The OA Coach app was co-designed with a multidisciplinary team including osteoarthritis researchers, rheumatologists, physiotherapists, and consumers living with osteoarthritis. The app prototype was developed by Vulsen, a developer with expertise in developing mobile apps for chronic disease management [23,24]. The study team (VD, NB, JE, SK, DJH) tested the mobile app prototype for approximately 2 weeks. A 1-hour co-design workshop was also held with 6 consumers with lived experience of osteoarthritis to gather initial feedback regarding the features and visual aspects of the app prototype. Co-design participants were given access to the mobile app to download onto their personal devices 1 week prior to the workshop. In the workshop, a brief overview of the app was presented, followed by the participants mapping likes or dislikes and potential areas for improvement for the app and a group discussion. Following alpha-testing, issues with the functionality or content of the app were relayed to the developer for resolution prior to the 6-week feasibility study.

Details on the main features of the app are summarized in Textbox 1. An overview of the Home, Progress, and Learning pages is displayed in Figure 1. The app’s log-in page required participants to enter their date of birth and answer questions about their knee pain (pain severity, stiffness duration). This information was used to make a clinical diagnosis based on the National Institute for Health and Care Excellence guidelines [25], which then determined access to the app. Following this, participants were prompted to complete questions about their weight, height, sleep, mood, and current activity levels in order to trigger corresponding educational modules targeted to the users’ needs. For example, if a participant’s BMI was above a healthy range or the questionnaire responses indicated poor sleep, participants would be assigned the “Weight Management” or “Sleep” modules, respectively.

Exercise plans for the app were displayed under the “Learning” page. The page included a brief introduction to knee-strengthening exercises, including the benefits of exercise, how to complete the exercise programs (warm up, joint specific exercises, cardiovascular endurance exercises [if possible], warm down), how to assess difficulty and progression, and frequency of program. Separate plans for individuals with good mobility, moderate mobility, and limited mobility were presented, with 3 progressions within each plan. Each exercise plan included instructional photos and descriptions as well as suggested dosage. Participants were free to choose a plan which best aligned with their current mobility status.

Participants were also provided with an activity tracker to wear for the duration of the study. The app requires that participants wear the Fitbit tracker to collect baseline data on sleep and physical activity levels in order to generate appropriate goals. The step and sleep count data are displayed on the “Home” and “Progress” pages of the app. The generated goals are displayed under the “Goals” heading on the “Home” page.

Behavior change theory was used to help motivate users to increase their physical activity and sleep levels, as well as complete tasks such as logging pain scores or completing surveys. BCTs, such as goal setting, feedback on behaviors, and graded tasks, were integrated into the setup process and push notifications. Participants were asked to set goals during the initial setup process of the app, where they were asked about their current exercise and physical activity levels in a questionnaire. Their responses were used to create weekly goals. For example, if a participant enjoyed playing a sport once a week, a goal would automatically be created to increase the frequency of the sport to align with the World Health Organization physical activity guidelines for older adults [26]. Feedback on behaviors, as well as graded tasks, was provided to participants as push notifications for meeting their step goal at the end of the day. Self-monitoring of pain levels and body weight was also encouraged through notifications and logged in the “Progress” page of the app, along with sleep and step count data. Further details of various OA Coach features mapped to the Behavior Change Taxonomy are outlined in Multimedia Appendix 1.

The feasibility study involved providing the OA Coach mobile app to study participants for 6 weeks to assess the usability of the developed app. The protocol for the feasibility study was registered on the Australian and New Zealand Clinical Trials Registry before commencement (registry number: 12623001223628). The results of this study are reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) checklist (Checklist 1).

Ethical approval was granted by the Northern Sydney Local Health District Human Research Ethics Committee (2023/ETH02288). Participant eligibility was assessed using an online survey, and participants who passed the screening survey provided electronic consent through a Research Electronic Data Capture e-consent form.

Potentially suitable participants were identified by rheumatologists and physiotherapists within the Osteoarthritis Chronic Care Program at Royal North Shore Hospital, Sydney, Australia, and were provided with a link to the eligibility screening survey. Participants from past clinical trials who had consented to be contacted about future research studies were also emailed and invited to participate in the study.

Individuals were eligible for the study if they were aged 45 years or older, had knee pain ≥4 out of 10 on an 11-point Numerical Pain Rating Scale (NPRS; where 0 represents “no pain” and 10 represents “worst imaginable pain”), and experienced knee stiffness lasting no longer than 30 minutes after rising from bed in the morning. Participants were also eligible to participate in the study if they had knee stiffness >30 minutes but had been diagnosed with knee osteoarthritis through radiographs or a health care practitioner. This inclusion criterion was added as an amendment to the study, as a number of participants with a confirmed diagnosis of osteoarthritis were being excluded due to stiffness slightly exceeding 30 minutes. These participants were contacted either from either our database of previous participants or were recruited from the Osteoarthritis Chronic Care Program at Royal North Shore Hospital thus already having a confirmed diagnosis of osteoarthritis. Participants also required ownership of a smartphone with internet access in order to download the OA Coach app. The exclusion criteria for the study were as follows: a diagnosis of inflammatory arthritis, such as rheumatoid arthritis or gout, and any comorbidities or medical conditions that would prevent participation in the physical activity component of the app.

All participants received a Fitbit tracker (Fitbit Inspire 2) in the post and instructions on how to download the OA Coach app on their personal smartphones and link the Fitbit tracker. The OA Coach app is freely available on app stores; however, its access remained password-protected during the evaluation period. Participants were instructed to wear the Fitbit tracker for a baseline period of 1 week and were then asked to interact with the app daily for 6 weeks. Participants were instructed to contact the study coordinator through email if they experienced any technical issues in setting up the OA Coach app or Fitbit tracker. The study coordinator aimed to resolve any technical issues through email, prior to contacting participants by phone.

Three sets of data were obtained from the study including baseline and 6-week online questionnaires, data from the mobile app (eg, step count, hours slept, the 21-item Depression, Anxiety and Stress Scale [DASS-21] and Pittsburgh Sleep Quality Index [PSQI] questionnaires), and qualitative feedback from surveys at 6 weeks.

The mHealth App Usability Questionnaire (MAUQ) for Standalone mHealth Apps Used by Patients was used to assess the usability of the OA Coach app for the primary outcome [27]. The questionnaire consists of 18 statements, and participants were asked to rate their agreement with each statement on a 7-point Likert scale, ranging from strongly disagree to strongly agree [27]. To determine the usability of the app, the total score was calculated and averaged across responses, with higher scores representing higher usability.

The Modified Computer Self-Efficacy Scale (mCSES) and the Osteoarthritis Knowledge Scale (OAKS) were used to assess technology self-efficacy and osteoarthritis knowledge, respectively, at baseline and 6 weeks through surveys. The mCSES asks respondents to rate their confidence using a new technology on a scale from 1 to 10, where 1 represents “not confident at all” and 10 “totally confident” [28]. The OAKS is an 11-item scale assessing important knee osteoarthritis knowledge, with scores ranging from 11 to 55, with higher scores indicating greater knowledge about osteoarthritis [29].

Secondary outcomes collected through the app included changes in knee pain (NRS), BMI, mood, sleep quality and disturbances, physical activity (average step count over the last 7 days), and time asleep (average sleep duration over the last 7 days). The DASS-21 was used to measure the emotional states of depression, anxiety, and stress. The questionnaire consists of 21 statements, with 7 statements pertaining to each subscale. Participants were asked to rate their agreement with each statement from 0 to 3 (where 0 indicates “Never,” and 3 indicates “Almost always”) [30]. Scores for each subscale were summed and multiplied by 2 for interpretation. The PSQI is a 19-item self-reported questionnaire assessing various aspects of sleep quality (eg, duration, latency, disturbances) over the past month. Scores for each question range from 0 to 3, with higher scores indicating more acute sleep disturbances [31].

Knee pain (NRS) and body weight (kg) were self-reported by participants. Physical activity and time asleep were automatically collected from participants’ Fitbit devices, which synced to the mobile app.

Participants were also asked to provide any feedback regarding the app through open text fields with the option of including their contact details should they wish to discuss the feedback in person with a study team member.

The outcome measures collected in this study were exploratory in nature and were not used for progression, modification, or stopping decisions within the study protocol.

As this study was a feasibility study, no formal sample size calculations were conducted. We aimed to recruit a total of 30 participants for this study, which is an acceptable size for a feasibility study [32].

Participant baseline demographics and clinical characteristics were expressed using means and SDs or medians and IQRs if not normally distributed. The primary outcome (MAUQ) was summarized descriptively. The secondary outcome measures at baseline and 6 weeks were summarized using median and IQR. The change scores (baseline minus 6 weeks) were summarized using median and IQR. Analyses were conducted using STATA (version 18.0). Qualitative data from participant feedback were collated into an Excel spreadsheet, and each comment was mapped to a general category (eg, App download, Fitbit integration, goals, instructions, learning materials, notifications, design, or other) by 1 researcher (NB). These were reviewed by a second researcher (VD) and suggestions on how to improve the app were discussed and documented. The suggestions were then reviewed within the wider study team, including the app developer. Changes that the researchers or developer deemed reasonable and appropriate informed the next iteration of the OA Coach mobile app. Changes that required additional resources and funding were temporarily put on hold.

A total of 30 participants were recruited between March 2023 and April 2023. There was an even split of female and male participants (n=15 per sex), and participants had an average age of 66.9 (SD 9.1) years and a median pain level of 6.0 (IQR 5.0-6.8) on an 11-point NRS. The participant baseline demographics are summarized in Table 1. Due to technical issues, there were some participants who did not have complete step, and sleep data at baseline and demographics in Table 1 are based on a complete case analysis.

Overall, participants found the app usable, with the majority of scores greater than 5 out of 7 on the MAUQ. Most participants found the app easy to use (mean 5.8, SD 1.5), easy to learn how to use (mean 6.1, SD 1.2), and were satisfied with the amount of time involved in using the app (mean 6.0, SD 0.9). The lowest scoring statements were regarding improving access to health care services (mean 3.6, SD 1.8) and the app functionalities and capabilities (mean 4.1, SD 1.9), which are in alignment with the qualitative data collected. The results of the MAUQ statements are summarized in Table 2.

The NRS median pain reduced from a median of 5.0 (IQR 4.0‐6.0) at baseline to 4.0 (IQR 2.0‐5.0) at 6 weeks, giving a within-patient median reduction of 1 point (IQR 0.0‐3.0). The modified Computer Self-Efficacy Scale (mCSES) moved in the opposite direction: the median score declined from 89 (IQR 74‐100) to 81 (IQR 67‐93), reflecting a median change (reduction) score of 2.5 points (IQR –1.0 to  7.0). The OAKS score did not indicate a substantial change in knowledge (median change of 0.0 points, IQR –2.0  to  3.0).

The magnitude of change in BMI was very small—change from a median 26.7 (IQR 24.4‐31.2) to 26.6  (IQR 24.5‐29.1)—corresponding to a median within-patient change of 0.02 (IQR –0.15  to  0.05). There was a small improvement in the psychological distress measure. The median depression scores falling from 6.0 (IQR 2.0‐8.0) to 4.0 (IQR 2.0‐8.0), anxiety from 4.0 (IQR 0.0‐6.0) to 2.0 (IQR 2.0‐6.0), and stress from 10.0 (IQR 4.0‐16.0) to 6.0 (IQR 2.0‐12.0). Sleep quality improved slightly, with the median PSQI score decreasing from 6.5 (IQR 5.0‐10.5) to 6.0 (IQR 4.0‐9.0).

The median 7-day step count increased from 8073 (IQR 5392‐10,726) to 8616 (IQR 6028‐10,057), a median gain of 446 steps (IQR –1433  to  2,322), while the average nightly sleep duration remained almost equal at median 7.02 (IQR 5.56‐7.68) hours and 6.90 (IQR 5.40‐7.65) hours. These changes are minimal and may be partially explained by the cohort that was recruited. Most participants were either previous participants of our clinical trials or had been receiving care from an osteoarthritis management team and therefore were more likely to be knowledgeable about their condition, already be engaged in some form of physical activity and have sufficient sleep hygiene. Secondary outcomes are summarized in Table 3.

Overall, when comparing outcomes at 6 weeks to baseline, there was a small reduction in pain and psychological distress; slight improvements in sleep quality and physical activity; and stability in knowledge, BMI, and sleep duration. These outcomes were expected in a small, uncontrolled feasibility study that was not designed to detect statistically significant changes within the group.

The qualitative feedback largely focused on usability and suggestions for improving the app, such as better integration of the exercise programs and improved personalization of goals and activity tracking. Additionally, participants highlighted the need for improved onboarding support. Participants also noted confusion over the labeling of certain features. The main areas for improvement and updates made to the next iteration of the OA Coach app are outlined in Table 4.