Biostats Flashcards
What is a case control study?
Select pts with dz (cases) and those wo dz (control) and RETROSPECTIVELY determine exposure
What is a cross sectional study?
Prevalence study. Ie a survey.
Of looking at ppl w heart disease and then making them go thru img test to calculate their calcification score
What types of prevention are there? Primary, secondary, tertiary? Primordial?
Primordial- before risk factor develops
Primary - for those at risk, before disease develops
Secondary halt disease process
Tertiary when disease has become advanced and now try to limit disability- ie cabg or pulm rehab
Quarternary preventions - HEALTH SYSTEM- using EMR to prevent extra catheterization
95% of pts lie within how many standard deviations?
99.7%?
68%?
2 standard deviations from the mean
3 SD
1 SD
Funnel plot is used to asses what type of bias?
Publication bias
X axis has treatment effect, y axis is standard error
No bias means 95% of the studies lie in triangle
Larger studies are at top and narrow spread bc more powerful. Smaller studies wide and on the base of triangle
Funnel plot is used to asses what type of bias?
Publication bias
X axis has treatment effect, y axis is standard error
No bias means 95% of the studies lie in triangle
Larger studies are at top and narrow spread bc more powerful. Smaller studies wide and on the base of triangle
What is best answer for the NET CLINICAL BENEFIT of a drug??
Answer should include the beneft and harm (ie intracranial bleed). Not just the benefit.
Important features of randomized controlled trials include randomization, blinding, and intention to treat analysis. ITT is to preserve what
Randomization. Regardless of if they drop out or die or didnt finish tx
What is external validity?
How applicable the results are or how representative they are to the desired population
What is type II error?
When a study FAILs to reject null hypothesis that is false. Depend on power of the study to detect a difference when a difference exists.
Which is better measure of central tendency in a strongly skewed distribution? Mean or median?
Median
If normally distributed then mean is best.
What is type I error?
False positive
Study incorrectly rejects null hypothesis that is true
Reflects the significance
Define: Case fatality
Mortality
Case fatality is porportion of ppl on specific gp who die
Mortality is prob dying from condition in general population
What would using one tailed test do to the power?
Incr power bc only look at effect in one direction
When you see alpha, that is the p value or statistical significance
How to calculate incidence ratio?
Number of observed cases/expected cases
What test uses p value and F to study mean values between groups?
Anova - analysis of variable
Fischers test only when small sample
Mcnemar test comparing two proportions where pts serve as own control
Case control studies measure what?
Association
How to calculate relative risk?
Risk in exposed/risk in unexposed
What is attributable risk? Ie what percent of ischemic strokes observed in SMOKER is attributed to smoking status?
Risk in exposed - risk in unexposed / risk in exposed
What is population attributable risk percent ? Ie what percent of ischemic stroke observed in the POPULATIon is attributed to smoking?
Risk in total population - risk in unexposed / risk in total population
When does a result become NOT statistically significant if there is CI?
If included null value of 1
In order to calculate Ntt, how to calculate arr?
Risk in group-risk jn placebo. Then ntt is 1/arr
When asked to find group with selection bias, what do you look at?
The N in both the observed group and treatment group should all be accounted for. If only 33% of the ppl had their ferritin levels measured then that is not representative of the target population
A study shows significance in use of new inhaled agent with mean number of rescue uses 22.6 vs 24.1 withh p value 0.04. What does this study lack?
Clinical significance!!
Just like how a new antihtn which lowers significantly by 0.5 mmhg doenst matter
How to calculate
Positive likelihood ratio
Neg likelihood ratio
Positive likelihood = sensitivity/(1-spec)
Neg likelihood= 1-sense/spec
These do not depend on prevalence
When a study uses a test ie colposcopy on pap abn women and also performs pap on women w normal pap what is this trying to reduce
Verification bias
What is smr the standardized mortality ratio?
Number observed deaths/expected number deaths
What is this kind of study
Crossover study
Cross sectional
Factorial study
Crossover is experimental where expose to diff tx sequentially
Cross sectional is observational
Factorial study is experiment where use more than 2 interventions in all combinations. Je randomize pts to glutamine, antioxidants, both or nothing
sensitivity vs ppv
Sensitivity is percent of predicting dz if it is present
Vs ppv is ratio of having the disease given positive result
Rr of 0.4 can be thought of as what percent
40%
Odds ratio
Odds are lower!!
Ie 0.25
How to calculate NNH
How to calc attributable risk
Need artributable risk incidence of dm in exposed minus incidence in nonexposed
18.9/1000-16.9/1000=0.002
NNH = 1/Atrributable risk=500
Screening test should have high
SeNsitivity!!!! To rule out
If the tail (less side) is to the right then the orders of mean median node are from left to right
Median mode mean
