The methodology of this pilot study aligns with previous research into the therapeutic affordances of social media in PWCP [25-27]. This approach recognizes that each patient has individualized needs, building on the principles of participatory health care [28].

The setting chosen for the study was the Royal Melbourne Hospital—Pain Management Services (RMH-PMS). The Royal Melbourne Hospital (RMH) is a large publicly funded hospital in Melbourne, Australia. It offers one of the most comprehensive outpatient pain management services in the country. RMH-PMS has between 900 and 1100 outpatient referrals per year. At any one time, there are from 200 to 300 PWCP on the waiting list, and they have a wait time of 6 to 8 months for a first appointment. RMH-PMS provides an ideal context in which to test a social media intervention.

The Human Research Ethics Committees of Melbourne Health and the University of Melbourne approved this study (ID No. 2014.043). During the review of the project, the ethics committee specifically considered the following areas of study design:

1. The sample: comments were made about representativeness and selection of the sample, as well as the impact using social media may have on expectations of patients for management of their condition.

2. Social media resources: feedback centered on anticipated outcomes and what impact social media would have on participants’ conditions. Queries related to the appropriateness of the resources, who would be responsible for the content, and monitoring the resources.

3. Logistics and staff allocation: this covered staff and time required to recruit, to manage participant numbers during the study, to collect data, and to conduct follow-up.

The sample pool consisted of patients with chronic pain—pain of greater than 3-months duration [29,30]—on the RMH-PMS outpatient waiting list. Sampling was sequential and inclusion criteria were applied (see Table 1). Exclusion criteria were also applied, including the following: a change in priority status (ie, immediate intervention required) or discharge from the waiting list.

A clinically relevant sample was sought. Desired cohort size for this study was estimated based on the following factors: participant numbers used in pilot studies within a similar domain [32-38], the number of referrals to the RMH-PMS program within a 3- to 4-month period, and the number of patients managed in the pain service at any particular time. Based on a combination of these factors, the aim was to enroll approximately 20 patients.

Potential patients classified as priority 2 were sourced from the RMH-PMS Direct Access Unit referrals database. Priority 2 includes patients receiving support via a primary care practitioner while waiting for a consult at RMH-PMS. Referrals are received from general practitioners in the local area, community health centers, other medical specialists, and internal referrals from RMH. Each referral was reviewed for preliminary suitability by two of the study’s investigators who met nine times between April and September 2014. Potentially suitable participants were flagged based on matching information recorded in their referral with study inclusion/exclusion criteria (ie, English language ability, current interventions/treatments, any significant conditions requiring treatment noted in the history). Recruitment was conducted via phone calls from July to September 2014. The people reached were advised of typical waiting list times and introduced to this study as a self-management strategy available in the meantime. The phone transcript can be found in Multimedia Appendix 1.

Interested patients were emailed with a link to the study website and a unique identifier. The link led to SurveyMonkey—an online survey site—where the study information, consent form, and pretest questionnaire were hosted. After 1 week, follow-up phone calls were made to discuss the study, determine suitability, confirm Internet competence, and to ensure the prestudy questionnaire was submitted to confirm registration in the study. Once registered, contact was made to discuss final details and educational material was emailed that contained links to the suggested social media resources and instructional videos—filmed by the study’s primary investigator—to act as an introduction to the resources and provide instructional information for using the resources. The educational material can be found in Multimedia Appendix 2.

Patients who were ineligible to participate in the study were provided with information about online resources (ie, painHEALTH [12]) if they desired for their future reference.

Each enrolled patient completed the intervention over 12 weeks and because recruitment was staggered, some commenced at different points in time. The entire intervention period ran from July 24 to December 5, 2014. Waiting list time is typically between 6 and 8 months with specialist preclinic pain education provided close to the medical admission appointment. Therefore, treatment was not delayed by completing this study. No incentives were offered to participate. It was also explained that the study was a pilot study and that participant feedback may help to shape future consideration of social media resources for the RMH-PMS.

Patients were informed that the intervention consisted of using the suggested social media-based pain management resources in an unrestrained fashion over the allotted time period. Patients were asked to interact with these at their own pace. They were given the autonomy to be selective as to which they interacted with and how. Of interest was the impact that use might have on their condition and on their understanding about pain management, and finally, to know how they interacted with the resources—this would be examined during monthly semistructured interviews that will be described further.

The social media resources incorporated into this study included a large chronic pain support community on Facebook, a selection of chronic pain blogs, and pain management YouTube videos filmed by painHEALTH. Further detail about these resources and their selection for this study can be found in Multimedia Appendix 3. These particular platforms were suggested based on social media used by PWCP from a global online survey investigating social media use in chronic pain self-management. Also, platforms were selected based on each platform’s ability to foster various therapeutic affordances that appear conducive to positively impacting health effects in chronic pain [26,27]. The resources were reviewed and agreed on by all study investigators from both the Health and Biomedical Informatics Centre at the University and RMH-PMS.

Pre- and postintervention data were collected, as well as data collected at monthly intervals during the study. To start, participants completed a questionnaire on SurveyMonkey that amalgamated demographic information, chronic pain status, and patient-reported outcome measures (PROMs)—pain interference and pain self-efficacy—into one survey (see Multimedia Appendix 4). Patient-reported outcomes (PROs) provide insight into the patient’s perception of the impact that interventions have on their health [18].

Pain interference (PI) is an example of a regularly examined standardized outcome that measures the burden on an individual across a wide range of health-related quality-of-life (HRQOL) measures [39,40]. PI was measured using 16 items from the Patient-Reported Outcomes Measurement Information System—Pain Interference (PROMIS-PI) item bank and included one item from the pain behavior item bank to measure pain severity. Unlike common legacy outcome measures used to measure chronic pain, the PROMIS-PI item bank demonstrates good reliability and validity across a variety of chronic diseases, including chronic pain, and shows strong correlations to other common outcome measures, allowing findings from this study to be compared and generalized in the future [39,40].

Pain self-efficacy (PSE), or confidence in one’s ability to perform certain tasks in the face of pain, was measured using the 10-item Pain Self-Efficacy Questionnaire (PSEQ) [41]. Social cognitive theory underlying PSE was discussed briefly in the introduction [1]. Based on social media’s ability to foster peer modeling and support, it was logical to include the PSEQ as a primary measure. The PSEQ has frequently been cited as a standardized outcome measure in chronic pain management. Accordingly, efficacy determines the effort and persistence an individual will apply when faced with hurdles and adverse experiences. Like PROMIS-PI, the PSEQ provides a means to overcome inconsistency and generalizability among previous existing legacy measures covering a range of general behaviors and activities, and directly contextualizes self-efficacy in regard to living with pain [41,42]. Similar to the argument made for the use of PROMIS-PI, the PSEQ is favored over legacy self-efficacy measures because it more effectively overcomes the variability in measurement presented by other instruments.

At the end of the study, patients were guided to a second SurveyMonkey questionnaire that amalgamated all PROMs (PROMIS-PI and PSEQ) and a social media-use instrument into one survey (see Multimedia Appendix 5). The social media-use instrument sought to examine which resources were used, the amount used, features of each used, positive/negative aspects of using the resources, and perceptions toward various therapeutic affordances. The same line of questioning was used for each resource—Facebook page, blogs, and YouTube videos. Questions about therapeutic affordances were phrased to better understand the degree to which each affordance is present and impacts PROs. Five therapeutic affordances of social media, refined through a global online survey of PWCP [26], were examined through 15 statements, each consisting of three exploratory components. These measured the following: (1) self-presentation—preferences regarding one’s identity, (2) connection—using social media to connect with others, (3) exploration—guidance toward useful information, (4) narration—sharing experiences of chronic pain, and finally (5) adaptation—motivation, frequency, and type of use.

At monthly intervals during the intervention period, patients were also contacted on the phone by the primary investigator to complete a brief 10- to 15-minute, semistructured interview (see Multimedia Appendix 6). Semistructured interviews are well suited to small-scale studies with small participant numbers, and their utility lies in their ability to collect rich qualitative data that can be analyzed in a variety of ways to supplement pre- and posttest survey data [43,44]. Data collection gave people a chance to discuss participation and study progress, but more importantly offered patients the opportunity to help shape social media use in future studies and interventions by collecting information regarding social media use and perceptions toward the five therapeutic affordances free of coercion. These phone calls were not medical consultations and patients were advised of this. Any queries of a medical nature were flagged for follow-up by a member of the clinical team.

Major study design procedures and the study process are the focus of the results described. Brief descriptive statistics of pre- and posttest survey data are also presented. Pertinent to the study was the examination of barriers to engagement with the resources and intervention. To explore these, thematic analysis was employed using a grounded theory inductive phenomenological approach on the semistructured interview data to uncover any emergent themes.

Deductive coding of semistructured interviews was also conducted to categorize patient comments against our list of five therapeutic affordances: self-presentation, connection, exploration, narration, and adaptation. These results will be the subject of a separate manuscript.

Furthermore, the posttest PROMs were analyzed using paired t tests and Fisher's exact test. In order to compute the PROMIS-PI data, raw scores were first translated into a T score for each patient. The T score rescales the raw scores into a calibrated and standardized score with a mean of 50 and standard deviation of 10. The value of hypothesis testing is limited due to the small sample size and given that potential covariates such as gender, age, and educational level were not able to be stratified [45].

Coding of the semistructured interview data yielded eight separate engagement themes (see Table 5). Based on the number of times each barrier was identified, time poor and low specificity of resources to one’s condition were most noted. Effects of medication was also highlighted; patients noted that this was due to their sedative effect and/or impact on concentration. Only 2 out of 17 patients (12%) suggested access to the Internet was an issue.

Some of these barriers were also identified in the free-text responses of the posttest survey, for example, low specificity of resources. According to Patient SM021, “I have a more localized pain condition and most of the resources were for generalized conditions so I had trouble relating to many of them.” This was echoed by Patient SM112, who stated, “I think that because they weren’t about my pain and there were none about the pain I was going through I found it not very useful.”

The aim of this study was to examine the pilot-testing of social media use with patients in the clinical setting, focusing on major aspects of the study design. The study also produced evidence to support social media use and PROs from earlier studies [26,27]. However, these findings are not the focus of this protocol paper.

The use of mixed methods in eHealth research is often advocated [46-48]. This study also presents a strong case for the complementarity of such a research design. The richness of both quantitative and qualitative datasets and their value to examining the effectiveness of pilot studies has been discussed within a similar research context to this study [32,36,49].

By combining empirical PROs data from PROMIS-PI, PSEQ, and from the social media-use instrument with the qualitative semistructured interview data and free-text responses about social media use from the posttest survey, a rich dataset was obtained. This becomes particularly useful with a small sample size, such as the one in this study. It helps to cross-validate findings and strengthen converging inferences [46].

Collecting empirical data about social media use was able to corroborate semistructured interview data. Similarly, data collected regarding patient perceptions toward each of the five therapeutic affordances of social media helped validate findings from a global online survey of PWCP [26,27]. This enhances the scope to generalize study findings in future research and across different study areas and conditions within similar domains.

The recruitment procedure used in this study was extensive and demonstrates a targeted method aimed at recruiting the most suitable clinical cohort that might benefit from the intervention. It is the belief of the research team that the method for recruiting such a targeted cohort in this study is appropriate and would be scalable in a larger-scale randomized controlled trial (RCT) study. This is because the aim was to refine the most suitable cohort required to test the effectiveness of social media to positively impact PROs. This was not a study of readiness to adopt social media. A similar targeted approach has been seen in other pilot studies within a similar study domain [32,36,50].

However, targeted recruitment does highlight pertinent ethical considerations, such as “representativeness and selection.” Internet access and English language problems were two considerations noted from the RMH-PMS population. Similarly to this study, most social media interventions target Internet-competent, social media-using participants and/or people thinking about engaging in health behavior change [23,51,52]. While some may argue that this may limit the generalizability of findings, it has also been argued that chronic pain patients almost always self-select their own management (ie, medication, physiotherapy, exercises, surgery, etc). Hence, a self-selecting study design is still ecologically valid [11].

Demographic characteristics predicting participation in social media or other Web-based interventions for chronic disease management are also interesting and relevant. This has traditionally been skewed towards well-educated females of a relatively high socioeconomic status. Marital status and age are more varied but studies are typically slanted toward those in relationships/married and aged 30 to 60 years [53-55]. Literature has also previously reported that people living with chronic illness are typically more representative of lower socioeconomic status and educational level [51,56,57]. What was found in the present study is that the clinical setting is representative of a greater demographic spread and it was this clinical PWCP population that was of interest for examination. Patients were an equal mix of males and females, education level was generally lower (in this study, 9/17 [53%] completed high school or less), more than half were unemployed due to ill health (9/17, 53%), and age range varied from approximately 18 to 59 years. Earlier studies, such as that of Berman et al [33], reported significant improvements in PROs for older chronic pain patients (55+ years). They also reported minimal issues for participants navigating the online resources, suggesting that older adults in the chronic pain setting may also derive benefit from online interventions. In this study 19 out of 138 (13.8%) potentially suitable study recruits were aged 55 years or over.

According to Sheaves et al [36], the study completion rate of 53% (9/17) is comparable with other Web-based health studies. Reasons for attrition according to interview data about barriers to engagement, may have been low specificity of resources to one’s condition and lack of time to use the resources.

The semistructured interviews conducted monthly by telephone were an important tool in understanding reasons for engagement or nonengagement above. In contrast to overall completion rate, results showed that 94% (16/17) of patients were able to be engaged by phone. This enhanced collaboration between patients and the research team and placed greater emphasis on patient preferences and perceptions about using social media for their self-management. In other studies, such as that of Sheaves et al [36], email was used to collect data. The authors reported that lack of response to feedback emails was high. Hence, conducting semistructured interviews by telephone rather than via email may be a more successful data collection and patient engagement method in social media studies.

As noted from patient feedback, low specificity of resources to one’s condition negatively influenced engagement. It has been reported that often those living with chronic conditions have multiple comorbid manifestations and, hence, Web-based resources can fail to cover enough of their individualized needs for them to be deemed of significant value [58]. As could be seen from the research design, general chronic pain management social media resources were suggested to patients, not those specific to any one chronic condition (eg, low back pain, rheumatoid arthritis). This was decided based on current evidence-based approaches to chronic pain management, focusing on holistic multi-faceted, versus disease-specific, management [10]. The conflict observed between a desire to deliver evidence-based practice (EBP) and patient preferences for their own management highlights the need for greater emphasis on shared decision-making models between patient and clinician to achieve success in larger-scale studies in the future [18].

Suggesting certain social media resources as a starting point was done to provide some element of quality control and uphold ethical standards of care. However, freedom was encouraged for patients to use the resources as they wished and explore different avenues using the social media resources. Unlike many earlier studies conducted in this space, the aim of the present study was to more closely replicate social media use in day-to-day life (ie, open, engaging, collaborative, participatory) rather than to create a specific online intervention that dictates exactly what patients engage with and how they engage with it.

Finally, facilitated engagement of social media interventions, or involvement of clinicians, has been reported as another positive way to improve participation and intervention adherence. In turn, this has a positive flow onto PROs [9,20,53-55]. In Hoch et al [32], the entire pilot intervention was guided by a clinical nurse. This led to a completion rate of 24 out of 28 (86%). However, unlike this study, that of Hoch et al [32] was a social media intervention that translated a traditional face-to-face intervention into one delivered in a virtual world. The nurse had an integral role in delivering each session, thus mimicking offline management. Autonomous self-management was the goal for this study, not translating an intervention into an online one. Hence, using social media resources cannot be truly facilitated in the same manner. Empowering patients to make their own decisions about using social media in this manner also has positive connotations for study logistics, suggesting that fewer clinical staff would be required to run future larger-scale studies.

This pilot study has outlined key considerations for conducting social media interventional research in the clinical setting, in particular, study design, recruitment, and engagement.

Targeted recruitment of social media users indicates that enthusiastic, competent social media users may be still largely underrepresented on pain management services’ waiting lists. Therefore, these interventions may not yet be appropriate for all PWCP. Further work is required to ensure that those in need of online support will also be catered to by social media’s integration into clinical service models. In line with recommendations by Sheaves et al [36], future studies would be well advised to (1) include general practice sites in research, where patients may have more need for information (given that they are more likely to be earlier in the course of managing their pain) and (2) include eHealth literacy education, training, and support as part of care models for those who have low computer literacy skills, but may stand to benefit from online resources [35].

As patients become increasingly connected and active shared decision makers in their self-management, researchers would be advised to pay close attention to study designs that give patients greater flexibility and empower them to make decisions about the online resources they use. This is where the potential for social media sits above traditional Web-based interventions. Social media resources must actively engage patients as seen in this study. Finally, acknowledging patient preferences for resources that adequately address disease-specific needs is also a consideration.

Future research into the effectiveness and potential for social media use in the clinical management of chronic pain is warranted. While this study cannot ascertain what effect the use of social media resources can ultimately have on hospital waiting lists, a considered approach to conducting this type of research has been offered. Future studies need to focus on larger-scale, multicenter RCTs and involve patients in the intervention design in order to achieve desired effect sizes and for evidence to be sound.