Evidence Based Medicine Flashcards
Specificity
proportion of pts who have a negative test among the those who truly don’t have the disease (true negative rate)
OR
proportion of people who don’t have the disease who test negative. or probability that a test indicates non disease when disease is absent.
Sensitivity
the people who tested positive who actually have the disease or true positive
probability of a test detects disease when disease is present.
Positive predictive value PPV
proportion of the positive test that are true positives.
Given a positive test, what is the probability that the person has a disease
chances of having the disease given you tested positive.
Negative predictive value (NPV)
proportion of negative test results that are true negatives.
chances that you don’t have disease given you tested negative.
prevalence
number of existing cases of dx (PREVAIL — to hang around)
prevalence = # of existing cases / population at risk
prevalence ≈ incidence time x average dx duration (time)
incidence takes into account time
bathtub example
incidence (water going into the tub)
prevalance ( water in the tub because it prevails)
incidence
number of new cases of dx
incidence rate = # new cases in a specified time / population at risk at that same time.
incidence looks at new cases (incidents or happenings)
example: bathtub example
incidence (into the tub)
prevalence (water in the tub)
what is higher: incidence or prevalence in highly communicable dx with a short period of illness
Influenza - incidence is higher than prevalence
more people are getting influenza and than having influenza so incidence will be higher than prevalence.
bathtub example:
water INto the tub (incidence)
water already present in tube (prevalence)
Prevalence is a measurement of number cases/population at risk or people with dx
incidence is number of new cases in specified time frame / population at risk in that same time frame
what is higher incidence or prevalence in dx that last a long time (ie HIV or DM)
prevalence is higher than incidence
prevalance will be greater than incidence in chronic dx
higher duration of disease length and fewer people getting the dx
bathtub example:
water INTO the tube is incidence
water already present in tube is prevalance
what kind of dx’s will have roughly the same incidence and prevalance?
incidence and prevalence will be about the same in dx that are highly virulent and have short incubation periods and relatively long dx persistence.
95% Confidence interval means
if study were repeated 100 times the result obtained would be in the CI 95% of cases. NOT the same thing as the same result or the same CI would be obtained 95% of the time.
tests with high sensitivity are helpful because they
SnOUT help rule out a disease
so screening tests should have a high sensitivity
tests with high specificity help to
SpIN or better rule in a disease. so confirmatory tests should have a high specificity
Higher specificity means
fewer false positive and increases the PPV for a test
higher sensitivity tests
fewer false negatives and increases NPV of a test
accuracy or validity is
tests ability to measure what is supposed to measure.
false discovery rate (FDR)
expected proportion of false discoveries (type 1 errors) among all statistically significant discoveries (rejected null hypotheses) during simultaneously testing of multiple hypotheses. FDR correct methods are designed to reduce the proportion of incorrectly rejected null hypotheses (false discoveries) while maintaining acceptable statistical power.
NNT
estimate of effectivness of a treatment or intervention
calculation of NNT
1/ARR (absolute risk reduction)
Odds ratio
measure of association between exposure and an outcome. represents the odds that an outcome will occur given a particular exposure as compared to the odds of the outcome occurring in absence of the exposure
case controlled studies use
ods ratios
confounding factors
things that can affect outcome that make it seem like there’s a relation between the outcome and exposure when in fact there isn’t
adjusted odds ratio
strips away the potential confounders and reveals true association between outcome and exposure.
LR +
sensitivity / 1- specificity
probability of pt with disease testing positive depited by probability of a pt without dx testing positive. equivalent of the true positive rate / false positive rate.
0.6 / (1-0.8) = 3
LR -
1-sensitivity / specificity
probability of pt with the disease testing negative divided by the probability of the pt without disease testing negative.
Equivalent to the false negative rate / true negative rate.
LR can range from
zero to infinity
test result of LR >1 means
presence of disease and higher the LR the more likely the disease is present and the lower the LR the less likely disease is present
LR is used
to assess the value of a diagnostic test if that will help us to diagnose a disease present
It is only used for diagnostics, should I get this test? will it provide evidence to rule in or out the disease?
Then need to use LR to trace out a normagram that the patient has this.
It doesn’t speak on treatment efficacy.
These do not change with prevalence of a disease
if you don’t know your pretest probability with the normagram, then go with prevalence of dx.
LR ratio of a test that would provide strong evidence to rule in the disease
>10
remember likelihood ratios are only meant for diagnostic tests and do not speak about treatment.
LR ratio that has moderate evidence to rule in the disease
5-10
LR ratio of a test that would provide strong evidence to rule out the disease
<0.1
Type 1 error is
alpha error- stating there is an effect or a difference when none exists (null hypothesis incorrectly rejected in favor of alternative hypothesis)
FALSE POSITIVE ERROR
alpha - you sAw a difference that did not exist.
alpha is the probability of making a type 1 error. p is judged against a present alpha level of significance (p<0.05)
so that there is less than 5% chance that the data doesn’t show something that isn’t there.
Type 2 error is
beta - stating there is not an effect or difference when one exists (null hypothesis is not rejected when it is in fact false)
beta = you were Blind to the difference that did exist. Set a guilty man free. no association with lung cancer and smoking.
beta is probability of making a type 2 error. related to power (1-beta) which is the probability or rejecting null hypothesis when it is false.
Increase power by decreasing beta: increase sample size, increase expected size effect, increase precision of measurement.
Levels of evidence
what is the best way to make a change in the hospital?
“model for improvement”
identifies a specific goal to be accomplished or a change.
determines how the results of the change will be measured
deciding on the changes that will bring about improvement.
this has a PDSA cycle or plan-do-study-act
lead time bias is a
bias occurs when survival time (time from diagnosis to death) appears to be lengthened because the screened patient is diagnosed earlier during the preclinical phase but does not actually live longer.
to prevent this bias, disease specific mortality rates rather than survival time should be used as an outcome derived from randomized clinical trials.
recall bias is
pts with disease of interest are more likely to recall past exposures compared with controls.
affects observational retrospective studie deisgns.
selection bias
study participates do not accurately reflect the population being studied
usually because of choice to participate is influenced by the clinical question.
length time biasis
screening is more likely to detect indolent dx which has long latent period than aggressive dx which has short latent period and is most often deleted with onset of symptoms
Length time bias- screened detected cohort will have overrepresentation of indolent dx where as a symptom detected cohort will have overrepresentation of aggressive dx.
thus the screened detected cohort falsely appears to have better prognosis.
