A purposive sample of expert clinicians were recruited to participate in the e-Delphi survey. An expert clinician was defined as someone with clinical experience treating females with pelvic pain, and who fit the following criteria: (1) held a relevant tertiary qualification in their clinical speciality (e.g., physiotherapy, medicine, psychology); (2) had > two years full-time equivalent experience treating females with pelvic pain associated with benign gynaecological and urological conditions; (3) had practiced clinically within the past two years; (4) had additional training in the contemporary understanding of persistent pain science; (5) was proficient in the English language.

We aimed to recruited a diverse range of clinicians to seek a variety of opinions, thereby increasing consensus validity (18). To ensure diversity, panel members were recruited from a range of international geographical locations, rural and metropolitan areas, age groups, professional groups (training and clinical experience), and healthcare settings (e.g., public, private). Based on the expected heterogeneity of the panel and a likely drop-out rate of 20%–30% between survey rounds (19), we decided a priori to invite 20 expert clinicians. We used a purposive, snowballing sampling approach to recruit eligible panel members. First, we identified potential panel members using the professional network of the authorship team and Internet searches. Panel members were also asked to identify further potential panel members through their networks (i.e., snowballing). Panel members were invited by the primary researcher (AKM) to participate by a personalised email, which contained detailed information about the study. Informed consent from each panel member was obtained electronically at the start of the Round One survey.

Qualtrics software (2019, SAP, Provo, UT, USA) was used to develop, conduct, and distribute the e-Delphi survey rounds. Participation was anonymous – a key characteristic of the Delphi process (19). Prior to distribution, the Round One survey was piloted by two local experts to collate feedback on readability, relevance, and usability; edits were made accordingly.

The Round One survey link was open from 28 August 2022 to 18 September 2022, with email reminders sent twice during this period. The Round One survey comprised of three parts: (1) study information and participant consent; (2) panel members' demographics (e.g., sex, gender, country of residence, profession); and (3) open-ended questions asking their opinion on what content should be included in PSE for females with benign gynaecological and/or urological pelvic pain. Subsequent questions asked panel members to describe additional educational information specific to pelvic pain conditions (e.g., endometriosis) or life stages (e.g., adolescence).

The Round Two survey included the concepts identified in Round One survey responses. The survey link was open from 4 October 2022 to 18 October 2022 to participants who completed the Round One survey (due to consent purposes). Email reminders were sent twice during this period. Participants were invited to rate the importance of each concept in a randomised order using a six-point Likert rating scale (“Not at all important” - “Very important”). Concepts were presented in three parts: (1) PSE concepts important for females with pelvic pain; (2) PSE concepts important for specific pelvic pain conditions; (3) PSE concepts important for specific life stages. Open-ended questions were provided following each survey section to allow panel members to provide further comments on each concept (e.g., suggested rewording of items, condensing multiple statements into one) and identify any further concepts for consideration into the Round Three survey.

Following Rounds Two and Three, panel members were given controlled feedback of survey responses, including their individual responses and the group median and interquartile range (IQR) on each survey item. Feedback was also provided for open-ended responses, which included clear delineation of reworded or removed statements and reasoning behind the changes and labelling of any new statements generated. The purpose of this feedback was to allow panel members to reflect and potentially revise their responses when compared to the groups response, with the aim of reaching consensus by the end of Round Three (19) (see data analysis section for consensus criteria).

The Round Three survey comprised of concepts rated in the Round Two survey, including those that were re-worded and new concepts suggested. The Round Three survey link was open from 31 October 2022 to 14 November 2022 to panel members who completed the Round One survey. Email reminders were sent twice during this period. The Round Three survey was conducted in the same manner as Round Two.

Although Delphi survey rounds can be performed until consensus is reached, panel members' responses are unlikely to change following three rounds of rating statements (19). Thus, we decided a priori that we would conduct a maximum of three survey rounds after generation of statements (a total of four rounds for this study) unless stability of survey responses was reached in earlier rounds. Following termination of the e-Delphi process, the final list of concepts (derived from Round Three survey) was grouped into categories by the first author (AKM) before being discussed and refined with the wider research team. Panel members were then invited via email to provide anonymous feedback on the grouping of concepts and the developed categories using Google Docs software (California, USA) or email. Feedback was collected from 12 January 2023 to 26 January 2023; two email reminders were sent during this time, which was a deviation from the protocol to allow panel members more time to provide feedback.

Quantitative data were analysed using IBM SPSS v26 (20). Descriptive statistics were used to analyse demographic data and survey response rates. Survey item scores were analysed by median and IQR because they are robust and objective measures of consensus, which can be given pre-defined cut-off values (19, 21). For all survey rounds, items were retained if they were considered “important” (a median >3.0 - “slightly important”); items were excluded if they were rated “unimportant” (median value of <3.0) and had reached consensus (IQR <1.0). Consensus was considered reached between panel members for items with an IQR <1.0. Stability of responses between survey Rounds Two and Three were calculated using the nonparametric Wilcoxon matched pairs signed rank test (19, 21). Panel members' responses were considered stable when no statistically significant change (p ≥ .05) was detected between the rounds (21). The final list of PSE concepts (“items”) were ranked within their defined categories based on their median Likert scale rating of importance.

Qualitative data were analysed using QSR International's NVivo software (release 1.6.1). For Round One, open-ended responses were analysed using a simple inductive, content analysis method (22, 23). Two researchers (AKM and MVW) independently open-coded responses and collated their codes into representative “concepts”, taking into consideration wording used by participants (22). Responses that were the same, or very similar, between participants were combined and collapsed into a single statement, whilst ensuring to stay close to the meaning of the original suggestions. To ensure consistency and reliability between coders, inter-coder reliability was determined using Cohen's kappa. Kappa values were categorised as having no agreement (<0.20), minimal agreement (0.21–0.39), weak agreement (0.40–0.59), moderate agreement (0.60–0.79), strong agreement (0.80–0.90), and almost perfect agreement (>0.90) (24). Inter-coder reliability was piloted in a random sample of 20% of survey responses. When agreement was moderate (Cohen's kappa >0.60), coders continued open-coding the remaining responses. Open-ended responses from subsequent survey rounds, including proposed rewording of concepts and suggested new statements, were analysed by AKM and discussed among the wider research team.

Twenty panel members responded to the Round One survey. From the open-ended responses, 110 statements were identified relating to PSE concepts for pelvic pain. For PSE concepts specific to pelvic pain diagnoses, eight concepts were identified for endometriosis and adenomyosis, two for bladder pain syndrome (BPS), four for vulvodynia and vulvar pain, and one for sexual pain. For PSE concepts specific to life stages, four concepts were identified for adolescents, one for reproductive years, and two for post-menopausal pelvic pain. See Supplementary File 1 for content analysis inter-coder reliability.

Seventeen panel members responded to the Round Two survey. Panellists were provided 132 concepts to review. All concepts were rated as “important” (median >3.0) and 109 (82.6%) reached consensus on their importance rating (IQR >1). No concepts met the exclusion criteria (i.e., unimportant). Therefore, all concepts were retained for inclusion in the Round Three survey.

Panel members suggested rewording 16 concepts and added 14 concepts for the Round Three survey. The research team also excluded two concepts that were originally included in the Round Two survey because they did not meet the study aims.

Fifteen panel members responded to the Round Three survey. Panellists were provided 144 concepts to review, and all were rated as “important” (median >3.0); 104 (72.2%) concepts reached consensus on their importance rating. One concept did not meet stability between rounds two and three (p-value <0.05). Due to the high rate of stability, the e-Delphi process was terminated, and a further round (Round Four) was not required. Stability could not be calculated for the 14 concepts suggested in the Round Two survey.

Panel members condensed 30 concepts into 13. Panel members also suggested the rewording of 11 concepts; no additional concepts were suggested. Three concepts were removed from the survey by the research team because they did not meet the study aims.

There were 125 final concepts, of which, 92 (73.6%) reached consensus on their importance rating (IQR ≤1.0). One hundred and twenty-four (99.2%) concepts reached stability between Rounds Two and Three. Nine statements were re-worded by the research team for clarity and to reflect consistency with wording of other concepts.

Of the final concepts, 102 were generated for female pelvic pain in general; concepts were grouped into 13 categories representing overarching PSE concepts (see Table 2 and Figure 2). Sixteen concepts were generated for specific pelvic pain conditions, including endometriosis and adenomyosis, bladder pain, vulvodynia/vulva pain, and sexual pain. Seven PSE concepts were generated for specific life stages, including “adolescent pelvic pain”, “pelvic pain during the reproductive years”, and “post-menopausal pelvic pain” (Table 3).

Two (10%) panel members provided feedback on concept grouping and category names. One category name (“persistent pelvic pain doesn't always equate to tissue pathology”) was refined for clarity (“persistent pelvic pain and tissue pathology rarely correlate”). The full list of final concepts can be found in Supplementary File 2.

Future research may consider co-creating PSE curricula with both clinicians and females with pelvic pain, in particular across the life stages identified (e.g., adolescence). Whilst this study proposes important PSE concepts, investigations are also needed into their wording and delivery to improve clinical applicability. Recent work emphasised the importance of enhancing clinicians' awareness of educational strategies for effectively implementing PSE concepts, along with providing simple resources and case scenario examples to support their application (45). Co-design research methods (46) with clinicians and consumers will be important for improving the clinical applicability of the concepts generated in this study. It would also be imperative to investigate the views of females with persistent pelvic pain to understand what educational content they value. The development of PSE resources for other pain conditions found that clinicians and consumers valued different learning concepts (14). Females with pelvic pain have highlighted the importance of acknowledging pathological contributions to pain (25, 47), whereas clinicians de-emphasised its importance in this study. These might be important to consider when developing PSE curricula for pelvic pain. There is also the need to develop a validated tool to assess pain knowledge specific to pelvic pain. Although tools have been developed to assess the knowledge of adults (48, 49) and children (50), none assess knowledge on concepts specific to pelvic pain. Finally, the efficacy of PSE could be tested in randomised clinical trials. Given that females with pelvic pain have expressed the importance of providing education alongside pain management interventions and the current evidence-base supports this, it would be beneficial to test its efficacy within a complex care package [e.g., alongside physiotherapy (51)], that could improve the outcomes of females with pelvic pain.