A hybrid type 1 trial of a digital behaviour change intervention for frailty prevention in community-dwelling older adults aged 60 years and older

Abstract

Aims

This hybrid type 1 randomised controlled trial evaluated the effectiveness and implementation of a digital behaviour change intervention (DBCI) designed to empower older adults (60+) who are nonfrail and community-dwelling to adopt activity, vaccination, optimising medication, interaction and socialisation, diet and nutrition (AVOID) behaviours to prevent frailty.

Methods

Sixty adults, from one South Australian local council, were randomised to receive the DBCI (n = 31) or wait (control; n = 29) for 6 months to receive limited DBCI access. Repeated-measures mixed models were used to assess between-group differences in changes in the primary outcome, the frailty index (FI), and the secondary outcome, quality-of-life (QoL; EQ-5D-5L) scores, from baseline to 6 months. Content analysis of survey and focus group data from the intervention group assessed technology acceptance, perceived knowledge and behaviour change. DBCI behaviour change techniques included baseline health and three-month behaviour change reports, education, a library of community resources, goal-setting and action planning and gamification and nudging. Participants in the intervention group received newsletters and were invited to face-to-face expert talks.

Results

The median age was 75.4 years, and 63.3% of participants were female. In the intervention group, the mean FI score decreased from 0.135 (SD 0.053) to 0.115 (SD 0.071) over 6 months, while in the control group it increased from 0.134 (SD 0.048) to 0.160 (SD 0.087). The intervention group had a mean FI change of −0.044 [95% confidence interval (CI) −0.076 to −0.012] and a mean QoL change of 0.032 (95% CI 0.016 to 0.058) compared with the control group. Qualitative feedback indicated that the DBCI was acceptable, improved knowledge and supported behaviour change across AVOID behaviours.

Conclusion

A DBCI promoting the adoption of AVOID behaviours among nonfrail, community-dwelling older people is effective in preventing frailty and improving QoL when implemented. Solving technical barriers and incorporating intelligent technologies could enhance impact.

Key Points.

• A digital intervention implementing behaviour change techniques empowers older adults to AVOID frailty.

• The DBCI was usable, acceptable and impactful in supporting behaviour change.

• The DBCI was effective in reversing frailty risk and improving quality of life when compared to the control group.

• Findings have relevance for the use of similar digital approaches to support healthy ageing. Increasing digital literacy, greater acceptance of technology and advancing technology will create opportunities for scale-up.

Introduction

Frailty becomes more common as people age. The international prevalence is 4% in individuals aged 65–69 years, increasing to 16% in individuals aged 80–84 years [1]. One in four people aged 85 years or older is frail [1], experiencing loss of muscle mass and strength, decreased mobility and endurance and reduced activity. These changes inevitably result in poor health outcomes, characterised by loss of independence, falls, hospitalisation and a poor overall quality of life (QoL) [2].

It is possible to reduce frailty risk by intervening before deficits aggregate to the point that reversibility is lost. A recent review identified the need for implementation research focused on health promotion strategies, including those leveraging technology, to prevent frailty in those not yet frail and to delay progression to frailty. Reducing frailty risk is a research priority identified by community-dwelling older people [3].

A synthesis of evidence from 12 guidelines (including the 2017 Asia Pacific [4] and the 2019 International Conference of Frailty and Sarcopenia Research guidelines [5], five best practices and four expert consensuses) confirmed that personalised multicomponent interventions are effective for reducing frailty risk. These interventions typically include risk assessment, health education, physical activity, healthy nutrition, social opportunities and medication management [6]. While studies have reported on the effectiveness of frailty prevention interventions [7], few interventions have leveraged technology for this purpose or examined their implementation.

Digital behaviour change interventions (DBCIs) leverage technologies and strategies to support behaviour change to improve health. DBCIs support self-management and behaviour change, eventuating in habit formation [8]. A recent multicomponent DBCI focused on physical activity, nutrition and social support was shown to reduce psychological distress among Australian adults aged ≥60 years [9]. However, a study is required to determine the utility of DBCIs in preventing frailty.

This project utilises the DBCI (https://healthyagingcentres.ca/) developed by the Canadian Frailty Network (CFN) in partnership with the community to facilitate behaviour change in older people through the adoption of the healthy lifestyles of activity, vaccination, optimising medications, interaction and socialisation, diet and nutrition (AVOID) [10]. The Canadian DBCI was implemented in four cities across Canada. However, it was unknown whether the AVOID DBCI would be effective and feasible for implementation outside Canada.

Therefore, in partnership with one South Australian metropolitan local council (with a population similar to the Canadian trial cities) and its citizens, we adapted the CFN DBCI for implementation. As part of a hybrid type 1 randomised controlled study aligned with the integrated, design, assess and share (IDEAS) framework [11] for the development of effective DBCIs, we tested intervention effectiveness using a wait-list, parallel arm, single-blinded, randomised controlled trial (RCT) on the frailty index (FI) and QoL. We also evaluated the implementation of the DBCI for feasibility and scale-up readiness, focusing on outcomes such as reach, engagement, technology acceptance, impact and maintenance.

Methods

Standards for reporting implementation studies [12] were followed. The trial protocol was prospectively registered at Open Science Framework on 19 December 2024 (identifier: DOI 10.17605/OSF.IO/J2YSU), off embargo on the 30 March 2025, and updated on 13 August 2025 and 12 October 2025.

Design

A two-arm, single-blind RCT was undertaken, with participants randomised to either the intervention (DBCI) or the control (usual care).

Participants were not blinded to the intervention. An independent statistician generated a variable-block schedule and programmed it into REDCap after baseline data collection, ensuring that research staff remained blinded. Trial outcome assessments at 3 and 6 months were conducted via REDCap-generated email links or by telephone interviews. Research staff conducting the interviews were blinded to group allocation. Participants could choose their preferred assessment method. They were asked not to discuss allocations during phone contact with staff. No unintentional unblinding occurred.

At the start, two participants continued after being inadvertently assigned to the intervention group outside the randomisation process. Analyses included all participants. Additionally, for outcome effectiveness analyses, results with the two non-randomised participants excluded are also presented.

Context/setting

The study enrolled community-dwelling people aged ≥60 years who were not frail (FI score ≤ 0.21) and living within the City of Charles Sturt Council, a metropolitan council with a population of 129 196 people, of whom a quarter were aged ≥60 years.

Other inclusion criteria were the ability to use a web platform or have someone assist them, regular access to a computer and the internet, the ability to read, understand and converse in English, and being the only person in the household participating in the study. Individuals involved in the earlier platform adaptation phase were excluded.

Recruitment and baseline assessment

The study was promoted to community groups through ‘pop-up talks’ and dissemination of flyers to the community, including through brief articles in partner newsletters.

The flow of study assessment is presented in Appendix 1. Recruiting began on 15 January 2025 and ceased on 28 March 2025. Staff sought verbal consent to confirm eligibility (age, living circumstances and location) and to explain the research. The participant information sheet and two consent forms were posted to interested and eligible participants. Participants were contacted a week later to address any queries. Those who consented returned a signed consent form via a reply-paid envelope and retained a second signed copy. One follow-up phone call was made at 2 weeks if the consent form had not been returned.

A modified 36-item self-assessed report of personal capacity and healthy ageing (SEARCH [13]; Appendix 2) questionnaire was administered by phone to consenting participants. SEARCH assesses across the domains of locomotion, nutrition and oral health, cognition, psychology, sensory, continence, sleep, chronic diseases, pain and medications, for which suggested actions (Table 1) are provided as part of the health assessment report. Activities of daily living are not included. In this study, SEARCH was modified to remove two questions on mental health and fatigue, and one on walking speed. We also used the EQ-5D-5L visual analogue scale score in SEARCH to reduce survey burden. In accordance with the methods by Searle [14], the FI score (range 0–1) was derived by dividing the number of deficits by the number of deficits assessed. The FI method was chosen because it supports tracking frailty over time and may be more sensitive to interventions than the phenotypic method [15].

Table 1.

Behaviour change techniques used within the intervention platform.

COM-Ba domains	Function	Behaviour change techniques (version 1)	Description
	User interface		Best viewed via desktop or laptop. To support engagement.
Capability	Health assessment reports	2.2 Feedback on behaviour 2.7 Feedback on outcomes of behaviour 1.6 Discrepancy between current behaviour and goal 5.1 Information about health consequences	Baseline health assessment report outlining risk domains and suggested action. Mini report at 3 months presenting information about behaviour change achieved.
			Rationale—increases awareness and understanding of personal health risks to motivate towards action and support reflection on progress.
Capability	Education	4.1 Instruction on how to perform behaviour 5.1 Information about health consequences 5.3 Information about social and environmental consequences 6.1 Demonstration of behaviour	Education included videos, printables and links to relevant publicly available educational resources. Topics with video modules developed for AVOID DBCI were ingredients for change, activity, vaccination, optimise medications, interact, diet and nutrition, sleep and falls prevention, ranging from 11 to 21 minutes to view (excluding quizzes). Other topics included chronic disease, memory, pain and urinary incontinence.
			Rationale—improves knowledge and skills related to AVOID behaviours, increasing the capability to act.
Opportunity	Community resources library	3.1 Social support (unspecified) 3.2 Social support (practical) 3.3 Social support (emotional)	Signposting—directory of local programs, coaching, care navigators, and crowdsourced information.
			Rationale- Signposting towards available resources that can enable behaviour change.
Motivation	Goal setting, Action planning and tracking progress	1.1 Goal setting (behaviour) 1.3 Goal setting (outcome) 1.4 Action planning 1.5 Review behaviour goals 2.3 Self-monitoring of behaviour	A tool to set simple, measurable, achievable, relevant, time-bound (SMART) goals and action plans and tracking progress.
			Rationale—a process to support commitment and planning towards action, mitigating barriers and monitoring progress.
Motivation	Gamification	10.3 Nonspecific reward 2.2 Feedback on behaviour	Points are earned for platform use, with highest participant points displayed next to participant points.
			Rationale—to motivate and support habit formation.
Motivation	Nudging	7.1 Prompts/cues	E-mail prompts to encourage platform use where participants did not log in for ≥28 days. Up to three nudges over 6 months.
			Rationale—supports action without deliberate reflection.

^aBehaviours relating to activity, vaccination, optimising medication, interaction and socialisation, diet and nutrition

Those scoring >0.21 [16] on the FI were ineligible to participate in the study. They were posted a copy of their responses to the SEARCH questionnaire and encouraged to discuss this with their general practitioner.

Participants with an FI score ≤ 0.21 were considered not frail, and the following additional baseline data were collected during a face-to-face clinic assessment:

1. Additional demographic details—form of transportation, household income, living arrangement, education level and employment status.

2. Technology—use of technology to monitor health, wellbeing and comfort and use of technology in everyday life, including frequency of use and experience with internet connectivity issues.

3. Independence—basic activities of daily living (five items), instrumental activities of daily living (eight items) and use of an assistive device

4. Body mass index.

Sample size

Based on previous research [17, 18], the change in the intervention group to be detected was set to the minimally important difference in FI score of 0.11 (anchor-based method) for the wait-list trial. Assuming a correlation of 0.5 between the reported measures in our study, with 90% power and an error variance of 0.008, the required sample size was 20 per group. A final sample size of 30 per group allowed for 30% attrition between measurements.

Allocation

Digital behaviour change intervention (or intervention) group

Those allocated to the intervention group had access to the DBCI (TIDieR checklist, Appendix 3) for 6 months. The implementation strategy aligned with Proctor’s reporting guidelines (Appendix 4) [19]. The Behavioural Change Wheel [20] underpins the DBCI, which draws on the COM-B model (comprising three components—capability, opportunity and motivation-behaviour [20]) and the taxonomy of behaviour change techniques [21] to guide the development of the digital intervention, including behaviour change techniques (Table 1).

Intervention group participants were invited to five face-to-face expert talks on activity, nutrition, intrinsic capacity, sarcopenia and dementia, with attendance ranging from 7 to 11 participants per talk. Participants in the intervention group also received four newsletters over the study period. Simple modifications to the DBCI, based on intervention participant feedback during this study, are outlined in the TIDieR checklist (Appendix 3).

Wait-list (or control) group

This group did not have access to their health assessments and received 6 months of access to the DBCI education modules after completing their six-month assessments.

Outcomes

Randomised controlled trial effectiveness

Change in scores over 6 months

Primary outcome

• FI scores.

Secondary outcome

• QoL assessed using the EQ-5D-5L [22], assessing domains of mobility, self-care, usual activities, pain/discomfort and anxiety/depression and with five response categories of no problems, slight problems, moderate problems, severe problems and extreme problems. The Australian value set was used to score EQ-5D-5L using Stata V15.1 [23].

Implementation outcomes

Table 2, aligned with Proctor’s framework [24], describes implementation outcomes, measurements, time points and study groups investigated, with additional methods provided below.

Table 2.

Description of implementation outcomes.

Implementation outcomes		Description	Measurement	Time point	Group investigated
Study reach and feasibility		Of the interested participants, the number eligible. Representativeness of participants of the community. Attrition rate and sample size achieved.	Recruitment, sample size and attrition. Baseline characteristics.	Baseline 6 months	Intervention and control
Study withdrawals		Number of withdrawals and reason for withdrawals.	Withdrawal surveys	Continuous	Intervention and control
Study safety		Falls, falls injury and unplanned hospital admission related to DBCI use.	As reported to staff and through questionnaire at 6 months.	Continuous 6 months	Intervention and control
Use	Adoption	Proportion of participants who logged into DBCI two times or more and viewed the health assessments once or more.	DBCI data	6 months	Intervention only
		Reason for non-adoption.	Adoption and Maintenance Survey	After 6 months	Intervention only
	Maintenance	Proportion of participants who used DBCI ≥ 4 months and completed 6 months assessment.	DBCI data	6 months	Intervention only
		Reason for non-maintenance.	Adoption and Maintenance Survey	After 6 months	Intervention only
	Engagement	Use of DBCI user interface and behaviour change techniques.	DBCI data	6 months	Intervention only
Usability		Participant satisfaction with usability of DBCI.	System Usability Scale	6 months	Intervention only
Technology acceptance		Perceived ease of use, perceived usefulness, attitude towards using DBCI and intention to continue to use DBCI. Identifies intervention factors for future improvement.	Focus Groups	After 6 months	Intervention only, who adopted DBCI use (subset)
Impact	Readiness	Self-reported readiness for change.	Readiness for Change Questionnaire	Baseline, 3 months, 6 months	Intervention and control
	Knowledge	Self-reported knowledge change for AVOID behaviours.	Knowledge and Behaviour Change Survey	After 6 months	Intervention only, who adopted or maintained DBCI use
	Behaviour	Self-reported behaviour change relating to AVOID behaviours.	Knowledge and Behaviour Change Survey	After 6 months	Intervention only, who adopted or maintained DBCI use
			Behaviour Change Questions	Baseline 3 months, 6 months	Intervention and control

DBCI, digital behaviour change intervention

• Technology acceptance—This was evaluated through two focus groups (7th and 8th of October 2025) involving a subset of intervention participants who had adopted the platform and were invited after completing their six-month assessment. The focus group discussion was guided by the interview schedule (Appendix 5), developed in alignment with the Technology Acceptance Model (TAM [25]) and with input from the research team and citizen scientists from the earlier platform adaptation phase.

• Usability—This was assessed using the System Usability Scale (SUS) [26], which consists of 10 items rated on a 5-point scale from strongly disagree to strongly agree. SUS was completed by intervention group participants after completing their six-month assessment. Where SUS was collected via telephone, research staff were unblinded. The SUS was modified by replacing the word ‘product’ in each statement with ‘AVOID Frailty platform’. Higher scores indicate greater usability, with a total score ranging from 0 to 100. A score of ≥68 indicates that the platform is usable.

• Adoption and maintenance—Adoption was measured as the proportion of intervention participants who logged into the platform more than two times and viewed the health assessment more than once (using recorded platform usage data). Maintenance was measured as the proportion of intervention participants who used the web intervention for ≥4 months. Participants’ reasons for not adopting or maintaining use [27] of the platform in the intervention group were evaluated from self-reported responses to the Adoption and Maintenance Survey (Appendix 6). The survey was conducted by telephone and offered to participants upon completion of their 6-month assessment.

• Engagement—Participant engagement with the platform was recorded in real time through the platform, exported into Excel format and described using the framework for engagement by Cole-Lewis et al. [28].

• Impact—Readiness to change and behaviour change were evaluated quantitatively through assessments administered to participants (intervention and control groups) at baseline, 3 and 6 months. Self-reported knowledge and behaviour change in the intervention group were evaluated using responses to the Knowledge and Behaviour Change Survey, which we developed based on the Transtheoretical Model of Health Behaviour Change [29] (Appendix 7). This survey was administered to those who adopted or maintained platform use after completing the Adoption and Maintenance Survey.

Analysis

Quantitative

Data analyses used SPSS V29.0.1.0. Descriptive data were provided in numbers (n) and proportions (%), mean and standard deviation (SD) (e.g. platform engagement), standard error of mean (SEM) (e.g. effectiveness) or median and interquartile range (IQR). In the intervention group, one item of the SEARCH questionnaire was not completed by five participants at 6 months, requiring re-estimation of FI using 35 rather than 36 items. Continuous variables were tested for normality using the Shapiro–Wilk test and a normal quantile plot. Differences in means between groups were tested using a t-test for normally distributed continuous variables, and a Mann–Whitney U test for non-normally distributed continuous variables. Because of small cell sizes, binary variables were tested using a Fisher’s exact test, and polytomous variables were tested using the exact significance probabilities for the test of independence.

A multilevel repeated-measures model was used to assess differences in change over time between groups for the outcomes of frailty and QoL. The main effect of group (treatment vs control) allows for differences between groups at baseline, and the main effect of time allows for change over time in the outcome variable in the reference group. The interaction between group and time estimates (and tests) the treatment effect. The multilevel model was fitted using restricted maximum likelihood with Kenward–Roger estimated degrees of freedom, the preferred model when the data are unbalanced, and the sample size is small.

To test the associations over time for dichotomous behavioural change outcomes, the Cochran’s Q test was used. As linear scales (interaction and diet scores) showed non-normal distributions when examined using the Shapiro–Wilk test, and for the ordinal outcomes of readiness to change, the Friedman’s nonparametric 2-way ANOVA test was used to assess association across three time points. For the ordinal variables of healthy diet and readiness to change, the overall treatment effects were assessed using multilevel multinomial regression models, thereby avoiding the need to assume proportional odds. The overall treatment effect was assessed using multilevel repeated-measures models, with logistic regression for dichotomous variables and linear regression for continuous variables.

For all analyses, P < .05 was considered to be statistically significant.

Qualitative

Conventional content analysis [30] was used to analyse responses to open-ended Adoption and Maintenance Survey questions. The coder read through all survey responses and inductively developed the initial coding frame. Codes were developed as salient to implementation outcomes (Table 2). The unit of analysis was a phrase coded to a single code. Related codes were grouped together into subcategories and categories. The initial coding frame was refined after independent application to subsets of participants by two researchers and, following discussion with other researchers, further refined [29]. The final coding frame was then applied to all surveys. Data within each code were checked for consistency. Categories were summarised as frequencies. NVivo 15.0.0 was used to manage data analysis.

Focus group audio recordings were transcribed by a professional transcriber using the Intelligent Verbatim style (edited for ease of analysis and readability). Research staff listened to audio recordings to check and correct the accuracy of the transcription and de-identify data. Directed content analysis [30] was used to analyse focus group data using Technology Acceptance Model constructs of perceived usefulness, perceived ease of use, attitude and intention towards using the DBCI, as well as to develop categories inductively. Intervention factors that enabled or hindered uptake of AVOID behaviours were generated inductively. After familiarisation with the data, two research team members independently coded transcriptions using predetermined codes, compared their coding and then identified subcategories before comparing their coding. After the coding book refinement, one researcher coded both focus groups. Data within each code were checked for consistency. Researchers met frequently to discuss, review and refine coding.

Results

Feasibility, reach and withdrawals

Of the 137 expressions of interest (EOI; Fig. 1) received, 42 (30.7%) were excluded. Following the eligibility assessment, a further 21 (15.3%) were excluded after eligibility step 1. Of the 74 (54.0%) assessed following consent, 14 (10.2%) were frail and excluded, leaving 31(22.6%) nonfrail participants allocated to the intervention and 29 (21.2%) to the control. The median age of participants was 75.35 years; 63.3% (n = 38) were female, and almost half (n = 26; 43.3%) had university-level education (Table 3). No differences were noted in the characteristics of the control and intervention groups. The required sample size was achieved, and the attrition rate was lower than expected. At the end of 6 months, 6 (19.4% of 31) withdrew from the intervention group, one was the nonrandomised participant.

Figure 1.

Flow diagram relating to participant recruitment and follow-up in the study.

Table 3.

Baseline socio-demographic and health data.

	Total (n = 60)	Intervention (n = 31)	Control (n = 29)	P-value
Socio-demographic data
Age in years [median (IQR)]	75.35 (70.00–78.90)	76.00 (70.20–81.10)	73.20 (69.20–76.80)	.109
Gender female n (%)	38 (63.3)	17 (54.8)	21 (72.4)	.188
Highest level of education n (%)				.507
Secondary or less	20 (33.3)	12 (38.7)	8 (27.6)
Apprenticeship/university certificate	14 (23.3)	8 (25.8)	6 (20.7)
University bachelor or higher	26 (43.3)	11 (35.5)	15 (51.7)
Household income n (%)				.147
$26 000 or less	6 (10.0)	2 (6.5)	4 (13.8)
$26 000–$51 999	10 (16.7)	6 (19.4)	4 (13.8)
$52 000–$103 999	26 (43.3)	10 (32.3)	16 (55.2)
$104 000 or higher	16 (26.7)	11 (35.5)	5 (17.2)
Prefer not to say	2 (3.3)	2 (6.5)	0 (0.0)
Living arrangement n (%)
Alone	16 (26.7)	10 (32.3)	6 (20.7)	.387
In past 12 months, technology use to monitor health/wellbeing? (% yes)	46 (76.7)	22 (71.0)	24 (82.8)	.365
SEARCH score [median (IQR)]	5.00 (3.50–6.00)	5.00 (4.00–6.50)	4.75 (3.25–6.00)	.806
EQ-5D-5L [median (IQR)]	0.956 (0.932–1.000)	0.968 (0.924–1.000)	0.956 (0.956–0.984)	.392
Health rating (0 is poorest and 100 best is best health) [median (IQR)]	85 (75–90)	85 (75–90)	85 (75–90)	.923
Health data
Proportion with normal or brisk walking speed n (%)	41 (68.3)	18 (58.1)	23 (79.3)	.100
Proportion memory rating fair or poor n (%)	23 (38.3)	10 (32.3)	13 (44.8)	.427
Proportion with little interest or pleasure in doing things over the previous 2 weeks for several days or more n (%)	8 (13.3)	5 (16.1)	3 (10.3)	.708
Proportion with food intake decreasing over the previous 3 months due to loss of appetite, digestive problems, chewing or swallowing difficulties n (%)	3 (5.0)	3 (9.7)	0 (0.0)	.238
Proportion with heart troubles n (%)	8 (13.3)	6 (19.4)	2 (6.9)	.257
Proportion with stroke history n (%)	0 (0.0)	0 (0.0)	0 (0.0)
Proportion with diabetes n (%)	3 (5.0)	2 (6.5)	1 (3.4)	1.000
Proportion with bad sleep quality n (%)	16 (26.7)	10 (32.3)	6 (20.7)	.387
Proportion with ≥1 fall in past year n (%)	15 (25.0)	5 (16.1)	10 (34.5)	.139
Proportion with eyesight interfering a little or fair amount with life in general n (%)	14 (23.3)	6 (19.4)	8 (27.6)	.547
Proportion with hearing interfering a little or fair amount with life in general n (%)	27 (45.0)	14 (45.2)	13 (44.8)	1.000

SD, standard deviation; IQR, interquartile range; n, number; SEARCH, self-assessed report of personal capacity and healthy ageing

Effectiveness

Primary outcome

There was a reduction in frailty risk over 6 months from a mean of 0.135 (SD 0.053) to 0.115 (0.071) in the intervention group compared to an increase from a mean of 0.134 (0.048) to 0.160 (0.087) in the control group (Table 4). The difference in mean change was −0.044 [95% confidence interval (CI) −0.076 to −0.012; P = .008, effect size 0.38].

Table 4.

Between-group changes in the intervention and control group with all participants included.

		Descriptive statistics		Multilevel repeated measures
Outcome variable	Time point	Mean (SD) Intervention Baseline n = 31 6 months n = 24a	Mean (SD) Control Baseline n = 29 6 months n = 29a	Difference in mean change (95% CI)	P-value	Effect size
FI	Baseline	0.135 (0.053)	0.134 (0.048)	−0.044 (−0.076 to −0.012)	.008	0.38
	6 months	0.115 (0.071)	0.160 (0.087)
QoL	Baseline	0.958 (0.044)	0.958 (0.032)	0.032 (0.006 to 0.058)	.016	0.34
	6 months	0.962 (0.042)	0.924 (0.053)

FI, frailty index; QoL, quality of life (EQ-5D-5L); SD, standard deviation

^aAt 6 months, there were six withdrawals and one participant who did not complete the assessment. P-value significant <.05

Table 5.

Between-group changes in the intervention and control group with the two intervention participants who were not randomised excluded.

		Descriptive statistics		Multilevel repeated measures
Outcome variable	Time point	Mean (SD)	Mean (SD)	Difference in mean change (95% CI)	P-value	Effect size
		Intervention	Control
		Baseline n = 29	Baseline n = 29
		6 months n = 22a	6 months n = 29a
FI	Baseline	0.133	0.134	−0.042
		(0.054)	(0.048)	(−0.074 to −0.009)	.013	0.36
	6 Months	0.115	0.160
		(0.072)	(0.087)
QoL	Baseline	0.959	0.958
		(0.045)	(0.032)	0.028	.030	0.31
	6 Months	0.960	0.924	(0.003 to 0.054)
		(0.042)	(0.053)

FI, frailty index, QoL—quality of life (EQ-5D-5L), SD, standard deviation

^aAt 6 months, there were six withdrawals and one participant who did not complete the assessment. P-value significant <.05

Secondary Outcome.

There was an increase in QoL over 6 months from 0.958 (0.044) to 0.962 (0.042) in the intervention group, compared with a reduction from 0.958 (0.032) to 0.924 (0.053) in the control group (Table 4). The difference in mean change was 0.032 (95% CI 0.016 to 0.058; P = .016, effect size 0.34).

Effect of protocol deviation

When the two participants who failed to be randomised were removed from the analysis, the findings were unchanged (Table 5) with the difference in mean change −0.042 (−0.074 to −0.009; P = .013, effect size 0.36) for frailty and 0.028 (0.003 to 0.054; P = .030; effect size 0.31) for QoL.

Other implementation outcomes

One participant in the intervention group did not complete the six-month assessment, with SUS scores obtained for 24 (77.4%) participants (Fig. 1). Twenty-five (80.6%) intervention participants were invited to complete the Adoption and Maintenance Survey, but two declined, leaving 23 (74.2%). Of these, two had not adopted the platform. Therefore, 21 (67.7%) responded to the Knowledge and Behaviour Change Survey.

Invitations were offered to 13 (41.9%) participants who had adopted the DBCI and completed their six-month assessment prior to the focus group dates. Seven declined, and six (46.1%) participated (Appendix 8).

Safety

No adverse events related to the DBCI were recorded.

Use—adoption, maintenance and engagement

Twenty-eight participants (90.3%) adopted the intervention, and 22 (71.0%) maintained DBCI use. Two participants were asked why they did not adopt the platform, and three were asked why they did not maintain use of it. Reasons for not adopting the platform related to life constraints (e.g. health issues, travel or time) and lack of digital literacy. Two participants did not maintain use of the DBCI due to life constraints, while another did not want to regularly interact with a digital tool. Participants had varied engagement with the DBCI (Table 6).

Table 6.

Participant engagement with the DBCI techniques or elements over 6 months.

DBCI techniques or elements	Intervention participants; n = 31
Number of logins; mean (SD)	27.00 (37.41)
Gamification
Points scored; mean (SD)	146.00 (131.02)	Score range (20–413)
Number of participants with points scored	28
Assessment reports
Baseline health assessment report	Average number of times viewed; n (SD)	9.00 (13.00)
3 months mini report	Average number of times viewed; n (SD)	2.00 (2.15)
Educational modules (videos)
	Number of participants engaged; n (%)	27 (87.1)
	Average total time (minutes) per participant spent; mean (SD)	116.50 (148.08)
Average time (minutes) spent per education module; mean (SD)
Ingredient for change 19.50 (22.78)	Activity 17.26 (16.54)	Vaccination 18.67 (61.89)
Optimising medications 11.03 (30.80)	Interaction and Socialisation 11.14 (54.13)	Diet and Nutrition 13.78 (14.84)
Sleep 15.97 (22.78)	Falls prevention 9.16 (11.44)
Action plans
	Number of participants creating action plans, n (%)	20 (64.5)
	Number of action plan per participant; mean (SD)	6.71 (9.85)
	Total action plans created; n	222 (some covering more than one category) Activity—134 Falls Prevention—90 Interaction and Socialisation—83 Diet and Nutrition—59 Sleep—53 Brain Health—51
Community resources
	Number of participants engaging with community resources; n (%)	23 (74.2)
	Number of views per participant; mean (SD)	31.23 (41.23)

Usability

The SUS median score of 76.25 (IQR 58.75–87.50) indicated the DBCI was perceived as usable (Appendix 9).

Technology acceptance ( Table 7 )

Table 7.

Technology acceptance with the digital behaviour change intervention (DBCI) using technology acceptance model constructs.

Perceived ease of use of DBCI
Category	Summary	Quote
Mixed perceptions of ease of use	There were mixed perceptions relating to ease of use, ranging from the DBCI being very easy for a participant with high computer literacy, to it being harder to use than what participants expected, and for some becoming easier to use over time.	FG2 Speaker 1: I taught people how to use a website, and to me, it's a piece of cake, absolute piece of cake. FG1 Speaker 4: You have to understand the program, so … took a little bit of time to understand it and work out the jargon, what it all meant. … So, that was sort of a settling down, and then I found it quite easy to use. … I found the experience, overall, once you go through the learning side – which shouldn’t be unexpected you’d have to learn. This is something new. FG1 Speaker 1: I assumed it would be quite straightforward – and I consider myself relatively competent – and it wasn’t. I didn’t find it easy to work my way through it at all.
Design issues negatively impacting on ease of use and engagement with DBCI and features of DBCI behaviour change techniques	Participants identified design issues with suggestions to overcome these, such as using an application that functioned across both computer and mobile devices, and where participants could reset their passwords.	FG2 Speaker 2: I lost my password once. Remember, I come and you set it up again. FG1 Speaker 3: I’d like to see an app form. … something simplified that an older person could get onto, easy to navigate, an app where they could just look up stuff in an instance … I found it quite hard, first, to log in every time. I kept thinking it was my laptop or my phone … I travelled a lot this year, so I had to do anything … on my phone… and I found it quite frustrating to navigate. Like the gentlemen have said, it was very hard. FG1 Speaker 2: … I still used mine when I was away [travelling], just to set the goals every week, but … I wasn’t reading up on information, because of the screen size on the phone.
	Participants identified insufficient design features of Behaviour change techniques (BCT action planning and education), which impacted on ease of use. It was not intuitive to implement action planning for most participants who tried using action planning, affecting their using it as intended and or abandoning use. Examples given for viewing education was the progress bar not displaying 100% on completion of AVOID modules, education modules not resuming at the last point of viewing (i.e. participants had to view module again from the start) and no indication when education resources (printables) were already accessed, which required extra effort.	FG1 Speaker 2 action plans, same trouble there. But I got around it by, ah, I’d fill a whole week in. I’ve done it every day, did it beforehand, and then submitted it. So, then it was all submitted, so I didn’t have to go back at the end of the week and submit it. FG2 Speaker 2 The only one I had the problem was the [Action Planning] - I could set my goals, but whether I was doing - sometimes it would submit, sometimes it would not. FG1 Speaker 1 I had the same sort of initial difficulties with trying to do an action plan and then trying to go back to it to fill in detail, and I couldn’t. … I didn’t do them on the system. FG1 Speaker 4 You went through it, saw all the videos, and … a quiz at the end, and completed it. It was meant to show 100 per cent completed. Well, it wouldn’t; it didn’t. A minor thing, but it sort of frustrated me a little bit, and I thought, that’s not right. So, I did it two or three times, the whole thing again, repeated, and it still only came – not being completed.
Perceived usefulness of DBCI
Category	Summary	Quote
Perceived useful to enhance their ability to change frailty behaviours. Perception of usefulness varied across components.	Most participants perceived using the DBCI as useful in enhancing their ability to change their AVOID behaviours, giving examples of their changed behaviours (and sometimes changed behaviour of others who were not enrolled in the study). Education was highly valued (credible, organised relevant topics, sufficient amount, different modes). The usefulness of Action Planning, Community Library and Assessment Reports (including face-to-face Baseline Assessment) varied between participants.	FG1 Speaker 4: The most useful part for me was the resources, the modules, and the reading … all the material that was there, … I read them all, and then reread some … I found that more valuable than the action plans …. Well, falls prevention certainly did. I’ve read them thoroughly a couple of times, and then moved on to increasing my exercise physiology to twice … a week. FG2 Speaker 1: We [husband and participant] do … five sit-stands, … do five of each … leg raises, … three times a day. … I actually put that down to being able to give up the walking stick, because that has helped my strength and my balance to the extent that I'm now walking straighter and stronger. … I found the setting the goals really helpful … to focus on doing it and doing it religiously, because I wanted to get the seven ticks every … So I found that part extremely valuable. FG2 Speaker 2:I researched around [Community Library] … it got me going every day, … it led me to the Council a lot … They offer lots and lots of things there. … Chair yoga and I've got two lots of walking group, and I said meditation. FG1 Speaker 1 That was about the most useful part of this whole exercise, that very first, initial assessment, and the resulting health assessment [Report] that you could go through. ‘Yes, I’m doing okay in that area, so I won’t focus on that, but here are some, like, you know, vaccinations’.
Perceived usefulness of DBCI negatively impacted by design and content issues Suggestions to increase usefulness	Participants identified DBCI design issues and content (e.g. static content over time) limiting usefulness. New content, with DBCI-based updates would increase usefulness. Other relevant Education topics were suggested. For action planning, have an option to set multiple action plans and to intuitively develop and monitor them. For community library, more resources (with accurate contact details) were suggested. A notification to repeat baseline assessment would be useful.	FG2 Speaker 1: New information would make it better to last the six months, but otherwise, I agree with [Speaker 2], because I've gone off the boil, so to speak. FG1 Speaker 2: You could just have a homepage, news page, … ‘this has changed’… Then it makes you aware of it, so then you don’t have to trawl through the whole thing. FG1 Speaker 4. I thought ‘I’ll get into this, do a whole lot of action plans’. FG1 Speaker 2: But I thought there probably could even be more on there [Community Library]. FG1 Speaker 1: Well, what I would like is a notification to say, look, ‘it’s been two years since our last assessment; it’s time to have another one’. That would get me focused.
Attitude towards using DBCI
Category	Summary	Quote
Mixed attitudes towards using the DBCI, with negative attitudes reducing as participants became more familiar with using the DBCI, except for one participant	Attitudes ranged from positive to negative and included mixed attitudes. Participants enjoying using the DBCI and viewing its usefulness in helping them change AVOID behaviour positively influenced their attitude towards using the DBCI. Participants’ experiences of technical difficulties negatively influenced their attitude towards using the DBCI. As participants became more familiar with using the DBCI, attitudes improved except for one participant who did not feel positive towards DBCI use.	FG2 Speaker 1 I found the whole thing extremely positive. I do not have any negatives. … And because I'm hooked on playing with computers, I really got stuck into it, and I had all those tests done in a week and all the rest of it, which was totally ridiculous. … I loved it. … It was colourful. It was easy to navigate. The pages were well set out, easy to access. Yeah, no problem. … I cannot speak highly enough of the whole program. I think it's fabulous. FG1 Speaker 2: I found the system all right. Like everyone else, I had the initial difficulty. It probably wasn’t too bad, but you just sort of had to take your time. But once you’ve done it – it’s the same as … First time doing anything, especially for us, we’re sort of technically challenged, most of us, because we’re not working, … so the last training I would’ve had on a computer through work would be 20 years ago. … I think it’s a useful tool. FG2 Speaker 2: Yeah, okay. I'm not a computer whiz there, but I only used it to what I wanted to know. … It was exciting when you could go on the links and see what it was and you could read about it and everything like that. Then, sometimes, if you could not get on them, you'd think, ‘oh, what have I done wrong, and you go back again.’ But that's probably more human error … FG1 Speaker 1: I would prefer a non-computer-based system, if that’s at all possible. … I tried to use the action plan … but, obviously, I did it wrong and couldn’t get back into it. But, meanwhile, I was still mindful of what I was trying to achieve with the action plan, but I was doing it mentally
Intention to continue to use DBCI
Category	Summary	Quote
Positive intention to continue to use the DBCI, except for one participant	Participants expressed positive intention to continue use of the DBCI only with updated content, with users alerted to updates via the DBCI and where time permitted, except for one participant.	FG1 Speaker 2: I’d be quite happy to have access to it for the rest of my life, really, because as long as the information was updated … [and] you knew it was updated, it would be a – it’s a useful tool. FG2 Speaker 2: I'd give it a break for a while, though, because I'm busy doing everything. FG1 Speaker 1: I’d continue to have an occasional assessment with you personally.

Perceived ease of use

There was a mixed response to the ease of use of the DBCI, ranging from very easy for those with high digital literacy to harder than expected for others, though it became easier to use over time for some. Participants identified some technical issues with the Education and Action Planning tool. It was suggested that a DBCI application functional on computers and handheld devices could overcome some difficulties, and that an easier method to reset their password would be helpful.

Perceived usefulness

Most participants perceived using the DBCI as useful in enhancing their ability to change frailty behaviours. Participants used behavioural change techniques variably, with most reporting their usefulness. Participants reported the initial face-to-face health assessments and reports as motivating, identified areas of health risk (Table 8), and one participant reported seeing improvement at 3 months as positively reinforcing. Viewing education provided relevant and credible information that improved knowledge on a range of modifiable behaviours and how to change them, leading to informed, decisive action. There was a mixed response to earning points through DBCI engagement, ranging from being very motivated by gamification to disliking it. Goal setting and action planning prompted focus and behaviour change for some, and some integrated it into their weekly lives to sustain it. Two participants abandoned their use due to technical difficulties, while another used their own method to monitor progress. Two participants who received nudges perceived them as valuable. Some participants believed they already had access to the information in the community resources library. Others found the library a useful guide to resources they were unaware of, and one participant reported using library resources to make changes.

Table 8.

Intervention factors (listed on TIDieR) that enabled or hindered behaviour change.

Intervention factors	Enabled or hindered behaviour change	Example quotes
Education modules, online printable and other online resources	Education on the DBCI enabled behaviour change by providing relevant, credible and useful information that raised participants awareness on a range of behaviours that could be changed, improved knowledge on how to change them and provided online resources that participants could view in their own time. There were no findings that education hindered behaviour	FG1 Speaker 3: just like the immunisation … a bit out of that to get myself up to date FG1 Speaker 4: The most useful part for me was the resources, the modules, and the reading … all the material that was there, … I read them all, and then reread some…. I’ve read them thoroughly a couple of times, and then moved on to increasing my exercise physiology to twice … a week FG1 Speaker 2: I found the resources really good too. Yes, I agree, that was the most benefit I got.
Self-monitoring of health at commencement and behaviour modification at 3 months	At least one participant found the Baseline Report enabled behaviour change through providing focus, insight into their vulnerability and intrinsic motivation to act. Similarly, the face-to-face Baseline Assessment also enabled behaviour change. One participant found that viewing the 3-month report showing progress on behaviour change was positively reinforcing. There were no findings that Baseline Report hindered behaviour change	FG1 Speaker 1: That was about the most useful part of this whole exercise, that very first, initial assessment, and the resulting health assessment [Report] that you could go through. ‘Yes, I’m doing okay in that area, so I won’t focus on that, but here are some, like, you know, vaccination’s, I thought I was totally bulletproof when it came to vaccinations. I wasn’t vaccinated for anything. FG2 Speaker 1: I've done a couple of checks over the time. … I'm quite happy with what it shows. Facilitator 1: Was that useful? FG2 Speaker 1: Yeah …the initial one, and then the later one showed that I'd made some improvements, and that's encouraging.
Community database of library resources, including access to health coaching	Community library enabled behaviour change for one participant as it provided them an opportunity to search for information about local physical and social activities, of which they joined. There were no findings that Community Library hindered behaviour change. Community library was perceived useful for some participants noting resources they were previously unaware. Others found they already had access to such information or could easily search for it. No participants mentioned accessing health coaching.	FG2 Speaker 2: With all the other activities, I researched around and I did things through the Council. … The website, it led me to the Council a lot. That was good, … I must admit … they are very good. They offer lots and lots of things there. FG1 Speaker 4: Once I scanned all of … the community stuff, it was really, well, apart from picking up a few things at the beginning, I didn’t look at it hardly ever again. FG1 Speaker 2: I’m … quite aware of what is on in the area. … I was probably doing four things on that list already. FG1 Speaker 3: I didn’t really look at anything on there, really. But if I wanted to go on a walking group or whatever, basically, do my own research, and find out which one would suit me most.
Goal setting, action planning and progress tracking	Goal setting and action planning prompted participants to focus on behaviour change. Some participants found this technique initially enabling to change behaviour, while others integrated it into weekly life as a prompt to sustain behaviour change. Some used the DBCI to monitor while another used their own monitoring method. There were no findings that goal setting and action planning hindered behaviour change. Abandoned by two participants after experiencing design issues with using it and another used their diary/calendar to schedule once off or ongoing behaviours. One participant did not see the need to use this function.	FG1 Speaker 4: Initially, I tried it, and then they worked ... some things … like, see the pharmacist about my medications. I might’ve done that, and I thought, well, I’ve already put it on my phone, so I’m just sort of doubling up again. FG1 Speaker 2: Most of my action plans had to do probably with walking and the amount of walking I did each week. I’ve got a step thing on my phone, which I can review the week. … That was my monitoring of it. I didn’t use the system to monitor it. FG2 Speaker 1: I found the setting the goals really helpful to … focus on doing it and doing it religiously. FG1 Speaker 1: I had the same sort of initial difficulties with trying to do an action plan and then trying to go back to it to fill in detail, and I couldn’t. FG1 Speaker 4: I stopped using action plans, because I thought this is unnecessary.
Gamification—earning points for DBCI use; comparative lead score displayed next to participant score.	Two participants enjoyed earning points as an acknowledgement of progress. This included one participant who viewed earning points and viewing the comparative lead score as reinforcement to complete action plans, which enabled behaviour change. There were no findings that gamification hindered behaviour change. Some participants were not encouraged to use the DBCI by earning points and one participant would prefer not to view their points. Two participants thought the lead score was made up.	FG1 Speaker 4: I found it as a useful tool. … I used to enjoy, … But I didn’t do it to amass points. But I just felt, okay, well, … the system’s saying, ‘good on you.’ FG2 Speaker 1: I am [action planning], because I'm a competitive person ... I need to make it see my points go up every week. It's [comparative lead score] just something that you look at and think, ‘yeah, well, I'm not doing too bad’. FG1 Speaker 1: Just on the point score thing … I thought that was a little bit childish, … trying to incentivise you to use the thing. FG1 Speaker 2: I thought it [comparative lead score] was probably bogus. … I’m a bit of a cynic. FG2 Speaker 2: They did not mean anything. … I'm not competitive.
Nudging—email reminder (up to three times) to revisit the DBCI after not using it for 4 weeks	There were no findings that nudging was an enabler for behaviour change. There were no findings that nudging was a barrier for behaviour change. Two participants receiving a ‘check-in’ email were both receptive to receiving the reminder email.	FG1 Speaker 3: Probably, you’ve emailed us and reminded us to look at it.

Attitude

Attitudes towards the platform varied, with negative attitudes decreasing as participants became more familiar with the DBCI, except for one who preferred a nontechnology-based system.

Intention to continue to use

Participants expressed an intention to continue using the DBCI if new content was added over time and if time permitted.

Implementation strategy factors

Some participants valued the baseline assessment being face-to-face, and one preferred a face-to-face re-evaluation of health at 6 months (Appendix 10). Attending Ask an Expert sessions enhanced motivation, providing valuable opportunities to learn from and interact with experts and other participants and to check in with research staff about technology issues. Suggestions for other implementation strategies included a preference for a hands-on orientation to the use of the DBCI, either individually or in groups, with participants followed up to address technology issues. Another suggestion was to provide opportunities for participants to connect with each other to share knowledge about resources and provide support from early adopters using DBCI.

Impact

There were no significant differences in readiness to change (Appendix 11) between the control and intervention groups over time. A reduction in readiness for change was seen with vaccination (P = .025) and optimising medication (P = .010) in the control group over time. There was no significant difference in behaviour change between the control and intervention groups over time (Appendix 12). In the intervention group, improvements in behaviour change were observed for vaccination (P = .004) and optimising medications (P = .050). For the control group, there were improvements in vaccination (P = .009) and sedentary behaviour (P = .050) but deterioration in interaction and socialisation (P = .017) over time (Appendix 12).

For the intervention group, almost all participants (n = 19, 90.5%) who responded to the knowledge and behaviour survey reported learning something new in at least one AVOID domain, with learning reported across all AVOID domains and frailty in general. The most frequently self-reported changes in knowledge were for activity (n = 11) and diet and nutrition (n = 10) (Table 9). Almost all participants (n = 18, 85.7%) reported changing at least one AVOID behaviour. Behaviour change was reported across all AVOID domains, with the most self-reported AVOID behaviours being activity (n = 14, 66.7%) and diet and nutrition (n = 12; 57.1%) (Table 10). The most frequently reported behaviour not changed with the perception that they were already achieving those healthy lifestyle behaviours was also activity (n = 6; 28.6%) and diet (n = 4; 19.0%). Two participants reported advocating to others.

Table 9.

AVOID behaviour domains and frequencies for self-reported knowledge change for intervention group.

Healthy action	Frequency (n = 21)	Knowledge change example quote/s (maximum three examples provided)
Learnt Something New
Activity	11	How important it is to keep moving and to do it regularly. Sitting down not doing much is not good [ID03 woman, 65–69]. One of the things was sarcopenia. I hadn’t heard of that before. … that simple set of exercises he suggested 3 times a week [ID114, woman, 75–79]. I have learnt that exercise is age related: osteoporosis, frailty, falls, as well as strength training and cardio training [ID88, W, 75–79].
Vaccination	8	I hadn’t thought about those vaccinations before – the ones you get every decade or so [ID135, man, 70–74]. I am more aware that as you get to a certain age that you are more susceptible. I had not been aware of shingles per se [ID18, man, 75–79]. I learnt to go to your doctor and ask about vaccinations [ID51, woman, 75–79].
Optimising medication	2	Interesting to learn ... that a more regular review is needed [ID03 woman, 65–69]. Asking the GP to check my medication and asking the GP questions ‘do I need to take 20mg?’ [ID88, woman, 75–79]
Interaction and socialisation	7	Really opened my eyes what a big problem it [isolation] is [ID03 woman, 65–69]. I learnt …to keep mingling with people I know and like [ID18, man, 75–79]. I learnt it was important to keep up with people [ID59, woman, 70–74].
Diet and nutrition	10	Protein is important and you do not want to lose weight or lose muscle condition. Having a balanced diet is important [ID18, man, 75–79]. Proper nutrition. … to have the right amount of vegies and fruit [ID51, woman, 75–79]. Learnt how nutrition does impact on ageing well. Like more protein and to have protein at each meal. And have various protein. To have water and herbal tea to keep body and brain hydrated [ID88, woman, 75–79].
Frailty	8	How important it is to care for your body and overall health, … everything – diet, movement, resistance exercises, vaccination, interaction [ID117, woman, 60–64]. There is a pathway to frailty that if you do not do anything and it never occurred to me that we can make changes and we are looking at this a bit differently now [ID53, man, 80–84]. I learnt that it was multi factorial. … I wouldn’t have thought about including socialising and vaccinations; nutrition yes but I would not looked at it more widely [ID56, woman, 65–69].
Behaviour change	2	I’m coming to terms that there are little incremental changes that are more likely to be lasting rather than big changes, which often are not. I think of making small improvements than big ones [ID46, man, 70–74]. Keeping a record of your activities is useful, so when you are slacking off a bit or not doing something that you should do. For example, ‘say you set a goal for walking 5 times a week, and you only do it a couple of times, it is useful to review my progress and see where you are sticking to your goals you have set yourself. Just being aware that you need to do these things for your own good and then having a baseline assessment – to compare yourself against. … Before [the study] I was doing the exercise but I wasn’t recording the activity. I find it useful to keep tabs on myself, just for exercise [ID97, man, 80–84]
Other topics.	1 (sleep)	Interesting to learn about sleep and how you can improve it [ID135, man, 70–74].
	4 (brain health)	Using your mind and being open to things [ID23, woman, 65–69] I learnt dementia is terminal. I learnt that dementia is quite common and … dignity, to treat people with understanding and patience [ID46, man, 70–74]. I picked up from the memory section in AVOID was to ensure you keep your brain active [ID77, woman, 80–84].
	2 (falls)	Ask an Expert: falls and managing them and some reference to a judo club and how to manage a fall and to minimize the injuries from a fall and to how to roll … being aware that there were alternatives to putting out your hands to fall or try to minimise the fall. I had been of the opinion that you would try to stop it rather than accept it [ID46, man, 70–74]. The falls – keep consistently exercising … how to get up if you have a fall [ID59, woman, 70–74].
Community resources that could support behaviour change	5	The Community Library was helpful to know what was out there [ID51, woman, 75–79. The Strength for Life ones [activity videos] in the Falls Prevention included those particular exercises [ID77, woman, 80–84]. Looking at the Community library, there was a gym called Viva not far from us and I didn’t know about it. … I found ‘This Way Up’ on NSW Health Dept - from the AVOID platform [ID88, woman, 75–79].
Did Not Learn Anything New
	3	I didn’t learn anything that I didn’t know. (for frailty) [ID01, man, 80–84] I didn’t get anything like that from the program at all. I visit my GP regularly and I am with an EP group. (for frailty) [ID72, male, 70–74]. Not really [ID03 woman, 65–69]. (stated only for vaccination)

Table 10.

AVOID behaviour domains and frequencies for self-reported stage of behaviour change for intervention group.

Healthy action	Frequency (n = 21)	Stage of behaviour change example quote/s (maximum three examples provided)
Activity	Behaviour changed
	14	I have incorporated those simple sarcopenia exercises 5 days a week. … I am now walking greater distances [ID114, woman, 75–79]. More strength and weight bearing exercises ... I now hold a 5kg weight when I do squats - before … was … 3kg …. Balance exercises … are new exercises that I include in the gym on my own [ID50, woman, 75–79]. I used to be a runner and play football … I go out running now with my daughter, so that is rewarding … most days it is 5kms, 2 or 3 times a week [ID61, male, 60–64].
	Behaviour unchanged as perceived to be sufficient
	6	Was fairly active before the study. I was doing keep fit class, golf twice a week, balance class, cross fit class, aqua aerobics [ID03 woman, 65–69]. Pre … AVOID Frailty I did see an exercise physiologist and they gave me a couple of exercises to add to my home exercises. AVOID Frailty was more affirmation that … I was doing … the right thing [ID137, woman, 70–74]. I haven’t changed anything. … I already walk 3-4 time a week; I do Pilates once a week [ID23, woman, 65–69]
	Advocating to others
	1	I have been preaching that [sarcopenia exercises] to everyone who will listen... I taught others there and various friends about those exercises. I am a good advocate for it [ID114, woman, 75–79].
Vaccination	Behaviour changed
	6	I finished the vaccinations and completed the second shingles course [ID114, woman, 75–79]. I organised to have all of those vaccinations with GP [ID18, man, 75–79]. I had a pneumonia vaccination [ID51, woman, 75–79].
	Planning to make change
	3	I am seeing the GP in December for a check-up and I will see then about the vaccinations - shingles and the pneumovax [ID117, woman, 60–64]. Next time I see the GP, I will speak to him about tetanus and diphtheria vaccination [ID137, woman, 70–74]. I have booked in to have shingles [ID61, male, 60–64].
	Behaviour unchanged as perceived to be sufficient
	2	There wasn’t more that I could do than I was doing. … Vaccinations are completely up to date [ID01, man, 80–84]. I haven’t needed to make changes [ID03 woman, 65–69].
Optimising medications	Behaviour changed
	3	I had the pharmacist look at the medication [ID53, man, 80–84]. I am on a lot of medications. … I asked my GP if anything can be changed and they said ‘not really, if it is working, leave it as it is’ [ID56, woman, 65–69]. Melatonin - always have taken the same dose every night. I talked to the GP and we cut it back to 1/2 mg. That was a big change [ID88, woman, 75–79].
	Planning to make change
	1	I go to the GP once or twice a year. I will talk to the GP then about medications [ID117, woman, 60–64].
Interaction and socialisation	Behaviour changed
	4	I have reconnected with a couple of cousins that I had not been in contact with for at least several years … I have reconnected with friends of my wife … I will be visiting them interstate [ID18, man, 75–79]. I joined a meditation group, a walking group, and I am thinking about joining a book club. You meet other people in these groups [ID51, woman, 75–79]. I make sure I contact at least 3 people each week. …. You make the effort and it is more rewarding [ID59, woman, 70–74].
	Recognised need for change
	1	I would like to be able to do more of that but I can’t because of my partner’s health [ID77, woman, 80–84].
	Planning to make change
	1	Try to find community groups for walking which we are interested in [ID135, man, 70–74].
	Behaviour unchanged as perceived to be sufficient
	2	I still catch up with family and friends … I am still social [ID03 woman, 65–69]. I have a lot of social interaction, so I haven’t increased or decreased it [ID56, woman, 65–69].
	Advocating to others
	2	I suggested to them to look up/Google groups or join Probus or exercise groups, especially when people are newly retired [ID03 woman, 65–69]. Have been encouraging other people in my age group to become involved with other people and maintain social contacts [ID114, woman, 75–79].
Diet and nutrition	Behaviour changed
	12	When I was cooking just for myself, I lost about 7kgs over 15 months. I am now very conscious of having sufficient protein in my diet; … a good dose of cheese and diary; as well as fruit and vegies [ID18, man, 75–79]. We are now getting professionally fresh meals delivered …. We are finding meal preparation harder. … The food we eat now is better balanced. … we were cooking on a ‘as need basis’ without a lot of thought given to the balance of the nutritional value of the meal [ID53, man, 80–84]. On your own it is sometimes easy just to have a bit of toast. I try to have a bit of schedule for the week to cover a lot of main dietary requirements [ID70, woman, 70–74].
	Recognised need for change
	1	I don’t have a very heathy diet. I am aware of what I should be doing but I lack motivation [ID46, man, 70–74].
	Behaviour unchanged as perceived to be sufficient
	4	I am pretty well on top of nutrition because I am vegetarian and you can’t just skip the sausages [ID56, woman, 65–69]. I didn’t need to update my diet and nutrition [ID135, man, 70–74]. Our diet is satisfactory. There is no need to change that [ID97, man, 80–84].
Other topics	Behaviour changed
	4 (sleep)	‘Relaxing bedtime stretches for better sleep.’ I saw that and thought that might work … I do it religiously every night. Even the other night when I went out to a show, I still did it. I don’t sleep perfectly every night but I sleep better than before [ID77, woman, 80–84]. Trying to slow down at nighttime, because I was having trouble sleeping, not regularly, so I learnt to put my earphones on to listen to the mediation with the slow down programs they have on there [ID70, woman, 70–74]. I did the online meditation course … (via the AVOID platform) … Some of the techniques … I have adopted to go to sleep each night [ID114, woman, 75–79].
	3 (brain health)	The healthy brain; the presentation was fabulous and I am eating a lot more fish oil (3 capsules) because of attending that session [ID18, man, 75–79]. More religious to taking fish oil every day … I am mad keen on wordle and sudoku and the luminosity program (series of activities that exercise your brain) … but I was hit and miss and did it when I felt like it. Now I am committed to doing ten a day. I can’t start the day without doing some brain exercises [ID77, woman, 80–84]. I have increased my brain training games online. I was doing this before AVOID but I am doing them more now [ID88, woman, 75–79].
	1 (falls)	I did have a fall. … I was able to roll and not put my arm out and as a result I had minimal injuries [ID46, man, 70–74].
	Behaviour unchanged as perceived to be sufficient
	1	It gave me confidence that I was doing everything that I needed to do [ID97, man, 80–84].

^aCoded only once per participant per AVOID behaviour i.e. if participant states multiple levels of behaviour change within one AVOID behaviour, highest level of behaviour change coded. Stages of behaviour change that were reported by no participants are not listed in the table.

Discussion

Principal results

This study demonstrated that a DBCI incorporating multiple behavioural change techniques supported behaviour change in older community-dwelling nonfrail participants aged 60 years and older. Compared with a control group, the DBCI was effective. Frailty risk was reversed (primary outcome), and QoL improved (secondary outcome) at 6 months. Furthermore, the DBCI was adopted by almost all participants, and engagement was maintained by three-quarters, who found the DBCI usable, acceptable and impactful despite technical barriers. While there was no difference between the intervention and control groups in terms of readiness to change and behaviour change, qualitatively, there was evidence of self-reported knowledge gain and positive behaviour change across all AVOID domains.

Comparison with prior work

Participants in this study started with a mean FI score in the pre-frail range [31]. Those in the control group had a 0.03 increase in FI score over 6 months, consistent with the observation of a 0.02 increase in FI score over 12 months in another study [14]. A recent systematic review concluded that frailty risk commonly worsened over time. Our intervention was effective in reducing frailty risk over 6 months, with a 0.02 reduction in FI score in the intervention group and an average difference in frailty risk of 0.044 when the intervention and control groups were compared, within a previously identified range for minimally important difference over 12 months of 0.02 (small) to 0.06 (large) using the distribution method [32]. These findings are also consistent with findings from a recent meta-analysis of seven studies (only two studies used FI) where digital health interventions were said to improve frailty status when compared to control (random-effects model: standardised mean differences = 0.42, 95% CI −0.72 to −0.12 [33]) in frail older people. The same meta-analysis also reported that digital health interventions resulted in better QoL, where only two of the six studies focused on the EQ-5D with Danish tariffs, when compared to control (standardised mean difference 0.3; 95% CI 0.11to 0.50 [33]) as a secondary outcome measure, in line with our findings. Therefore, when older people are empowered to take action by building their capability, increasing their awareness of opportunities and available community resources, and influencing motivation to act, improvement in healthy life expectancy is demonstrated.

Addressing a gap in the literature, we explored the association between DBCI use and self-reported knowledge gain or behaviour change to address the reported gap between engagement and change [34]. Qualitatively, behaviour changes were most reported for activity and diet, confirming the effectiveness of our DBCI strategy in increasing the uptake of interventions effective for both frailty prevention and management. Engagement with behaviour change techniques in our study similarly demonstrated a focus on activity-related goal setting and action planning, including for falls prevention. Our findings add to research showing that DBCIs can be effective in improving physical activity among healthier older people and provide additional evidence that such measures contribute to reduced frailty risk and improved QoL [7, 35]. Notably, these qualitative findings were derived from the two-thirds of participants who remained engaged and so may have been positively biased. With one fifth of participants withdrawing from the intervention arm by 6 months, there is a clear need for DBCI research focused on sustaining engagement, an issue also highlighted in a recent scoping review of mobile health applications for frailty [36].

Our observation that the DBCI use dropped off over time is consistent with findings from the literature. However, our measure of maintained engagement relied on information captured from DBCI use and therefore participant use of other digital or nondigital interventions arising as a direct result of engagement with our DBCI was not captured [34] and hence could be under-reported. Even though technical challenges were identified, participants demonstrated willingness to continue using DBCIs where new content was added to the intervention. Identifying engagement drop-off in real time is an important next step to reduce intervention drop-off through automated nudges and deploying surveys to understand reasons, supporting DBCI adaptive responses to encourage engagement. Some participants were unable to initiate or maintain use of the goal-setting technique because of technical difficulties pointing to the need for a more engaging digital tool as goal setting and action planning are key ingredients that empower people [37] to change behaviour. Participants pointed out that the DBCI needed to be mobile device-friendly. In Australia, internet access for older people increased from 68% in 2017 to 93% by June 2020, with the use of mobile phones, tablets and laptops for internet access increasing from 51% to 78%, 42% to 59% and 50% to 58%, respectively [38]. At the same time, the use of desktops had reduced from 52% to 47%. In the adaptation phase, citizen scientists indicated a desire for voice-interactive systems. In addition to addressing the technical barriers identified through this research, incorporating newer technology such as Agentic Artificial Intelligence [36] to improve the effectiveness of the DBCI while addressing concerns around data privacy, security and decision independence [39] is an important direction for further research.

Strengths and limitations

The involvement of older people as citizen scientists in an earlier phase to inform the implementation strategy and DBCI adaptation for Australia was a major contributor to success. The involvement of consumers in research can improve trust between participants and researchers, improve the design of the research and ensure that digital interventions are built for older people. Another significant factor was our partnership with local government, the government arm best placed to support a ‘bottom-up’ engagement with local stakeholders and a ‘top-down’ policy and intervention approach [40, 41]. Given their connections with the community and local stakeholders, local councils are well-positioned to promote the research to the community and enable the necessary trust for research success as well as provide social infrastructure for Healthy Ageing. Finally, as direct beneficiaries of the research, councils are well-positioned to act quickly, planning for changes in response to the research.

Improvements in vaccination and sedentary behaviour were noted over time in the control group. Because those recruited had an interest in improving their health, the provision of SEARCH results with encouragement for discussion with their general practitioners could have prompted behaviour change, independent of the DBCI. This may have contributed to the lack of observed quantitative differences between groups in addition to the possibility that the method used to measure behaviour change may not have been sufficiently sensitive to detect more subtle between-group differences.

This research was limited to one local council, and effectiveness and experience in implementation are likely to vary across councils and countries. To support generalisability, translational research, including multilocality and multinational studies, is required to better understand the barriers and enablers to the widespread adoption, uptake and use of a frailty DBCI. This research involved only English-speaking participants who were well educated and well resourced, with three-quarters having used digital health technology in the 12 months prior. For greater impact, it will be important to better understand how to engage those from culturally and linguistically diverse communities and those with less digital literacy. DBCIs are designed with complex techniques and not considering the needs of those with less education or digital health literacy could worsen health inequities [42]. We only included nonfrail individuals, and further research is recommended to investigate the utility of DBCI in frail individuals. For those who are less motivated or capable, other methods, such as facilitation by peer volunteers or health professionals, may support the implementation of the DBCI. Finally, no health economics analysis of the cost-effectiveness of this intervention was undertaken, and such research is important if policy makers and funders are to be convinced.

Conclusions

This study suggests that a DBCI that uses evidence-based behavioural change techniques to encourage healthy lifestyle behaviours contributed to frailty prevention and improved QoL. The intervention was usable, acceptable, impactful and effective, with implications for the prevention and management of other noncommunicable conditions, such as dementia [43]. As digital literacy increases among older people and as the acceptance and use of technology grow, DBCIs may offer a scalable approach to health longevity at the population level, and this requires further research. Solving technical barriers and incorporating emerging intelligent technologies could enhance impact, while future work should consider expanded implementation across jurisdictions and include diverse populations, including those already frail.

Declaration of Conflicts of Interest

Professor Visvanathan is a co-founder and investor in Atian.ai, a technology start-up focused on edge-enabled AI automation, monitoring and analytics for emergency and hospital care.

Declaration of Sources of Funding

This work was supported by The Hospital Research Foundation (grant number 2023-CP-IAW-012).