Untitled spreadsheet - Sheet1 Flashcards
Topic
Discussion
Lead-time Bias
earlier diagnosis does not = extended life expectancy
Female incidence & mortality of cancer
Incidence of cancer in women: breast > lung > colon; Mortality of cancer in women: lung > breast > colon
Confounding
blurs the exposure & outcome association when a THIRD variable correlates with both (EtOH use & bladder cancer, no link bc smoking is assoc. w/ bladder cancer & smokers are more likrely to drink)
Sensitivity & Specificity
Sn = A/(A+C) Sp = D/(D+B)
Positive-Skew
Mean > Median > Mode
Attributable Risk Percent
ARP = 100 x (RR-1)/RR
Prevalence vs Incidence
Incidence = new cases/at risk pop; Prevalence = # of cases/total pop; Prevalence = incidence x duration of disease
Relative Risk Reduction
RRR = ((AR:control)-(AR:tx))/AR:control
Positive Predictive Value
PPV = A/(A + B)
Cohort Study
Odds ratio = Odds of exposure in diseased/Odds of exposure in non-diseased
T-test
test compares the mean of 2 groups of subjects (regardless if groups are different or same)
Chi-Square test
Categorical data comparison; comparing >/= 3 groups
If RR b/w 2 studies is the same but the p-value varies
sample size may have been too small
Correlation coefficient
0: no correlation b/w the 2; 1: absolute correlation, -1: perfect inverse correlation
Hawthorne Effect
subjects change behavior knowing they are being studied
Observer Bias
double-blind studies reduce observer bias
Relative Risk
RR = (a/a+b) / (c/c+d)
Power
1-type II error (b)
Negative Predictive Value
Likelihood that a negative test means the patient does not have a disease NPV= D/(C+D)
Reliable tests
Reproducibility of a test
Reduction of risks
smoking is almost always the #1 reducible mortality risk
Statistical significance
if the 95% CI does not include 0 the Ho can be rejected
PPV is higher when the prevalence of ds increases
NPV is lower when the prevalence of ds increases