Intro Flashcards
What percentage of diagnoses rely to some extent on a lab result?
60-70%
What are some of the general reasons to order a lab test?
diagnose a disease screen for disease risk assessment of future disease monitor disease progression monitor response to treatment
What are the general principles of lab diagnosis?
Generally speaking, labs look for changes from normal in regards to:
- Damage to cells (leakage of itnracellular contents)
- Failure of excretory processes from organ damage
- Increased amount of tissue-of-origin (tumor markers)
- Co-incident patterns of biochemical markers
- Genetic factors rapidly emerging
In general, how is the reference range made?
population mean plus or minus 2 std deviations
so includes 95% of observations in the healthy population
Based on how the reference range is made, 5% of values will be outside the range even in a healthy person. If n independent tests are performed on a healthy person, what number are predicted to be abnormal?
1 - (0.95)^n
if you run a test 20 times, you have a 64% likelihood of finding at least one abnormal value
In some cases we use a more stringest criterion to establish the abnormal level. What’s the main example of this?
troponin needs to be over the 99th percentile of the upper reference limit of a normal population to define AMI
How was the cholesterol range defined?
by a committee (national cholesterol education program)
in fact, when you look at the actual range for cholesterol, the mean is 201 (above “normal” and about 20% of the population is above 240!)
What is the diagnostic definition of a test’s sensitivity?
TP/(TP+FN)
How many people with the disease test positive?
What is the diagnostic definition of a test’s specificity?
TN/TN+FP
How many people without the disease will test negative
Usually normal and diseased values overlap, so if you increase sensitivity, you decreased specificity and vice versa. What is the graphical representation of this called?
a receiver-operator curve
plot of sensitivity vs. 1-specificity
What do you look for on a receiver-operator curve to indicate the ability fo a test to separate the groups of interest (healthy vs. diseased)?
You measure the area under the ROC curve
higher values will indicate better ability to separate the groups
What is the definition of positive predictive value?
If someone tests positive for a disease, how likely is he/she to have the disease?
TP/TP+FP
What is the definition of negative predictive value?
If someone tests negative for disease, how likely is he/she actually healthy
TN/TN+FN
True or false: disease prevalence affects tests sensitivity and specificity.
false - it affects the predictive values, not the sensitivity/specificity
higher prevalence = fewer FP, so higher positive predictive value
lower prevalence = more false positives, so lower predictive value
Screening in low prevalence groups is therefore inherently limited. For example, what is the PPV of the minnesota CAH (newborn screen)
only 8%!!
