Table 2.

Presentation of the themes and subthemes of the one-to-one interviews.

Themes				Examples of participant narratives
Theme (1): previous opinion of a prototype concept library
	Positive			“If there’s a way of doing that already that is set up and is validated and is consistently applied that would be an amazingly useful resource” (researcher 2).
	Neutral			“It will be helpful, but it needs to be extended. If they want to build something like this, and it is effectively working as a library, you need two things to be happened: (1) people are happy to feed in their constructs so it builds up, and (2) a useful library, easy to go, to browse, and to borrow phenotypes definitions” (a clinician).
	Negative			None
Theme (2): requirements of a prototype concept library
	Usability
		Simplicity	“Simple plain English not in SQL or python” (a clinician).
		Searching ability	“What is the type of search engine? Is it a search engine that just does disease phenotypes or also does the health status phenotypes or risk factor phenotypes, symptoms phenotypes?” (a clinician).
		Data quality	“It’s really just about transparency and documentation. So, anybody can effectively do anything that can be turned into a reproducible research output. The barriers are usually not enough time to comment and document it properly and then not enough quality assurance” (a senior research manager).
		Sharing ability	“It would be very useful to share the knowledge about codes such as read codes, ICD 10 codes, or OPCS codes, and share ideas and concepts between other users that will save lots of time” (researcher 3).
	Sustainability
		Interoperability	“How interoperable it is with other systems because the major failure of most of these systems is that they’re not interoperable, so people don’t use them” (a senior research manager).
		Accessibility	“So, from a group like myself, or me as a user, we would probably like direct access to the underlying data it stores. So, whether that’s through something like SQL directly, or something like that through a statistical package, because where we do lots of bulk type work” (a senior research manager).
		Analyzability	“I wanted to look at all health codes of my study population. Then, through machine learning, like feature selection, I tried to identify the most important list of codes, which are associated with the popular health conditions” (researcher 1).
Theme (3): user experience of existing concept libraries
	Aware (used them)			“Yes, so with QOF, we definitely used QOF codes a lot, because obviously going back to the quality assurance question, they’d been assured so that the NHS can use them for remuneration of money and payments. With other systems, we tend to look online to see CALIBER of things with us, then yes we have used outputs from those systems before” (a senior research manager).
	Aware (not used them)			“No. I have not used any of these things before so I think there is CALIBER and I think, is that part of what was set up within the previous Farr institute? so I am aware that some of these exist but I haven’t looked into them before” (researcher 2).
	Not aware			None
Theme (4): user’s recommendation to improve repeatable research				“If we want reproducible research, we have to all be using these resources in a similar way or at least we need to be able to understand what previous projects have done. It is about setting things out clearly. Clear definitions, clear sets of codes that people can then either use themselves or build on I think” (researcher 2).