Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
107 Cards in this Set
- Front
- Back
4 step method to decide if evidence reported in the literature is applicable to a patient's condition
|
1. ask - the right question
2. access - the right information 3. assess - validity and utility of the information 4. apply - skillfully to patient care |
|
what does the pico model apply to?
|
a format to asking questions very specific to a patients situations.
P = patient specific I - intervention and C = comparison O = outcome |
|
decision support tools including "guidelines" and care support systems
|
systems
|
|
address broader care, integrating multiple sources
|
synopses
|
|
integrat multiple studeies addressing single question.
uses explicit criteria to select studies, quantitatively summarize results |
systematic reviews
|
|
a quantitiive summary of results from multiple carefully selected studies examining the same hypothesis
|
meta-analysis
|
|
a way to take multiple studies the impercisions to make them collectively more precise
|
meta-analysis
|
|
what ways can be applied to Pubmed to look for "pre-reviewed" literature
|
limits
-> setting limits to search clinical queries ->clinical study category = randomized trials ->systemic reviews = meta-analysis |
|
1. not all imp clin question have been addressed
2. results may not be generaliizable to pt like yours 3. comparable txmt may not be acheived in clincial settings 4. rct even if appropraite may not be feasible for reasons of sample size, interest,cost and time required |
real world limits of rct
|
|
what are three blindedness in a rct?
|
1. assigner of txmt - allocation concealment
->biased assignment to txmt group 2. subject ->compliance, placebo, and hawthorne effect 3. investigator -> differencial intensity |
|
when looking at
-blinded delivery -compliance -contamination of controls -conintervention within either group what question is being asked? |
was the txmt fairly applied?
|
|
what 3 concepts are looked at when studying the validity of a study?
|
1. architecture of the study, is the study a randomized study
2. delivery - was the txmt fairly applied 3. outcome - was the outcome fairly assessed |
|
what question is being asked when looking at the following:
-blinded assessment -follow up complete -analysis by intention to tx |
Was the outcome fairly assessed?
|
|
how to solve for relative risk reduction (rrr)?
|
ARD / risk control group = RRR
ARD = difference btwn placebo and exp group |
|
what is absolute realative risk (ARR)?
|
difference between the the placebo and treated group (% difference)
|
|
how is number needed to tx (NNT) calculated?
|
NNT = 100/ARR(as %)
or NNT = 1/ARR (as a proportion) ARR = absolute rick reduction or risk difference ARR = pt w/o txmnt - pt w/txmt each divided by total # of pt in the study. |
|
what does the number need to treat depend on?
|
1. what proportion of the population is at risk
2. how effect the txmt is |
|
falsely rejecting the null hypothesis (inferring txmt has a real effect when it does not)
|
type 1 error (alpha)
|
|
level of tolerance for type 1 error (often 5%)
|
alpha
|
|
the probability that a difference as large (or larager) than the one observed could have occured by chance alone
|
p - value
|
|
when is the null hypothesis rejeted?
|
when the p-value is smaller than alpha
|
|
falsely failing to reject the null hypothesis (inferring there is no txmt effect when there is one) missing a difference
|
type 2 error (beta)
|
|
= 1 - B
depends upon sample size, our tolerance for type 1 error (alpha), and magnitude of the effect size we want to be able to detect |
power
|
|
if infer no difference
|
power
|
|
if infer difference
|
p-value
|
|
txmt will be "acceptablly close" in effectiveness to standard
|
non-inferior txmt
-must offer other benifit (cost, harm, other factor) |
|
comparing the new and standard txmt ( with a difference of 5% being acceptable)
|
non-inferior txmt
|
|
txmt with all results above zero line -> mean difference of 95% CI
|
superior txmt
|
|
txmt with all results below zero line -> mean difference of 95% CI
-and below acceptable margin of inferiority |
inferior txmt
|
|
when a study applies information to a sub-groups after a randomized study has been performed, not randomized study
|
post hoc
|
|
3 key concepts to address when questioning therapy?
|
architecture - randomized trial or not
delivery - was txmt fairly applied outcome - was the outcome fairly assessed |
|
number of people to tx to save one person
|
NNT = 100/ARR
absolute relative risk = placebo group - txmt group |
|
does p-value prove txmt works?
|
p-value has absolutely nothing to do with magnitude of effect, only the probablity of drawing fals inference about its existence
|
|
what is compliance?
|
was txmt received as intended?
|
|
what group is effected by contamination?
|
control group
|
|
what group is evidence of co-intervention a concern?
|
experimental group
|
|
by how many percentage points does probability of improvement inc in exp compared to control
|
absolute benefit improvement
|
|
how many fold higher is probability of improvement in the experimental as compared to control group
|
relative "risk" for improvement
risk (exp) / risk (control) |
|
reduce risk of adverse outcome
|
therapy
|
|
reduce risk of developing disease
|
prevention
|
|
type of prevention?
-prevent asymptomatic illness |
primary prevention
|
|
type of prevention?
-prevent symptomatic illness |
secondary prevention
|
|
type of prevention?
-prevent complication or death |
tertiary prevention
|
|
knowing about a disease longer bc dx earlier due to screening but dont change likelihood of death
|
lead time bias
|
|
2 paralell phenomenon that appear related but that may not be causally linked
|
ecologic fallacy
|
|
what does compliance, contamination and co-intervention assess in a study?
|
was txmt delivered as intended
|
|
what are 3 aspects of validity of a study?
|
1. is this a randomized, controlled study?
2. was txmt fairly applied? 3. was the outcome fairly assessed |
|
what are 3 components of the actually observed (summary of results)?
|
1. how large was teh txmnt effect
2. what is a plausable range of the effective size 3. how likely is it that my conclustion is in error |
|
can change explain results
|
yes
|
|
does a highly statistically significant number indicate to you the magnitude of benifine is high, low or you can't tell
|
can't tell , bc has nothing to do with how good the p-value is
|
|
what are 3 key questions about utility (will the results help me in caring for my pt?
|
1. do the results apply to my pt.
2. is the effect large enough to matter 3. were all clincially imp our come considered 4. are the likely benefits worth the potential harms and costs |
|
appropriate ranges for probabilities and values can be substituted for the baseline estimate and alternatives compared
|
sensitivity analysis
|
|
predicted outcome per alternative (outcome may be assigned a value but costs of acheivement are not included)
|
decision analysis
|
|
cost per health outcome
-ex, cost per year oflife saved |
cost effectiveness
|
|
cost per quality adjusted health outcome
-ex, cost per quality adjused life year |
cost utility
|
|
cost ($) to acheive outcome compared to value to outcome ($)
|
cost benefit
|
|
primary
secondary and tertiary prevention |
focus of prevention policy
|
|
-burden
-tests for early detection -effectiveness of early txmt -"wisdom" which may include cost-effectiveness |
criterias of prevention policy
|
|
what do all 4 of these represent:
1. patient description 2. intervention 3. comparison 4. outcome |
the summary of methods in a study PICO
|
|
not pre-specified hypotheses that can be tested and are subject to multiple sources of bias (sub-groups were not randomly assigned to treatment status)
|
post hoc
|
|
the capacity of , on average, being correct
|
accuracy
|
|
if disease is present, test is correct - positive
|
sensitivity
|
|
if disease is absent, test is correct - negative
|
specificity
|
|
the capacity to consistently provide the same answer
|
reliability
|
|
test-retest results are the same
|
precision
|
|
a/(a + c)
|
sensitivity
|
|
d/(b + d)
|
specificity
|
|
the ability of the test to correctly id those with disease
-erroded by false negatives |
sensitivity
|
|
the ability to correctly id those without (free) of disease
-erroded by false positives |
specificity
|
|
1 - specificity =
|
false positive "rate"
|
|
a very sensitive test when neg rules out disease
why? |
bc no fals negatives - did not miss anyone
-reduce specifity |
|
graph that plots trade off btwn sensitivity and specificity
-a method to quantify the amount of info captured by a test |
ROC - receiver operating characteristic curve
|
|
where would the best point be on the receiver operating characteristic (ROC)curve and why?
|
upper left corner bc both sensitivity and specificity would be 100%
|
|
provides an estimate of teh area under the ROC curve
|
c-statistic or concordance statistic
|
|
implication of a diagnoistic test
|
predictive value
|
|
characteristics of a diagnoistic test
|
specificity and sensitivity
|
|
a/(a + b)
|
postive predictive value = ppv
-proportion of people who test postitive for the diseae |
|
d/(d + c)
|
normal predictive value
-of those who are negative how many are actually free of disease |
|
c/(c + d)
|
missed - possiblity the remaing people who tested negative are actually positive for the disease
|
|
the estimated prevalance of disease in a similar population of individuals
|
best "guess" of pre-test population for a pt
|
|
correct amoung those WITH disease
|
sensitivity
|
|
correct amoung those WITHOUT disease
|
specificity
|
|
describe the performance (accuracy) of a test
|
sensitivity and specificty
|
|
best guess likelihood of disease in our patient before testing
|
pre-test probablilty
|
|
estimate of likelihood of disease following testing
|
post test probability (PPV or NPV)
|
|
depends on both test charactieristics AND the patient tested
|
post test probablility
|
|
-classic epidemiology and clinical research
-test score -substatial data |
harder data
|
|
clinical judgement - particularly applying epidemiolgy to individual pt
-expert opinion |
soft data
|
|
the probability that a given test result came from an individual with rather than without the disease
|
likelihood ratio (LR)
|
|
the test result provides no new info, disease is no more likely or less likely than it appeared before the test
|
Likelihood ratio of 1
|
|
sensitivity / (1-specificity) =
|
likelihood ration (LR)
|
|
what equations applies to the following statement?
What we know about something depends on what we know before and what we've learned from the test |
post-test odds = pre-test odds x LR
|
|
- proportion "correct" for a diagnostic test (true positivies pluse true negatives among all tested)
-calibration of a predictive model (comparing numbers observed verses expected for levels of predicted risk) |
Accuracy (average) can be assessed for diagnostic tests and predictive models
|
|
comparing numbers observed verses expected for levels of predicted risk
|
calibration of a predictive model
|
|
true positivies pluse true negatives among all tested
|
proportion "correct" for a diagnostic test
|
|
the ability to distinguish amoung individuals as affected or not
|
discrimination
-for both dianostic tests and predictive models this is assessed as area under the ROC curve (c-statistic) |
|
with this question what is being measured and what is the focus?
How close (accurate) is the average level of risk predicted for a group to the level actually observed? |
measure: calibration
focus: group |
|
with this question what is being measured and what is the focus?
How often are we correct (accurate) in distinguishing individuals at high risk vs low risk? |
measure: discrimination
focus: individual |
|
with this question what is being measured and what is the overall measure is?
"correctness on average" in the population |
measure: calibraion
measure: observed-expected across sub-groups |
|
with this question what is being measured and what is the overall measure is?
"correctness" in ranking for pairs of individuals (one with and one without outcome) |
discrimination
concordance statistic (c-statistic) and roc curve |
|
if a patient has ascites but no ankle swelling what are the chances he/she has ascities
|
absence of ankle swelling is going to substantially reduce the likelihood of ascities (5 to 10%)
|
|
if the patient does have ascites and ankle swelling does that automatically mean that he/she has ascites?
|
it supports but not confirm the suspicion of ascities
|
|
how likely is a pt have ascities if the pt does have a fluid wave?
|
presence of fluid wave ismost specific and is particularly informative when prsent, thus its present inc the probablility of ascities by about 30 - 40 %
|
|
what lab measure absolutely rules out portal htn
|
low serum - ascities albumin gradient (SAAG) - the odds of portal HTN dec by 15 fold when the SAAG < 1.1
|
|
what virtually confirms the dx of CHF?
|
S3
|
|
What lab test if elevated is associated with CHF?
|
BNP - a polypeptide secreted by cardiac ventricules in response to "stretching" of myocyte
->inc levels associated with CHF |
|
how does presence or absence of BNP effect dx of CHF?
|
if elevated BNP cant be used to confirm dx of CHF bc with a positive LR of 4.1, only modest inc the odds of chf (inc absolute probability by perhaps 25%)
a negative BNP is quite helpful in excluding CHF bc negative LR of < 0.1 the odds are reduced 10 fold and the absolute probability by 50% |