Patients’ interpretations of individual PROMIS-10 Global Health items: a cognitive interview study in high-impact chronic pain

The study’s first objective was analyzed using two approaches. The part concerning the correspondence between participants’ interpretations and the intended meanings of the PROMIS-10 GH items was analyzed mainly deductively using Coding Reliability Thematic Analysis [31], based on systematic coding through predefined categories. A coding framework inspired by Bunzli et al. [32] was applied, consisting of four categories: 1) the participant’s response was consistent with the intention of the item (congruent response), 2) the participant’s response was not consistent with the intention of the item (incongruent response), 3) the participant misunderstood the content of the response scale (confusion), and 4) the participant’s response was both incongruent and congruent (ambiguous) [32]. To ensure consistency within the research team, a coding manual was developed, outlining characteristics of congruent responses for each item based on literature from PROMIS development [33, 34] and WHO definitions [35]. The manual was piloted by having researchers from the team independently code two interviews, followed by group discussions to resolve discrepancies.

The analysis primarily drew on data from the first and second steps of the TSTI, which provided more direct insight into participants’ immediate reflections [29]. When needed, the third step was used to inform coding. All interviews were independently double-coded: ESM coded all interviews, while DB (PhD student) and LMSR (research assistant) each coded half. Discrepancies were resolved through discussion, and, if needed, by consulting the wider research group. After analysis, codes were quantified for congruency, incongruency, ambiguity, and confusion, and items with > 50% non-congruent responses were considered problematic.

The other part of the first objective regarding causes for non-congruency was analyzed using a more inductive approach, where ESM systematically analyzed responses to elucidate the underlying causes linked to responses coded as non-congruent. This was done by gathering all non-congruent responses for each item, reviewing them, and making notes on the reasons for non-congruence. These reasons were then analyzed to identify specific underlying causes that led to responses being coded as non-congruent.

The second objective, regarding participants’ interpretations and understandings of PROMIS-10 GH items in relation to their own life experiences, was analyzed by ESM using a mainly inductive approach with Reflexive Thematic Analysis (RTA) [36]. The analysis followed the six steps of RTA, emphasizing the iterative nature of theme development. As a novice in RTA, ESM approached the process with an awareness of its reflexive character, continuously engaging with the data to refine interpretations and remain open to unexpected patterns. In line with RTA principles, themes were not merely identified but actively constructed through iterative engagement with the data [36].

The analysis primarily drew on data from the third step of the TSTI, which focused on participants’ reflections and interpretations, providing more in-depth and nuanced insights [29]. All transcripts were read and re-read to ensure thorough familiarity with the dataset, then organized by each PROMIS-10 GH item to facilitate a systematic analysis. Once organized, the formal coding process began, conducted sequentially for each item. For each item, steps two to five of RTA were completed before proceeding to the next. Following these steps, transcripts were systematically reviewed and coded inductively, with codes serving as concise labels capturing both explicit participant statements (semantic) and underlying meanings (latent). Patterns across codes were then analyzed to identify themes representing shared interpretations. These preliminary themes were subsequently reviewed and refined against the original transcripts for accuracy and coherence. In the final steps, themes were clearly defined, named, and illustrated with representative quotes to enhance clarity and readability [36].

While no qualitative studies to our knowledge exist for direct comparison of PROMIS-10 GH, the findings from this study support existing quantitative research, showing that PROMIS-10 GH aligns well with other questionnaires measuring HRQOL, including both general and condition-specific tools [12, 14].

While no items were found to be broadly problematic, some participants overlooked the time frame in items 8r (fatigue) and 7r (pain intensity). This appeared to occur primarily because the time frame instruction was presented only above item 10r, rather than being repeated for items 8r and 7r. Even when the time frame was noticed, some participants found it irrelevant or had difficulty summarizing their experiences within such a short period, with several reporting that a seven-day average did not reflect their fluctuating pain or fatigue. This may have implications for the clinical interpretation and use of the PROMIS-10 GH. These findings are consistent with a report by the Danish Health Data Authority, which noted that patients with chronic pain may miss the “past 7 days” reference in PROMIS-10 GH and recommended emphasizing the time frame more clearly within each item [37]. Also, the report noted similar concerns for item 10r (emotional problems), where recent events, such as holidays or personal loss, could distort responses [37]. To solve this challenge, the report proposed extending the time frame [37]. However, doing so may introduce recall bias [38]. The report also suggested allowing patients to report their minimum and maximum pain levels over seven days, which participants considered more meaningful [37]. This perspective was shared by some participants in the present study, who suggested that such a revision would better capture the fluctuating nature of pain. Previous research supports this, showing that individuals with chronic pain often find it difficult to summarize their experiences over short recall periods due to day-to-day symptom variability [39]. Moreover, studies comparing real-time assessments with recall ratings over a week or longer have found discrepancies in reported average fatigue and pain intensity [40]. In addition, studies report that recall ratings for periods of a week or more may be influenced by cognitive heuristics, such as recalling peak or recent experiences [41], and tend to indicate higher symptom levels than real-time assessments [39].

To address the challenges related to the timeframe in the PROMIS-10 GH, further research is needed. Based on the findings and literature, potential revisions could include: 1) explicitly integrating the timeframe into each item, 2) rephrasing to reflect real-time assessment, 3) allowing reports of minimum and maximum pain, or 4) removing the timeframe altogether to better capture typical pain and fatigue experiences and align these items with the rest of the items in PROMIS-10 GH. The latter option is supported by participants in this study, who noted that their responses, especially to item 7r, would likely have been the same if the timeframe had covered the past year.

One response regarding item 9r (social activities and roles) was coded as “ambiguous,” as the participant answered positively based on their partner’s ability to manage the roles on their behalf. As this issue was identified in only one participant and has not otherwise been highlighted as problematic, we do not see sufficient reason to revise this item.

For the second objective, findings largely aligned with the Danish Health Data Authority report [37], while also revealing some differences. Both showed that pain strongly influenced responses to item 1, with participants focusing either solely on pain or also including other conditions. This study added that mental health and social participation also shaped interpretations. For item 2 (quality of life), both studies emphasized the role of meaningful relationships [37], and this study further identified “freedom in life” as an additional theme. Regarding item 3 (physical health), both studies found that chronic pain and other conditions impacted perceptions, with participants linking physical health to both functionality and bodily well-being [37]. Item 9r (social activities and roles) posed similar challenges in both the report [37] and this study due to the guiding effect of the provided examples, which influenced how participants evaluated their abilities. The same challenge applied to item 6 (daily physical activities). These findings suggest that items 9r and 6 could benefit from clarifying that the assessment should include broader aspects beyond the examples. Differences emerged in item 4 (mental health), where only the report found participants struggled to distinguish between mood and cognitive function [37]—an issue not present in this study. Similarly, patients found item 5 (social activities and relationships) difficult to interpret in the report [37], whereas participants in this study did not. However, both the report [37] and this study emphasized low energy for social activities and the importance of relationships with friends and family.

This study found thematic overlap among items 5 (social activities and relationships), 9r (social activities and roles), and 6 (daily physical activities), particularly regarding social participation, relationships, and duties. This raises questions about the necessity of item 9r, which appears to address themes already covered by items 5 and 6. Since item 9r is not included in the physical or mental health scoring of PROMIS-10 GH [23], its inclusion in the measure could be reconsidered. Participants also interpreted fatigue in item 8r as both physical and mental exhaustion, with mental fatigue influencing cognitive and emotional states. As Hays et al. found that this item correlated more with physical health, it was included in the physical health score [23]. However, the present findings suggest reconsidering whether item 8r should solely reflect physical health.

Purposive sampling ensured diverse interpretations by including participants varying in sex, age, and chronic pain characteristics [42], thereby enhancing the study’s information power [28]. To ensure linguistic validity, only patients proficient in Danish were included. In several domains, participants reflected the broader HICP population, which is predominantly women with an average age of 50 years and chronic pain duration of approximately 120 months [5]. However, participants in this study reported slightly poorer physical health and slightly better mental health compared to the study population [5]. The low representation of ethnic minority participants (one person) may limit the representativeness of the Danish sample and should be considered when interpreting the findings [40]. Finally, recruitment via PainData may have introduced selection bias [38], as the comprehensive and cognitively demanding questionnaire could discourage less resourceful patients from participating.

Cognitive interviewing provides insights into the overall quality and cross-cultural validity of PROMs by identifying issues in item interpretation, relevance, and comprehensibility [15, 16, 18], and is therefore essential part of establishing content validity [17]. In this study, cognitive interviews provided in-depth insights into responses to the PROMIS-10 GH, following recommended guidelines [15, 16]. Using the structured TSTI method enhanced data collection by capturing immediate reactions and clarifying response behaviors [29]. However, TSTI also poses challenges, as it creates an artificial setting that differs from normal questionnaire completion. Thinking aloud during the first step may feel unnatural and lead to incomplete reflections [15], and the interviewer’s presence might have influenced responses. The method is also resource-intensive, requiring lengthy interviews and deep reflection, which can cause cognitive fatigue and less reflective answers toward the end [29]. Finally, the online format of one interview limited observation of non-verbal cues, and the presence of relatives in three interviews may have influenced responses.

Overall, we argue that methodological triangulation [43] and double-coding of the deductive coding parts enhanced the credibility and consistency of the results, supported by a collaboratively developed and piloted coding manual. Also, applying the RTA method enabled a flexible and nuanced analysis of subjective experiences while maintaining researcher reflexivity [36]. However, the lack of explicit definitions for all PROMIS-10 GH items by the PROMIS group required referencing external literature, which may have led to discrepancies between the coding framework and the PROMIS framework. Furthermore, conducting a deductive analysis first may have influenced the inductive approach, introducing potential researcher bias [44].