Two of us (RM & LBD) independently coded the data by drawing on constant comparison methods, wherein preliminary categories were revised and emerging categories were identified and expanded through constant comparison to the data [ 42, 43].
Within a feedback loop those categories are revised, and are eventually reduced to main categories and checked in respect to their reliability.
Categorization of concepts involved an open-ended constructivist process in which initial categories were revised so as to compare across data and to literature; namely, the categories identified in past research efforts.
Almost all categories were revised lower, with the exception of home construction and government.
Today, these categories are revised as Type 1 diabetes and Type 2 diabetes.
Codes and categories were revised and compared between participants within the same triad and social support network.
These categories were revised, elaborated, refined, and validated through data triangulation, where multiple interview sources were considered to develop a robust understanding of the phenomenon of interest.
Analysis was conducted as a group process in which the researchers (LR, RM and KM) read all transcripts and independently prepared analytical notes that were discussed in regular meetings where the development of key categories was revised and refined [ 30, 31].
This means that figures for the following categories may be revised upwards when all the data is available; ambulances queuing, cancelled operations, beds blocked and bed days lost due to norovirus.
Content-related index codes were grouped together into categories, which were revised as the analysis progressed.
