Biostats definitions and equations Flashcards
Compares people with DISEASE to people without disease
Case-control study
Patients serve as their own control
Crossover study (reduced confounding bias)
CI between two groups overlaps
No difference, fail to reject HO
Hawthorne effect
Groups who know they’re being studied behave differently (measurement bias)
PPV =
TP/TP+FP
Measured by a cross-sectional study
Prevalence
1 standard deviation = __%
68
Attributable risk
Difference in risk between exposed and unexposed group
Mean > median > mode
Positively skewed
(Tail to the right)
Mean < median < mode
Negatively skewed
(Tail to the left)
Incorrectly rejecting the null hypothesis
Type I error
(Saying there is a difference when there isn’t)
Information is gathered in a way that distorts it
Measurement bias
Collects data from a group of people to asses disease prevalence
Cross-sectional study
Which bias? Patients in treatment group spend more time in in-patient treatment
Procedure bias
Mean=median=mode
Normal distribution
AR =
% risk in exposed group - % risk in unexposed group
NNT =
1/risk difference
Specificity =
TN/TN+FP
1 - false positive rate
Increases with increasing prevalence
PPV
3 standard deviations = ___%
99.7
Coefficient of correlation =
r2
Nonrandom assignment to participate in a study group
Selection bias
95% CI includes 0
HO not rejected
RRR =
1 - RR
Berkson bias
Study looking only at inpatients (a selection bias)
Two means
T-test
Ways to increase power and reduce ß
Increase sample size
Increase expected effect size
Increase precision of measurement
Subjects in different groups are not treated the same
Procedure bias
ß
Probability of making a type II error
(Failing to reject the null hypothesis)
Which bias? Study over effectiveness of CRC screening concludes that their screening method leads to increased survival
Lead-time bias
Berkson bias
Loss to follow-up
Healthy worker and volunteer biases
Selection bias
Power =
1 - ß
(ß = type II error)
Alpha
Probability of making a type I error
(Falsely rejecting the null hypothesis)
Failing to reject the null hypothesis
Type II error
(Stating there isn’t a difference when there is one)
RR =
Incidence in exposed group/incidence in non-exposed
a/(a+b) / c/(c+d)
SEM =
σ/√n
Which bias? Researcher expects those receiving drug to show more signs of improvement, so he documents those signs in the treatment group more frequently than the placebo group
Observer-expectancy bias
Which bias? Coal miners more likely to have lung disease but also more likely to smoke
Confounding bias
Z of 95% CI? 99% CI?
95% CI Z = 2
99% CI Z = 2.5
Categorical differences
Chi-square
(“Chi-tegorical”)
Early detection is confused with increased survival
Lead-time bias
Relative risk reduction
Proportion of disease reduction attributable to intervention compared to control
Decreases with increasing prevalence
NPV
Hawthorne effect is an example of:
Measurement bias
Bias reduced via randomization and ensuring the right comparison group
Selection bias
3+ means
ANOVA
(Remember: ANOVA = 3 words)
Bias reduced via multiple studies, crossover studies, and matching
Confounding bias
CI =
[mean - Z*SEM] to [mean + Z*SEM]
CI between two groups doesn’t overlap
Difference exists, reject HO
OR =
ad/bc
Don’t change if sample size increases
Specificity and sensitivity
95% CI includes 1
Reject H0
Sensitivty =
TP/TP+FN
1 - false negative rate
Measured by a cohort study
RR
Bias reduced via “back-end” survival (measuring survival based on severity of disease at diagnosis)
Lead-time bias
2 standard deviations = __%
95
Prevalence =
Incidence x disease duration
(TP + FN)/(TP+TN+FP+FN)
(a+c)/(a+b+c+d)
NPV =
TN/TN+FN
Bias reduced via placebo and blinding
Measurement bias (makes sure participants and researchers don’t change their behavior)
A factor is related to both exposure and outcome, but not on the causal pathway, so it distorts the effect of the exposure on the outcome
Confounding bias
Compares people with an EXPOSURE to those without to see if exposure increases likelihood of a disease
Cohort study
Increases with decreasing cut-off value of a test
Sensitivity
NPV
Measured by a case-control study
OR
