Description

Buck and Gart developed formulae showing how a test's sensitivity and specificity might be affected by the sensitivity and specificity of the reference method ("gold standard"). Sometimes a diagnostic gold standard is imperfect, and a new test's performance may be distorted by this fact.


 

Parameters:

(1) disease prevalence (p)

(2) actual sensitivity of the screening test (S1)

(3) sensitivity of the reference method (S2)

(4) apparent sensitivity of the screening test (S3)

(5) actual specificity of the screening test (E1)

(6) specificity of the reference method (E2)

(7) apparent specificity of the screening test (E3)

 

apparent sensitivity of the screening test = S3 =

= ((p * S2 * S1) + ((1 - p) * (1 - E2) * (1 - E1))) / ((p * S2) + ((1 - p) * (1 - E2)))

 

apparent specificity of the screening test = E3 =

= ((p * (1 - S2) * (1 - S1)) + ((1 - p) * E2 * E1)) / ((p * (1 - S2)) + ((1 - p) * E2))

 

These can be rearranged to the following:

 

actual sensitivity of the screening test = S1 =

= ((S3 * ((p * S2) - p - E2 + (p * E2) + 1) + (E2 * (1 - E1 - p + (p * E1))) + (E1 * (1 - p)) + p - 1) / (p * S2)

 

actual specificity of the screening test = E1 =

= ((E3 * (p - (p * S2) + (E2 * (1 - p)) + (S1 * p * (1 -S2)) - (p * (1 - S2))) / (E2 * (1 - p))

 

Unfortunately the actual sensitivity cannot be determined until the actual specificity is known, and vice versa.

 


To read more or access our algorithms and calculators, please log in or register.