Protocol for the development of a core outcome set for pelvic girdle pain, including methods for measuring the outcomes: the PGP-COS study

We will conduct a systematic review of the literature to identify outcomes measured in intervention studies and systematic reviews related to PGP. We chose to limit the review to these study types as our PGP-COS will be developed for effectiveness trials, and follows COMET’s mission. We will search the following databases from inception: PubMed, the Cochrane Library, PEDro and Embase. Two reviewers will independently select relevant records by title/abstract and full-text, using the selection criteria outlined in Table 1. Disagreement will be resolved through discussion or if necessary by involving a third reviewer. Health, clinical, social and economic outcomes (and their verbatim definitions) will be extracted and their corresponding outcome measurement instruments, where reported, will be recorded. The quality of outcome reporting will be assessed using the six questions proposed by Harman et al. [16]. Subsequently we will group the outcomes into outcome domains using the OMERACT Filter 2.0 framework [17], but all identified outcomes will be presented for rating in the Delphi survey.

We will conduct semi-structured interviews using open-ended questions with a purposive sample of approximately 15 pregnant/postpartum women with PGP (5 in Ireland, 5 in Sweden, 5 in Mexico), because those experiencing PGP directly, might indicate different outcomes of importance than those measured by clinicians/researchers in studies [18]. Participants must be aged 18 years or more and have PGP (as primary complaint) at the time of the interview. PGP is defined as pain between the posterior iliac crest and the inferior gluteal fold and/or at the symphysis pubis [3], and confirmed by a qualified clinician following a physical examination. A pragmatic approach will be used in that the clinician will identify patients with PGP at his/her discretion based on current guidelines [3]. Following a complete history, the examination will include PGP pain provocation tests and functional tests, in particular, but not exclusively, the posterior pelvic pain provocation (P4) test and the functional active straight leg raise (ASLR) test [19‐22], which have been shown to have high sensitivity and specificity. In addition, other body areas that might be the source of a patient complaint will be examined to rule out other diagnoses. Participants must also be able to fluently speak and understand the local language (English, Spanish, or Swedish). Interviews will be transcribed and analysed using thematic analysis.

When examining existing COS development studies, some authors include patient interviews in establishing the initial list of potential outcomes but others only search the literature [23]. Current guidance recommends that potential relevant outcomes are identified from several sources: systematic reviews of published studies, reviews of qualitative work, examination of items collected in national audit data sets, and interviews/focus groups with key stakeholders [14]. However, no studies have examined the impact of including patient interviews compared to conducting a systematic review only to identify the initial list of potential outcomes, on the final COS. Including interviews with participants presents additional costs and workload to the COS development team, including additional ethics applications, time and resources. It may be that the inclusion of patient representatives in the Delphi survey and in the consensus meeting makes patient interviews at the outcome identification stage redundant, but this is unknown. We will assess how the outcomes that are derived from the interviews only (if any) are rated in each round of the survey, and how many/type (if any) are subsequently included in the final PGP-COS. We will also examine the extent to which additional outcomes provided by patient representatives in round 1 of the Delphi survey overlap with the outcomes obtained from the interviews.

We will invite people from all stakeholder groups, aiming for approximately 40 experts from each stakeholder group (patients, clinicians, researchers, service providers/policy makers) to participate (which will provide 20 people randomised to each scale; see methodological study 2). We will recruit participants for the Delphi survey through professional organisations, electronic discussion lists, and patient organisations. We will also encourage snowball sampling, asking participants to forward the invitation appropriately.

Eligible participants will be: (a) Researchers who are actively involved in research related to PGP/lumbopelvic pain and have at least one peer-reviewed publication on the topic; (b) Clinicians, currently treating patients with PGP (including chiropractors, physiotherapists, osteopaths, orthopaedic surgeons etc.); (c) Patients with PGP; (d) Policy-makers, involved in formulating policies related to the management of musculoskeletal conditions including PGP; and (e) Service providers of services for musculoskeletal conditions including PGP. Participants must be able to read and understand English in order to participate.

Round 1 will collect demographic data including nationality, age, gender, stakeholder group and profession. Participants will rate the importance of each listed outcome. Outcomes will be rated in each round on a 9-point scale [16] or on a 5-point scale (see methodological study 2 below). Participants will also be given the opportunity to add any additional outcomes not already in the list that they think are important. All outcomes from round 1 will be forwarded to round 2.

Participants who responded to round 1 will be presented with feedback (descriptive statistics) on their own and others’ scores from round 1 and asked to rescore all outcomes and score new outcomes added in round 1. The reason for providing feedback from the previous round to participants is to increase the degree of consensus [14]. Participants will rate outcomes using the same scale they used in round 1. Outcomes from round 2 will be forwarded to round 3 where 70% or more of participants score the outcome as ‘important’ (7 or more on the 9-point scale, or 4 or more on the 5-point scale), and less than 15% of participants scoring an outcome as ‘not important’ (3 or less on the 9-point scale, or 2 or less on the 5-point scale) (Tables 2 and 3).

Round 3 will further prioritise outcomes by rescoring the outcomes based on round 2 feedback. Only participants who completed round 2 will be invited to round 3. Outcomes scored by 70% or more of participants as ‘important’ (7 or more on the 9-point scale, or as 4 or more on the 5-point scale), and less than 15% of participants scoring an outcome as ‘not important’ (3 or less on the 9-point scale, or 2 or less on the 5-point scale) will be included in the preliminary PGP-COS. At the end of the round 3 survey we will also include a question to ask if they would be willing to take part in the face-to-face consensus meeting.

The 9-point scale, whereby a score of 1–3 means limited importance, 4–6 means maybe important but not critical, and 7–9 means critically important, is commonly used in Delphi surveys for COS development [14] and is based on the recommendations of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group to assess the importance of evidence. However, other scales have been used [24, 25], and comments from some participants in round 1 of an in-progress recent Delphi survey described the 9-points scale as somewhat confusing and that a smaller scale might have been more user friendly [26]. Currently, no studies have examined the impact of using different likert scales on the final COS. We aim to compare the use of a 9-point scale (Table 2) versus a 5-point scale (Table 3). Participants of the Delphi survey, when clicking on the link to the survey, will be randomised to completing the survey using either a 9-point or 5-point scale. Specific software has been developed to enable this [27]. We will compare the two groups, with regards to the following outcomes: (a) the proportion of outcomes rated as ‘important’ (7 or above on 9-point scale, and 4 or 5 on 5-point scale) in each round and any differences between the scales, in the outcomes that are included in the preliminary COS after round 3. We will use z-scores to test the differences in proportions between the groups, and calculate overall mean scale differences on each outcome to assess differences. Recognising that differences can occur due to chance, and allowing for a 20% possibility of this, we would anticipate balance between the groups due to random allocation and thus similarity in overall scale scores for each outcome by population group. We will also ask participants to rate the ‘ease of use’ of the rating scale to assess utility.

The constructs (outcomes) to be measured will be identified in phase 2 and the target population has been described in the scope. Initially, we will conduct a systematic search of the literature to identify systematic reviews of outcome measurement instruments for the outcomes included in the preliminary PGP-COS derived from phase 2. We will search PubMed, Embase and the COSMIN database from inception. If a high quality systematic review has been conducted for the measurement of a particular outcome, including a quality assessment of the outcome measurement instruments, we will use this. If no high quality systematic review is identified for a particular outcome, then we will conduct a systematic review of outcome measurement instruments for that outcome, according to the COSMIN guidelines [15]. The systematic review(s) will be registered with PROSPERO (International Prospective Register of Systematic Reviews).

We will search PubMed, Embase, and PEDro databases from inception for studies of outcome measurement instruments. The search strategy will be based on the search filter for finding measurement properties of measurement instruments, developed by Terwee et al. [28]. In addition, we will screen reference lists and contact experts. Eligibility of studies will be determine by two independent reviewers and in case of disagreement, consensus will be reached through discussion.

Quality assessment will be conducted using the COSMIN checklist [29]. The feasibility of use in clinical practice and in research of the outcome measurement instrument(s) will also be assessed and further discussed in the consensus meeting (phase 4).

We will make recommendations on selecting outcome measurement instruments for the preliminary PGP-COS based on the quality of evidence and feasibility of outcome measurement instruments, which will be put forward to the consensus meeting (phase 4).