improvements in the vitality sub-scale were noted by four other reviews [29,30,40,45]. Other sub-scales with improved scores included: physical functioning, social functioning, mental health [29,30,34,40], general health perceptions [29,34,45], bodily pain and energy/fatigue [29,30,34].

Outcomes related to depression and general ‘mental outlook’ were reported by three reviews [29,30,40]. Decreased depressive symptoms were noted in three studies at six and 12 months [29] and in one study no change was observed at 10 weeks [40].

Narrative reviews: Participation and adherence.

Initial and sustained participation in PA or adherence to interventions promoting PA were reported in five reviews [29,36,39,46,47]. Participation rates were reported in at least 40 studies included in these reviews and varied substantially from 36% to 100%. Some reviews reported the mean participation rate, which ranged from 72% to 79.2%.

Narrative reviews: Intervention characteristics.

The majority of narrative reviews attempted to establish the characteristics of most effective interventions and those that do not seem to affect outcomes. Four reviews highlighted the importance of tailoring the interventions to participants [29,37,40,44], particularly by environmental mediators, personal readiness and interests and type of activity available locally [37,44]. Environmental suggestions may include tailored information about opportunities in the local environment [37], for example maps of walking or cycling routes, information about upcoming events, neighbourhood gyms and exercise that could be done at home [44]; it seems particularly important that these environmental mediators match individual interests of older adults [44].

Mode of delivery did not appear to be a differentiating factor [37,44], as both face-to-face [40] and remote interventions [41] and group and individual delivery [33] led to generally positive outcomes. However, one review indicated that a supervised home-based format, or a combination of group- and home-based formats could be more effective than a class or group format only [39]–finding not consistent with results of meta-analyses discussed earlier. One review highlighted particular effectiveness of interventions promoting low- to moderate- intensity activity [40], while another concluded that outcomes were unrelated to intensity of frequency of PA [46]. Frequency of intervention (meaning number of intervention contacts) did not seem to affect PA engagement in another review [37].

Behavioural or cognitive-behavioural strategies were reported to have led to more positive outcomes than health education or instruction alone [39]. Interventions using remote feedback were more effective than treatment as usual and equally effective as supervised exercise without feedback [36]. Positive impact of having an activity partner if preferred [44] was also highlighted.