Biostats Flashcards
Specificity
Diagnostic test; probability a test shows non-disease when no disease is present. d/(b+d)
Sensitivity
screening test; probaility that test detects disease when disease is present. a/(a+c)
Positive predictive value
Proportion of positive test results that are true positives (given a positive test result the chance the person actually has disease). Increases with increased prevalence. a/(b+a)
Negative predictive value
proportion of negative test results that are true negatives (probability that person actually is disease free given a negative result). Decreases with decreased prevalence. d/(d+c)
case-control study
comparing a group with a disease to a group without a disease to identify a risk factor. It measures an Odds Ratio = (a/b)/(c/d)
Cohort
Study of an identified risk factor and wether it is associated with a disease. Compares group with exposure of risk factor to a group without exposure. Determines relative Risk (RR) = [a/(a+b)]/[c/(c+d)]; you can see that if the prevalence of the disease is not high the OD is a good estimate of the RR
Cross-sectional study
looks at people to asses prevalence of disease and related risk factors at a particular point in time. It determinse Disease prevalence and can show risk factor but NOT causality.
Twin concordance
Compares frequency with which both monozygotic or both dizygotic twins develop a disease. Measures heritability
Adoption Study
Compares siblings raised by biologic vs. adoptive parents. Measures heritability and influence of environmental factors.
Phase I clinical trial
Tests healthy volunteers to assess safety, toxicity, and kinetics
Phase II clinical trials
Small number of patients with disease are tested for efficacy, dosing, and adverse effects
Phase III clinical trials
Large number of patients randomly assigned to treatment or placebo. Measures new treatment agains current standard of care.
Crossover Study
Subjects are randomly allocated to a sequence of two or more treatments given consecutively (ie initially a placebo or initially treatment then placebo in different orders and comparing the two groups.) Allows for subjects to be own control.
Prevalence
total cases in population at a given time divided by the total population. It approximately equals (incidence)*(time), incidence = prevalence for acute and > for chronic diseases
Incidence
new cases in population over given time divided by total population AT RISK (people who already have disease are not at risk)
Odds Ratio
Odds of having disease in exposed goup over odds of having disease in unexposed. (a/b)/(c/d)
Relative Risk
Probability of getting a disease in the exposed group divided by probability in unexposed group. [a/(a+b)]/[c/(c+d)]
Attributable risk
Difference in risk between exposed and unexposed groups. a/(a+b) - c/(c+d)
Absolute risk reduction
Reduction in risk associated with a treatment compared to standard treatment or placebo. c/(c+d) - a/(a+b)
Numer needed to treat
1/absolute risk reduction. (number of patients need to treat to get beneficial effect)
Number needed to harm
1/attributable risk
Case fatality rate
fatal cases / total population w disease
Precision
RELIABILITY. Reproducibility of a test. Absence of random variation
Accuracy
VALIDITY. Trueness of test measurement
Random error
reduces PRECISION
Systematic error
Reduces accuracy
Selection bias
nonrandom assignment to study group (ie Berksons bias)
Recall bias
knowledge of presence of disorder alters recall by subjects
Sampling bias
subjects are not representative to general population
Late-Look bias
Information gathered at inappropriate time like using a survey to study a fatal disease (patients who are dead can’t answer)
Procedure bias
subjects in different groups are not treated the same- ie. More care and attention is given to treatment group which leads to more compliace and better overall health compared to control
Confounding bias
occurs with two closely related factors. The effect of one factor distorts or confuses the effect of the other. This can be controlled for with Crossover studies while each subject acts as own control
lead-time bias
early detection confused with increased survival. Seen with improved screening
Pygmalion effect
occurs when a researcher’s belief in efficacy changes outcome
Hawthorne effect
occurs when group behaves different owing to the knowledge of being studied
Observer bias
occurs when investigators decision is affected by prior knowledge of exposure status
Effect modification
occurs when the effect of a main exposure on an outcome is modified by another variable
Positive skew statistical distribution
mode<mean
negative skew distribution
mode>median>mean
Type I error (alpha)
investigator falsely rejects the null hypothesis by stating there is an effect or difference when none exists
Type II error (beta)
Stating there is NOT an effect when in fact one exists (failure to reject null hypothesis).
Power
probability of rejecting null hypothesis when it is in fact false (ie finding a difference if one truly exists). Power = 1 - beta. Power increases with sample size
Meta-analysis
pools data to increase statistical power. Limited by quality of individual studies or bias in study selection
Confidence interval
estimated range of reproducibility. Values in which specified probability of the means of repeated samples would be expected to fall. CI = mean + Z(SEM); SEM = st’d Dev/(sqrt(n)). For 95% CI, Z = 1.96. For 99% CI, Z = 2.58. If the 95% CI for a mean difference between 2 variables includes 0 then there is no significant difference and null hypothesis is accepted. If 95% odds ratio or relative risk includes 1 then null is not rejected. If CI between 2 groups overlaps then there groups are not significantly different.
t-test
difference between two means
ANOVA
Analysis of Variance. Checks difference between means of 3 or more groups
Chi-squared
compares percentages or proportions between two ro more categorical outcomes (not mean values)
correlation coefficient
always between -1 and 1. the coser to 1 the stronger the correlation between 2 variables