Biostats 3- Analysis of Screening Tests Flashcards
What is the eqn for a sample size needed for comparing differences in means between 2 groups with an alpha = 0.05?
n = 16s^2/d^2
s= standard deviation d= difference btwn the means
True or False: you use the same eqn for difference among more than 2 means.
True
What is the eqn needed for doing a correlation between 2 continuous variables?
n = 4 + 8/r
r= estimated correlation coefficient
What does “ballparkness” give you?
the order of magnitude of people for a study (multiple of 10)
How much should u multiply your calculated sample size by to reduce migration bias?
1.25
attrition
Why is screening a prevetnative intervention?
cuz u do it on asymptomatic individuals that allows for early detection, therapy, and ↓ mortality from the disease.
What are the 4 selection criteria for a disease that can be screened?
- Disease should be an important public health problem
- Early detection in asymptomatic individuals is possible and improves outcome
- Screenign tests are acceptable/inexpensive/accurate
- Treatment is available for disease
For Bayes Theorem, what is the tool you use to describe screening test outcomes within a popualtion?
2x2 contigency table
This is a proportion of true positives correctly identified by the test over those wwho do have the disease
(test + | pt present)
Sensitivity
What is the SE eqn for sensitivity?
SE = a/(a + c)
What is the opposite of sensitivity, where its the proportion of true negatives over those who DONT have the disease?
(test - | no disease)
Specificity
What is the eqn for specificity?
Specificity = d/(b + d)
True or False: the higher the sensitivity, the fewer real cases of diseases go undetected.
True
True or False: True higher the specificity, the fewer cases of normal patients go undetected .
True
This is the # diseased per 1,000.
Prevalence
This is the probability that pts with a + test result were correctly diagnosed as having the disease.
Positive Predictive Value (PPV)
What is the eqn for PPV?
PPV = a/(a+ b)
AKA”everyone who tested +”
use rows
This is the proportion of pts with a negative test result that did in fact NOT have the disease.
Negative Predictive Value (NPV)
What is the eqn for NPV?
d/(c + d)
notice we use d here cuz d is the true negative.
Accoring to beyes theorem, waht happens to PPV and NPV as the prevalence of the disease increases?
PPV ↑
NPV ↓
This is a measure of how likely it is that an event will occur.
Probability (P)
In mutually exclusive probabilities, how do you do measure probabilities?
Additive rule
roll a 2 OR a 5…
If 2 events are independent, how do u measure probabilities?
Multiplication rule
roll a 6 and THEN a 5…
What is the % chance a chem-6 panel comes up normal for everything, if each variable has a P=0.95?
74%
This is the proportion of pts who were diagnosed with disease who are alive 5 years later.
Survival rate
This is the proportion of ppl who died within 1 year.
Mortality rate
This is the type of bias which depends on the TIME of Dx, which can sway the survival rate.
Lead time bias
This is the type of bias where screening detects less aggressive cancers with long preclinical phases but not likely to cause death. Can sway the survival rate.
Overdiagnosis bias
This is the bias wher eindividuals who participate in screening trials are fundamentally different from those who do not.
Selection bias
What is the goal of screening: reduce mortality rates or increase survival rates?
Reduce mortality rates
