Reading...
Play button
Play button
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;
164 Cards in this Set
- Front
- Back
|
There is a Hep A outbreak and you interview people who did and did not develop the disease. What type of study is this?
|
case-control
|
|
|
You interview prostitutes in Las Vegas as an initial step in your investigation to an STD outbreak. What type of study is this?
|
cross-sectional
|
|
|
What is the type of study that will gain the highest level of evidence?
|
randomized control trial
|
|
|
What is the main weakness of case-control studies?
|
recall bias
|
|
|
In Atlanta, you compare 17 days of the Olympics to 4 weeks before and 4 weeks after for asthma events. What type of study is this?
|
-Ecological study (comparing characteristic and some outcome with geographic association; treats Atlanta at these times as different geographic areas)
|
|
|
Give patients estrogen/progesterone or placebo to see if there is an altered risk of CHD events. What type of study is this?
|
controlled trial
|
|
|
You look at 249 vertebroplasty interventions and analyze demographics, treatment, pain alleviation, QoL improvement. What type of study is this?
|
case series
|
|
|
Define probability
|
-a number expressing the likelihood that a specific event will occur
-expressed as the ratio of ACTUAL occurances to the number of POSSIBLE occurances -risk=incidence=probability |
|
|
Define risk
|
-usually expressed as the # of new cases per population per time period
|
|
|
Clinical trial (experimental study)
|
-prospective, interventional study designed to answer a focused question, usually comparing 2 or more treatment alternatives (experimental and control groups)
|
|
|
Observational studies
|
-include case series, case-control, cohort, cross-sectional, ecological studies
-useful when it's impractical or unethical to conduct a clinical trial -noninterventional |
|
|
Case series
|
-group or series of cases involving patients given a similar treatment
-detailed info about patients -typically no control group |
|
|
Case-control study
|
-start with an outcome and work back
-looking for possible exposure/risk factors -outcome is measured as an ODDS ratio -can't calculate incidence |
|
|
What can and can't you measure with a case-control study
|
-measure outcome as an odds ratio
-can't measure incidence (can't do risk) |
|
|
Cohort study
|
-start with exposure to a risk factor and follow them to see if an outcome develops
-outcome measured as incidence |
|
|
cross-sectional study
|
-measure both exposure and outcome simultaneously at one point in time
-measures PREVALENCE, not incidence |
|
|
ecological studies
|
-relate the freq with which some characteristics and some outcome occur in the same geographic area
|
|
|
ecologic fallacy
|
-assuming because the characteristic and the outcome are geographically associated, one causes the other
|
|
|
bias
|
any systematic error in the design, conduct or analysis of a study that results in a mistaken estimate of exposure's effect on the risk of disease
|
|
|
selection bias
|
-one group is over/underrepresented in a study
-also called "allocation bias" |
|
|
healthy user bias
|
-pts who adhere to preventative therapies may be more likely to engage in a broad spectrum of behaviors consistent with a healthy lifestyle
|
|
|
late-look bias
|
-cross-sect surveys tend to find longer-lasting and more indolent diseases
-diseases that resolve quickly or that are rapidly fatal are less likely to be found |
|
|
recall bias
|
-case-control studies, those who have an effect are more likely to remember details than those that do not have an effect
|
|
|
misclassification
|
-assigning participants to the wrong group
-if non-random, it may be bias |
|
|
measurement bias
|
-results not measured the same way in exposed/unexposed groups
|
|
|
confounders
|
-stealthy variables that are associated with the input variable and have an effect on the output variable
-randomization decreases their effect but does not eliminate them completely |
|
|
blinding (masking)
|
-avoiding bias by keeping the group assignment secret from either the investigators or the subjects (single blind) or both (double blind)
|
|
|
evidence-based medicine
|
the integration of individual clinical expertise with the best available clinical evidence from systematic research and the specific circumstances of the patient
|
|
|
foreground question
|
a question that asks for specialized and distinct knowledge needed for specific and relevant clinical decision-making
-what you would find in an original research study |
|
|
background question
|
-a general question about conditions, illnesses, syndromes and anything based on background knowledge
-what you find in a textbook or review |
|
|
patient oriented outcomes (POEMs)
|
how the results of a study relate to your patients (do they live longer, happier lives now?)
|
|
|
disease oriented outcomes (DOEs)
|
outcomes that may change or control an aspect of a disease but may or may not improve morbidity or mortality
|
|
|
criteria for causation
|
-temporal relationship
-strength of association -dose-response relationship -replication of findings -biological plausibility -consideration of alternate explanations -consistency with other knowledge -specificity of association |
|
|
What kind of error is failure to reject the null hypothesis when it is false?
|
type II error
|
|
|
If you want to know if a treatment is both better than placebo or worse, use a....
|
2-tailed t-test
|
|
|
Comparing USMLE scores between group that took vit E and placebo group, use a....
|
t-test
|
|
|
What kind of variables are income, education level, current marital status?
|
education=continuous
education level=ordinal variable current marital status=dichotomous |
|
|
What does a BMI percentil skewed to the right mean?
|
-explained by selection bias
and -BMI comparison scale is not representative of the population |
|
|
A hypothesis should be....
|
-simple, specific, testable, and stated in advance
|
|
|
null hypothesis
|
-hypothesis which states that there is no difference or existence of some aspect
|
|
|
what is the p-value?
|
-probability of a given experimental observation (or more extreme) under a null hypoth
-if P value is small, we reject the null as unlikely -so, it is the probability of rejecting the null hypoth when it is true |
|
|
One vs. two-sided hypotheses
|
-one-sided: need smaller sample size to prove result
-two-sided: allows for results to go either way and investigator can still reject null |
|
|
does a p-value give us a probability about the truth of the null hypothesis?
|
NO
it gives us the probability of rejecting the null hypothesis when it is true |
|
|
Type I (alpha) error
|
-when you reject the null hypothesis, but the null hypoth is true
-usually use .05 |
|
|
Type II (beta) error
|
-failing to reject the null hypoth when it is in fact false
-don't detect a difference that actually exists -1-beta=power of the study -usually use its upper limit: 0.2 |
|
|
what does a 95% confidence interval tell us?
|
that 95% of the intervals constructed in this manner will cover the true parameter value
4 SDs in width |
|
|
If a 95% CI for a difference btw two means spans zero, then...
|
-we may accept null hypoth (that the difference is 0)
|
|
|
If a 99% CI for a difference btw two means does not span zero, then...
|
-we may reject the null hypoth
|
|
|
interval variable
|
-size diff btw two values has consistent meaning
-ie degrees when telling temp |
|
|
ordinal variable
|
-values are ordered but difference may not be consistent or quantitative
-ie H/HP/P/LP/F |
|
|
what is the p-value?
|
-probability of a given experimental observation (or more extreme) under a null hypoth
-if P value is small, we reject the null as unlikely -so, it is the probability of rejecting the null hypoth when it is true |
|
|
One vs. two-sided hypotheses
|
-one-sided: need smaller sample size to prove result
-two-sided: allows for results to go either way and investigator can still reject null |
|
|
does a p-value give us a probability about the truth of the null hypothesis?
|
NO
it gives us the probability of rejecting the null hypothesis when it is true |
|
|
Type I (alpha) error
|
-when you reject the null hypothesis, but the null hypoth is true
-usually use .05 |
|
|
Type II (beta) error
|
-failing to reject the null hypoth when it is in fact false
-don't detect a difference that actually exists -1-beta=power of the study -usually use its upper limit: 0.2 |
|
|
what does a 95% confidence interval tell us?
|
that 95% of the intervals constructed in this manner will cover the true parameter value
4 SDs in width |
|
|
If a 95% CI for a difference btw two means spans zero, then...
|
-we may accept null hypoth (that the difference is 0)
|
|
|
If a 99% CI for a difference btw two means does not span zero, then...
|
-we may reject the null hypoth
|
|
|
interval variable
|
-size diff btw two values has consistent meaning
-ie degrees when telling temp |
|
|
ordinal variable
|
-values are ordered but difference may not be consistent or quantitative
-ie H/HP/P/LP/F |
|
|
nominal variables
|
-individuals grouped but not ordered
-ie eye color, sex, race, etc |
|
|
effect size
|
-how much of a difference you expect (in planning a study) or find (in reviewing a result) btw study groups
|
|
|
sample
|
-subset of the population
-measurements taken from samples are estimates of population values |
|
|
population
|
everyone in whom you could be interested
|
|
|
alpha
|
-threshold P value below which a difference is statistically significant
-usually .05 |
|
|
beta
|
-probability of making a type II error
-usually set at 0.2 -important in determining sample size |
|
|
what is the defining characteristic of an "intention to treat" analysis?
|
subjects are analyzed according to their original group
|
|
|
what are the most important factors in calculating sample size?
|
-expected effect size
-variance in population -alpha and beta levels |
|
|
when is it necessary to use survival analysis?
|
when subjects are enrolled for varying periods of time
|
|
|
what is one cause of type II (beta) error?
|
inadequate sample size
|
|
|
what is the intention of randomization
|
the study groups will be free of selection or allocation bias
|
|
|
incidence
|
-the number of new cases of a disease that occur during a period of time in a population at risk for developing the disease
-#new cases/ total pop at risk per time period |
|
|
prevalence
|
-the number of existing cases of a disease at a point in time
-#existing cases/ total pop at risk at that point |
|
|
probability
|
-a number expressing the likelihood that a specific event will occur
-expressed as the ratio of the number of actual occurrences to the number of possible occurrences -risk=incidence=probability |
|
|
risk
|
-usually expressed as the number of NEW cases per pop per time period
|
|
|
relative risk
|
-the ratio of risk in exposed persons to risk in unexposed persons
|
|
|
relative risk reduction
|
(risk exposed - risk unexposed)/ (risk unexposed)
|
|
|
absolute risk difference (attributable risk)
|
(risk exposed) - (risk unexposed)
|
|
|
number needed to treat
|
-the number of people needed to treat to produce 1 outcome
-1/ARR (ARR=risk exposed-risk unexposed) |
|
|
odds
|
-the probability of an event divided by the probability of a nonevent
-p/(1-p) |
|
|
odds ratio
|
-ratio of two odds
-good est of relative risk when outcome is relatively rare -correct measure of association in a case-control study (true pop denominator is unknown) |
|
|
what analysis gives you an odds ratio?
|
a logistic regression
|
|
|
clinical trial (experimental study)
|
-prospective, interventional study designed to answer a focused question
-usually compares 2 or more treatment alternatives |
|
|
survival analysis
|
-based on fact that probability of multiple indep events is the product of the prob of each event
-losses to follow-up are censored -useful to compare survival and any time-to-event data or when subjects are followed for different amounts of time |
|
|
hazard ratio
|
-uses Cox regression
-HR=1 means no effect on survival -HR<1 predictor is associated with increased survival -HR>1 predictor is associated with decreased survival -the greater the number, the greater the hazard |
|
|
intention-to-treat analysis
|
-when analyzing outcomes, keep participants in the groups to which they were originally assigned, even if they don't continue a treatment
|
|
|
logistic regression
|
-can be used to analyze multiple input variables that may be associated with the outcome in question
-adjusts for confounders -used when output variable is dichotomous |
|
|
effect size
|
-how much of a difference you expect (in planning the study) or find (in reviewing the results) btw study groups
|
|
|
which values are estimates of relative risk?
|
-risk ratio (cohort, some RCTs)
-odds ratio (case-control) -hazard ratio (survival analysis) |
|
|
Randomization is:
|
-unbiased allocation of study subjects to experimental and control groups
-a process that takes place AFTER subjects are included in the study |
|
|
calculating relative risk
|
(a/a+b)/(c/c+d)
risk (exposed) / risk (unexposed) |
|
|
odds ratio
|
(a/c)/(b/d)
(odds of exposure in diseased) / (odds of exposure in undiseased) |
|
|
what is a major problem in cohort studies
|
loss to follow up
|
|
|
what kind of study can and what cannot measure incidence?
|
-cohort CAN
-case-control and cross-sectional CANNOT |
|
|
what kind of study can measure population risk?
|
cohort (case-control canNOT)
|
|
|
what kind of study can measure prevalence?
|
cohort (starts with a baseline prevalance)
|
|
|
In case-control studies, subjects are classified first by:
|
outcome of interest
|
|
|
In a case-control study, what is the appropriate outcome measure?
|
odds ratio
|
|
|
What are the advantages of cohort studies?
|
-can study multiple risk factors simultaneously
-can study multiple outcomes simultaneously -are not affected by recall bias (but are affected by selection bias) |
|
|
what can cohort studies measure
|
-prevalence
-incidence -risk ratio |
|
|
what can case-control studies measure
|
-odds ratio
|
|
|
what can cross-sectional studies measure
|
-prevalence
|
|
|
Kaplan-Meier method
|
-method for survival analysis
-new interval begins whenever there is a death (or other outcome) -find probability of surviving each interval |
|
|
regression
|
-methods used to analyze the contributions of several variables ot outcome
-used to adjust for the effects of confounding variables |
|
|
logistic regression
|
-adjusts for multiple confounders
-used for dichotomous variables -outcome=odds ratio |
|
|
cox proportional hazards regression
|
-calculates effects of several variables on a survival curve (using time-event data)
-produces a hazard ratio |
|
|
what does relative risk mean
|
-risk ratio, odds ratio and hazard ratios are all measures of it
-when it =1, risk in exposed and control group is the same -when the CI around a ratio includes 1, no signif difference in risk btw the exposed and control group |
|
|
criteria for causation
|
-temporal relationship
-strength of association -dose-response relationship -replication of findings -biologic plausibility -consideration of alternate explanations -consistency with other knowledge -specificity of association |
|
|
confidence intervals
|
-calculated around the observed measurement (which may be a risk or odds ratio)
-related to the p-value but give more info -provides an estimate of the precision or lack that can be attributed to sampling variation |
|
|
The closer the RR is to 1.0....
|
-the less clinically significant it is
-even if it is statistically significant |
|
|
when is odds ratio a good measure of relative risk?
|
-when the cases are representative, with regard to the hx of exposure of people with the disease
-when the controls are representative, with regard to the history of exposure without the disease -when outcome is not so common that odds ratio is very diff from risk ratio |
|
|
the likelihood ratio positive (LR+) is used
|
-multiply the pre-test odds to get post-test odds
|
|
|
in a diagnostic testing study, what is the "gold standard?"
|
an established test used to determine the prevalence of a disease
|
|
|
what is the interpretation of a LR- of 0.16?
|
-an individual with a negative test has 0.16 times the odds of having the disease
|
|
|
sensitivity vs specificity to rule in/out
|
"spin-snout"
SPecificity rules IN SeNsitivity rules OUT |
|
|
pre-test probability
|
-probability estimate of disease before performing a diagnostic test
-may be based on community prevalence or observational studies in clinic |
|
|
sensitivity
|
-how often a test is positive when disease is present
-(true positives)/(true positives + false negatives) |
|
|
specificity
|
how often a test is negative when the disease is absent
|
|
|
positive predictive value
|
-the probability that someone with a + test has the disease
=TP/(TP+FN) |
|
|
negative predictive value
|
-the probability that someone with a - test does not have the disease
=TN/(TN+FN) |
|
|
how are likelihood ratios useful in practice?
|
-when the result is positive, a test with a large LR+ can increase the probability of a diagnosis from intermed to high
|
|
|
likelihood ratio of a positive test (LR+)
|
-how many times as likely a + test result is found in diseased, as compared to healthy individuals
=sensitivity/(1-specificity) |
|
|
likelihood ratio of a negative test (LR-)
|
-how many times as likely a negative test result is found in diseased as compared to healthy individuals
=(1-sensitivity)/specificity |
|
|
workup bias
|
-aka verification bias
-those who have negative tests don't receive the gold standard test bc it's painful, risky, expensive, etc |
|
|
spectrum bias
|
-test may be more accurate when given to patients with more severe or well-developed disease
-less likely to identify patients with earlier or occult forms |
|
|
ROC curve
|
-shows tradeoff btw sensitivity and specificity at diff cutoff points
-close the curve is to upper left corner, the more accurate the test |
|
|
probability of disease given a positive test (Bayes' Theorem)
|
(sensitivity*prevalence)/(s*p + (1-s)(1-P))
|
|
|
with higher prevalence of disease, PPV ___ and NPV ____
|
PPV increases
NPV decreases |
|
|
with lower prevalence of disease, PPV ___ and NPV ___
|
PPV decreases
NPV increases |
|
|
converting btw probability and odds
|
o=p(1-p)
p=o/(1+o) |
|
|
What measurements are inherent to the test and don't change with the population?
|
-sensitivity
-specificity -LR (theoretically) |
|
|
what measurements are dependent upon population prevalence of the disease?
|
PPV and NPV
|
|
|
calculating LR+
|
(a/a+c) / (1- d/b+d)
|
|
|
calculating LR-
|
(1-a/a+c)/(d/b+d)
|
|
|
calculating PPV
|
a/a+b
|
|
|
calculating NPV
|
c/c+d
|
|
|
calculating sensitivity
|
a/a+c
|
|
|
calculating specificity
|
d/b+d
|
|
|
screening is what kind of prevention
|
secondary prevention
|
|
|
criteria for screening strategy
|
intervention in pre-symptomatic (screening) phase proven to improve outcome
|
|
|
proof that screening improves outcomes is hard to establish because...
|
-long lead times
-large sample needed -"contamination" |
|
|
systematic review
|
-comprehensive search for available research
-quality assessment -synthesis of results |
|
|
what are the components of quality assessments as part of systematic review?
|
-comparable groups
-follow up -use of equal, valid, reliable outcome measures -clearly defined, comparable interventions |
|
|
what are the components of the synthesis of results in systematic reviews?
|
-meta-analysis: data pooled, results mathematically combined
-or analyzed without pooling data |
|
|
number needed to screen (NNS)
|
-Mortality: 1/ (risk of death in control - risk in screened group)
-Case finding: 1/ (incidence in control - incidence in screened) the number needed to screen in order to detect one case or prevent one negative outcome |
|
|
intention-to-screen analysis
|
-when analyzing data, keep participants in the groups to which they were originally assigned even if they change groups
-effectiveness measure more realistic to practice |
|
|
length bias (length-time bias)
|
-applies to screening programs
-screening tends to find less aggressive disease because people with less aggressive disease live longer and over time are more available for screening |
|
|
what does length bias tend to exaggerate?
|
the effectiveness of screening asymptomatic symptoms
|
|
|
lead-time bias
|
-applies to screening programs
-screening may lead to earlier diagnosis without altering course of disease -survival from time of diagnosis will be longer, but survival may not be |
|
|
what does lead-time bias tend to overestimate?
|
-the benefit of the screening
(survival from time of diagnosis will be longer, but survival may not be) |
|
|
publication bias
|
-tendency to preferentially publish studies that show an effect and not to publish those that "fail to reject the null"
|
|
|
how do you check for publication bias
|
-statistical tools (eg Funnel plot)
-used is all good systematic review |
|
|
primary prevention
|
-intervention which prevents a disease altogether
|
|
|
secondary prevention
|
-early detection of asymptomatic disease when early intervention improves outcome
-ie screening |
|
|
tertiary prevention
|
-intervention which slows progression, prevents recurrence or complications, or limits damage from asymptomatic disease
-ie treating acute coronary syndrome with stents |
|
|
meta-analysis
|
-a type of systematic review
-pools data form several studies to yield a larger sample size and more precise estimate of effect -considered the highest level of evidence |
|
|
good screening tests
|
-sensitive and specific
-acceptable to the target group -treat for fairly prevalent disease with significant mortality/morbidity -test in presymptomatic or early stage -intervention will improve outcomes |
|
|
prevalence increases with:
|
-increased incidence
-decreased recovery -decreased death rate |
|
|
spectrum bias
|
-study population has higher disease prevalence (or greater severity) than typical populations
|
|
|
a pap smear is an example of what kind of prevention?
|
secondary
|
|
|
the more screening tests performed on an asymptomatic individual...
|
the greater the chance of at least 1 false positive
|
|
|
An increase in the length of survival b/c of screening is exemplified by what kind of bias?
|
lead-time bias
|
|
|
A successful HTN campaign was run one year ago, but the next year's campaign find very few new cases. What is the explanation?
|
only incident cases are detected
|
|
|
What info do you want to look at to see if there is lead-time bias?
|
-mortality form time of enrollment (not from time of diagnosis)
|
