Study design
The experiment consisted of a factorial 2 (gain-frame vs loss-frame) x 2 (testimonial present vs testimonial absent) between subject design. After having given informed consent digitally, participants were randomized (computer-generated via CastorEDC) into one of the four interventions containing audiotaped video messages that were specifically created for the present study. Randomization was stratified by gender as females are over represented in OA and literature shows they can process messages differently than males (24). Subsequent to seeing the video, participants were asked to complete an electronic questionnaire (CastorEDC). The flowchart of the study design is presented in Fig. 1. This paper was written according to the Consolidated Standards of Reporting Trials - Social and Psychological Intervention (CONSORT-SPI) statement (25) and the Checklist for Reporting Results of Internet E-Surveys (CHERRIES)(26).
Materials
Four messages (average 1073 words, SD = 129) were audiotaped by the same female speaker (maximum duration of 10 minutes), subtitled, and complemented with stock photos. The four different combinations with an exemplary key message are shown in Fig. 2. The gain-framed messages focused on positive outcomes of adhering to treatment, while the loss-framed messages focused on negative outcomes of not adhering to treatment. Loss and gain-framed messages contained the same sentences and stock photos, and only differed by the use of antonyms (see Fig. 2 for an example sentence). When the patient testimonial was absent, the message emphasized facts, whereas when the patient testimonial was present, it depicted an individual’s personal story. The message content was held constant across interventions. The patient testimonial was embedded in a storytelling format describing experiences in the daily life of ‘Babs’, a fictional patient with OA. The four messages can be found in appendix 1. Messages were developed through an iterative process consisting of focus groups with two patient representatives (JM, SA) and several specialists in the field (JO, HD, CE), as well as a brainstorm session within the Persuasive Communication research group (Centre for Language Studies, Radboud University, Nijmegen, The Netherlands.
The Dutch Knee Panel
Recruitment took place through the ongoing Dutch Knee Panel, founded in February 2019 by the Department of Rheumatology, Sint Maartenskliniek, Nijmegen, the Netherlands, a specialized hospital in disabilities of posture and movement. Respondents were recruited via social media and via informational material within the hospital. People could enter their contact details via a website (www.kniepanel.nl) after which they would receive an enrolment questionnaire. The panel consists of people with a self-reported clinical diagnosis or with suspicion of KOA. Inclusion criteria for the panel consisted of the clinical classification criteria for KOA set up by the American College of Rheumatology (clinically diagnosed with KOA or experiencing knee pain for most days of the month, over a period of at least three consecutive months), Dutch language proficient in word and writing, living in the Netherlands, and aged 40 years or over. As people with OA often suffer from multiple diseases and causes of OA differ widely, no specific exclusion criteria have been set. The Dutch Knee Panel was established for research purposes. All members gave permission to be invited for research studies, as well as the re-use and merge of data from multiple research studies. The Dutch medical research ethics committee of Arnhem-Nijmegen waived ethical approval since the panel is not subject to the medical research involving human subjects act (file number 2018–4832). This description of the Dutch Knee Panel is used in publications of multiple studies (see (27)).
Panel enrolment questionnaire
Participants were obliged to complete a questionnaire to initially enrol in the panel. The average time difference between the enrolment questionnaire and the present survey was 21 months. Participants were asked for sociodemographic characteristics in terms of gender (male/female), age (years), education (secondary education or lower (i.e., elementary, high school, and technical or vocational training); higher education (i.e., universities and universities of applied sciences)), and height (cm) and weight (kg). Furthermore, self-reported health status was assessed using the EQ-5D consisting of a short descriptive system questionnaire (EQ-5D-5L (28)) and a visual analogue scale (EQ-VAS) ranging from 0 (“worst imaginable health”) to 100 (“best imaginable health”). In addition, daily functioning was assessed with the Dutch version of the Knee Osteoarthritis Outcome Score (KOOS)(29). The subscale concerning activities of daily living (ADL) consist of 17 items with 5-point Likert scales addressing the extent of problems (no problems – extreme problems) experienced in daily living over the past week. Severity of OA symptoms was assessed through the level of KOA related pain (VAS; 0–10), duration of symptoms (less than one year, between one and five years, more than 5 years), and the presence of OA symptoms in other joints (“Have you received a diagnosis from a doctor or other care provider for osteoarthritis in joints other than the knee?”): hands, wrists, elbows, shoulders, neck, back, hips, ankles, feet. Participants were also asked whether they had received a knee arthroplasty (yes/no; left, right, both) and if so, when (in the last two years, longer than two years ago). Furthermore, the presence and number of comorbidities was assessed by means of a list of 20 predetermined categories (including “other, namely…”).
Additional measurements
Sociodemographic and disease characteristics
Several of the items at enrolment in the Dutch Knee Panel were used for the present study, including sociodemographic characteristics in terms of gender, age, living situation, education, height and weight, as well as self-reported health status (EQ-VAS), and disease characteristics in terms of duration of KOA complaints, whether they experience complaints in one or both knees, which specific complaints they experience, the presence and number of other joints with OA symptoms, comorbidities, and daily functioning (KOOS-ADL). Two items previously measured at panel enrolment where re-assessed, including whether respondents had received or are scheduled for a knee arthroplasty, and KOA related pain complaints (VAS). Time difference between this panel enrolment questionnaire and the questionnaire of the present study was 21 (SD 5.6) months on average.
Pain Coping Inventory
To control for dominant behavioral tendencies to restrict physical activity when in pain as a passive pain coping behavior, the ‘resting’ subscale of the pain-coping inventory (PCI)(30) was used. It comprises 5 items consisting of 4-point Likert scales (hardly ever – very often).
Message perception items
To check whether the interventions had the intended effects, a total of 10 items on 7-point scales (‘strongly disagree' to ‘strongly agree') were presented on counter argumentation (2 items), appreciation (5 items), and credibility (3 items) based on previous studies in the field of narrative persuasion (e.g. (31–34)). Reliability of the counter argumentation, appreciation, and credibility items was good (r = 0.67, p < .001, α = 0.88 and α = 0.85 respectively).
Attitudes, beliefs and intentions
Attitudes towards physical activity and pain medication were measured with six items, based on the Theory of Planned Behavior questionnaire (35) (“If I am physically active for at least 30 minutes at least 2 times a week for the coming three months, that is…” (good/pleasant/wise), and “If I use pain medication when the pain is preventing me from being physically active for at least 30 minutes at least 2 times a week for the coming three months, that is…” (good/pleasant/wise)) consisting of seven-point adjective scales (completely disagree – completely agree). Reliability for the attitude scale about physical activity was adequate (α = 0.68), and good for the attitude scale about medication (α = 0.77).
Behavioral beliefs were measured through two subscales of the Treatment beliefs in knee and hip OsteoArthritis (TOA), developed and validated by members of our research group (4). It assesses patient’s behavioral beliefs about various treatment modalities of hip and knee OA and comprises a total of 25 items consisting of 5-point Likert scales (agree – disagree). Examples of items are “I think that doing exercise activities involves risks” and “I am afraid to use pain medication”.
Intentions were measured with two single items for physical activity and pain medication intentions, based on the Theory of Planned Behavior questionnaire (35)(“I am physically active for at least 30 minutes at least 2 times a week for the coming three months”, and “I use pain medication when the pain is preventing me from being physically active for at least 30 minutes at least 2 times a week for the coming three months”). Answering options of these two items consisted of 7-point Likert scales derived from the Multidimensional Pain Readiness to Change Questionnaire (MPRCQ) ranging from 1=“I'm not doing this now, and I'm not interested in ever doing this” to 7=“I've been doing this for a long time (at least six months)”)(36).
Procedure
Each member of the panel received an e-mail containing information on the study including aim and a unique hyperlink to an informed consent form. After two weeks, a reminder was sent to those who did not respond to the initial survey request. Participation in the survey was voluntary and no incentives were given. After having given informed consent, participants were randomized into one of four interventions (computer-generated via CastorEDC). Respondents were presented with one of four video messages via a link to YouTube. Subsequent to seeing the video, participants were provided a link to the online questionnaire. Participants were asked to declare having seen the video in its totality before continuing to the questionnaire. The questionnaire started with sociodemographic and disease related questions (i.e., whether they have received or are scheduled for a knee arthroplasty, KOA related pain complaints, and PCI). Next, questions were presented on attitudes, beliefs, and intentions, followed by message perception items to control whether the experimental interventions had the desired effects. The survey, including watching the video, took about 30 minutes to complete and included only close-ended questions. The questionnaire consisted of 14 electronic pages, and each page consisted of between 2 and 13 items. No randomization of items or adaptive questioning techniques have been used. Participants were required to answer each question in order to continue, and when applicable items were provided with a non-response option. Respondents were able to review and change their answers via a “Back” button. After completing the survey, respondents were thanked for their time and debriefed. The survey was locked after submission.
Based on the findings of a study with a similar factorial design and outcome measures (37) we expected a difference of 0.5 on intentions to adhere to physical activity between the condition positive framing (mean 4.5) and negative framing (mean 4.0). Assuming alpha = 0.05, 1-beta = 0.80, and SD = 1.5 a sample size of 286 participants is needed to detect a statistically significant difference of 0.5 between groups. Taking a response rate of 50 to 60% in account we decided to invite all members of The Dutch Knee Panel (n = 622) to participate in this study.