Epidemiology Glossary

A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z

A

Absolute Risk Reduction (ARR):

Absolute difference in the rate of events between the control and intervention group.
ARR = events in control group minus events in intervention group

B

Bayes Theorem:

The estimate of the likelihood of a disease is increased or decreased by the results of a test.

Pre-test odds of a disease x Likelihood Ratio = posttest odds of a disease
See nomogram for applying likelihood rations.

Bias:

A systematic error that leads to results that do not represent the true findings. See lead time bias and length bias.

Blinding:

Single Blind Study Design:
The subjects do not know whether they are in the placebo or treatment group.

Double Blind Study Design: Neither the subjects nor the researchers know whether the subject is in the placebo or treatment groups.

C

Case-control Studies:

Retrospective studies beginning with people who have a disease and controls who are similar but do not have the disease. The investigator then looks retrospectively to see what proportion in each group were exposed to the same risk factors.

case control study formula

Strengths Include:	Weaknesses Include:
useful for study of rare diseases	participants may have a recall bias
less expensive	selection bias possible in choice of controls

Findings expressed as Odds Ratio.

Causality:

Determination of causality

Suspected cause precedes disease.
Association is strong.
No likely non-causal basis for the association.
Association makes biological sense.
Magnitude of association is strongest when it is predicted to be so

Increased Exposure = Increased Risk
Increased Susceptibility = Increased Risk

Cohort Studies:

Observational studies, prospective or retrospective, in which groups are assembled according to exposure or lack of exposure, then followed longitudinally to determine outcomes (e.g., fractions who develop a disease).

formula

Strengths Include:	Weaknesses Include:
direct measurement of incidence	cohorts may differ at baseline and, therefore, relies heavily on multivariate analysis to adjust for potential confounders
able to note temporal relationship between exposure and outcome	expensive

Findings expressed as Relative Risk.

Confidence Interval (CI):

Usually calculated as "95% confidence intervals", indicating that there is a 95% probability that the effect of treatment in the whole population lies within the stated range. The CI is affected by sample size and by variability among subjects

Confounders:

Patient characteristics that may affect the results a study is trying to measure. Because these factors may be unevenly distributed (non-random) between study groups, they can decrease the validity of the study.

Cost-effectiveness:

Evaluates the outcomes and costs of interventions intended to improve health. Often framed as the resources required to achieve a set quantity of a specific health outcome.

D

E

Effectiveness:

The ability of an intervention to achieve the desired results under usual conditions.

Efficacy

The ability of an intervention to achieve the desired results under ideal conditions.

F

G

H

I

Incidence

The number of new cases of disease within a given time period.

J

K

Kappa Value:

Kappa value is a chance-corrected measure of agreement between pairs of observers. It reflects the degree of agreement for a particular physical finding. In general, a high level of agreement occurs when kappa values are above 0.5. Agreement is poor when kappa values are less than 0.3.

See formulas for calculating kappa.

L

Lead Time Bias:

Apparent lengthening of survival due to earlier diagnosis in the course of disease without any actual prolongation of life.

Length Bias:

Bias due to the tendency of screening to detect a larger number of cases of slowly progressing disease and miss aggressive disease due to its rapid progression.
length time bias

Likelihood Ratios:

The odds of disease given a specified test value divided by the odds of disease in the study population. The odds that a given finding on history or physical examination would occur in a patient with the target disorder as opposed to a patient without the target disorder.

See nomogram for applying likelihood rations.

Positive Likelihood Ratio:
What is the proportion of patients with a target disorder who have a positive test compared with the proportion of healthy patients who have a positive test?

Mnemonic:
WIWO: The fraction of patients with a disease who have a positive test divided by the fraction of patients without a disease who have a positive test.

indicates how much the odds of disease is increased if the test result is positive.
the ratio of something that clinicians do want in a test (sensitivity) divided by something they do not want (false-positive error rate).

How big is a big LR+?

1.0	The test is useless
1.0-2.0	Rarely important change from pretest to post test probability
2.0-5.0	Small change
5.0-10	Moderate change
>10	Large change

Negative Likelihood Ratio:
What is the proportion of patients with a target disorder who have a negative test compared with the proportion of healthy patients who have a negative test?

Mnemonic:
WIWO: The fraction of patients with a disease who have a negative test divided by the fraction of patients without a disease who have a negative test.

formula

shows how much the odds of disease is decreased if the test result is negative.
the ratio of something clinicians do not want (false-negative error rate) divided by something they do want (specificity).

How small is a small LR-?

0-0.1	Large change from pretest to post test probability
0.1-0.2	Moderate change
0.2-0.5	Small, but sometimes important changes in probability
0.5-1.0 Rar	Rarely important change

M

N

Negative Predictive Value:

See Predictive Value

Number Needed to Treat (NNT):

Number of people needed to treat to prevent one bad outcome.

formula for number needed to treat

O

Odds Ratio:

To calculate a, b, c, d: see the 2x2 table

P

Predictive Value:

For both PPV and NPV, use table for calculations:
2x2 table

Positive Predictive Value (PPV):
Proportion of persons with positive test who have condition.
formula=true positive divided by all positive

Negative Predictive Value (NPV):
Proportion of persons with negative test who do not have condition.
formula=true negative divided by all negative

Prevelance

The number of cases of disease in a population at a given time (frequency).
formula = all patients with disease divided by all patients tested

Use table to calculate:
2x2 table

Probability:

Pre-Test Probability:
The prevalence of disease in a specified group of subjects. While the overall prevalence may be known, for each individual the disease status may be unknown before a diagnostic test and is, thus, call the pre-test probability.

Post-Test Probability:
The probability of a disease after a given test result. Its value depends on the pre-test probability of the disease and the test's sensitivity and specificity. See Positive Predictive Value and Negative Predictive Value

Q

R

Randomized Control Studies:

Experimental study where people are randomly assigned to an intervention or control group. the randomization refers to allocation among study groups, not to eligibility for the trial.

randomized control study formula

Strengths Include: Weaknesses Include:

randomizing makes group comparable for even unknown factors may not be appropriate for ethical or practical reasons

double blind design minimizes observer expensive

Findings expressed as Relative Risk.

Recall Bias:

Subjects may have inaccurately recalled past medical history or previous exposures. This is especially a concern in retrospective studies in which the presence of a disease (or a particular outcome) may bias subjects' recall of their exposure to a putative risk.

Receiver Operator Curves (ROC):

A graph plotting the true positive rate (1-specificity) over a series of cutoffs for defining a positive test. A diagonal line indicates no ability to distinguish persons with and without the condition. The farther the curve reaches toward the upper left corner of the graph the better the test in discriminating diseased from non-diseased persons. ROC curve

Historical note: ROC are derived from British radar operators during WWII. The goal was to find the correct technique to identify all incoming planes without giving too many false warnings. By plotting the sensitivity and specificity cutoffs of the more and less successful radar operators, the best combination of sensitivity and specificity could be determined.

Relative Risk:

To calculate a, b, c, d: see the 2x2 table

Reliability:

The ability of a test to obtain the same results under the same conditions.

S

Sensitivity:

Proportion of persons with condition who test positive

sensitivity formula = true postive test results divided by all patients with disease

To calculate a and c, see the 2x2 table

Specificity

Proportion of persons without condition who test negative:

specificity formula = true negative test results divided by all patients without disease

To calculate b and d, see the 2x2 table

Surrogate Marker:

An alternate marker used in studies as a replacement for a true clinical outcome. Recent studies highlight the possible discrepancies between intermediate and endpoint markers. For example, eccanide and fleccanide, anti-arrhythmic medications, were shown to reduce ventricular arrhythmia (surrogate endpoint) but increase mortality (true clinical outcome).

T

Table ("2x2 table):

Goal of epidemiologic investigation

2x2 table

See Relative Risk, Odds Ratio, Sensitivity, and Specificity for formulas.

U

V

Variability

Intraobserver Variability:
Variation in test results during repeat testing by the same observer.

Interobserver Variability:
Variation in test results by different observers. The variability is often quantified with the Kappa value.

W

X

y

Z

A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | Top