Exam 1 Material Flashcards
Crude morbidity rate
diseased people / # total population
Crude mortality rate
deaths (all causes) / # total population
Incidence / risk/ attach rate
of new cases of a disease / # of people at risk for that disease
Prevalence
(# of existing + new cases of a disease) / total population
~morbidity
3 factors of descriptive epidemiology
Who / When / Where
Passive surveillance system
Relies on public healthcare system to bring in information by following regulations concerning required reportable disease
Must passively wait for information to be collected
Active surveillance system
Public health officials go into communities to actively search for new disease/condition cases
Syndromic surveillance system
Officials look for pre-defined signs/symptoms related to a trackable but rare disease/occurrence
CON: can be non-specific
Biosurveillance
Mix of passive surveillance and syndromic surveillance
Look for specific symptoms/signs in collected data for a population
Induction / incubation
Time between exposure and onset of symptoms of a disease
Latency period
Time between onset of symptoms of a disease and disease detection (via symptoms or diagnosis)
What is the most critical element that must be defined before any descriptive epidemiology information can be acquired?
A case definition
Why are case definitions important?
They define how a disease is detected/diagnosed and are guidelines for how we can accurately verify (confirmed or probable) cases of a disease
Epidemic
A “more than normal” occurrence of disease
Outbreak
A localized epidemic; limited by size of geographical area or population
Endemic
The constant presence of disease within a given area or population; the “normal” amount of disease
Pandemic
A world-wide/multi-national/multi-continent epidemic
Emergency of international concern
An epidemic requiring high vigilance
Pre-pandemic labeling
Cause-specific morbidity
of diseased people with a specific disease / # total population
3 factors in comparing measures of disease frequency between groups
- # of people affected
- # of people at risk
- Period of time that people are followed
Proportion
Part / whole
Division of 2 related numbers
Ratio
Whole / whole
Division of 2 unrelated numbers
Rate
Part / [whole*time]
A proportion over time
Cumulative incidence
Sum of incidence over (multiple time periods?)
Incidence density
Sum of incidence rate (over multiple time periods?)
Cause-specific mortality
deaths specifically from a disease / # total population
Case-fatality rate
deaths specifically cause by a disease / # cases of that disease
Cause-specific survival rate
survival cases from disease / # cases of the disease
Proportional mortality rate
cause-specific deaths / # total deaths in population
Live birth rate
live births / 1,000 population
Fertility rate
live births / 1,000 women who are fertile
Neonatal mortality rate
neonatal deaths / 1,000 live births
Neonatal: <28 days of age
Postnatal mortality rate
postnatal deaths / 1,000 live births
Postnatal: 28days < x < 1year of age
Infant mortality rate
infant deaths / 1,000 live births
Infant: <1yr of age
Maternal mortality ratio
maternal deaths / 100,000 live births
Maternal deaths: any death relating to pregnancy
Counter factual theory
It would be ideal, but impossible, to have the same person placed in different study groups for a disease.
In order for disease to be studied, we assume exchangeability (comparability) between subjects - that populations can generally be standardized, or made equal, in all else except the exposure.
Absolute differences
The highest difference between frequencies
A subtraction between 2 frequencies
Relative difference
Division between 2 frequencies or proportions
Typically larger than absolute differences; better for pharmacy
Probability outcome / incidence risk (IR)
Percentage / proportion of risk
Absolute risk reduction (ARR)
Risk difference attributed to exposure
A subtraction
Relative risk reduction (RRR)
(ARR) / R unexposed
Division of the difference between 2 risks by a baseline risk
Number needed to Treat (NNT)
Number of patients you need to treat before 1 person gets the benefit; smaller NNT (larger ARR) means less people need to be treated in order for a benefit to occur.
1 / ARR
Risk Ratio
Risk of outcome (exposed pop.) / Risk of outcome (non exposed pop.)
=1 … outcome is equally likely to occur
>1 … outcome is more likely to occur in comparison group (exposed)
<1 … outcome is less likely to occur in comparison group (exposed)
How do you interpret ratios for measures of association?
- Find difference from 1.0
- Convert difference to percent
- Interpret difference
What are 3 things to look for in interpreting measures of association?
- Comparison groups (comparing x to y is x / y)
- Direction of comparison (increase, decrease, _ times greater, etc.)
- Magnitude (# value)
Confounding
Type of bias where there is a distortion of the association between an exposure and outcome
What 3 aspects must be studied before inferring a real, true association?
Confounding and effect modification / Bias / Statistical significance
3 aspects of a confounder
- Independently associated with the exposure
- Independently associated with the outcome
- Not linked to causal pathway between the exposure and outcome
Steps for calculating presence of confounding variable or effect modification (*difference between the two)
- Calculate CRUDE value
- Calculate for individual strata of the proposed “confounder” (adjusted value)
- Compare (look for 15% difference)
Confounding variable compares CRUDE vs ADJUSTED
Effect modification compares between different strata (ADJUSTED)
What effects doe confounder have on a value?
Can change magnitude (strength) of association
Can change direction of association
Ways to control for confounding at study design stage?
Randomization
Restriction
Matching
Ways to control for confounding at the analysis of data stage?
Stratification
Multivariate statistical analysis
Randomization
Way to control for confounding in study design stage
Ideally equal allotment of subjects with known confounder (and unassessed) into each group
S-good if large population
W-bad if small popultion; used mainly for interventional studies
Restriction
Way to control for confounding in study design stage
Study participants are restricted based on if they present a confounding variable
S-no impact on internal validity
W-negatively impacts external validity (limits generalizability)
Matching
Way to control for confounding in study design stage
Subjects are selected in matched-pairs in relation to the confounding variable
S-greater analytical efficiency
W-difficult to do; can mask effect of other cofounders
Stratification
Way to control for confounding in analysis of data stage
Separating out strata bases on confounding variable
S-enhances understanding of data
W-impractical for simultaneous control of multiple confounders
Multivariate analysis
Way to control for confounding in analysis of data stage
Mathematically factoring out effects of confounding variable
S-simultaneous control of multiple confounding variables; interpret OR
W-time consuming; researcher must have good understanding of data
Effect modification
A 3rd variable that modifies the effect of a true association by variations within strata
If present, must be described at each level of the strata
Bias
Systematic, non-random error in study design or conduct that distorts an association
Cannot be “fixed” once study has been complete
3 elements that bias impact
- Source/type
- Magnitude/strength
- Direction
Selection-related bias
Bias as a result of the way in which researchers select their study subjects, thereby creating systematic differences between the groups
Measurement-related bias
Bias as a result of the way in which data is collected/observed that would lead to systematic difference between the groups
Can be subject, observer, or screening related
Healthy worker bias
A form of selection bias in which subjects are taken from a population restricted to healthy individuals (ignores the sick)
Easily seen in prospective Cohort studies
Self-selection/participation bias
A form of selection bias in which voluntary involvement dictates sampling group; there may be a difference between subjects that volunteer and those that do not wish to volunteer
Recall (reporting) bias
A form of (subject-related) measurement bias in which subjects memory/recollection impacts study results; diseased subjects tend to have a higher sensitivity for recollection or may exaggerate
Hawthorne (observation) effect
A form of (subject-related) measurement bias in which study subjects act differently from how they would normally because they know they are part of a study
Contamination bias
A form of (subject-related) measurement bias in which subjects in the control group are exposed to the exposure being studied
Compliance/adherence bias
A form of (subject-related) measurement bias in which subjects do not adhere to the protocols of the study
Lost to follow-up bias
A form of (subject-related) measurement bias in which subjects are lost track of or drop out of a study; aka. Differential attribution bias
Ex. Subject moves away
Interviewer bias
A form of (observer-related) measurement bias in which researchers solicit, record, or interpret data unequally;
includes how the treatment is administered and how research treat subjects during data collection
Diagnosis/surveillance bias
A form of (observer-related) measurement bias in which the researcher evaluates data while being affected by preconceived notions
“Hawthorne effect” for researchers
Lead-time bias
A form of (screening-related) measurement bias in which early detection causes beneficial outcome, not necessarily the exposure
Misclassification bias
A source of measurement bias in which exposure or outcome/disease status is erroneously attributed
Can lead to differential or non-differential error
Non-differential bias
Misclassification bias that lessens the association between variable so that the OR/RR/HR are closer to (attenuates to) 1.0
Error is in both groups equally; misclassification shifts value to make exposure seem UNRELATED to outcome
Differential bias
Misclassification bias in which association is amplified
Error is in one group over another; misclassification shifts value to make exposure seem even more RELATED outcome
Controlling for bias
- blinding/masking
- using multiple sources to gather info
- random allocation of observers/interviewers
- minimize loss-to-follow up bias
Cause
Precursor event required for the outcome of a disease
3 types of associations
- Artifactual (false)
- Non-causal
- Causal
Artifactual (false) association
An effect of confounding and effect modification
Non-causal association
A. Disease may cause the exposure, instead of other way around
B. Disease and exposure both affected by 3rd factor
Causal association
Exposure causes the outcome
Sufficient cause
A causal association in which a minimum condition leads to a disease
Necessary cause
A causal association in which the cause must be present for the disease to occur, but can also exist without leading to the disease
Ex. TB
Component cause
A causal association in which the cause increases the likelihood of disease; aka risk factor
Synergism
When factors work together so that the combined measure of effect is greater than the sum of their individual measures of effect
Parallelism
When factors interact so that the measure of effect is greater if either is present
Explain Hill’s Criteria
Set of guidelines that help determine if an exposure is a CAUSE for an outcome (strong association)
- Strength = large magnitude/value means stronger association
- Consistency = high reproducability means strong association
- Temporality = timing of effect (cause precedes outcome)
- Biological gradient = more of an exposure leads to more of an outcome
- Plausibility = does the relationship make physiological/biological sense