Flashcards in Biostats Deck (81):
1 and 5 min
>7 good, 4-6 assist, < 4 resuscitate
-set up a 2x2
Sensitivity rules out - If negative, very likely not going have. Too sensitive of a test you are going to have many false positives.
Sp -in rules in. If positive likely positive test
1-statisitcal power =
Type II error
Odds ratios are calculated in what type of study
Relative risks are calculated in what type of study
How does odds ratio relate to Relative risk?
Odds ratio is an approximation of relative risk because you cannot know the entire population in a case control. In a cohort the population is defined and risks can be calculated
Case control study is looking for what?
looking for risk factors and exposures that are in common in a group of people with a disease in comparison to those without the disease
it is retrospective and observational
Cohort study is looking for what
it is comparing those that have an exposure or risk to those without and seeing what becomes of it
It is observational and can be prospective or retrospective (form a cohort after an exposure and follow then)
Cross sectional study gives what about a disease
a snapshot in time of those that have the disease, can start looking at associations w/ risks to determine correlations
Twin concordance studies measures what?
Clinical trials can be improved w/ what 3 design strengtheners
Phase I of a drug trial asks
Is it safe
Tests on healthy people for toxicity and pharmacodinamics
Phase II of a drug trial asks
Does it work
Test on sick people for efficacy, dosing and adverse effects
Phase III of a drug trial asks
Does it work better
Large #s of people and clinical trials
Phase IV of a drug trial asks
What else can go wrong?
Post market surveillance
Pooling together of different studies based in inclusion criteria to gain statistical power
-some selection bias and concern for individual merits of the studies
Sensitivity tests what?
Sensitivity tests proportion of those that test positive to the disease and compares to ALL that have the disease
TP/(TP + FN)
Specificity Tests what?
Specificity tests the proportion those that test negative to the disease compared to ALL those that do not have the disease
Positive predictive value tests what?
PPV compares those that test positive and actually have the disease to the the proportion of ALL those that test positive to the disease
TP/(TP and FP)
negative predictive value tests what?
NPV compares those that are actually disease free to the proportion of ALL those that test negative to the disease
TN/(TN + FP)
How does Specificity and sensitivity change with prevalence?
They don't. Specificity and specificity are limited to the categories of those that either have the disease or don't.
How does PPV and NPV change with prevalence?
With increased prevalence the PPV increases and NPV decreases.
-testing in an area where there is a lot of disease you are more likely to believe a positive test over a negative test
With a decrease in prevalence, the PPV decreases and NPV increases.
-Testing in an area where there is not a lot of disease. more likely to believe the negative test than the positive test
Given only the false negative %, what is the sensitivity?
1 - FN
1 - False Positive% =
Which test do you use to screen for and which do you use to confirm?
you screen with a highly sensitive test because you want to catch all the positives. May have some FP.
You you then run a high specific test which is looking for those that are negative(Specific tests are those comparing those that are truly neg compared to all those that test negative); if you test negative you are negative in the disease and eliminates FP from the sensitive screen
what is incidence?
Incidence is the amount of new cases found in a population at risk
# of cases/those at risk (minus those w/ disease)
what is prevalence?
prevalence is the amount of people with a disease (all cases) in a population at risk
# of existing cases /population at risk
How does time course of an illness affect prevalence and incidnce
incidence = prevelence x time course
The shorter the disease duration the smaller the prevalence. Ex: the flu
Chronic disease may have a low incidence but a high prevalence (COPD, DM)
odds ratio is the odds that a group exposed to a risk will develop the disease compared to those that are not exposes - case studies
RR is the incidence of a disease in a population exposed to a risk divided by the incidence of a disease in those not exposed - Cohort
[a/(a+b)] / [c/(c+d)]
RR -> OR when the prevalence of the disease is really small ( a and c are minuscule)
Attributable Risk is ?
the difference between incidence of those that are exposed to a risk with a disease compared to those that are not exposed
a/(a+b) - c/(c+d)
Associated w/ Number needed to harm which is 1/AR
Absolute risk reduction is?
the difference between the incidence of the disease in those NOT exposed to a Rx to compared to those that are
c/(c+d) - a/(a+b)
Associated with number needed to treat
Number needed to treat
Determines the cost effectiveness of a treatment. Number of people that get treatment before having one positive effect
1/Absolute risk reduction
1 / [c/(c+d) - a/(a+b)]
Number needed to harm
determines the number of people needed to be exposed to a risk before before disease
1/ Attributable Risk
1 / [a/(a+b) - c/(c+d)]
the reproducibility/repeatability of a test
May not be accurate
the truthfulness of a test
may not be precise/reproducible
nonrandom assignment to participation in study
-ex may be those reffered to a study are closer to end stage disease
subtype of selection bias where selection of patents are more likely to be hospitalized
- selection skewed since those selected were more likely to be hospitalized anyways
knowledge of a disorder impacts the ability to recall detail for a study
Ex parents of autistic kids remember more than parents of kids w/o autism
subjects are not representative of the target population
ex - avg age of participants is less than avg age of target group
Late look bias
information gathered in a timeframe that preferentially selects for this w/ more indolent course of disease; severe cases dead and limits generizability
Subjects are not treated the same in a clinical trial
ex: experimental drug rx may receive extra couching and disease education compared to controls
Lead time bias
esp pertinent to screening studies
- falsely attributing increased survivability when in fact early screening only found the disease sooner
natural course of the disease has not changes
observer expectancy effect/pygmalian effect
the confidence of the researchers assumption of the results of a study influences the study to match expectations.
Not malignant in change, maybe participants conform to thought
disruption of truthfulness of the study due participants knowledge of being observed
ex: hand washing frequency
when there is a 3rd variable related to the variable tested and the measure result that either modifies or magnifies the result
Something else may be going on
Cross over study
patient serves as their own control. Take both the experimental and control substance unknowingly to determine effect
positive skew in a distribution
tail to the right
meaning the mean>median>mode
negative skew in a distribution
tail is to the left
meaning the Mean< mode
maintain Ho when H1 is true
False negative or type II error (beta)
falsely saying nothing is going on with 2 variables when in fact there is a relationship
alpha in biostatistics
is type I error - making a statement that there is an association when in fact there isn't; rejecting Ho for H1 when Ho is true
alpha is a usually selected at 0.05 for the cut off of risk of making a type 1 error- 5 % chance of saying there is something going on when there is not. p needs to be less than 0.05
beta in biostatistics
type II error - making a statement that there is nothing connecting 2 variables when there is something
a false negative
is probability calculated from the data the the results are due to chance alone.
Compared to alpha value and usually needs to be less than 0.05 If so then, you are safe to reject the null hypothesis, for there it is unlikely that there is nothing going on and an alternative hypothesis must be accepted
Power in biostats
1- beta(or type II error)
the probability of falsely rejecting a null hypothesis when it is in fact false.
-or the likelihood of finding a difference if one in fact exists.
empowering your decision to reject null, increase power, increase rationality that it is ok to reject null
Increase power how?(3)
increase sample size
increase expected effect size
increase precision of measurement
% of population w/in?
3SD - 99.7%
SD/ square root (n)
n = population
- as n increases SEM decreases (less error)
used in calculating Confidence intervals
Confidence interval -
Mean +/- Z x (SEM)
Z= 1.64 w/90%
Z= 1.96 w/95%
Z= 2.5 w/ 99%
When to not reject Ho with a confidence interval
When confidence interval includes 0
When CI for an OR or RR includes 1
checks difference between the means of 2 groups of continuous variables (i.e weight)
checks the difference between the means of 3+ groups of continuous variables (i.e. weight)
tests the difference between 2 or more percents or proportions of categorical incomes
ex: yes/no; age 50-59 vs 60-69; men vs females
is pearsons correlation which can range from -1 to 1
the closer the value is to the 1 the stronger the positive correlation that exists between 2 variables
0 = no correlation
Coefficient of the determinent
the goodness of bit and variance within the data pts. reflects the proportion of variance in y that is due to variance in x
Car seat recommendations
Vaccines in kids 0-6(10)
Vaccines in kids 7-18 (4)
TDaP -> Td q 10yrs
Vaccines in adults
Herpes zoster > 50
Hep A and B if liver damage
Screening recommendations in adults age 50
Colon q 10 yrs
Tobacco and alcohol
ASA use for preventative measures?(2)
45-79 - daily in Men for MI
55-79 daily in women for Stroke
Kubler Ross grief stages (5)
When does grief become pathological
depression criteria med at least 2 weeks after the 1st 2 months
generalized feelings of hopelessness, helplessness, worthlessness, and guile
Distreessed feeling > 6 months
inability to move on > 6 months
3 leading causes of death in adults > 65
Odd recommendation for AAA screening
1 US for men 65-75 who have ever smoked
DEXA scan q 2 yrs older than 65
PAP smears age
start age 21 end age 65
Mammogram screening age
Reportable diseases (5)
STDs - Chlamydia, gonorrhea, Shyphilis HIV
Hep A, B and C
Diarrheal - salmonella and shigella
vaccine preventable diseases like mumps
What does not change in the elderly (2)
Sexual interest (may have mech issues though)