Final Study Guide Flashcards
The 3 pronged approach to EBM
1) Clinical Problem Solving
2) Medical Informatics
3) Critical Appraisal
Sources of info for patient care
Medical informatics
-PubMed, Cochrine Library, TRIP, DynaMed
Examples of POE
Patient Oriented Evidence (POE)
- mortality
- morbidity
- quality of life
EBM prescription includes
Formulate + ASK a question ACCESS the evidence Critically APPRAISE the evidence APPLY the evidence ASSESS the use of information in practice
Highest quality literature
Meta-Analysis/ statistical systemic review
Uses for Meta-Analysis
Done to reconcile studies w/ different results
Looks at multiple negative studies to uncover Type II errors
Looks at clinical problems where there are some negative & positive studies to uncover Type I or II errors
Helps uncover a single study w/ totally different results d/t systematic error or research bias
May narrow a large CI in some studies by combining them
What is homogeneity in a meta analysis?
the results from a set of independently performed studies on a particular question are similar enough to make statistical pooling valid
Advantages to EBM
improves confidence through decision making, assists in communication with pt/provider, decreases time going through literature, fosters reading habits, lessons lag time for applying findings, dovetails with technology
Disadvantages to EBM
requires commitment & time, not everyone is skilled at database research, not everyone can afford resources, not everyone is equally skilled in appraising the literature, better known & chosen reliable filters, good evidence is not always there, risks “misinterpretation”
PICO question
1) Patient
2) Intervention
3) Comparison intervention
4) Outcome of interest
5) (T)ime
What does EBM serve to accomplish in genearl
1)Standardize practice while maintaining pt.-centered care
2)Promote life-long learning
3)Response to practice variability
4)Provide granularity on complex questions & gray areas
•Four elements: best evidence, clinical situation, clinical experience, pt. values
What is causation
one variable DOES cause the other
Hill’s Guidelines of Causation includes
i. Strength of association
ii. Consistency
iii. Specificity – cause should lead to a single effect
iv. Temporality
v. Biological gradient
vi. Biological plausibility
vii. Coherence with other data
viii. Analogy – similar relationships with other processes
What model of clinical reasoning is done by beginner clinicians
Exhaustive model
-comprehensive H&P, all things possible
good for zebras
time consuming
What is the best style of clinical reasoning?
Hypotheticoeductive
What model of clinical reasoning involves branching decision tree pathways?
Algorithmic
What model of clinical reasoning involves genitive shortcuts in prioritizing dx through patter recognition and informal methods?
Heuristic Model
What is anchoring bias?
getting stuck on one possibility
What is the best way to minimize selection bias?
random sampling
Clinical decision making tree
Info > Reasoning > Judgment > Communication > Shared Decisoin
Internal influences include
your own risk-taking nature, what questions you ask, jumping to conclusions, assumptions of objective findings
External influences include
Anatomical differences, different therapeutic responses, pt. biases (language, education, pain, etc.), personality traits, barriers (language, education, pain, coping mechanisms), co-worker biases
What EBM recommendations require strict adherence to correct etiquette?
Protocols
What EBM recommendations provide official recommendations of how something should be done?
Guidelines
What study is done for a descriptive study when there is little known about a disease
cross-sectional
What study is for prevalence studies?
cross-sectional
Pros of case control studies
easy, cheap, and quick
cons of case control studies
open to interpreter and referral bias
what is a prospective study done from a certain period in the past to the present
Cohort study
What study is best for answering etiology or harm and progress questions?
Cohort study
What study is good for rare exposures?
cohort study
what is the gold standard for studies?
RCT
What study is good for prognosis questions?
Cohort studies
Sections of a research paper
Introduction Review of related medical literature (inferences) Methodology Results Summary/decisions
What defines qualitative study designs?
analyzing non-numerical data to answer questions
-Hypothesis generating
good precursor to quantitative
examples: narrative, research questions, interviews, focus groups
What defines a quantitative study design?
analyzing numerical data to answer questions
- hypothesis answering
- very structured
What is nominal data?
Represents categories with no order
What is ordinal data?
Categories that can be ordered
What is interval data
Ordered set of values with equal distances between but there is no absolute zero
What data can you perform descriptive stats on (mean, median, mode, etc)
Interval and ratio
What is ratio data
Ordered values with an absolute zero
What is validity?
accuracy of a study
What is internal validity?
The degree that the test measures the content intending to be measured and how well it follows gold standard of criterion-related data.
What is external validity?
extent the study can be generalized to the population. Can it be generalized?
Trying to avoid the Hawthorne effect by being blinded
What does stability mean in terms of reliability of data
scores are consistent over time
test-retest
Mean
average
median
middle value in set of points
mode
most common value
range
highest to lowest, measure of dispersion
parameter
numerical value that describes a population (N)
variance
measure of variation
standard deviation
square root of variance
What is the most stable/best measure of variability?
Standard deviation
Central tendency measurements include
mean, median, mode
dispersion measurements include
range, standard deviation, variance
relative position measurements include
percentile ranks, standardized scores (t & z scores)
What was enacted in 1947 that set the stage for future laws regarding research on humans?
Nuremberg Code
What did the Tuskegee Syphilis Study lead to?
National Research Act of 1974
Define a confounding variable
third variable that is actually responsible for the outcome but wasn’t intended in the study. It influences both independent and dependent variables
Control for confounding variables by
random selection
subject matching
increasing the power
What is informed consent?
Written permission by the subject verifying that he/she agrees to participate with full understanding of what said participation would entail
OR of 1=
>1=
<1=
1= no impact >1= significant impact <1= possible protective effect
What is relative risk?
probability of outcome in exposure group divided by probability of outcome in non-exposure group
Ho=
Null Hypothesis
researcher hopes to reject this
Ha=
Alternative hypothesis, any hypothesis that does not conform to the one being tested
If p-value is greater than alpha you
fail to reject Ho, therefore results fall within two standard deviations of mean
If p-value is< alpha (usually 0.05) you
reject the null hypothesis and accept alternative
Who determines the critical value? What is it generally?
The researcher
Generally 0.05
sometimes less
Type 1 error is
falsely rejecting a correct null hypothesis
What kind of error fails to reject a false null hypothesis?
Type 2 error/ beta/ acceptance error
How can you decrease type 1 error?
lower the critical value
How can you decrease type 2 error?
increase sample size
Continuous data is not significant if _____ is in the confidence interval
0
Ratio data is not significant if _____ is in the CI
1
A CI of 95% (α = 0.05) means if we
repeat the trial 100 times, 95/100 times the actual value will lie within the range (CI)
Does correlation determine cause?
NO
What does correlation measure?
strength of the relationship between 2 variables
What is Pearson’s Correlation Coefficient (R)?
linear regression for continuous data
Range of correlation coefficient is
-1 to 1
How to control for confounding variables
randomization is best, can also match subject for subject and stratifying groups
use multivariate analysis calculations
What is the gold standard for recommendation
Oxford University grading scheme
Grading for level of recommendation is based on
quality of evidence A-D
strength of recommendation I-IV
Strongest design for evaluation of clinical question
systemic review
strongest single design to prove causation
RCT (then observational)
strongest support of causation
cohort study (then case-control)
What is used to test for null hypothesis and two categorical variables
chi-squared
What is used to predict outcome of 1 variable from another
simple linear regression
What is multiple regression used for
determines relationship between 2+ explanatory variables and a response variable
What is Tukey’s test
used in conjunction with ANOVA to determine if the means are significantly different from each other
What is Bonferroni test
used when multiple dependent or independent statistical tests are being performed simultaneously
As the # of hypotheses increases, the chance of witnessing a rare event leading to a type I error increases
Establish a more strict α level to obtain significance
What is a non-parametric analogue to ANOVA
Kruskall-Wallace
1 normal and 1 measureable variable
positive likelihood ratio
indicated how good a test is at RULING IN disease
negative likelihood ratio
indicates how good a test is at ruling OUT disease
positive predictive value
proportion of positive test that are true positives
negative predictive value
proportion of negative tests that are TRUE negatives
pre-test probability
initial intuition about a disease based on your pt.’s presentation + risk
post-test probability
estimated likelihood of a disease after a test result
treatment threshold
probability above which the dx is so likely you would treat the pt. without further testing (high suspicion)
parallel testing
many tests run at once
ANY positive test is diagnostic
increases sensitivity
ex: CKMB, troponin, ECG for MI
serial testing
ALL tests must be positive for a dx
increases specificity
reduces risks of unnecessary interventions
clinical breast exam, mammogram, bx
In general what makes up a good screening test?
- High true negatives & true positives
- Low false positives & false negatives
oMore important to prevent false negatives person would be walking around with undiagnosed illness/disease
oFalse negatives can be r/o with further testing
Define gold standard broadly
1) An accepted test that is assumed to be able to determine the true disease state of a patient regardless of positive or negative test findings or sensitivities or specificities of other diagnostic tests used.
2) An acknowledged measure of comparison of the superior effectiveness or value of a particular medication or other therapy as compared with that of other drugs or treatments
Why is an adequate sample size important?
1) Decreases risk of type 2 error
2) Compliance & drop-out need to be recognized (adequate numbers)
3) Stratify/sub-grouping data (adequate numbers)
