Biostatistics Flashcards
LR > 1
Disease you’re worried about is more likely after test result
(Likelihood ratio > 1)
LR < 1
Disease of concern is less likely after test result
LR = 1
Pt likelihood of having disease hasn’t moved
(e.g. still “30%” before and after test result)
LR = 10 is a very good _____
Rule-in test
(e.g. b-hCG, CT PE, “Gold Standard” tests)
Higher LR = better for rule-ins
Closer to 0 = better for rule-outs
Closer to 1 = less useful
LR = 0.1 is a very good _____
Rule-out test (e.g. D-Dimer)
Higher LR = better for rule-ins
Closer to 0 = better for rule-outs
Closer to 1 = less useful
How often a test is right when people have the disease
Sensitivity
(True positive rate)
- High sensitivity = low false negatives
How often the test is negative when people don’t have it
Specificity
(True negative rate)
- High specificity = low false positives
Probability of the outcome of interest occuring in the exposed group compared to the probability of it occuring in the non-exposed group
Relative Risk (RR)
- RR of 1.0 = null value (outcome occurs w/ equal frequency in both groups; no association btw exposure and outcome)
- RR > 1.0 = positive association (outcome occurs more frequently in exposed group)
- RR says nothing about the significnace of the study
95% confidence interval p value
p < 0.05 (Statistically significant)
- Note: if p > 0.05, then the 95% confidence interval contains the null value (1.0)
99% confidence interval p value
p < 0.01
Proportion of pts with a negative test result who truly do not have the disease
NPV
Chance that a positive test is truly negative (# true negatives / total negative tests)
Proportion of pts with a positive test result who actually have the disease
PPV
Chance that a positive test is truly positive (# true positives / total positive tests)
true positives / # true positives + # false positives
The positive and negative predictive values of a diagnostic test are highly dependent on the ____
Prevalence of disease in the population
Sensitive tests rule in/out
Out
(SnOUT / SpIN)
- High sensitivity = low false negatives = Negative test is reliable
Specific tests rule in/out
In
(SnOUT / SpIN)
High specificity = low false positives = Positive test is reliable
LR(+) values of 2, 5, and 10 correspond to an increase in disease probability by __, __, and __ respectively
15%, 30%, 45% increase in probability
LR(+) values of 0.5, 0.2, and 0.1 correspond to a decrease in disease probability by __, __, and __ respectively
15%, 30%, 45% decrease in disease probability
The difference in risk attributable to the intervention when comparing placebo and intervention groups
(Placebo rate - intervention rate)
Absolute Risk Reduction (ARR)
Absolute Risk of A - Absolute Risk of B
- NNT = 1/ARR
NNT
1/ARR = 1/(Absolute Risk of A - Absolute Risk of B)
The absence of a control group would affect a study’s ____
Validity
The post-test probability is dependent on these 4 things
- Sensitivity
- Specificity
- Pre-test probability
- PPV
Cases are selected from the entire disease population instead of just those who are newly diagnosed
Selective survival bias
(because study not limited to newly diagnosed pts will contain a higher proportion of relatively benign disease as these pts generally live longer)
People in a study who have suffered an adverse event are more likely to recall risk factors than those without such experiences
Recall Bias
Randomization is used to control for ____
Confounding
Methods to control confounding
- Randomization
- Matching (levels of a confounder between groups)
- Restriction (e.g. only men)
- Stratified analysis
- Statistical modeling (e.g. multivariate analysis)
Tight confidence interval
Precision
(low random error)
False Positive Rate
1 - Specificity
- High specificity = Low false positive rate
False Negative Rate
1 - Sensitivity
- High sensitivity = low false negative rate
Indicates how well a test can screen for disease
Sensitivity
Indicates how well a test can confirm the diagnosis
Specificity
The proportion of true results (true positives and negatives) out of all the results of a given diagnostic test
Accuracy
Increasing a diagnostic test’s cutoff value will ___ the sensitivity and ___ the specificity
- Decrease the sensitivity (more false negatives)
- Increase the specificity (less false positives)
A study’s ability to detect a difference between groups, if one exists
Power
The number of people that need to receive a treatment to prevent 1 additional adverse event
Number Needed to Treat (NNT)
1/ARR
Comparing two means: T test vs. Z test
Two-sample T test - compares means using sample variances
Two-sample Z test - compares means using population variances
The alteration of behavior by the subjects of a study due to their awareness of being observed
Hawthorne Effect
Unlike sensitivity & specificity, PPV and NPV depend on ____
The prevalence of disease in the population being tested
However, changing cutoff point of a diagnostic test affects all four
Atrributable Risk Percent (ARP) is easily derived from
Relative Risk
ARP = (RR-1)/RR
Generalizability
External validity
Applicability of study results to other populations
“A new screening test” - Think of this type of bias
Lead-time bias
Prolongation of apparent survival in patients to whom a test is applied, without changing the prognosis of the disease
Statistical test used to compare 3 or more groups using discrete or nominal (categorical) variables
Chi-Squared Test
Statistical test used to compare 3 or more groups using continuous (numerical) variables
ANOVA
