Definition of disease
Revised: 2001-10-28
Question: The definition of disease is essential in evaluating a test. How do you define disease? Please use the sore throat as a model although it may be applied to other diagnoses as well.
We may want to use a test to:
Test characteristics like sensitivity and specificity might be obtained in the scenario A and B. In scenario C and D we would obtain other measures of agreement rather than sensitivity and specificity. When we calculate sensitivity and specificity in scenario A and B we need a gold standard as comparison. Every test result is then compared to this gold standard. This webpage focus on scenario A and B.
In the scenario A our gold standard must predict presence of an etiologic agent. In the strep throat model it would predict presence of group A beta-hemolytic streptococci (GABHS) in the throat. A common example of such a gold standard is two conventional throat cultures. If any of them are positive then the gold standard is considered positive, else it is negative.
In the scenario B we need to compare our test results with a gold standard predicting disease. How do we define disease?
Disease may be defined in many different ways, for example:
Different definitions of disease: |
|
Definition | Example from the strep throat |
|
Presence of GABHS in the throat |
|
Rise in titers of antibodies directed towards antigens from GABHS |
|
An example could be the Centor criteria1 which is a sore throat and:
|
|
----- |
|
A sore throat caused by GABHS |
The first four definitions of disease are often used when evaluating the usefulness of a throat sample used in patients having a sore throat.
The fifth definition of disease is interesting. It could be used to evaluate the ability of a test to predict a disease caused by a specific etiologic agent. However, to evaluate a test using this definition of disease requires a gold standard separating individuals with a sore throat caused by GABHS from those having a sore throat not caused by GABHS. Unfortunately such a gold standard does not exists.
Some of the patients with a non-GABHS-caused tonsillopharyngitis will be carriers of GABHS. The example with tonsillopharyngitis illustrate the fact that many tests may classify symptomatic carriers as truly ill from the disease our test is designed to detect. This problem is likely to occur whenever a symptom can be caused by other factors than the specified etiologic agent and when the etiologic agent might be present without causing symptoms.
Etiologic predictive value (EPV) use the fifth definition of a disease and evaluates a tests ability to predict that disease is caused by a specified etiology. It is important to remember that this definition of disease is always linked to some specific symptom (or combination of symptoms and clinical findings) that can be used to easily differentiate individuals having this symptom from "healthy" individuals. Since EPV does not need a gold standard it can be used in situations, like the example with evaluating a throat sample, where a proper gold standard is lacking.
Other pages with subjects that might be of interest is:
(You can click on these links to quickly see the pages)
Ronny Gunnarsson MD PhD
Department of Primary Health Care
Göteborg University
SWEDEN