Bias and Causality

Random error

Error that results from the measurement of data due to imperfections in the measurement system.

Inherent Variability

Natural differences that exist between individuals in a population.

Bias

Systematic errors that disproportionately affect the data in one direction or another.

Internal validity

The methodology design of the study affected by bias (variables, controls, samples, tools, etc.).

External validity

The generalizability of the study affected by bias (whether our sample is reflective of the larger population).

Incidence

The number of new cases during a period of time / by the number of people at risk at the start of the time period.

Prevalence

The number of cases at a certain time / by the total population at that time.

Selection bias

Impacts external validity (can manage by changing the target population) and internal validity (cannot manage, must start the study over).

Selection bias example

People who work are generally healthier than the overall population

Misclassification bias

Measuring things incorrectly, leading to participants being placed in the wrong group.

Misclassification bias example

Putting the exposed or unexposed in the wrong group

Causation

A causes B.

Correlation

A and B share a relationship.

Correlation vs Causation

• Look at surveillance data to determine the issue

• Identify if there are any potential correlations

• Get statistics to help develop causation, and then conduct a controlled study

• Check for association (is the association real or why isn’t it real?)

Good practices for study design

Include making the study short, asking broad questions first, avoiding leading questions, not combining two questions, and including a variety of answers.

Reduce random error

Use statistics to understand error

Minimize bias

Can’t use statistics to account for, can only try to minimize it

Strength

strong correlation between the variables

Temporality

cause comes before the outcome

biological agents

as the level of exposure increases, so does the amount of disease

consistency

relationship is consistent across different studies, populations, times, etc.

specificity

single cause and effect

plausality

cause/effect relationship is biologically reasonable

coherence

relationship is consistent with previous knowledge

analogy

similar relationships observed with similar exposure/disease

experiment

interventions modify outcomes