Study sets, textbooks, questions
Upgrade to remove ads
Medical Evidence Exam 1
Terms in this set (89)
What are the basic questions that we have to answer for patients?
1. Therapy: What is the correct intervention for my problem?
2. Prognosis: How am I going to do over time?
3. Harm: Will this exposure hurt or help me?
4. Diagnosis: How will you diagnose the cause of my problem?
What are the three ways we can answer patient's questions?
1. mechanistic explanation
2. Experience (personal or handed-down)
What is evidence?
The result of studies that compare populations
What questions can randomized control trials answer?
1. For a given patient population, what outcome will a certain intervention yield?
2. For a given patient population, what outcome will a certain standard of care have?
What kind of questions do cohort studies answer?
What outcomes are associated with specific habits, behaviors, exposures, or risk factors?
What questions do case control studies answer?
What types of questions do diagnostic test studies answer?
How good is this diagnostic test at diagnosing disease of interest?
Assess burdens of disease (e.g. there were 60 cases of West Nile Encephalitis at UCMC last summer)
-numerator included in denominator
-reported as value or percentage
-time is denominator
-change in one quantity per unit change in another
Numerator not included in denominator
the proportion of persons in a fixed population who develop the disease
-a proportion or percentage
-denominator is persons at risk
number of new cases for a specified population in a given time period
-number of persons with a condition present at a specific time divided by number of persons in population at that time
-snapshot of disease burden in population at a given time
cumulative incidence in an outbreak
# live births/# females aged 15-44 years at midyear
crude (all-cause) death rate (mortality)
# deaths/midyear population
cause-specific death rate (mortality)
# deaths from specific cause/midyear population
case fatality rate
# persons with disease A who die/# persons with disease A
-measure of severity
(rate of disease in unvaccinated - rate of disease in vaccinated)/(rate in unvaccinated)
a quantitative variable that has an infinite number of possible values
-age, BP, weight
categorical variable (nominal)
a variable that names categories (whether with words or numerals)
-gender, marital status, race
What is an ordered categorical variable (ordinal)? What are some examples?
a categorical variable that has organizes data into hierarchical categories (e.g. education level, self-rated health)
What is an unordered categorical variable? What are some examples?
a categorical variable that organizes data into categories that are not hierarchical
-e.g. gender, HIV positive, relationship status, race, zip code
What is a randomized control trial?
A study in which people are randomized to receive one of several clinical interventions (i.e. new drug, standard treatment, placebo, no intervention)
What is a Phase 0 Clinical Trial?
first human trial
-single subtherapeutic doses of study drug are given to a small number of subjects (~10) to gather data on tolerability and pharmacology
What is a phase 1 clinical trial?
treatment tested in a small # of healthy volunteers (20-80) to evaluate safety and at subtherapeutic (often ascending) doses?
What is a phase 2 clinical trial?
Treatment given to a larger group with disease of interest to evaluate efficacy at therapeutic doses (100-300 patients)
What is a Phase 3 clinical trial?
RCT of 1000-2000 patients with disease
What is a Phase 4 clinical trial?
Post-marketing studies to determine risks and benefits in larger, more diverse populations and treatment settings and discover rare or long-term side effects
What are the 5 design elements of RCTs?
1. Sample of patients is selected (affects generalizability/external validity)
2. Patients in sample are randomly assigned to one of two groups (reduces selection bias)
3. Intervention group receives therapy and control group receives standard therapy or placebo
4. Two groups are observed for differences in outcome
5. Every aspect of "randomized, placebo-controlled, double-blind trial"
What is equipoise?
general uncertainty within the expert medical community about the preferred treatment
-must apply to all phases of RCT
What effects does equipoise have on likely RCT results?
What 2 biases does blinding prevent?
-reporting bias (patients are more/less likely to report events of adverse events)
-measurement bias (outcomes may be sought or measured differently)
What is a triple blind study?
Patient, physician, and statistician are blinded
What is a clinical endpoint? Give examples.
-things that the patient would notice (subjective or objective)
-e.g. death, MI, stroke, joint pain, severity of depression, general feeling of wellbeing
What is a surrogate endpoint? Give examples.
-stand-in for clinical endpoints
-invisible to patient
-e.g. blood pressure, average glucose level, bone density, carotid intimal thickness, pulmonary function
What are the risks of surrogate endpoints?
-might not be good representations of real clinical outcome (might not be the sole contributor to a clinical outcome)
-easy to select wrong surrogate endpoint if the hypothesis itself was inaccurate
What is intention to treat analysis?
Subjects are analyzed according to the group to which they were assigned, not according to the treatment they ultimately receive
What are the pros of intention to treat analysis?
-most closely mimics the real world
What are the cons of intention to treat analysis?
-biases towards null
-not appropriate for non-inferiority trials or analyzing adverse outcomes
What is relative risk?
-exposure to risk factors increases the risk of getting disease
-individuals with a history of [exposure] are [RR] times more likely to have a [outcome] than those without that history
What is attributable risk?
For every [incidence,exposed] with a history of [exposure], [AR] could be attributed to their history of [exposure]
What are the ranges for relative risk?
0 to <1: decreased risk
1: no effect
>1 to infinity: increased risk
What is a type 1 error (alpha)?
What is a type 2 error (beta)?
error that occurs when one fails to reject a null hypothesis that is actually false
probability of finding a statistically significant result when one really exists (1-beta)
If we repeated study an infinite number of times, calculating a CI each time, 95% of CIs would contain the true value
What 2 things do confidence intervals tell you?
-significance (<1 = significant)
-precision; small range is more precise
What are the 4 guidelines for appraising studies?
1. surrogate endpoints (a surrogate endpoint is a measure of effect of a specific treatment that may correlate with a real clinical endpoint but does not necessarily have a guaranteed relationship - i.e. using lipid/BP measures to assess cardiovascular health)
2. What is the pre-study probability that the treatment is effective?
3. inadequate controls for subjective endpoints
4. publication bias and p-hacking (positive studies are more likely to be published than negative ones)
What are the 4 observational study types?
3. case control
What is a cross-sectional study?
select individuals and determine outcome and exposure status simultaneously
What is an ecological study?
Select groups and determine outcome and exposure status (i.e. clinic, state, country)
What is a case control study?
Select individuals with disease and without disease and determine exposure history
What is a cohort study?
Select individuals with exposure and without exposure history and follow overtime to observe outcome
What are the pros of cohort studies?
-hypothetical harmful exposures may be studied (in contrast to RCTs)
-relatively inexpensive compared to RCTs
-can determine incidence and prognosis
-hypothesis generation (associations of benefit)
What are the cons of cohort studies?
-require long periods of follow-up
-inappropriate for rare outcomes
-high potential for confounding (use propensity matching to overcome selection bias)
what is a confounding variable?
A variable that distorts the measured association between an exposure and outcome
What are the 3 properties of confounding variables?
1. Associated with exposure
2. Associated with outcome
3. Not part of causal pathway
What is a positive confounder?
Observed association is biased away form null; once adjusted for confounder the RR is farther from 1
What is a negative confounder?
Observed association is biased toward the null; once adjusted for confounder the RR is closer to 1
What is a qualitative confounder?
RRcrude and RRadjusted are on opposite sides of 1
What is effect modulation? What are some examples?
-magnitude of association between an exposure and outcome differs depending upon another variable
-due to real biological process
-e.g. age, genetics
What is survival analysis?
What is quantification of risk and when should it be used?
-using odds ratio in case control studies to quantify risk
-Cases were [OR] as likely to have history of [exposure] than controls
Factors supporting causation
-strength of association
-consistency across studies
-temporality (cause precedes effect)
As prevalence increases, what happens to the predictive value?
-the predictive value of a positive result goes down
-the predictive value of a negative result goes up
What is a likelihood ratio?
-How much should I shift my suspicion that a patient has a disease based on a given test result?
-most important thing to look for in a diagnostic test
What is a positive likelihood ratio?
How much more likely is a person with the disease to have a positive result than a person without the disease?
What is a negative likelihood ratio?
identifies how much the odds of the disease decrease when a test is negative
What is the test threshold?
The disease probability above which you would employ a diagnostic test to rule in or out disease and below which you would feel comfortable concluding that the disease is not present
What is the treatment threshold?
The disease probability above which you would feel comfortable treating without further diagnostic information
What is spectrum bias?
Test characteristics are influenced by spectrum disease (higher severity increases sensitivity; other disease decreases specificity)
How can you avoid spectrum bias?
A study evaluating a diagnostic test should include a broad range of diseased and non-diseased patients
What is verification bias?
Outcome of the test in question influences application of the gold standard
How do you prevent verification bias?
-results of the test and the status of disease must be determined independently
-results of one test must in no way effect the results of the other
What 4 things make a good screening test?
1. sufficient burden of disease (common, significant morbidity/mortality)
2. detects disease in preclinical stage (no symptoms/signs)
3. high sensitivity and specificity
4. improves survival
What is lead time bias?
Screening study bias in which screening detects disease earlier but survival not really improved; patients appear to live longer just because they are detected at an earlier stage
What is length time bias?
Screening study bias in which aggressive disease does not remain asymptomiatic for long enough to be detected by screening, so screening mainly detects indolent disease
What is overdiagnosis?
Screening study bias in which screening detects disease that is unlikely to actually harm or kill people
What are 4 clinical decision rules?
1. Quantitatively integrate elements of history, physical exam, and labs towards diagnosis or prognosis
2. Derivation: nominate potential predictors, perform logistic regression and convert to point system
3. Should be cross validated on set of patients not used for derivation
4. Clinical impact: Does it change physician behavior? Reduce costs? Improve outcomes?
What does qualitative research do?
-collects data and extrapolates theory to generate hyptohesis
What is saturation?
Sample size at which you stop getting new themes
What is positive predictive value?
Probability that a patient with a positive result truly has a disease (Is a positive result a true positive?)
What is negative predictive value?
Probability that a patient with a negative test truly does not have the disease? (is a negative result a true negative?)
What is sensitivity?
Probability that a patient with the disease has a positive result. Used to rule out
What is specificity?
Probability that a patient without a disease has a negative result. Used to rule in
Sets with similar terms
Clin Med - Exam 1 Study Guide Part IV - Epidemiolo…
Other sets by this creator
Cell Path Midterm Disorders
Genetics Exam 2
ACS Exam Prep