Clinical Effectiveness: Interpreting test results

1
Clinical EffectivenessInterpreting test results

• Nick Price 17th October 2006

2
Aims

• to reflect on the implications of a study of
  health professional's interpretation of a test
  result
• to develop skills in interpreting test results

3
Objectives

• By the end of the session you should be able to
• Define sensitivity in ordinary language
• Define specificity in ordinary language
• Understand how the prevalence of a condition in
  your test population influences the significance
  of a positive test result in a particular
  patient.
• Understand how 'testing more patients, just in
  case' will influence the likelihood of a patient
  with a positive result having the condition.
• Understand to term 'positive predictive value'.
• Have an opportunity to try explaining the result
  of a test to your peers.

4
Sensitivity

• How many true positives in comparison to the
  gold standard.
• Or (most accurately)
• The chance of having a positive test, assuming
  that you do have the condition.
• Or
• So with a very Sensitive Test a Negative will
  rule Out the condition SnNOut
• Or
• So a sensitive test is likely to pick up the
  condition.

5
Sensitivity 2

• Can you think of some tests with very high
  sensitivity in comparison to a gold standard?
• e.g. D-dimer (99), Leucocytes on Multistix
  (87), random blood sugar

6
Specificity

• (most accurately)
• The chance of having a negative test given that
  you do not have the disease.
• Or
• How many false negatives.
• Or
• With a very Specific test a Positive result rules
  the condition IN -SpPin
• So with a specific test a positive test is likely
  to mean you have the condition.

7
Specificity 2

• Can you think of some very specific tests?
• 3 of glucose and ketones on multistix?
• A hard craggy breast lump?
• A yes score of 3 on CAGE (99.8)
• Some not very specific ones
• Moderately raised random blood sugar in general
  population

8
The Truth Table
Sensitivity is the probability a / (a c) in
the table that a true positive has been
correctly classified as positive by the
test. Specificity is the probability d / (b
d) that a true negative is correctly classified
negative by the test
9
Example

• With leukocyte esterase dipstix (LED) for
  chlamydia vs gold standard
• In a GUM clinic 500 patients were tested, 100
  tested positive with gold standard, 90 tested
  positive with LED. Of these 90, 5 were in fact
  negative with the gold standard.
• What is the sensitivity and specificity of LED

10
Example 2Sensitivity 85/100 85Specificity
395/400 98
11
So what is the chance that a positive LED test
means you have chalmydia?

• Aka what is the positive predictive value
  (PPV).
• This is the true positives / true positives and
  the false positives
• PPV a/ac 85/90 94.
• Excellent, so this is a good test to use in GP
  e.g. routinely when taking smears!

12
PPV 1

• So the incidence of chlamydia in the general
  population of all women having smears in GP is
  say 5.
• We do 500 smears a year
• We have a test that has sensitivity of 85 and a
  marvellous specificity of 98.
• What chance the patient with a positive test
  actually has chlamydia in this context?

13
Example 3Sensitivity 85Specificity 98PPV
21/31 67NPV 465/469 99
14
So the incidence of the disease greatly effects
the PPV or how many patients you will see with
false positive test result
15
So what about the case in the experimental study?

• 1 of babies have Downs
• If the baby has Downs 90 will have ve test.
• If the baby does not have Downs 1 chance the
  result will be positive
• With a ve result what is the chance baby has
  Downs?

16
So what about the case in the experimental study?
2

• 1 of babies have Downs (incidence)
• If the baby has Downs 90 will have ve test.
  (90 sensitivity)
• If the baby does not have Downs 1 chance the
  result will be positive (99 specificity)
• With a ve result what is the chance baby has
  Downs? (PPV)

17
Example 4 Maths solutionSensitivity
90Specificity 99PPV 90/190 47NPV
9800/9810 99.9
18
Example 4 narrative solution

• Read the paper!
• Now practice explaining one of these example in
  trios, then rotate.

19
Objectives

• By the end of the session you should be able to
• Define sensitivity in ordinary language
• Define specificity in ordinary language
• Understand how the prevalence of a condition in
  your test population influences the significance
  of a positive test result in a particular
  patient.
• Understand how 'testing more patients, just in
  case' will influence the likelihood of a patient
  with a positive result having the condition.
• Understand to term 'positive predictive value'.
• Have an opportunity to try explaining the result
  of a test to your peers.