Study design

This study employs an exploratory qualitative approach and was designed in accordance with the Consolidated Criteria for Reporting Qualitative Research [39] and the Standards for Reporting Qualitative Research [40]. These criteria were specified and applied at each stage of the study, ensuring completeness and transparency in data reporting and analysis. The complete semi-structured interview guide is provided in Supplementary S1 File.

A qualitative, cross-sectional methodology was adopted using Interpretative Phenomenological Analysis (IPA), as described by Smith [41]. This flexible and versatile qualitative method assesses the meaning that individuals attribute to their lived experiences, allowing the exploration, description, interpretation, and contextualization of these experiences from the perspective of a small number of participants [42–44]. We focused specifically on perceptions of social-media exercise/motor-action videos used for action observation (AO), and reporting follows COREQ.

Ethics

The study was approved by the bioethics committee of the La Salle University Center for Advanced Studies (CSEULS-PI-005/2020). The procedures performed in this study complied with the ethical principles of the Declaration of Helsinki for research on human subjects [45].

Prior to the semi-structured interviews, participants provided written informed consent, agreeing to participate, allowing audio recording, and acknowledging confidentiality terms that ensured their anonymity throughout the study. Specific measures for data anonymization included removing any personal identifiers from interview transcripts (e.g., names, contact details, contextual references) and replacing them with a unique alphanumeric code assigned to each participant. These codes were used at all stages of the analysis to maintain confidentiality. Additionally, all audio recordings and transcripts were stored on a password-protected server, accessible only to the core research team. Participants were informed that they could request and review their transcripts for accuracy if they wished.

Participants

Patients with chronic nonspecific low back pain (CNSLBP) were recruited from two rehabilitation clinics in Madrid. The inclusion criteria were: a) having low back pain for at least 6 months; b) nonspecific low back pain; c) pain intensity of 3 or higher on the numerical pain scale; d) age between 18 and 65 years; e) having undergone a therapeutic exercise treatment that included instruction and learning through the observation of actions via videos; and (f) having accessed these exercise videos through social media at least three times in the last two months. The exclusion criteria included: a) having undergone back surgery in the past year; b) having specific spinal diseases such as malignant or inflammatory diseases of the joints and bones; c) having experienced spinal trauma in the past year; d) having infectious or tumorous diseases; e) presenting uncorrected visual impairments and/or cognitive disabilities that would prevent adequate understanding of audiovisual material; f) having difficulties in comprehension or communication; and g) having insufficient knowledge of Spanish to follow the instructions related to the collected variables. To enhance heterogeneity, purposive sampling sought variation in sex, age strata, duration/severity of CLBP, and level of digital literacy/usage of social-media exercise content.

For the recruitment of asymptomatic participants, a sample was drawn from the CSEULS campus through social media postings, faculty announcements, direct contact with participants, and posters. The participants recruited included students, administrative or teaching staff, and others in the area who wished to participate in the study. Inclusion criteria for asymptomatic participants were (a) no pathology or pain in the last 3 months; (b) age between 18 and 65 years; (c) previous experience in performing exercise with instruction and learning the exercises through videos; and (f) having accessed these exercise videos through social media at least three times in the last two months. The exclusion criteria for this group were (a) having uncorrected visual disabilities and/or cognitive disabilities that prevented adequate comprehension of the audiovisual material; (b) having difficulties in comprehension or communication; or (c) having insufficient knowledge of Spanish to follow the instructions of the variables collected. All participants provided written informed consent prior to the start of the research.

Study context

To ensure an atmosphere of confidentiality and comfort, interviews and data collection were conducted in private offices within the facilities of the clinics where the study took place. During the interview sessions, only the researcher and the participant were present, without the intervention of third parties. This ensured an atmosphere of trust that facilitated openness and honesty in participants’ responses. When necessary, scheduling accommodated participant preference to minimize burden.

Procedure

Semi-structured in-depth interviews were employed, applying an interview guide developed on the basis of research team meetings and the scientific literature on action observation in patients with pain. The interview protocol was designed according to the criteria set out by DeJonckheere and Vaughn [46].

The interview questions were tested on a pilot group of patients with CNSLBP not included in the study, allowing for adjustments such as simplifying certain terms and including additional questions to further explore related emotional experiences. The finalized guide (items and probes) appears verbatim in Supplementary S1 File. Pilot feedback also led to adding probes on explicit safety cues, graded options, pacing/tempo, and point-of-view (egocentric/allocentric). Pilot data were not included in the analysis. These additions are marked in the guide and reflected in the versioning notes of the audit trail (S2 File).

Additionally, participants were asked the following question to assess their engagement with exercise videos: “Which social media or online platform did you use to watch the exercise videos, and how many times did you do so?”. This preliminary phase eliminated the need for repeated interviews.

Audio recording devices were employed to capture the interviews, with the informed consent of the participants. Interviews lasted between 29 and 55 minutes, with an average length of approximately 42 minutes. Recordings were transcribed literally.

To ensure fuller responses or in cases of brief responses from participants, follow-up questions were used, such as “Could you expand on that?”, “What do you mean by that?”, or “Could you explain more about...?” These additional questions were applied flexibly to gain a deeper understanding of the patients’ experiences.

During the interviews, the researcher took field notes to ensure full coverage of the topics discussed. Immediately after each interview, reflective notes were taken on the relevance of the topics of discussion to facilitate further analysis. This notebook was consulted during the transcription and analysis phases to incorporate reflective comments on the researcher’s insights and the interpretative process. Although interview transcripts were not automatically returned to participants for comments or corrections, participants were informed that they could request and review their transcripts if they wished. Reflexive notes were revisited during coding meetings to make explicit assumptions and to support consensus decisions.

Sample size

A non-probability purposive sampling method was employed to select participants for both phases of the study. This approach is based on choosing participants according to the research question and the objectives of the study [47]. The sample consisted of 10 patients with chronic low back pain and 10 asymptomatic participants. This sample size is in line with recommendations that suggest including 2–12 participants for interpretative phenomenological analysis [48]. The final size (n = 20) was further justified by information power and the idiographic depth sought in IPA. Saturation was operationalized as the absence of novel subthemes toward the end of data collection, confirmed by team consensus.

Qualitative analysis

The qualitative analysis of the data was carried out in seven distinct phases to ensure accurate and consistent interpretation. A step-by-step description of the seven-phase analytic flow, with “evidence hooks” and versioning, is summarized in the Audit Trail (Supplementary S2 File).

In the first phase, transcripts were read and re-read to achieve deep immersion in the data. This was complemented by listening repeatedly to the interview recordings, capturing the emotions, tone, and emphasis of the participants

In the second stage, exploratory notes and codes were recorded both manually (during the reading) and digitally through ATLAS.ti. [49,50]. This software was selected for its capacity to manage large volumes of qualitative data and to facilitate multi-stage coding. Operational boundary rules (e.g., Evaluation of instruction vs Action comprehension), concrete discrepancy examples, and the dated changelog of definitions are documented in Supplementary S2 File.

The third phase involved generating experiential statements by identifying the most salient elements from the previous steps and constructing conceptual meanings based on both transcript content and researchers’ interpretations. In the fourth phase, the emerging connections and patterns among these experiential statements were examined in light of the study’s objectives. The fifth phase mapped and named these connections, resulting in the final coding scheme. The sixth phase repeated these processes for each interview, and in the seventh phase, individual experiential themes were refined to produce a synthesis of group-level themes. The full codebook (themes, sub-themes, analytic definitions, exemplar quotations, and indicative counts by group) is provided in Supplementary information S1 Table.

To ensure reliability and validity within this analytic framework, data triangulation and investigator triangulation were employed. Specifically, two members of the research team coded the transcripts independently, and any inconsistencies in codes or interpretations were resolved through discussion to achieve consensus, with moderation by a third investigator when needed. This multi-researcher approach helped minimize potential biases and enhance credibility. Additionally, reflexivity was maintained through field notes and ongoing critical reflection, allowing the team to acknowledge and manage their own assumptions and perspectives. Coding proceeded primarily inductively within the IPA framework; limited deductive sensitization (e.g., fear of movement, motor simulation, instructional features) was used only to orient early memoing, with final themes emerging from iterative, data-driven interpretation. Independent dual coding was followed by scheduled consensus meetings (third-researcher moderation if required). All qualitative data were managed and coded in ATLAS.ti (Scientific Software Development GmbH).

Data saturation was considered achieved when no new themes or significant ideas emerged. This point was established through consensus among the researchers regarding the sufficiency and redundancy of the information collected. Themes were identified inductively from the data, facilitating an informed and authentic interpretation of participants’ experiences. Lastly, the selection of illustrative quotes for reporting the findings was made by consensus within the research team, ensuring that excerpts represented both typical and revealing perspectives. ATLAS.ti software was used for data management and analysis throughout the process [49].

Positionality and reflexivity

The interdisciplinary team comprised clinicians and researchers with prior experience in CLBP rehabilitation and AO. Anticipated assumptions (e.g., expectations regarding exercise safety and AO utility) were addressed through reflexive journaling, peer-debriefing sessions, iterative memoing, and negotiated consensus during analysis. Decisions about code boundaries and theme formation were documented and revisited across coding meetings. Illustrative memos (e.g., on safety messaging and viewpoint/pacing) are excerpted in Supplementary S2 File.

Trustworthiness

To enhance trustworthiness, we applied a focused set of procedures: (i) Credibility—independent dual coding with consensus meetings; (ii) Dependability—maintenance of an analytic decision log/audit trail documenting codebook evolution and theme clustering; (iii) Confirmability—reflexive memos and retention of the evolving codebook, code-application examples, and theme maps; and (iv) Transferability—maximum-variation purposive sampling and detailed description of setting, participants, and digital context to support informed judgments about applicability. Details of credibility checks, consensus procedures, reflexive memos, and storage of intermediate maps are compiled in the Audit Trail (Supplementary S2 File).

Sociodemographic data from the qualitative analysis

The qualitative analysis was conducted with 20 participants (10 patients with CLBP and 10 asymptomatic participants), adult women and men of similar age. Of the total participants included in the study, 16 (80%) had performed action observation of exercise videos on YouTube, whereas the remaining 4 (20%) utilized Instagram for this purpose. Overall, the frequency with which participants reported viewing these exercise videos ranged from 5 to 10 occasions. To streamline presentation, we report only core variables in Table 1. Table 1 shows the sociodemographic characteristics of the participants who took part in the qualitative study. Table 2 provides a compact summary (theme, sub-theme, analytic definition, exemplar quotation, and indicative counts). A compact codebook with sub-themes, analytic definitions, exemplar quotations, and indicative counts (CLBP/AP), plus the thematic map, is available in Supplementary File S1 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the qualitative study sample.

https://doi.org/10.1371/journal.pone.0326638.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of themes, subthemes, analytic definitions, exemplar quotations, and indicative counts.

https://doi.org/10.1371/journal.pone.0326638.t002

We report simple indicative counts where helpful (e.g., “n/10 CLBP; n/10 asymptomatic”) to convey prevalence while maintaining a qualitative emphasis. Subthemes are presented in order of observed prevalence/centrality across cases.

Results of the interpretative phenomenological analysis

Three main themes and several sub-themes were derived from the IPA (see Table 2): Emotional and Motivational Impact, Self-Assessment of Physical Capacity, and Cognitive Assessment of Movement. Each theme is introduced with a brief link to the guiding research questions and illustrated with a small number of non-redundant quotations selected to exemplify the analytic claims. Detailed descriptions and in-depth analyses of each theme and sub-theme are presented below, including literal quotes from participants to exemplify the analysis, as well as comparisons between the patients with pain (PPs) and the asymptomatic participants (APs). Additional exemplar quotes and cross-case notes are provided in Supplementary S1 Table.

Emotional and motivational impact

This topic focused on describing how watching exercise videos influenced participants’ emotions and motivation, highlighting the importance of this visual medium in the perception of and emotional response to physical activity. The visualization of movement through video can evoke emotions that could in turn influence willingness to participate in exercise, influencing the effectiveness of exercise programs:

PP3: “I am worried about watching the videos and then doing the movements incorrectly and causing more pain.”

PP7: “I pay particular attention to movements that require me to bend my back, I have to be careful with these exercises, I don’t find them very suitable for this pain.”

Fear of bending/load was commonly referenced by CLBP participants (6/10) and less frequently by asymptomatic participants (1/10).

Self-assessment of physical capacity

This theme was based on how participants perceived and evaluated their own ability to perform the exercises presented in the videos. Self-assessment of physical ability can influence confidence, motivation, and willingness to participate in physical activities. It is possible that this perception of physical ability is established by factors such as previous experiences, current health status, and understanding of the instructions.

Cognitive movement assessment

This topic addresses how participants processed and understood information related to the visualized exercises. It includes aspects such as attention to detail, understanding of instructions, and familiarity with the movements, all of which could have influenced the interpretation of the visualized movement and its possible execution.

Threat appraisal versus opportunity appraisal

Across accounts, CLBP participants framed AO through a vigilance–threat lens, especially when videos displayed movement features they associated with flare-ups (e.g., bending, loading, fast tempo). This pattern aligns with contemporary extensions of fear-avoidance perspectives in musculoskeletal pain, where anticipatory threat and protective rules constrain activity even before movement occurs [51,52]. This emotional response appears to be mediated by anticipation of pain and previous negative experiences with movement [53,54]. It also resonates with evidence that pain can perturb action perception/simulation, altering automatic imitation and motor resonance [30].

Our data add that video properties can serve as threat or safety cues within viewers’ interpretative schemas. Conversely, many participants—CLBP and asymptomatic—experienced AO as an opportunity for vicarious learning and preparation, consistent with social-cognitive models in which observational experience can enhance perceived capability and motivation [11,55].

In digitally delivered care, recent large trials indicate that supported, structured self-management can improve back-related disability and pain even when contact is remote, provided guidance is credible and actionable [56]. Our IPA adds that the form of delivery matters: viewers read pacing, camera framing, and explicit error-avoidance cues as signals of “safety to try,” which appears to be a prerequisite for reframing AO from threat to opportunity. Clinically, briefing scripts that normalise initial caution, plus curated video libraries that begin with low-threat demonstrations and progressions, may help patients move from protective avoidance toward exploratory practice. Consistent with this interpretation, randomized evaluations of digitally delivered back-pain programs have shown that, when guidance is structured and credible, outcomes can be comparable to usual in-person physiotherapy, supporting the clinical utility of remote AO-supported self-management [57].

Instructional delivery as the gateway to understanding

Across accounts, participants distinguished between understanding the movement and the quality of instruction. When instructions were concrete (goal clarity), richly cued (error-avoidance, breathing, posture), and paced for learning, action comprehension improved; when absent or rushed, comprehension faltered and avoidance scripts re-emerged. This dovetails with social–cognitive models of observational learning—attention, retention, and reproduction are instruction-sensitive processes [58]—and with neurocognitive evidence that AO engages distributed networks whose responses are tuned by task relevance and familiarity [20]. The practical takeaway is not that any AO will do, but that instructional craft (delivery and representation) is part of the clinical dose. Beyond clinical settings, platform dynamics also shape instructional uptake. Exposure to fitness content through social media can shift intentions and health behaviors via perceived credibility and modeling quality [59].

Our interpretative lens extends these results by showing that viewers map delivery features onto safety and competence signals. Where instructions were dense or rushed, CLBP participants reported breakdowns in “action comprehension,” prompting avoidance; where options and cues were explicit, they described an immediate uptick in “I can do this.” That mechanism plausibly mediates adherence in digital programs.

Contextual moderators: Prior pain, self-efficacy, and digital literacy

Participants’ meaning-making was shaped by their pain history, perceived competence, and comfort navigating online media. Supportive evidence from recent digital back pain programmes shows that guidance intensity and credibility signals matter for engagement and outcomes [56]. In social platforms, perceived creator credibility and parasocial connection can amplify or dampen intentions to act on health content [37]. Our accounts suggest that CLBP participants scrutinise source credibility and adaptiveness (e.g., presence of modifications) more than asymptomatic viewers. Clinically, this argues for (i) clinicians providing short “why this is safe” primers before video use; (ii) level-tagging (beginner/intermediate/advanced) to match perceived competence; and (iii) supporting digital literacy (how to use playlists, captions, playback speed, and save functions).

Cultural transferability and equity considerations

All interviews were conducted in Spain, and participants engaged primarily with Spanish-language or culturally familiar content. Video tone, body ideals, and instructional style vary by culture and may shift safety/credibility appraisals. Digital access and platform literacy also differ across age and socioeconomic groups, potentially widening gaps in who benefits from AO-based resources. We encourage cross-cultural adaptation (language, examples, models) and minimal-literacy design (onscreen captions, iconography, short segments), followed by local user-testing before broader rollout.

From a service-delivery perspective, health-system reviews of digital MSK services identify recurring implementation determinants—governance/guidelines, integration into clinical routines and information workflows, stakeholder acceptance/usability, and financing/incentives—that are directly pertinent to deploying AO videos in routine practice [60].

Practice guidance derived from the IPA

Translating these qualitative insights, we propose three implementable tactics for AO-based exercise content in CLBP rehabilitation:

1. Safety-first scaffolding: Open videos with a one-sentence safety rationale and show an immediately “doable” version before harder variants; keep early exposures short to promote successful first attempts. This matches patient preferences here and complements supported self-management evidence.
2. Instructional minimalism with explicit goals: State the functional aim (“stand up with less bracing”), break the task into two or three observable steps, and add one error cue to avoid. This directly targets the action-comprehension gap described by participants and is consistent with current learning-from-others frameworks.
3. Contextual tailoring and credibility: Where possible, host or curate content under the clinic’s channel/profile, add brief clinician voice-over or captions, and use platform tools (chapters, speed control) to reduce cognitive load. Evidence from digital health literacy and parasocial dynamics suggests these signals can increase intention to follow through. Moreover, parasocial bonds with creators can increase perceived self-efficacy and intention to enact exercise guidance, a mechanism especially relevant when clinicians curate or co-brand content [61]

Strengths and limitations

This study’s core strength is its IPA design, which privileges idiographic depth and participants’ sense-making. By attending to how people with and without CLBP interpret AO exercise videos, the analysis adds explanatory granularity to “what works” signals from intervention studies—clarifying why videos feel usable, for whom, and under what instructional conditions. Methodological rigor was supported through independent dual coding with negotiated consensus, maintenance of an audit trail, and reflexive practices (journaling, memoing, peer debriefing). Sampling targeted maximum variation (sex, age strata, CLBP duration/severity, digital literacy), and sample adequacy was justified via information power within an idiographic IPA frame.

Several limitations should be acknowledged. First, interview data capture subjective experience and intention; willingness to attempt exercises may not translate into behavior or functional change. Second, we did not directly observe motor performance alongside interviews, so interpretative shifts (e.g., reduced threat appraisal, better action understanding) cannot be causally linked to execution quality or outcomes. Third, the single-country sample and specific recruitment settings may limit cultural transferability and generalizability. Fourth, technology familiarity likely shaped participants’ experiences with core video features (e.g., pausing, replay, playback speed), echoing implementation challenges reported for digital back-pain programs and apps [62]. Finally, as is typical in chronic pain research, current emotional state and prior pain history may have influenced recall and interpretation, adding heterogeneity to accounts; our reflexive procedures mitigate but do not eliminate this risk.

Future directions

Design, evaluation, and implementation work should proceed in parallel. Design: co-design AO video libraries with patients and clinicians to operationalize a “safety-first, graded-options” template—clear goals, explicit safety messaging, beginner variants, pacing aligned with learning, and transparent credibility signals. Measurement: develop and validate a concise instrument that indexes the key experiential dimensions identified here (threat appraisal, perceived competence, instructional clarity, familiarity) to inform clinical tailoring and monitor change. Evaluation: test therapist-briefed, curated AO playlists against generic AO content in pragmatic hybrid trials; use mixed-methods to connect interpretative shifts (e.g., threat→opportunity reappraisal; comprehension gains) with downstream adherence, function, and pain-related disability, extending signals from supported self-management and digital implementations. Implementation: evaluate culturally localized curation strategies and clinician-supported onboarding to address digital-literacy variation and trust; examine implementation outcomes (acceptability, feasibility, fidelity, cost) and equity impacts across socioeconomic groups.

Longitudinal studies that embed AO resources into routine care—with periodic professional feedback, progression support, and data-governance safeguards—can test durability of engagement and clinical benefit. Where feasible, future qualitative work should be paired with direct observation of task performance or brief behavioral probes, allowing triangulation between sense-making and execution. Finally, given long-standing evidence that perspective and instructional delivery modulate observational learning, targeted experiments can isolate which delivery elements (e.g., safety cues, graded variants, pacing) most effectively translate interpretative clarity into confident enactment.. Furthermore, it would be relevant to investigate the impact of various types of videos and observational perspectives, given that the literature suggests that egocentric and allocentric perspectives can activate different neural networks and affect the efficacy of motor learning [63].