Sensitivity
- Rules Out
- Screening Test
- TP / (FN+TP)
Specificity
- Rules In
- Confirmatory Test
- TN / (TN + FP)
Case Control Study
- Observational
- Retrospective
- Compares those with disease to those without
- (Asks) "What happened?"
- Odds Ratio
Cohort Study
- Observational
- Longitudinal
- Studies risk from exposure
- Retrospective
- (Asks) "Who developed disease?"
- Prospective
- (Asks) "Who will develop disease?"
Cross-Sectional Study
- Observational
- (Asks) "What's Happening?"
- Data at One Specific Point in Time
- (Finds) Frequency of Disease
- Prevalence
Quantifying Risk: Contingency 2x2 Table Setup
- Disease
- Along Top (x axis)
- Exposure
- Along Side (y axis)
- A: Exposed and Diseased (Sick)
- B: Exposed and Healthy
- C: Not Exposed and Diseased (Sick)
- D: Not Exposed and Healthy
Odds Ratio (OR)
- Case Control Studies
- Measures Association Between an Exposure and an Outcome
- Odds of Exposure in Diseased (Sick) Group
- a / c
- Divided /
- Odds of Exposure in Healthy Group
- b / d
- =1 Exposure does not affect odds
- > 1 Exposure increases disease
- < 1 Exposure decreases disease
Attributable Risk (AR)
- Proportion Exposure Increases Disease
- Incidence of Diseased (Sick)
- Equals =
- Exposed Group
- a / (a + b)
- Minus -
- Unexposed Group
- c / (c + d)
Relative Risk (RR)
- Used in Cohort Studies
- Risk of Developing Disease
- Equals =
- Exposed Group
- a / (a + b)
- Divided by /
- Unexposed Group
- c / (c + d)
- When Low Prevalence Odds Ratio Approximates Relative Risk
Positive Predictive Value (PPV)
- Proportion of Positive Tests that are True Positive
- Probability that Person with Positive Test has Disease
- TP (True Positives)
- Divided by /
- All Positive Test Results
- TP + FP
- Varies Directly with Prevalence
Negative Predictive Value (NPV)
- Proportion of Negative Tests that are Truly Negative
- Probability that Person with Negative Test is Healthy
- (TN) True Negatives
- Divided by /
- All Negative Test Results
- (FN + TN)
- Varies Inversely with Prevalence
Common Statistical Tests
- Determine If Comparing Numerical Or Categorical Variables
- T-Test For Comparing Means Of Two Groups
- ANOVA For Comparing Means In Greater Than Three Groups
- Comparing Mean Values On Normal/Bell Distribution
- Chi-Squared Test For Comparing Categorical Outcomes In Large Sample Sizes
- Fisher's Exact Test For Comparing Categorical Outcomes In Small Sample Sizes
Scientific Bias Types
- Bias
- Selection Bias
- Berkson Bias
- Attrition Bias
- Recall Bias
- Measurement Bias
- Procedure Bias
- Observer-expectancy Bias
- Confounding Bias
- Lead-time Bias
- Length-time Bias