We designed this three-arm, parallel groups, randomized controlled trial (prospectively registered on ClinicalTrials.gov on 2023-01-26, NCT05698368) per the SPIRIT guidelines for randomized controlled protocols (Supplementary Tables 1, 2 and Supplementary Figs. 1–5)30. We reported the trial according to the Consolidated Standards of Reporting Trials (CONSORT) Outcomes 2022 Checklist31 (Supplementary Table 3). The Stanford University Institutional Review Board approved this study (Protocol ID: 69227).
Study design and recruitment
We recruited participants via social media (Supplementary Fig. 6), flyers (Supplementary Fig. 7), email (Supplementary Note 2), and word of mouth. Interested individuals accessed study information and provided electronic informed consent (Supplementary Note 1) on REDCap32. After passing a qualification questionnaire, participants completed a baseline evaluation (Survey 1) one to two weeks later (Supplementary Note 3). A week after Survey 1, participants received a link to Survey 2. After reading the introductory text of Survey 2 and choosing to continue, participants were immediately randomized using a randomizer function to one of three conditions: mindset intervention, education intervention, or no intervention. The assigned condition was then presented within the survey. Participants were considered enrolled upon randomization. One month later, participants received a follow-up survey (Survey 3). All surveys and interventions were delivered online (Qualtrics, Provo, UT, USA), and all questions required a response.
Participants
Individuals 45 years or older qualified for the study if they self-reported a clinical diagnosis of knee osteoarthritis or met clinical criteria33. We required that patients were not regularly physically active or undergoing other significant knee osteoarthritis treatments.
Randomization and blinding
We randomly assigned participants (1:1:1) to each group. The consent form shared with participants that the purpose of the study was to evaluate a new online osteoarthritis program and that they would be randomized into one of three groups. After randomization, the survey informed participants of their group assignment (i.e., participants were not blinded) but not about the specific nature of the other groups. The research team was aware of the condition assignment but had minimal interaction with participants, which was primarily done through automated online surveys. The biostatistician performed data analysis blind to group labels.
Rethinking Osteoarthritis mindset intervention
We deployed an iterative, mixed-methods approach to develop (Supplementary Note 4, Supplementary Table 4, and Supplementary Figs. 8, 9), pilot (Supplementary Note 5, Supplementary Tables 5–7, and Supplementary Fig. 10), and refine the digital mindset intervention, Rethinking Osteoarthritis (Fig. 2). This process incorporated mindset theory (Supplementary Fig. 9) and multidisciplinary expertise (i.e., bioengineering, psychology, narratology, and orthopedics). Knee osteoarthritis patients gave feedback throughout the design process and participated in the intervention videos.

The mixed-methods approach to develop, pilot test, refine, and evaluate Rethinking Osteoarthritis, showcasing the integration of various research methods and feedback loops.
Rethinking Osteoarthritis has four modules, each featuring short films and reflective activities (Fig. 3; additional details in Supplementary Note 6). The videos, narrated by subject-matter experts, provide education on knee osteoarthritis and misconceptions surrounding it34, the benefits of exercise, and mindsets about osteoarthritis and physical activity. Reflective questions helped individuals think deeply about the content in the context of their lives, such as, “What mindset do you think would be helpful in overcoming your challenges with osteoarthritis and why?” Each module is approximately 10–20 minutes, with a total intervention length of less than 2 h. Participants had one week to complete the program at their own pace.

The modules of Rethinking Osteoarthritis (mindset intervention, left) and Understanding Osteoarthritis (education intervention, right) are paired with their associated main learnings and, if applicable, targeted mindsets. Below the targeted mindset indicates whether the module aimed to change the mindset implicitly or explicitly called out mindsets. Key intervention elements (bottom) exist throughout both interventions. Modules consist of videos and reflective questions featuring topic experts, interviews with individuals with osteoarthritis, and supplemental animations and b-roll footage. Both interventions total approximately 1 hour of videos and 30–60 minutes of reflective questions.
Understanding Osteoarthritis education intervention
As an active comparison education intervention, we developed Understanding Osteoarthritis (Fig. 3; additional details in Supplementary Note 7). Understanding Osteoarthritis consisted of educational videos and reflective questions of the same duration and required attention as Rethinking Osteoarthritis. We sourced YouTube35,36,37,38,39,40,41,42 videos on topics such as osteoarthritis pathology, risks, symptoms, and treatment strategies. The educational videos were verified for factual content and similarity to the mindset intervention videos, including teaching from live experts and patient testimonials.
No intervention
This group completed the same surveys simultaneously as the other groups but received no additional video content or intervention-based reflective questions. We expected some benefit to participants in the education group; thus, the no-intervention group provided a comparison to quantify this effect.
Data collection
Participants received $10 electronic gift cards after completing Surveys 1 and 2 and a $30 gift card after completing Survey 3. We reminded participants to complete the surveys via email, with up to three reminder emails to complete their surveys. Data collection was completed in September 2023.
Participant characteristics
Survey 1 collected age, sex, gender, body mass index (BMI) from height and weight, race and ethnicity, education, employment, comorbid conditions, unilateral vs. bilateral osteoarthritis presentation, the time since knee pain started, and the time since knee osteoarthritis diagnosis.
Primary outcomes
The primary outcomes assessed at the end of Survey 2 (post-intervention) included Exercise and Osteoarthritis mindsets. Exercise mindset was measured using the Mindset about the Process of Health – Exercise scale. This one-factor, 4-point scale was developed and validated43 to assess mindset about the process of engaging in physical activity (e.g., physical activity is difficult/easy, unpleasant/pleasurable, boring/fun), with a higher score reflecting a more appeal-focused mindset about exercise. Osteoarthritis mindsets were assessed using the Illness Mindset Inventory, which measures three mindsets about the nature and meaning of illness: that it is a catastrophe, manageable, or an opportunity, each on a 6-point scale, with a higher score indicating greater agreement with the mindset. This Illness Mindset Inventory is valid and reliable in individuals with knee osteoarthritis44. We adapted the Illness Mindset Inventory to focus on mindsets about “knee osteoarthritis” as opposed to “chronic disease.”
The primary outcomes assessed at one-month follow-up included knee pain and physical activity. Knee pain was assessed using the question, “What was your average osteoarthritis-related pain over the past week?” and measured on an 11-point Numeric Rating Scale from 0 (no pain at all) to 10 (the worst pain imaginable). We assessed physical activity using the Physical Activity Scale for the Elderly45. The scale is scored from 0 to 793, with higher scores indicating higher physical activity levels.
Secondary outcomes
The secondary outcomes assessed post-intervention included the Knee Osteoarthritis Knowledge Scale46, the Illness Mindset Inventory for body mindsets44, and the Adequacy of Activity Mindset Measure47. The secondary outcomes assessed at one-month follow-up included (1) the knee pain and function subscales of the shortened version of the Western Ontario and McMaster Universities Arthritis Index, (2) the perceived need for surgery26, (3) chosen symptom management strategies19, (4) the Brief Fear of Movement Scale for Osteoarthritis48, (5) the pain and other symptoms subscales of the Arthritis-Self Efficacy Scale49, and (6) the physical and mental health subscales of the Patient-Reported Outcomes Measurement Information System v.1.1 Global Health Short Form50.
At the end of the study, we prompted qualitative insight as an exploratory measure: “We would like to learn more about your own experience with knee osteoarthritis and exercise since participating in this study. In your own words, please share your experience with knee osteoarthritis and exercise over the past three weeks.” Responses were reviewed to identify illustrative examples that contextualize the quantitative findings.
Additional details on measures are on ClinicalTrials.gov (NCT05698368).
Sample size
The mindset intervention was considered beneficial if we found improvements in one or both of the primary outcomes of pain or physical activity level. An a priori power analysis was performed at a 5% significance level and effect size of 0.3 to achieve 80% power in detecting an improvement of 26 points in the Physical Activity Scale for the Elderly (additional details in the Supplementary Information – SPIRIT checklist). This analysis led to a sample size of 139 per group (417 total), which was adequately powered to detect a clinically relevant difference of 2 points in NRS pain14. We aimed to recruit 501 individuals to account for a 20% dropout rate.
Statistical analysis
An author blinded to group assignment (D.G.) completed all analyses in R (v4.3.2)51. We summarized baseline characteristics with descriptive statistics. We used independent sample t-tests for continuous variables and Pearson χ2 or exact tests for categorical variables to test baseline group differences. We evaluated mean change scores within groups along with Cohen d (Eq. (1); where \(\barx_2\) denotes the mean of variable after intervention, \(\barx_1\) denotes the mean of variable before intervention, and \(s\) denotes the standard deviation of the change in the variable due to intervention):
$$cohen\,d=\frac\barx_2-\barx_1s$$
(1)
We evaluated mean differences in change scores between the groups using independent t-tests (Eq. (2); where \(\barX\) is the mean of changes in the first group and \(\barY\) is the mean of the changes in the second group, \(s_pooled\) is the pooled standard deviation from the two groups):
$$\frac\barX\,-\,\barYs_pooled\sqrt\frac1m+\,\frac1n$$
(2)
We present the results with accompanying 95% confidence intervals (CIs) and P values, as well as with and without Bonferroni correction. Our analysis included only participants who completed all surveys in full, as each survey question required a response, ensuring no missing data within completed surveys.
link