d & rsquo Performances; a diagnostic test

0
4579

I- PREAMBLE :

  • Before a reason for consultation, after questioning and clinical examination and on the basis of & rsquo; biological tests:
  • The doctor gets an idea about the possible diagnoses or II emits a priori (of the) hypothesis(s)
  • On this basis he prescribed (of the) exam(s) biological(s) and / or radiological(s).
  • Once the(s) result(s) got(s), the practitioner will be a posteriori idea about the plausibility of(s) l & rsquo; hypothesis(s) issued(s) a priori
  • Probability stronger if results(s) positive(s) and less if results(s) negative)

II- INTRODUCTION :

  • Decision making (formulation & rsquo; diagnosis) : based on a valid course of action in view of the reality of the situation d & rsquo; a patient
  • This diagnostic approach uses diagnostic tests
  • Unimaginable medicine without technical platform : significant and growing number of diagnostic tests under
  • Mass screening : based on tests
  • diagnostic tests : Their use requires knowledge of their: performance and limitations
  • diagnostic test ? Facing the sick

C & rsquo; is any means d & rsquo; get useful information to the practitioner, as part of & rsquo; a decision support

III- TYPES OF DIAGNOSTIC TEST :

  • Clinique exam : Blood Sugar, ASLO, Hémogramme, etc.
  • Medical imaging : Scanner, IRM, pulmonary radio face, coronarographie, etc.
  • Performance test : Test d & rsquo; stress, spiromètrie, etc.
  • physical sign : sign Koplick, eruption, nodule, etc.
  • functional sign : Triad evocative d & rsquo; pathology, persistent cough, etc.

IV- TYPES OF ANSWERS :

The test response can be

  • either binary : positive or negative
  • Either several modalities

Response to several conditions :

* Ordinal response to several conditions : SCORE Birad (mammography for breast cancer screening)

  1. : Normal
  2. : Benign tumor
  3. : probably benign tumor
  4. : Suspected malignancy
  5. : Malignant tumor

* quantitative response (countless ways) :

It has set a threshold or to return to a binary or ordinal response with several modalities

→ A good screening test should :
– be reliable
– reproducible
– easy to apply
– be accepted by the "healthy assets"
– n & rsquo; have little d & rsquo; s side effects
– be moderate cost

V- PERFORMANCE OF DIAGNOSTIC TEST :

Measurements (probabilities) validity of two types :

Experimental measurements of d & rsquo performance; test : Sensitivity, Specificity
Measures in real situations d & rsquo performance; test : positive predictive value, negative predictive value

1- Experimental Performance Measures :

Also tell Measures (probabilities) intrinsic validity d & rsquo; a diagnostic test in experimental situation :

  • Evaluate the quality of the method itself
  • Try the & rsquo; d & rsquo ability; a diagnostic test to identify the presence and & rsquo; absence of disease
  • Ability d & rsquo; a diagnostic test to identify the disease
  • Validate : estimate the intrinsic performance of & rsquo; a diagnostic test new compared to existing tests or existing, that the reference test (gold standard) can be put into practice : too expensive, risky, time alone can provide an answer and we can not wait

The measures aim to answer the following question :

The test he reacts correctly to the presence or the & rsquo; lack of what it is intended to highlight ?

A- Sensitivity :

C & rsquo; is the ability of & rsquo; a test to give a positive result when the phenomenon is present C & rsquo; is the probability that & rsquo; a subject is positive test knowing that & rsquo; he really is sick

B- Specificity :

C & rsquo; is the ability of & rsquo; a test to give a negative result when the phenomenon is absent C & rsquo; is the probability that & rsquo; a subject or a negative test knowing that & rsquo; he really is not sick

  • Sensitivity is the ability of & rsquo; a test to alert patients
  • Specificity is the ability of & rsquo; a test to not alert falsely non-sick

Calculation methods :

  • Sensitivity and Specificity : Estimated using an experimental type of study
  • Clinical test : validation study d & rsquo; a diagnostic test
  • Study based on the formation of two groups : "Sick" and "not sick"
  • The disease is defined from d & rsquo; a reference test (gold standard)
Sickness

Reference Test result

M NM
Test result T + VP FP
T- FN VN
Total VP+ FN VR+FP

X= VP+FN Y= VR+FP

M : Group Patients
NM : No group of Sick
VP + FN : Total sick (M) under test to validate
VN + FP : Total non-sick (NM) under test to validate
T + : Positive diagnostic test
T – : Negative Diagnostic Test

  • VP : true positives (number of positive tests in patients) : Diagnosed with success
  • FN : false Negatives (number of negative tests in patients) : diagnosis failed
  • VN : true Negatives (number of negative tests in non-diseased) : Diagnosis removed successfully
  • FP : false Positives (number of positive tests among non-patients) : false alarm

Estimated value of the → :

  • Sensitivity : From = VP / (VP + FN)
  • Specificity : Sp = VN / (VN + FP)

2- real situations performance measures :

Two measures of validity in real situations :
Positive predictive value
Negative predictive value

Also tell Measures (probabilities) predictive validity or extrinsic validities :
What confidence grant a result of & rsquo; test ?
– Ability to describe the actual situation of the disease in the population studied
– Ability d & rsquo; a diagnostic test to confirm the presence or absence of disease
– A good internal validity (sensitivity and specificity) d & rsquo; n & rsquo a test, is not necessarily a good diagnostic tool (or screening) !

C & rsquo; is the answer to the question :
– "Does this mean the positive test that & rsquo; individual is affected by the problem or not ? »

C & rsquo; is & rsquo; d & rsquo event; be sick once the result of the resulting diagnostic test.
– "Does this mean the positive test that & rsquo; individual is affected by the problem or not ? »

C & rsquo; is & rsquo; estimated positive predictive values ​​and negative

A- Positive predictive value :

  • A positive test there is a high probability of & rsquo; the problem exists ?
  • A probability of & rsquo; have the disease for a group about the positive test
  • real situations performance measures

B- Negative predictive value :

  • A negative test is there a high probability of not being affected by the problem ?
  • A probability of & rsquo; have the disease for about negative group test

❖ Calculation methods :

Sickness

Reference Test result

Total
M NM
Test result T + VP FP VP+FP
T- FN VN FN+VN

❖ Estimated value of the :

  • Positive predictive value : VPP = VP / (VP + FP)
  • Negative predictive value : VPN = VN / (VN + FN)

IV- CONCLUSION :

  • Performance Measures : probabilities between 0 and 100 % (0 and 1)
  • No diagnostic test = 100 % !
  • A test is d & rsquo; all the better that & rsquo; it is both sensitive and specific : Se and Sp tend to 100%
  • A test is d & rsquo; more interesting than its predictive values (VPP et VPN) tend to 100%
  • A sensitive test (Reaches out to 100 %) rarely misses subjects with disease
  • A specific test (Sp tend vers 100 %) rarely says that & rsquo; a subject is sick so that & rsquo; s it & rsquo; is not.
  • a sensitive test is used when :

– The disease is serious and that & rsquo; it must not be ignored
– The disease is curable

  • It uses a test spécifique.quand :

– The disease is difficult to cure
– incurable

  • It is important to know that & rsquo; it n & rsquo; is not ill : l & rsquo; existence of false positive results in serious problems

Cours du Dr LAKEHAK – Faculty of Constantine