DIAGNOSTIC STUDY PERFORMANCE
A diagnostic laboratory study is one that is
designed, or has been discovered, to improve the
clinician's ability to discriminate between persons
suffering from a disorder or condition of interest and
persons free from the disorder or condition. The
degree to which a study accomplishes this discrimi-
nation is referred to as its diagnostic performance.
Sensitivity, specificity, and ROC curves
The fundamental measures of the diagnostic
performance of a laboratory study are its sensitivity
and specificity. Sensitivity is the frequency with
which a study indicates the correct diagnosis in
persons with the disease. Specificity is the
frequency with which a study indicates the correct
diagnosis in individuals who are disease-free.
As an example, the data obtained in a clinical
investigation concerned with the laboratory diagnosis
of iron deficiency in infants (Dallman
et al.
1981)
can be used to quantify the performance of the
study, transferrin saturation (the ratio of plasma iron
concentration to plasma iron-binding capacity).
Transferrin saturation was determined in capillary
blood specimens from 165 1-year-olds who were
suspected of having iron-deficiency anemia. Infants
were classified as iron deficient if the transferrin
saturation was less than 10% and as iron replete if
the transferrin saturation was greater than 10%. The
study classifications, categorized according to the
final diagnostic classification, are presented in Table
3.1. The numerical entries indicate the number of
study subjects in each category. The sensitivity of
the study is calculated as the frequency of correct
diagnosis in the iron-deficient infants. In this case,
the frequency is 29 divided by 55 which equals 0.53.
A little better than one-half of the iron-deficient
infants are properly identified. The specificity is the
frequency of correct diagnosis in the iron-replete
subjects; here, it is 82 divided by 110 which equals
0.75. Three quarters of the iron-replete infants are
correctly identified. A table of classification catego-
ries, as used in the example, can be constructed for
any diagnostic study (Table 3.2). The designations
true or false and positive and negative are usually
assigned to the categories as shown. From the table,
sensitivity
=
true positives
true disease
and
specificity
=
true negatives
true disease free
Because a study's specificity is determined by
the frequency distribution of results in a stable refer-
ence population, it will remain constant. There will
be some variability in its measurement because the
composition of the sample of subjects will vary by
chance. However, as long as the subjects are chosen
at random from the same reference population,
estimates of the study's specificity will cluster
around the value that would be found were the entire
reference population to be studied. Constancy of
estimation is not true for sensitivity. The sensitivity
of a diagnostic study is usually greater in individuals
with more advanced or severe forms of a disease. In
the case of iron deficiency, as the condition persists,
and the iron deficit deepens, all of the diagnostic
studies used to identify iron deficiency, transferrin
saturation included, show increased sensitivity.
Diagnostic and Prognostic Classification
3-1
Chapter 3
DIAGNOSTIC AND PROGNOSTIC CLASSIFICATION
© 2001 Dennis A. Noe
Table 3.1
Classification Categories for Transferrin Saturation
Classification Using Final Diagnostic Classification
Transferrin Saturation
Iron-replete Iron-deficient
Iron-replete
82
26
Iron-deficient
28
29
Total
110
55
Table 3.2
Classification Categories for a Diagnostic Study
Classification Using Final Diagnostic Classification
Study Result
Disease-free
Disease
Disease free
true negative
false negative
Disease
false positive
true positive
Total
true disease free true disease
