Hello

I am trying to predict a binary disease outcome (AD). I have two 'tests' that I would like to compare, both of which are binary diagnoses in effect (ADNIaMCI and ThreeMethodaMCI) that are binarised as 0 or 1.

I have obtained odds ratios for each test individually using:
diagti AD ADNIaMCI
and
diagti AD ThreeMethodaMCI

The odds ratios are quite different (~25 vs ~60), but I would like to test that and derive a significance level. I understand I need to create an interaction term such as ADNIaMCI#ThreeMethodaMCI

I tried:
logistic AD ADNIaMCI#ThreeMethodaMCI

Logistic regression Number of obs = 717
LR chi2(3) = 163.83
Prob > chi2 = 0.0000
Log likelihood = -107.12996 Pseudo R2 = 0.4333

------------------------------------------------------------------------------------------
AD | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
ADNIaMCI#ThreeMethodaMCI |
0 1 | 37.16049 19.24927 6.98 0.000 13.46344 102.5668
1 0 | 3.185185 3.431128 1.08 0.282 .385663 26.3064
1 1 | 95.97101 41.45165 10.57 0.000 41.16143 223.7637
|
_cons | .0149502 .0050205 -12.52 0.000 .007741 .0288731
------------------------------------------------------------------------------------------

How do I interpret that output?

Is it that
- patients who test negative for ADNIaMCI but positive for ThreeMethodaMCI, the OR is 37.16
- patients who test positive for ADNIaMCI but negative for ThreeMethodaMCI, the OR is 3.19
- patients who test positive on both tests, the OR is 95.97?

If so, how do I conclude whether one test is better than the other?

Many thanks for your consideration.

Kind regards
Nick