1

I am a medical doctor not a statistician with some experience using STATA and SPSS.

I am currently analysing a dataset looking at the prognostic value of two diagnostic tests at predicting clinical outcomes.

The outcome is: Death Diagnostic Test: Test_A and Test_B

Test_A provides results from 0.1 to 1 A result of <= 0.80 is positive Test_B provides results from 0 to 60 A result of >=11.25 is positive

I have 200 patients. Each patient has 3 coronary artery blood vessels (600 vessels in total).

I aim to compare the predictive ability of a positive Test_A result with positive Test_B result in identifying patients who have a clinical outcome (i.e Death). I have compared them using an area-under-receiver operator curve (AUC) both as a dichotomous variables and continuous variables. The analysis is performed as a 'per-vessel' analysis (i.e 600 vessels). I.e each vessel underwent 2 diagnostic tests (Test_A and Test_B)

In STATA i performed this comparison as follows

roccomp OUTCOME Test_A Test_B

My questions are: 1) I would like to compare the predictive ability of Test_A and Test_B when the result of Test_A is between 0.75 to 0.80
2) In the patients with a Test_A result of 0.75 to 0.80 is there any incremental benefit of adding Test_B to the diagnostic assessment (i.e Test_A_Test_B)?

Any assistance with the above would be greatly appreciated.

user237031
  • 11
  • 1
  • If your goal is to identify patients - rather than vessels - who have a clinical outcome, doesn't it make more sense to compare the predictive ability of a positive Test A result across all 3 vessels for a patient against a positive Test B result across the same 3 vessels? After all, the patient may die if at least one of those vessels is clogged. In this case, you will need to define what a positive test result means for each test across those 3 vessels. For example, for Test A, it could mean a result of <= 0.80 for at least one of the three vessels (or for at least two of the three vessels). – Isabella Ghement Feb 09 '19 at 14:37
  • You could consider various alternate definitions of what it means for Test A to yield a positive result across all 3 vessels. Same for Test B. In my view, this approach, where you combine vessel-specific test results to get a patient-specific test result mimics how doctors will have to use the test in real practice: they will consider all 3 vessel-specific test results and make a combined decision based on those results. – Isabella Ghement Feb 09 '19 at 14:40

0 Answers0