Sensitivity, Specificity, and Predictive Values
What is the "gold standard"? what may be some characteristics of it?
reference standard for diagnosing a particular disease to which all other diagnostic tests are compared; invasive, expensive, risky, and painful
What are the properties of diagnostic tests that make it useful?
reliably classify patients into those who have disease and those who do not have the disesae
Review 2x2 table in written notes
...
What is the clinical context of an individual patient that guides the interpretation of diagnostic tests?
1) history, PE, prevalence of disease in population, epidemiologic pattern of diseaes, risk facctors, clinical judgment 2) suspicion 3) probability of diseae 4) pre-test probability of disease
What does a NPV value of 85% mean?
85% of patients with a negative test actually do not have the condition and 15% of patients with negative test are false negatives, they actually have the condition
What does a PPV value of 85% mean?
85% of patients with a positive test actually have the conditions and 15% of patients are false positives
What is the significance behind the acronyms SPpIN and SNnOUT?
SPpIN = a positive result in a test with high specificity is useful to rule in a diagnosis SNnOUT = a negative result in a high sensitivity test, is helpful to rule out the diagnosis
What is our job as clinicians?
interpret the test and decide if, when confronted with an individual patient, can the test help determine whether the patient has or does not have the disease if the test does not help answer the question - don't order it
What does a perfect test show?
all patients are correctly classified as either having or not have the disease; no false negatives or false positives
What are answers like of tests?
continuous = Hb, lipase, TSH, etc. titers of antigens/antibodies = anti-nuclear antibody interpretation: physical findings, mammogram, EKG
The take home point is that sensitivity and specificity are measures of what?
diagnostic accuracy of a test
What is the take home point of the FBG and diabetes?
diagnostic tests with continuous results require a cut-off value and above the cut-off value, results are positive
What is the goal when looking for cut-off value?
excludes most people who do not have a condition w/o missing an unacceptable number of people who do
If the prevalence of a disease is low, most positive test reusts will be what?
false positives
Right now what is the FBG standard to diagnose diabetes?
greater than or equal to 126 mg/dl
Diagnostic tests are interpreted how?
in the clinical context of an individual patient: one test in one patient is ordered for one reason and in another patient it may be for another reason
What if the FBG standard to diagnose diabetes was shifted to 100?
more patients would be labeled diabetic sensitivity would increase so all diabetics would be correctly classified and specificity would decrease = incorrectly classify some w/o
What if the FBG standard to diagnose diabetes was shifted to 300?
more patients would be labeled non-diabetic lower sensitivity = only a small percentage of diabetics would be correctly classified as diabetics; more would be incorrectly classified as nondiabetic higher specificity = all non-diabetics would be correctly classified
Is there such a thing as a perfect test?
no
What is the gold standard for diagnosis of type 2 DM?
oral glucose tolerance test: fast 8-16 hrs, drink 75g of glucose and serum glucose before and 3-5 times after over the next 3 hours, usually at 2 hours; 200 mg/dl or greater at two hours = positive test
What concepts are used to describe if the test is useful?
positive and negative predictive value
What is pre-test probability? what type of estimates are used?
probability of the target disorder before the results of a diagnostic test are known subjective estimates disease prevalence estimates clinical predication rules estimates
What is specificity?
proportion of patients who do not have the disease who have a negative test; probability that a test will be negative given a patient w/o the condition
What is sensitivity a measure of?
proportion of patients who have the disease who have a positive test; probability that a test will be positive given a patient with the condition
Review: what is prevalence?
proportion of population that has the condition or disease
What are types of tests in the clinical setting?
screening diagnostic prognostic monitor disease activity
What are the specific properties of a screening test that makes it useful?
screening test needs to be acceptable to the patient, inexpensive, and identify disease before it is clinically apparent and at a time when treatment is available and such treatment impacts survival
What are concepts used to describe if the test is accurate?
sensitivity and specificity
Unless a test is perfect what is there a trade off between?
sensitivty and specificity: in order to gain sensitivity you lose specificity and vice versa
What are some tests that help monitor disease activity?
serum ACE level for sarcoidosis PSA level for prostate cancer
What are some examples of prognostic tests? what are their properties that make them useful?
serum albumin in case of hepatic failure and scoring system such as GAIL model for breast cancer or PORT score of community acquired pneumonia; they need to reliably predict outcomes that help clinician decide how to communicate info or make decisions for a particular patient
If you want to assess the accuracy of a test, it is essential to compare it to what?
some established method of diagnosing the condition
What depends on the cut-off value?
some healthy patients may (falsely) test positive and some diseased patients may (falsely) test negative
Each type of test needs what?
specific properties to be useful
What happens when prevalence increases?
the PPV increases and NPV decreases
What happens when prevalence decreases?
the PPV value decreases and NPV value decreases, so there are more false positive than there are true positives if PPV values is decreased
PPV and NPV are measures of a test's _______. what does it vary with?
usefulness; prevalence of the condition for which you are testing
What does it mean that sensitivity and specificity are measures of a test's accuracy?
values should be consistent when test is used in similar patients in similar settings
Most clinical questions have what type of answer? Do tests have the same answer?
yes or no; no
Can these tests be used for more than one purpose?
yes, depending on the clinical situation
Can one test be classified as more than just a screen, diagnostic, prognostic, or disease monitoring test? give example.
yes, in a mammogram, it can be a screening test and a diagnostic test