Likert scores - how to deal with them?


We distributed a survey with 3 Likert questions that we meant to combine to a scale and use as a 'continious' outcome in linear regression. The 3 questions asked a very similar thing and inter-item reliability was high. However, one of the question phrasings was reversed in the survey. We have very strong reasons to believe that a lot of the survey participants did not catch that and essentially we can't use this question. This leaves us with 2 Likert questions and now I am unsure what to do...
Calculating a scale and treating it as a continuous variable seems not appropriate with only 2 items. I'd rather stay away from ordinal regression but I feel by dichotomizing the result we might lose information.
Any ideas would be very much appreciated!
Thanks :)


TS Contributor
In my experience, a dependent variable based on Likert items and with such a range usually is treated as if interval scaled.

With kind regards