Drug Test Reliability

A WORD ON TEST RELIABILITY

Published data indicate that a system of drug screening similar to that
used by most laboratories has a sensitivity of 76% and a specificity of
99%. This excellent specificity parameter means that of 100 persons who do
not use drugs, 99 would be expected to test negative by confirmation. This
is certainly an excellent specificity for any medical determination.
However, one should also be aware of another parameter, the predictive
value of a positive test. As applied to drug testing, this figure expresses
the probability that a subject that has tested positively has in fact used
the drug. Although a high specificity, such as 99%, optimizes the
predictive value, a more significant factor is the prevalence of drug use
in the population being tested. The more prevalent the usage of drugs in a
subject population, the greater the reliability of drug testing procedure.
Given the sensitivity and specificity values quoted above, the following
table indicates the predictive value for several levels of drug abuse
prevalence.

Percentage of tested population | Probability that a given

using drugs (the prevalence of | subject that tests positive

drug abuse) | has really taken the drug

| (the predictive value of a

| positive test)
______________________________________________________________________

0.1% | 7.1%
1.0% | 43.4%
10.0% | 89.4%
20.0% | 95.0%
50.0% | 98.7%

Therefore, in a population with a high incidence of drug use (200 per
thousand), the false positive rate on drug screens is only 5%, while in
a low-incidence population (1 per thousand) the false positive rate on
randomly screened individuals (i.e., those of whom there is no particular
suspicion of drug use) is expected to be a whopping 93%! For this reason,
it is my recommendation that drug screens not be applied on a random,
not-for-cause basis, except in situations where the prevalence of drug use
is known to be high (such as in substance abuse treatment programs).