General Definitions Flashcards
Accuracy
How close to the actual value
Aka validity
Precision
How reproducible are the results
Aka reliability
Sensitivity
how many people with the disease will have a positive test?
=true positives / those with disease
Specificity
How many people without the disease will have a negative test
=(true negatives) / (those without disease)
Pre-test probability
Probability you have the disease before test is done
Post-test probability
Probability that you have the disease after test results have been received
Predictive value of positive test
Chance that the positive result is correct
=(true positives) / (all positive tests)
Predictive value of negative test
Chance that negative result is correct
= (true negatives) / (all negative tests)
Threshold concept
- TESTING THRESHOLD: How sure must I be that individual does NOT have illness in order to feel comfortable not treatingn them?
- TREATMENT THRESHOLD: How sure must I be that individual DOES have illness in order to feel comfortable treating?
PICO
Patients
Intervention
Comparison
Outcome
Likelihood ratio
Ratio between likelihood of a particular test result in those patients with the disease to the likelihood of the same test result in those without the disease
Can have likelihood ratios for positive and negative tests
Likelihood ratio for positive test
(Likelihood positive test in diseased) / (likelihood positive test in healthy)
Sensitivity / (1-specificity)
Likelihood ratio for a negative test
(Likelihood of a negative test in diseased) / (likelihood of negative test in non-diseased)
(1-sensitivity) / specificity
Criteria for a reasonable screening test
- Substantial burden of suffering from the suspected disease
- Test is a fairly good predictor of the presence or absence of the disease
- Finding and treating the disease early leads to a better outcome than treating after symptoms appear
- Benefits of screening outweigh the harm
EBM definition
Integration of the best research evidence with our clinical expertise and our patient’s unique values and circumstance
Should be patient centric
Perfect study type for diagnosis and prognosis questions
Large prospective cohort with 100% capture of information and no loss to follow up
Diagnosis using a gold standard
Prognosis: everyone starting at a well-defined time in their illness
Perfect study for intervention/exposure questions
Large RCT with 100% adherence and followup, blinding, and similar patients
P-value
Probability of finding level of association when the null hypothesis is true, due completely to chance
Above p-value–>cannot reject the null hypothesis because association could be due to chance
Below p-value–>significant difference does exist; reject the null hypothesis
Confidence interval
Estimate of the precision of the results and statistical significance
Type I (alpha) error
Reject null hypothesis when the different is not significant
Type II error
Accept null hypothesis when there is a statistically significant difference
Internal validity
How well an experiment is done so that it avoids confounding variables
Degree to which results found can be attributed to the independent variable and not other confounding factors
External validity
How well can the study results be generalized
Common flaws associated with observational studies
Confounding by indication: does drug kill you or is drug prescribed to sicker ppl?
Selection bias: ppl get selected to have surgery because their physician thinks they can handle it
Recall bias
Ascertainment bias
