Feasibility of extension to ordinal models?

[timdisher]

Hello! Currently working on a project with an interesting MOD problem arising from response based re-randomized trials being mixed with treat-through studies where the non-responders are essentially informative missing data. The outcome itself is perhaps best treated as a two level ordinal outcome, but I'm not sure to what degree that complicates the underlying model. Is this something you have looked into/considered?