Study characteristics
There were 6489 participants (mean age 74 years) across 23 studies. Eleven of the 23 studies were conducted in the US and Canada,18–28 eight in Asia,29–36 two in Europe,37,38 and two in the UK.39,40 Most participants were female. Eleven of the studies focused on general older adult populations, six on participants with heart disease or angina,29,30,36,38–40 two on those with diabetes,23,31 two on stroke survivors,21,37 and two on nursing home residents.34,35 Five studies focused specifically on older people with Medicare insurance.22,24,25,27,28 Two studies focused specifically on older people with frailty, although frailty was not defined.18,20 Although more than half the trials explicitly mentioned behaviour change theory, details varied widely.
Delivery settings included participants’ homes (n = 9),18–24,27,28 hospitals (n = 5),26,29,30,33,38 primary care practices (n = 4),24,25,32,39 and a nursing home (n = 1).34 Modes of intervention delivery included online, via telephone, and face-to-face. Eleven interventions18–20,24,25,27–30,36,38 used both face-to-face and telephone delivery. Almost half of the interventions (10/23) were delivered by nurses.18,24,25,28–30,32,38–40 Other delivery agents were GPs,39 occupational therapists,37 volunteers,20 and researchers.36
The majority of interventions aimed to improve self-care or self-management of a disease. Others aimed to improve participants’ independence in their homes;21,37 ability to carry out activities of daily living,37 or engagement in therapy.26 Some also aimed to reduce use of health services.24 Supplementary Table S1 contains detailed information on study characteristics.
Risk of bias assessment
Eleven of 23 studies scored low on the majority of the criteria for risk of bias.18,20,23,26,27,29–31,33–37 Generally, there was insufficient information on method of randomisation, allocation concealment, and blinding, but a high risk of bias was observed on ‘blinding of participants and personnel’.17 Only one study18 had a green risk assessment in this area. All other studies were either amber or red. The red, amber, and green assessment of risk for each of the criterion is shown in Supplementary Table S2.
Findings
Seventeen of the 23 studies reported statistically significant findings on one or more outcome measure between groups23,30,33,35 or a within-group difference over time.31 There were significant findings relating to mortality and disease-specific outcomes in five studies.2,27,30,31,36 Five studies demonstrated significant improvements in mental health outcomes.23,32,34,35,40 Five also showed significant improvements in behavioural outcomes such as physical activity or attendance at fitness classes,25,26,28,29,33 and five for QoL outcomes.23,30,31,33,35
Behaviour change techniques
Intervention groups
Forty-seven of the 93 BCTs in the taxonomy were reported in the intervention groups. Supplementary Table S3 summarises the BCTs used in intervention and control groups.
Control groups
Twelve BCTs were identified in the control groups, with the most common ones being ‘social support’ (practical), ‘information about health consequences’, ‘credible source’, and ‘pharmacological support’.
Intervention functions
Six of the nine IFs were coded. Table 2 shows ‘persuasion’ coded for all interventions. ‘Enablement’ and ‘education’ were also prevalent. Most studies satisfied more than one IF. The mean number of IFs per study was three. ‘Incentivisation’, ‘coercion’, and ‘restriction’ were not coded. Many interventions provided some form of lifestyle information or education. For more detailed intervention descriptions, including significance, see Supplementary Table S4.
Table 2. Frequency of intervention functions
Promising behaviour change techniques in trials reporting quality of life outcomes
There were 11 trials that included QoL as an outcome. Of these, five trials reported significant improvements for QoL, either between groups,23,30,31,33,35 or within groups over time.31 One of the significant QoL interventions was web based.23 From these five trials, the authors identified 11 ‘promising’ BCTs. Of the 11 promising BCTs, six were present in all five of the trials. These were: ‘goal setting’, ‘action planning’, ‘problem solving’, ‘social support’, ‘instructions on how to perform a behaviour’, and ‘information on health consequences’. ‘Goal setting’, divided into ‘goal setting (behaviour)’ and ‘goal setting (outcome)’, aligned with the BCT taxonomy. ‘Unspecified social support’, usually including advice, encouragement, or coaching, was also a promising BCT; emotional social support was rarely identified. Examples of all 11 BCTs are shown in Table 2.