Lecture 7 - Biostats Part 3-1 Flashcards
Summarizes the same kind of information sensitivity and specificity and can be used to calculate the probability of disease in a low prevalence setting.
Likelihood ratio (LR)
provides indication of the test’s discriminatory power.
Likelihood ratio (LR)
Predictive values are lower with a low prevalence…. SO WE USE?
Likelihood ratio (LR)
______ can be defined for the entire range of test result values
Likelihood ratio (LR)
Low prevalence = Less reliable positive test result; therefore, use
Likelihood ratio (LR)
_____addresses: How much more likely are we to find that a test is positive among patients with disease compared with those without disease?
Likelihood ratio (LR)
LR = ?
Likelihood of the same result in someone WITHOUT the disease
(W/ WO)
How good the test is at “Ruling in” disease!
A positive LR (LR+)
____________ is the ratio of the proportion of diseased people with a positive test result (sensitivity) to the proportion of non-diseased people with a positive result (1-specificity).
A positive LR (LR+)
Range: 1.0 to infinity; Null value: 1.0 (no difference)
The bigger the better (Desirable: 5 or more)
How good the test is at “Ruling out” disease
A negative LR (LR-)
is the proportion of diseased people with a negative test result (1-sensitivity) divided by the proportion of non-diseased people with negative test results (specificity)
A negative LR (LR-)
The smaller the better (Desirable: 0.2 or less)
Range: 0.0 to 1.0; Null value: 1.0 (no difference)
Is one of the most common ways to examine relationships between two or more categorical variables.
Chi-square
___________ tests the null hypothesis that the variable are independent of each other, that there is no relationship between the two variable.
The chic-square of independence
does not give any information about the strength of the relationship.
Chi-square
Computed the same way as the chi-square test for independence, but instead tests the hypothesis that the distribution of some variable is the same in all populations.
Chi-square test for equality of proportion:
Is used to test they hypothesis that the distribution of a categorical variable within a population followed a specific pattern of proportion.
Chi-square test of goodness of fit:
A non-parametric test similar, similar to the chi-square tests, but can be used with small or sparsely distributed data sets.
Fisher’s exact test:
Is a type of chi-square test used when the data comes from paired samples.
McNemar’s Test for Matched Pairs:
Odds ratio (OR)
measures?
Measures the strength of association between an exposure and disease
Odds ratio (OR)
OR the effect of one intervention v. another
OR = (AD)/(BC)
Look at the slide on page 20
Odds ratio (OR)
If exposure does not affect (either cause or protect from) disease, the OR is ~ 1
If the exposure is related to the disease, the OR > 1
If the exposure is protective against the disease, the OR < 1
Keep in mind that odds of an event can be defined as the ratio of the number of ways the event can occur to number of ways the event cannot occur.
Just as we estimated the mean value of the response when the dependent variable was continuous, we would like to be able to estimate the probability of an outcome associated with a dichotomous response for a single or multiple variables.
What can be used for this purpose?
Logistic Regression
a single outcome (or set of outcomes) from an experiment.
Event
The proportion of subjects in a study group in whom the event is observed. Usually seen as a %.
Rate
A measure of how often a particular event (such as response to a drug, adverse event or death) occurs within the scientific control group of an experiment.
Control Event Rate (CER) %
A measure of how often a particular event (such as response to a drug, adverse event or death) occurs within the experimental group of an experiment.
Experimental Event Rate (EER) %
Basic risk statements express the likelihood that a particular event will occur within a particular population
What exposure is responsible for an illness or other outcome?
Identifies what in our environment can lead to beneficial or adverse medical outcomes
Relative risk
Must have incidence information to calculate
Cohort or clinical trials are conducted over time
Relative risk is calculated using cumulative incidence data to measure the probability of developing disease
______ measures the magnitude of an association between an exposed and non-exposed (control) group.
Relative risk
Makes insignificant findings appear significant!
Relative risk reduction
Not a good way to compare outcomes
Does not report the baseline risk of outcome
Relative risk reduction
Measures such as percent reduction in mortality, is selected because it gives a more optimistic view of the effectiveness of a preventive measure.
Relative risk reduction
The percentage difference in outcome between control (C) and experimental (E) groups
Formula?
Relative risk reduction
RRR= (CER-EER)/CER
The actual reduction in events in the treated group (EER)
Absolute risk reduction
The arithmetic difference in outcomes between treatment and control groups
Absolute risk reduction
The “true difference” between the experimental and control intervention
Formula
ARR = CER - EER
Absolute risk reduction
True difference
Absolute risk reduction
When is the odds ratio a good estimate of relative risk?
In a case-control study, only the odds ratio can be calculated as a measure of association ,whereas in a cohort study, either the relative risk or odds ratio can be calculated.
Odds ratio are an ________ of risk, not a direct measure of risk.
indirect estimate
Odds ratios calculated in a case-control study are a good approximation of relative risk in the population when the following conditions are met:
When cases studied are representative, with regard to history of exposure, of all people with the disease in the population from which the cases were drawn.
When the controls studied are representative, with regard to history of exposure, of all people without the disease in the population from which the controls were drawn.
When the disease being studied does not occur frequently.
Key point: The odds ratio is not a good estimate of the _____ when disease occurrence is frequent.
relative risk
Number needed to treat
The number of patients who need to receive the new intervention instead of the standard alternative in order for….
one additional patient to benefit
Number needed to treat
Formula?
NNT= 1/ARR
Expresses the likelihood of the treatment to benefit an individual patient
Number needed to treat
There is NO absolute value for NNT that defines whether something is effective or not.
NNTs for treatments are usually low because we expect large effects in small numbers of people
Yessiree
NNTs for very effective treatments are usually in the range of
2 to 4
Number needed to treat
Since few treatments are 100% effective, and few controls are without some effect (including placebo or no treatment)
Cool
Number needed to treat
Rule of thumb for therapy and prevention?
Rule of thumb:
NNT 10 or less for therapy
NNT 20 or less for prevention
Number needed to harm (NNH)
When an ______ is detrimental, the term number needed to harm (NNH) is often used.
experimental treatment
Number needed to harm (NNH)
The equations and approach are similar to those described above, except that NNH will have a negative
absolute risk reduction
AKA Student’s t-test
T-test
Generally is used to analyze continuous data
T-test AKA Student’s t-test
Compares the means and standard deviations of two populations
Data be must be normally distributed
T-test AKA Student’s t-test
Computes a p-value to test the null hypothesis
Null hypothesis: is that the difference between the two group means is 0 or no difference.
Alternative hypothesis: The difference between the two group means is >0 or there is a difference.
-test AKA Student’s t-testT
T-test
Assesses whether a difference between two groups’ averages is unlikely to have occurred because of random chance in sample selection. A difference is more likely to be meaningful and “real” if-
- The difference between the averages is large.
- The sample size is large.
- Responses are consistently close to the average values and not widely spread out (the standard deviation is low).
which is better for causation… odds ratio or relative risk?
relative risk!!