Thursday, 13 January 2011

Diagnostic Test Performance

Sensitivity: PIDPositive In Disease” = true positive / all who are really diseased (true positive + false negative by that test)

Specificity:  NIHNegative In Health” = true negative / all who are really disease free (true negative + false positive by that test)

Positive Predictive Value (PPV):  true positive / all +ve results by that test in other words true positive + false positive

Negative Predictive Value (NPV):  true negative / all –ve results by that test in other words true negative + false negative

In simple words:

Sensitivity: probability of a positive TEST result if the patient is diseased.

Specificity: probability of a negative TEST result if the patient is not diseased.

PPV: probability that a PATIENT tested positive is indeed diseased.

NPV: probability that a PATIENT tested negative is indeed not diseased.

The predictive value of a test not only depends on its sensitivity and specificity, but also on the disease prevalence.

The more prevalent the disease, the higher the positive predictive value of the test and the lower its negative predictive value and vice versa.

This may confuse some, as one might say, why on earth would the prevalence be affecting the whole issue in the population tested?

The following example will clarify it:

Using a test with a sensitivity of 100 % and a specificity of 99 % (i.e. 1 false positive in 100), and screening 1,000 patients, one would expect to see the following:

*In a setting with a high disease prevalence of 10 % (i.e. 100 per 1,000)
100 true positives per 1,000
10 false positives per 1,000 (constant)
Positive predictive value: 100 true positives/110 total positives = 91 %
100 (91 %) of the 110 positive test results are true.

*In a setting with a low disease prevalence of 0.1 % (i.e. 1 per 1,000)
1 true positive per 1,000 (this figure is now significantly lower as the disease is less prevalent)
10 false positives per 1,000 (constant)
Positive predictive value: 1 true positive/11 total positives = 9.1 %
Only 1 (9.1 %) out of 11 positive results is true.
In other words, >90 % of patients tested positive in a low-prevalence population will be actually false positives!

No comments:

Post a Comment

Note: only a member of this blog may post a comment.

Main Works of Reference List (The first eight are my top favourites)

  • British National Formulary
  • British National Formulary for Children
  • Guidelines (BAD - BASHH - BHIVA - Uroweb)
  • Oxford Handbook of Genitourinary Medicine, HIV, and Sexual Health
  • Oxford Handbook of Medical Dermatology
  • Rook's Textbook of Dermatology
  • Simple Skin Surgery
  • Weedon's Skin Pathology
  • A Concise Atlas of Dermatopathology (P Mckee)
  • Andrews' Diseases of the Skin
  • Andrology (Nieschlag E FRCP, Behre M and Nieschlag S)
  • Bailey and Love's Short Practice of Surgery
  • Davidson's Essentials of Medicine
  • Davidson's Principles and Practice of Medicine
  • Fitzpatrick's Colour Atlas and Synopsis of Clinical Dermatology (Klaus Wolff FRCP and Richard Allen Johnson)
  • Fitzpatrick’s Dermatology in General Medicine
  • Ganong's Review of Medical Physiology
  • Gray's Anatomy
  • Hamilton Bailey's Demonstrations of Physical Signs in Clinical Surgery
  • Hutchison's Clinical Methods
  • Lever's Histopathology of the Skin
  • Lever's Histopathology of the Skin (Atlas and Synopsis)
  • Macleod's Clinical Examination
  • Martindale: The Complete Drug Reference
  • Oxford Handbook of Clinical Examination and Practical Skills
  • Oxford Textbook of Medicine
  • Practical Dermatopathology (R Rapini)
  • Sexually Transmitted Diseases (Holmes K et al)
  • Statistics in Clinical Practice (D Coggon FRCP)
  • Stockley's Drug Interactions
  • Treatment of Skin Disease: Comprehensive Therapeutic Strategies
  • Yen & Jaffe's Reproductive Endocrinology