Table 2.

Summary of key themes and suggested solutions/future areas for development

Impacted area	Key message	Representative quote (ID)	Suggested solutions/future areas for development (when applicable)
Guideline development stages
Scoping	Scoping in diagnostic guideline is more extensive and resource intensive compared to intervention guidelines.	“It takes a fair chunk of an analyst's time over several months to go through scoping, yes. It's absolutely critical that the problem is well defined…to understand exactly what all the ins and outs of the problem are and it's not something you just throw together. It's one of the things that makes diagnostics different. It requires a vastly, a more complicated problem definition phase that you would typically have for treatment.” (ID 1)
Key question formulation and test—treatment pathway	PICO format for question formulation is not very useful for diagnostics. The panel needs to be educated and trained on how to develop focused questions that include patient outcomes.	“PICO is not very relevant for diagnostic questions…Educating the panel on test's downstream consequences helps define exact questions to be answered.” (ID 5) “The panel still need help with question formulation. There is a lack of appreciation of what's required in a question…because a lot of them don't know that, you know. You can call it education of the clinicians.” (ID 13)	A test-treatment pathway can help panelists develop focused questions that are patient outcome centred, but the awareness of the panel on its importance needs to be raised and it requires resource commitment in terms of time, money and training.
Impacted area	Key message	Representative quote (ID)	Proposed solutions/future areas for development (when applicable)
Searching and synthesising the evidence	Search filters are not well developed for test accuracy studies; meta-analysis is often complex due to complexity of the methods and the heterogeneous nature of test accuracy studies.	“Yes, we have a lot of them that show very heterogeneous data. That indicates there are a lot of things still to be done and that we should be very cautious of single studies and drawing conclusions.” (ID 10) “We do not do meta-analysis because we are not as familiar on how to do this compared to treatment.” (ID 3)	We need good search filters for test accuracy studies, training and more explicit guidance on meta-analysis methods. There is a need for better quality primary studies on test accuracy for meaningful data syntheses to occur.
Types of outcomes and evidence	Resource is a major consideration as to whether the panel includes outcomes other than test accuracy. This is compounded by the lack of availability of such data.	“Define the budget, get another team to bring the resource … I can't just extract diagnostic test accuracy studies. It's not a complete enough picture.” (ID 11) “Doing qualitative research as part of a guideline would be really useful, but is not possible because of time constraints.” (ID 15) “For questions which the panel feel there will be very little or no evidence, other methods should be explored such as Delphi or focus groups for gathering the information.” (ID 3)	Inclusion of qualitative data and/or methods (eg, Delphi method and focus groups) should be explored as alternative ways to include patient outcome-related evidence in a structured way in the guideline process.
Making recommendations	Expert opinion is important, but in the face of the lack of good quality evidence this can make the process unstructured, not transparent and political. Usefulness of modelling to overcome this lack of evidence has conflicting views.	“There's a discussion about the benefits and harms, about resources and about patient values and preferences. Do we know those? No. Again, people give you their opinions about it, but that's the best we can do at this point.” (ID 11) “It's political and then lack of evidence that's not there. You have to build on the expert opinion and that can be quite difficult.” (ID 10) “Modelling is the only other answer I know of. Is it probably more accurate most of the time than people just making individual guesses in clinical practice? I would say yes.” (ID 1) “Models need assumptions and the assumptions cannot be proved, so it's very uncertain…We don't want to accept this uncertainty. We think it's better to give some pressure on the community to perform such studies.” (ID 14)	Delphi processes, focus groups or the use of modelling could make the process more systematic and transparent, but there are contrasting views on this.
Awareness, education and training
Within the guideline panel	Educating the guideline panel about test accuracy statistics and how test accuracy can impact a range of patient outcomes prior to the start of guideline development is crucial.	“The panel finds it very hard to make a choice as to when high sensitivity is important and when is high specificity important. They do not understand the consequences of high sensitivity and low specificity … hence cannot guide the methodologist either on what are important characteristics for the tests.” (ID 3)	Guideline panels should consider investing, prior to starting the guideline development, training of the panelists on test accuracy and downstream test consequences. This can be in the form of developing a test-treatment pathway, for example.
Outside the guideline panel	General medical education of doctors in test accuracy was seen as inadequate. Intervention research was perceived as receiving greater focus and funding in the form of RCTs. Regulatory authorities and the medical testing industry need to recognise the importance for end-to-end studies that report on downstream testing consequences.	“Most attention goes to intervention studies in journals and in guidelines normally…in the education of medical professionals there is less focus on diagnostic accuracy. They're not used to it.” (ID 3) “RCTs that evaluate full strategies including tests and treatment—we need new funding mechanisms before we get them.” (ID 4) “If we change the regulatory process which should be similar to drugs, then I feel we will in a few years have much better studies.” (ID 14)	Having a regulatory framework that recognises the importance of tests' downstream consequences can help bring the needed attention to medical test evaluation at several levels that were identified as lacking.

RCTs, randomised controlled trials.