Prerequisites for EPV
 Revised: 2001-12-04

Some prerequisites must be met before EPV can be calculated (Table I).

Table I - Prerequisites for Etiologic Predictive Values (EPV)
    Pos. EPV Neg. EPV
1. P(M+|D+) = 1 x x
2. Sen is known and is between 0-1 x x
3. Theta (θ) is known and is >0 x x
4. Testresults from S+ and S- x x
5. P(T+|S+D-) < P(T+|S+) < Sen x x
6. P(T+|S-) > 0 x  

  1. This means that the disease cannot be present unless the etiologic agent is also present. For example, the patient cannot have the disease "a sore throat caused by group A beta-hemolytic streptococci (GABHS)" unless GABHS is present in the patient at some stage of the illness. This prerequisite is quite obvious and means that you must define disease to be a symptom or a clinical sign (or any combination of these) caused by the possible etiologic agent that the test is designed to detect. For more examples see the page "Suitable scenarios".

  2. Sensitivity of the test to detect the possible etiologic agent is estimated in previous studies. Sensitivity must of course be somewhere between 0-1 (0-100%).

  3. Theta is known or assumed to be known. It must be >0. A reasonable assumption is to assume theta to be 1.

  4. The number of positive and negative tests must be known in both the group of patients and in the healthy comparison group.

  5. For any reasonable test the proportion of positive tests will increase if disease prevalence increase. The highest disease prevalence is 100% and the proportion of positive tests in such a group of patients may be described as P(T+|D+), which is the same as P(T+|M+), and both are the same as Sen.

  6. The formula for positive EPV will not work if the number of positive tests in the control population is zero (division by zero).

The EPV-calculator provided on this website will check that requirement 2-6 are fulfilled. If not the calculator will object and refuse to do the calculation.

Ronny Gunnarsson MD PhD
Department of Primary Health Care
Göteborg University
SWEDEN

Back to Main menu