Methods

Study design

We undertook an modified version of the Delphi consensus method,[9] which is a way to structure a group to communicate about and deal with a complex problem, “based on the idea that it is possible and valuable to reach a consensus.”[10] It is particularly useful in areas of limited research and is suited to explore areas where controversy, debate or a lack of clarity exist. We have conducted and report this work in line with the conducting and REporting DElphi Studies (CREDES) guidance.[11]
The exercise, carried out over 11-weeks between June and September 2022, comprised three anonymous survey rounds and a final online workshop. The aim was to reach agreement on relevant allergy symptoms, SPT allergens and thresholds, and what dietary advice should be given, based on a given combination of symptoms and SPT results for four foods: cow’s milk, hen’s egg, wheat, soya.

Recruitment of panel members

The research team identified five relevant professional groups (allergists, dermatologists, dietitians, general practitioners with an interest in allergy, and paediatricians) and sought to recruit at least two panel members from each group, from varied geographical settings. Potential participants were approached through professional networks.
From 22 experts approached in May/June 2022, 19 expressed of interest, with 14 able to participate. Panel members provided consent and data on personal/professional characteristics. In recognition of the time commitment, participants were offered £50 for each of the three surveys and consensus meeting, with an additional payment of £50 for engaging with all four events.

Consensus process

Following the Delphi approach, panellists remained anonymous in the survey rounds. All surveys were completed online and after every round each panellist received a personalised summary detailing their own answers and the anonymised answers aggregated across the entire panel. The summaries confirmed where agreement had been reached, and previewed items to be revisited at the next round. The majority of survey items asked for opinions rather than whether the participant agreed or disagreed with a statement, and free text responses were invited, to allow respondent to explain their responses or to raise any issues.
Comments and suggestions by panel members were used to modify or add new survey items for subsequent rounds. For example, between rounds one and two, some symptoms were reworded or split into two symptoms. Items in rounds one and two that were new, had been substantially modified or with insufficient agreement were carried forward to the subsequent round (Table S1). Therefore, only those items which were unchanged and with consensus were removed between rounds.
Delphi studies require criteria for consensus to be clearly defineda priori, including the threshold for agreement. We sought agreement of 80% or above. Items with agreement below 80% were carried through to the next round or discussed at the workshop.

Online rounds

All surveys were first drafted offline, transferred onto onlinesurveys.ac.uk, and tested before release. Once live, the first respondent to the round two survey raised the absence of their preferred answer option to one of the questions. Their answer was noted for analysis, and this option added before the remaining panellists completed the survey.
Each survey was open for a minimum of two weeks. Panellists were asked to answer the survey questions in the context of a child less than two years old with mild, moderate or severe AD and no recorded allergy to the study foods.
Round one asked which symptoms (timing and type) should be considered as immediate- and/or delayed-type food allergy, which SPT reagents were most appropriate for the four study foods, and how the SPT results should be interpreted (Table S1). Round two asked respondents what dietary advice should be given according to 72 possible combinations of: immediate, delayed or no allergy symptoms; had or had not recently ingested the food; and negative, equivocal or positive skin prick tests (Table S1). The results were summarised for round three distinguishing only between “negative” and “positive” SPTs (no equivocal category). In round three, participants were also asked their opinion when a child had never ingested the food.

Workshop

The workshop aimed to reach consensus on areas where there was insufficient agreement and to pull the findings together in a dietary advice flowchart. It took place online in September 2022, lasted 1.5 hours, and was chaired by an academic general practitioner with an interest in dermatology, who was independent of the process. Minutes of the discussion and the final flowchart were circulated to all panel members and approved post-workshop.

Analysis

Data were exported from the online survey website and analysed using Stata (version 17.0). Free text comments were collated under themes by LG and reviewed by the research team (LG, MJR, RB, RM, IS and SJB).

Measures to prevent bias

To ensure there was critical reflection of outcomes within the team, the process leads (LG, primary care researcher from a quantitative social sciences background, and MJR, academic general practitioner with a research interest in skin and allergy) met monthly with the steering group (all clinical academics – RB paediatric consultant allergist, SJB consultant dermatologist, and RM and IS, dietitians).
All panel members were invited to be co-authors, to credit their intellectual contribution and to represent diversity of opinion.

Ethics

The study was reviewed and approved by the University of Bristol Faculty of Health Science Research Ethics Committee (reference 10819).

Funding

This study was funded by a joint award from the Rosetrees Trust and The Stoneygate Trust (OoR2021\100007), who had no input into design, delivery, analysis or reporting.