Biostats/Epi Flashcards
How do you calculate the accuracy of a test?
add the correct tests (TP +TN) and divide by all tested
What must be considered when comparing a trait or outcome in different populations?
is there another variable that isn’t being accounted for that may be the reason for the different
ex. gender make up, education level, socioeconomic status
What factors change the prevalence of a disease?
incidence, mortality, recovery
What is a crossover study?
One where all subjects recieve interventions at different points in time (use themselves as the control)
What is the attributable risk and how to calculate?
good for cohort studies
way to measure how many more cases there were in one group than the other
attributable risk = incidence of the exposed group to the incidence of the unexposed
ex. exposed incidence 10%, unexposed incidence 5%, attributable risk 10-5 = 5%
What are the characteristics of this type of study?
Cross-sectional study
looking at absence or presence of illness or variable at a particular time
observational
What are the characteristics of this type of study?
Case study
observational - no control, just seeing what happened to a group
What is sensitvity? How to calculate?
measures how many people who have the dz test positive
TP/(all with dz - (TP+FN)
*tells you - are you catching most of the people with the disease?
What are the testing phases that an intervention must go through to be approved by the FDA?
phase 1: tested in healthy volunteers
phase 2: protocol and dose levels tested in small group of pt volunteers
phase 3: tested for efficiacy and side effects in larger group of pt volunteers
What is relative risk and how to calculate?
good for cohort studies
likelihood that an exposed person will get the dz in comparision to an unexposed person
= incidence rate of exposed/incidence rate of not exposed
ex. if exposed group gets the illness 10% of the time and the unexposed group gets it 5% of the time, the relative risk is 10/5 = 2 (exposed group 2x as likely to get it)
How to calculate prevalence?
proportion of people in a population with the finding
people with finding/entire population
How do you calculate the chance of 2 nonindependent events happening at the same time?
have to change the equation to account for the previous event
How to set up a 2x2 for an odd’s ratio?
OR = (A/C)/(B/D)
What is a useful measure for analyzing cross-sectional studies?
Chi squared
What is lead-time bias?
when dz is picked up sooner (better screening) so it looks like improved survival, but actually not
How do you calculate the chance of 2 non-mutually exclussive events not happening at the same time?
the probabilty of having each added together minus the chance of them happening at the same time (product of the two)
What is the most scientifically rigorous study type?
randomized control trial (RCT)
what is tertiary prevention?
actions to prevent recurrence or slow progression of disease
ex physical activity for pt who had a MI
What are useful measures for analyzing cohort studies?
relative risk (chance of one over the other)
attributable risk (difference in incidence rates)
What sort of bias is seen in case-control studies?
retrospective usually so recall bias
How do you calculate the chance of 2 independent events happening at the same time?
multiple the porbablities of each
How do you calculate the positive predictive value?
this is the number of people who have the disease who test positive
PPV = TP/(all positive tests - TP + FP)
How to set up a 2x2 table for a screening test?
How to calculate the infant mortality rate?
infant deaths/live births
*usually kids who die at less than 1 yr of age
What is specificity and how to calculate?
specificity is how many people without disease test negative
TN/(all dz free or TN+FP)
*identifies the well people accurately if it is high!
What is the life expectancy defined as?
median survival (50% survival)
What is confounding bias?
when the factor being examined is related to other factors that actually explain the relationship
How do you calculate prevalence from a screening test?
look at TP and FN over the sum of all screened
*usually as prevalence increases, so does PPV (while NPV decrs)
*usually as prevalence decreases, so does PPV (while NPV incrs)
How to calculate the mortality rate?
deaths/population
How to calculate the proportionate mortality rate?
deaths from a specific cause/total deaths * 100
What are the characteristics of this type of study?
Cohort Study
this is forward looking, compare one exposed group to a non-exposed group and see what incidence of disease or outcome is
ex. kids who watch tv, kids who don’t, outcome: who needs glasses
What is a postive likelihood ratio?
proportion of dz people to nondiseased people with a positive result
LR+ = sensitivity/1-specificity
How to calculate the fertility rate?
live births/women of childbearing age
What is late-look bias?
where individuals with severe dz get missed beacuse they die first/aren’t included
What is experimenter bias?
experimenter’s expectations somehow communicated to subjects
*use double blind studies to combat this
What is the incidence?
number of new cases over a period of time in an at risk population
*does not include people who already have the dz
What is the number needed to treat?
number of people in the general population who need to be treated to prevent one case
inverse of incidence
= total population/#new cases
How do you calculate the nagative predictive value?
proportion of peole who do not have the disease who test negative
NPV = TN/(all negative or TN+FN)
What sort of bias is seen in cohort studies?
selection or sampling bias (group is not representative of the population of interest)
What is a negative likelihood ration?
proportion of dz people to nondiseased people with a negative result
LR- = 1-sensitivity/specificity
what is primary prevention?
actions that reduce the incidence of disease (new disease)
ex. teaching nutrition to children to prevent diabetes
what is secondary prevention?
actions that decrease the disease prevalance (current disease load)
ex. treatment for smoking cessation or treating TB
How to calculate the maternal mortality rate?
maternal deaths/live births
How to calculate the case fatality rate?
deaths/cases
What is an odds ratio and how to calculate?
good for case-control studies
looking from the disease perspective - odds that diseased people were exposed to the risk factor vs odds that nondiseased people were exposed to risk factors
ex. how many people with lung cancer were smokers vs how mnay people without lung cancer were smokers
= (disease w/ exposure/disease w/o exposure)/(no dz w/ exposure/no dz w/o exposure)
How to calculate the birth rate?
live births/population
What is a useful measures for analyzing case-control studies?
odds ration
What are the characteristics of this type of study?
Case-control study
comparing a group with dz to a group without
almost always retrospective
good for rare conditions
What is the mean?
average
What is the median?
number than divides the group in 2 equal parts - 50th percentile
What is the mode?
the most frequently occuring value in a group of data
What is a deviation score?
value from the mean of that number or data point (difference)
What is the range of a group of data points?
difference between the highest and lowest values
What is the standard deviation?
it is a measure of how widely spread the data set is
For a normal distribution, how much of the data set is captured in each standard deviation?
1 SD - 68%
2 SD - 95.5%
3 SD - 99.7 %
What is the chance of being 2 SD away from the mean?
either +/- = 5%
just + = 2.5%
How do you calculate confidence intervals for 95th and 99th CIs?
95th = mean +/- 2(standard error of the mean)
99th - mean +/- 2.5(standard error of mean)
standard error of the mean = stand deviation/sq root of population (N)
What is the clinical apilcability of confidence intervals?
if the range includes 1 - it is not significant
What is the null hypothesis?
that the findings occurred by chance
What is a typical p value used to determine significance?
p =0.05
What are the 2 types of error that can be made when using p values/doing statistical analysis?
type 1 - reject the null hypothesis when it is really true (say the intervention works when it doesn’t)
type 2 - accept the null hypothesis when it is not true (say the intervention doesn’t work when it does)
How do you interpret correlation values?
-1 means correlated in opposite ways (eye sight and age)
1 means correlated in same way (higher test score and overal grade)
0 means no correlation
When do you use this statistical test?
Pearson correlation
when you have 2 interval variables (numbers)
to see if there is a linear relationship
When do you use this statistical test?
Chi-square
to evalute 2 nominal (categories) variables
can do it with any number of groups
When do you use this statistical test?
t-test
with 1 nominal and 1 interval variable
2 groups only
When do you use this statistical test?
One-way anova
1 nominal and 1 interval variable
*2 or more groups
When do you use this statistical test?
matched t-test
1 interval and 1 nominal variable
*2 groups, good for paired data, before and after
When do you use this statistical test?
repeated measures ANOVA
1 interval and 1 nominal variable
more than 1 groups, linked data
