MRC Basic Science/Stats/Epi Flashcards
(79 cards)
- ) What is a Case-Control Study? What Level of Evidence?
2. ) What stat is calculated?
- ) Group of patients compared RETROSPECTIVELY (ie. look at disease and see what risk factors they had) = Level 3
- ) Odds Ratio = probability event will occur/probability event will not occur
- ) What is a Cohort Study? What Level of Evidence?
2. ) What stat is calculated?
1.) Group of patients compared PROSPECTIVELY (ie. look at risk factors 1st and see who gets dz) = Level 2
2.) Relative Risk = risk in exposed/risk in unexposed
(RISK = INCIDENCE!)…only a cohort study can tell you incidence b/c it is going forward in time!
RR = 1: No association
RR > 1: positive association (risk in exposed > risk in unexposed) CANNOT INFER CAUSAL!! - can’t tell!
RR < 1: negative association (risk in exposed < risk in unexposed) CANNOT INFER PROTECTIVE!!
**If 95% CI crosses 1 = NOT SIGNIFICANT
1.) What type of study are:
Case-Control, Cohort, Case Series, and Case Report?
2.) What type of study is a RCT?
- ) Observational
2. ) Experimental
- ) What Level of Evidence is a RCT typically?
2. ) What can downgrade it?
- ) Level 1
- ) Poor f/u (<80%), heterogeneous results, no blinding, concerns about randomization. ANY OF THESE WOULD MAKE IT A LEVEL 2 STUDY!!
What is the difference b/t a systematic review and a meta-analysis?
Systematic review just looks at a bunch of studies.
Meta-analysis uses fancy STATS to combine the results!
FDA Phases of Research - what happens in each Phase I-IV?
I -> “First in Man” to determine safety
II -> Determine if device/drug is effective (MOST COMMON phase of failure)
III -> Confirm efficacy through large trials
IV -> Postmarketing surveillance
Does QI study require IRB?
NO! Not research!!
Name the studies that are Levels 1-5 of evidence?
o Level 1 – Randomized controlled trial or Meta-analysis
o Level 2 – Cohort study (PROSPECTIVE)
o Level 3 – Case-control study (RETROSPECTIVE)
o Level 4 – Case series, cross sectional
o Level 5 – Expert opinion
What types of Bias occurs BEFORE a study (3)?
o Selection – improper recruitment of subjects with different features
*Prevent this with randomization that is blinded
o Channeling – subjects unequally given treatment based on their features
o Chronology – use of historical controls
What types of Bias occurs DURING a study (5)?
o Detection – looking harder at one group than another
o Recall – relying on patients to remember events
o Interviewer – influence the interviewer has on responders
o Performance – procedures not performed in uniform way
o Hawthorne Effect – alteration of behavior of subjects based on knowledge they are being observed
What types of Bias occurs AFTER a study (3)?
o Citation – more likely to believe study in top journal
o Publication – positive results more likely to be published
o Conflict of Interest – researcher personal conflicts
What is the difference b/t incidence and prevalence?
Incidence = risk over time (looked at with cohort studies b/c they are PROSPECTIVE) Prevalence = proportion of existing cases in the population being looked at - at a single moment (snapshot)
What statistical tests are used to evaluate categorical data? And how do you know which one to use?
Chi-squared (most common to use)
Fischer exact -> use for SMALL groups (remember that exact count easier with small group!)
What statistical tests are used to evaluate continuous data? And how do you know which one to use?
T-test, ANOVA, Pearson correlation co-efficient, Regression, Mann-Whitney U test.
PARAMETRIC DATA: (bell curve, data normally distributed)
T-test (two groups)
ANOVA (3 or more) “OVA two!
NONPARAMETRIC DATA -> pick Mann-Whitney U test
Determine relationships -> Pearson correlation co-efficient
Predict outcomes from variables -> Regression
What is sensitivity?
How do you calculate sensitivity?
Sensitivity = ability of a test to detect dz
TP/TP+FN
SnOUT = negative result rules OUT a diagnosis if you have high sensitivity
What is specificity?
How do you calculate specificity?
Specificity = ability to detect health
TN/TN+FP
SpIN = positive result rules IN a diagnosis if you have high specificity
What is PPV?
How do you calculate PPV?
How likely you are to have dz w/ a positive result.
TP/TP+FP
What is NPV?
How do you calculate NPV?
How likely you are to not have dz w/ a negative result.
TN/TN+FN
What is a Type I error?
Alpha error = False positive error
Incorrectly rejected the null hypothesis -> said there was a difference and there IS NOT! (Cried wolf!) MORE DEVASTATING!
Only willing to except this 5% of the time
p < 0.05 is a marker of certainty -> which means the likelihood of the results happening by random chance is 5/100 when no association really exists. *NOTE: A HIGH P-VALUE DOES NOT MEAN IS STATISTICALLY INSIGNIFICANT, IT MEANS THAT THERE IS A HIGH DEGREE OF UNCERTAINTY!!
What is a Type II error?
Beta error = False negative
Incorrectly accepted the null hypothesis -> said there was no difference/association and there was - you missed it!
LESS devastating
Willing to accept this 20% of the time.
What is the Power of a study?
Probability of finding a significant association if it exists -> ability to find a true positive or true negative.
Calculated by 1-beta = 80% chance of doing a study that finds p < 0.05 if true association exists.
What test gives you detection of publication bias in meta-analysis?
Funnel plot
- ) What is the equation of Stress?
2. ) What is the equation of Strain?
- ) Stress = Force/Area
2. ) Strain = change in height/original height
What is the definition of Hooke’s Law?
Stress is proportional to strain in the elastic zone of the stress strain curve (initial linear part of curve)