MMD I: Week 1 - Statistical Tests and Measures

Exam 1 content

Sensitivity “True Positive Rate”: What is it and when is it useful?

• Tests have the ability to detect almost all people who have the condition (SnOUT)

• This is relevant for screening tests, where we want to ensure we do not miss any positive results

Specificity “True Negative Rate”: What is it and when is it useful?

• Tests that have almost no false positives, therefore you can trust a (+) (SpIN)

• This is relevant for diagnostic tests, where treatment may be too risky if they do not have the condition

What are likelihood ratios and what are they used for?

• Combines sensitivity and specificity

• Used to quantify the shift in probability of the condition given a certain test result.

• Use when comparing different tests

What does a positive (+) LR mean?

A change in odds favors a positive test

• +LR (>5.0 = moderate / >10 = conclusive)

• Increases the odds of diagnosis by RULING IN the condition

What does a negative (-) LR mean?

A change in odds favors a negative test

• -LR (<.30 = moderate / <0.10 = conclusive)

• Changes the odds of diagnosis by RULING OUT the condition

MCID Definition:

"The smallest change score associated with a patient’s perception of a change in health status”

• It should be GREATER than MCD! (measurement error)

An LR of ____ is worthless

1

What is a confidence interval?

Uncertainty around point estimate

• If the test is repeated 100 times, 95% of that time, the answer will be in the range given (NOT 95% confident that the true mean is in the range)

When can we rule out a trivial benefit when talking about a CI with “significant” findings (positive trials)?

If the lower bound CI> MCID (excludes MCID)

• That means that the lowest value would still greater than MCID

What does it mean if the MCID is included in the CI of a study with “significant” findings (positive trial)?

Study is not completely convincing that it provides meaningful benefit in the population

What does it mean if the upper bound of CI>MCID of a study with “non-significant” findings (positive trial)?

We cannot rule out a clinically meaningful benefit since the “true mean” may be the same or > MCID!

What does it mean if the MCID is included in the CI of a study with “non-significant” findings (positive trial)?

We should not trust any conclusion; the trial is underpowered/imprecise