Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
89 Cards in this Set
- Front
- Back
Four measures of validity in diagnostic testing
|
sensitivity, specificity, predictive values, likelihood ratios
|
|
Sensitivity equation
|
# true positive (TP) / # with disease (TP+FN)
|
|
Specificity equation
|
# true negative (TN) / # without disease (TN +FP)
|
|
Positive Predictive Value (PPV)
|
likelihood that a person with a positive test result actually has the disease
|
|
Negative Predictive Value (NPV)
|
likelihood that a person with a negative result truly does not have the disease
|
|
PPV equation
|
TP/TP+FP
|
|
NPV equation
|
TN/TN+FN
|
|
Difference between sensitivity and specificity and predictive values?
|
predictive values are dependent on characteristics of the test AND the prevalence of the disease
|
|
SNOUT and SPIN
|
Sensitive tests have few FN and rule out the disease
Specific tests have few FP and rule in the disease |
|
LR
|
LR+ = probability of an individual with the condition having a positive test/ probability of an individual without the condition having a positive test
(can be negative) |
|
Post-test odds
|
= pre-test odds * LR
|
|
Post-test probability
|
= post test odds/(post test odds + pre test odds)
|
|
incidence
|
number of new cases occuring in a particular time period
|
|
incidence rate
|
# of new cases/
total # at risk/ unit time |
|
prevalence
|
number affected at a point in time
|
|
prevalence rate
|
# with disease/
# at risk |
|
attack rate
|
# contracting disease/
# at risk or exposed X100 variation of incidence used when looking at a specific short term risk or exposure (e.g. potato salad) |
|
mortality rate
|
# deaths/
# at risk/ time |
|
case fatality rate (CFR)
|
# dying of disease/
# with disease X100 |
|
redraw epidemiologists bathtub
|
check drawing
|
|
health disparities
|
between different groups, which are not avoidable nor unfair (e.g. black people get more diabetes due to genetics)
|
|
health inequities
|
health disparities which are avoidable and unfair
|
|
discrete variables
|
nominal and ordinal
|
|
continuous variables
|
interval and ratio
|
|
nominal variables
|
discrete groups
e.g. male/female, smoker/non-smoker |
|
ordinal variable
|
ordered, without meaningful intervals
e.g. class rank |
|
interval variables
|
ordered, with meaningful intervals
e.g. temperature in celcius |
|
ratio variables
|
interval data with an absolute zero (so a value that is double really is double)
100% really is twice as high as 50% on a test, but 20 degrees isn't really twice as hot as 10 degrees |
|
frequency
relative frequency cumulative frequency |
# in each catagory
% in each catagory cumulative % |
|
kaplan meier method
|
used to plot survival data, adjusts to reflect patients who drop out
|
|
mean
median mode |
average
50% above, 50% below value that occurs most |
|
graph mean, median, mode on skewed graph
|
check graph
|
|
standard deviation distributions
|
68% within 1 SD
95% within 2 SD 99.7% within 3 SD |
|
alpha level meaning
|
level of uncertainty you are willing to accept
generally .05 |
|
p-value meaning
|
probability that difference between groups in due to chance
|
|
paramentric tests
|
focused on population parameters
require an interval or ratio scale assumes normal distribution ex. t-test, paired t-test, ANOVA |
|
non-parametric tests
|
not focused on pop. parameters
can use ordinal or nominal scales few assumptions ex. chi-squared |
|
difference between parametric and non-parametric
|
parametric is more powerful
non-parametric used for small sample size |
|
paired t-test
|
compares means of single group before/after intervention
|
|
t-test
|
compares means of two groups
ex. drug and placebo groups |
|
ANOVA
|
compares means between more than two groups
|
|
chi-squared
|
test of proportions
ex. 70% of hs grads use seatbelts vs 85% of college grads, tests significance |
|
type I error
|
false positive
5% chance if alpha is .05 |
|
type II error
|
false negative
likelihood depends on power or study (often due to small sample size) |
|
power
|
probability of avoiding type II error
(1-Beta) so if power is .8, then there is an 80% probability of avoiding a type II error |
|
what is cut-off for beta level of well designed study?
|
0.2 or less (gives a power of 0.8 or higher)
|
|
confidence interval
|
measure of precision (normally 95%) gives range that we are 95% sure the actual population is in
lets us know if results are statistically significant small confidence interval - more precise |
|
what determines the width of a confidence interval?
|
variability of the sample
size of sample |
|
correlation coefficient (r)
|
1 = perfect correlation
> 0.5 = strong correlation 0 = no correlation |
|
coefficient of determination
|
r-squared, expressed as percentage
expresses proportion of variance in a variable explained by another variable |
|
randomized controlled trial
|
subjects randomly assigned to intervention and control groups
|
|
double blind
|
patient and researcher don't know which group they are in
|
|
Intention to Treat (ITT) anaysis
|
all patients initially assigned to a treatment group are analyized in that group, regardless of whether they received their assigned treatment
preserves value of randomization |
|
randomized control trial strengths and weaknesses
|
strenghts: most powerful study
cause and effect relationship weakness: expensive, laborious, possible ethical or practical issues |
|
cohort studies
|
a group without disease is observed noting exposure to risk factors and development of disease
|
|
cohort studies strengths and weaknesses
|
strengths: low bias, establishes risk of disease, suggests cause/effect relationship
weaknesses: time consuming, expensive, not practical for rare diseases |
|
case-control study
|
select a group with a disease and pair with group without disease
investigate past exposure to risk factors |
|
case-control study strengths and weaknesses
|
strengths: quick, cheap, no ethical issues, good for rare diseases
weaknesses: significant bias, cannot determine prevalence rates or risk of disease, weak indicator of causality |
|
cross-sectional study
|
same as prevalence
survey population for presence of disease and potential risk factors |
|
cross-sectional study strengths and weaknesses
|
strengths: quick, inexpensive if data available, gives prevalence
weaknesses: no causality, you can only suspect it |
|
case reports/case series
|
describes event in single patient or a series of patients
|
|
case reports/case series strengths and weaknesses
|
strengths: can by done by any clinician, no cost, used for preliminary hypotheses
weaknesses: no statistical validation, results may be coincidental |
|
systematic review (SR)
|
summary of medical liturature regarding clinical question
|
|
SR strengths and weaknesses
|
strengths: combiniation of multiple studies increases power, quick, inexpensive, strongest evidence
weaknesses: may lack adequate number of studies, heterogeneity of studies, prone to poor methodology |
|
research design evidence pyramid
|
systematic review
randomized controlled trial cohort case/control cross-section case report/series |
|
gold standard
|
test which is considered consistently correct (best test out there) which you compare other tests to
|
|
test reliability
|
level of agreement between repeated measures of same variable (repeatability)
|
|
test validity
|
extent to which a test actually tests what it claims to test
|
|
test sensitivity
|
ability to detect people with disease
=TP/(TP+FN) |
|
test specificity
|
ability to detect people who do not have disease
=TN/(TN+FP) |
|
positive predictive value (PPV)
|
likelihood that a person with a positive hest result actually has the disease
=TP/TP+FP |
|
negative predictive value (NPV)
|
likelihood that a person with negative result truely does have disease
=TN/TN+FN |
|
difference between sensitivity/specificity and PPV and NPV
|
predictive values depend on prevalence, sensitivity and specificity are characteristics of the the test alone
|
|
SNOUT and SPIN
|
sensitive tests have few false negatives
specific tests have few false positives |
|
likelihood ratio
|
use nomogram and estimate post-test probability
|
|
ROC curve ratings
|
0.9-1 excellent
0.8-0.9 good 0.7-0.8 fair 0.6-0.7 poor 0.5-0.6 failed |
|
odds ratio (OR)
|
measures degree of association of risk factor with a disease
= odds that case had risk factor/ odds that control had risk factor OR of 6 means a six times greater chance of person with case having risk factor weak indicator of causality |
|
absolute risk (AR)
|
same as incidence
risk over a given time =# developing disease/total # |
|
relative risk (RR)
|
measures how many times exposure to risk factor increases risk of disease
= AR among exposed to risk factor/ AR of among non-exposed -must have incidence -indicates causality |
|
relative risk reduction (RRR)
|
measures decrease in RR due to intervention
= 1-RR |
|
absolute relative risk (ARR)
|
decrease in absolute risk due to intervention
= AR(control) - AR(treated group) |
|
compare ARR and RRR
|
RRR is generally higher, more impresive
ARR is more clinically relevent since it incorporates incidence |
|
number needed to treat (NTT)
|
number required to treat in order to prevent one negative outcome
=1/ARR |
|
cost of intervention
|
=NNT X cost of one intervention
|
|
common research flaws
|
bias
internal validity confounding variables external validity power |
|
types of bias
|
selection
observer participant withdrawl recall instrument publication |
|
internal vs external validity
|
internal - are you measuring what you say you are?
external - is outcome relevant in real world? |
|
confounding variable
|
variable which research failed to measure, impacting outcome
|
|
level of evidence/grade of recommendation
|
1-5 (1 is strong, 5 is week)/A-D (A is strong)
level of evidence - quality of research grade of recommendation - based on level of evidence |