derive cut-off score at fixed sensitivity level (ROC analysis)

[Laura Kudlek]

Dear Statalist friends,

I am fairly new to stata (and ROC analysis), so please forgive me if I am slow to understand. My research involves deriving cut-off scores of the mean value of a questionnaire that is intended to be used as screening tool in practice (so my input variable is continuous and non-integer while the outcome variable is binary).

I am performing a non-parametric ROC analysis using the roctab command. I now want to derive a cut-off score for a fixed sensitivity level of 70%. I already used the cutpt command which gives me the Youden's and Liu's index and the point nearest to (0/1), however, I want to additionally provide the cut-off score at a fixed sensitivity level. Which command can I use to achieve this?

Thank you so much for your help and advice, it is very much appreciated.

[Leonardo Guizzetti]

You can use the -table- option of roctab to find the cutoff that provides at least 70% sensitivity.

[Laura Kudlek]

You can use the -table- option of roctab to find the cutoff that provides at least 70% sensitivity.

Thank you very much for answering so quickly! Do you know if there is another command aside from the -table- option that can provide a more specific estimate of the cut.off at 70% and potentially also gives the confidence interval?
Otherwise I will go ahead and just select a cutoff from the -table- option, thank you!

[Leonardo Guizzetti]

This empirical approach can't be more specific as it uses cutoffs defined by the level's of your measurement. I'm not aware of another command that will give you what you seek.