In naturalistic sampling, measures of diagnostic or prognostic accuracy,
such as sensitivity, specificity, and positive and negative predictive value,
are somewhat biased depending on the sample size, "N."1
This bias is only negligible if N is large enough relative to the prevalence
of the disorder, "P," and the level of test, "Q," (ie, the number of observations
in each marginal position of the 2 × 2 table must be reasonably large).
As a rule of thumb, Kraemer1 recommended
that "no test evaluation be undertaken unless one can expect to see at least
10 patients in each of the marginal positions of a table."1(p39)
Thus, based solely on the lifetime prevalence for schizophrenia of 1%, more
than 1000 persons had to be included; based on the incidence of 0.0001% or
1/10 000 and an average follow-up period of 10 years, even more than
10 000 persons had to be included for a naturalistic sampling of the
Klosterkotter J. Limitations of the Bonn Scale for the Assessment of Basic Symptoms as a Screening Measure. Arch Gen Psychiatry. 2002;59(5):471. doi:
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: