Epi Midterm Flashcards
Cohort Study
Measures of association/frequency: Risk, Rate, RR, RD, IRR, IRD
Advantages: good for rare exposures, good for multiple outcomes, known temporal association
Disadvantages: time consuming, expensive, not good for rare outcomes, lost to follow up
Prospective: Enrolling a cohort and observing to see who develops the disease
Retrospective: Enrolling a cohort and recording who had or did not have the disease
Intervention Study
Directionality: Forward
Measures of frequency/Association: Risk, Rate, RR, RD, IRR, IRD
Advantages : Causal inference, less opportunity for bias, participants may benefit
Disadvantages: Ethical challenges, expensive, generalizability
Ecological Study
Procedure: ID a set of populations, Classify by exposure, Classify by outcome
Measure of frequency: Correlation
Advantages: cheap if using existing data, Efficient, good if looking at broad trends
Disadvantage: ecological fallacy, missing data on confounders, might use more indirect measures
Good place to start to guide hypothesis
Case Control Study
Measures of association/frequency: Odds, OR
Cannot calculate risk so no RD or PRD
Advantages: Good for rare disease, good for multiple exposures, good for long induction/latent period
Disadvantages: temporality is not always clear, bias in exposure ascertainment since you are often relying on participants to tell you
Equipoise
The point at which an intervention study must stop either bc of discovering harms or advantages of the intervention that rise to a pre-determined level that would make it unethical to continue.
Key Features of Descriptive Epi
Person, Place, and Time
Great for generating hypothesis
Ecological Fallacy
Attempting to draw individual level conclusions based on population based data (mostly a problem with ecological study design)
Ethical Considerations
Comparison Group - who is your comparison group? Are participants in the comparison group being harmed by not receiving the intervention?
Equipoise - is there sufficient evidence that the intervention is not harmful? Is there enough doubt that it will be beneficial?
Sustainability/Feasibility - is the intervention feasible to continue if found beneficial to participants? Is scale-up possible?
PH surveillance
The ongoing systematic collection, analysis, and interpretation of health related data essential to planning, implementation, and evaluation of PH practice, closely integrated with timely dissemination of these data to those responsible for prevention and control.
Analysis of Case Control Studies
Odds and odds ratios are common but not as easily interpreted as a measure of risk or relative risk
Luckily the OR can approximate the RR given one of the following conditions: disease is rare OR you use population sampling for your controls
Selection of Cases
Cases: incident case of disease preferred because etiology is more clear but you can also use prevalent cases
Selection of Controls
Goal: estimate the distribution of the exposure in the general populations
Types of controls: population controls hospital controls neighborhood, family, etc (less common)
Things to consider: Feasibility, How representative
would the control population be of the population the controls came from?
Difference between retrospective cohort study and a case control study
Key differences: retrospective cohort we are determining exposure first and then determining outcome. With case control, we enroll based on outcome status to see the outcome
With retrospective cohort we have underlying population in case control, we don’t which is why we cannot calculate risk
Standardization
One method of adjusting for confounding or biased factors
Used frequently in descriptive epidemiology to allow for comparison between groups
We talk about this bc they use it a lot in descriptive epi (comparing between groups)
Interpretation Shells
Risk Difference –> The excess risk of the outcome associated with the exposure was XX%
Rate Difference –> The excess rate (or risk) of the outcome associated with exposure was XX per X person time
Risk Ratio –> The risk of outcome among exposed was XX times that of the unexposed.
Prevalence Ratio –> The Prevalence of the outcome among exposed was XX times that of the unexposed
Odds Ratio –> The odds of (disease or exposure) were xx times that of the exposed/cases compared to the unexposed/controls.
Cross Sectional Study
Procedure: ID population of interest, ID who has exposure, and ID who does not have exposure.
Measures of association/ frequency: Prevalence, PR
Advantages: Efficient, great for descriptive Epi
Disadvantages: difficult to determine temporarily, can be difficult for rare disease with short duration
Nested Case Control
Measures of Association
Absolute comparisons: Risk difference (RD), Population Risk Difference (PRD), Attributable proportion among exposed (APe), attributable population among unexposed (APtot).
Relative Comparisons: Relative Risk (RR)
Why? To understand the impact of exposures on population health (absolute) and to understand risk factors of health outcomes (relative)
Shoe Leather Epidemiology
You go door to door to get all the info you need
Two types of Cohorts
General Cohort : group enrolled as cohort without regard for exposure our outcome (nurses health study). Group is enrolled prior to classification according to exposure
Exposed Cohort: Enrolls a population based on sharing a particular exposure. Useful for rare exposures
Internal vs External Comparison group
You are comparing within your cohort
If you enrolled based on exposure, you would have to go outside to compare
Prevalence Ratio
(A/(A+B))/(C/(C+D))
from 2 X 2 table
Analysis of Ecological Studies
Cannot make a 2 x 2 table because you are generally working with aggregate (numerical/continuous) measurements. Instead, you use correlations
Positive correlation = r > 0 (exposure & outcome
increase or decrease together)
Negative correlation = r < 0 (exposure & outcome have
an inverse relationship)
Independent = r = 0 (no relationship)
Analysis of Intervention Studies
Risk type intervention study analysis : CI, Cumulative Incidence Difference, Cumulative Incidence Ratio
Rate type intervention study analysis : Incidence Rate (IR), incidence rate ratio (IRR), incidence rate difference (RD)
Intent to treat type analysis: “real world” measure of effectiveness. analyze if everyone followed the treatment regiment (regardless of actual compliance). This helps maintain comparability and statistical power
Efficacy (per-Protocol) analysis: efficacy under ideal conditions. only analyze those who followed the treatment or intervention regiment per protocol. This may help better analyze the effect but reduces comparability and statistical power.
Explanation of study design must include
-What study population are you going to enroll?
-How are you going to ascertain exposure status/assign
intervention?
-How are you going to determine disease outcome?
-How long are you going to follow your participants for?
-Measures of association?
-Advantages?
-Disadvantages?
How do we know randomization worked?
The intervention and control population should match the underlying population.
Measures of Association: Absolute/excess risk
Calculate the difference between a measure of disease frequency in an exposed group and an unexposed group
- This is a measure of excess risk.
- Absolute measures of association provide information about the public health impact of an exposure.
- “Simple Subtraction”
Measures of Association: Relative comparison
Calculate the ratio of a measure of disease frequency in an exposed group and an unexposed group
- -> Relative measures of association provide information about the magnitude of the association, which gives clues about risk factors.
- “Simple Division”
Attributable Proportions
Attributable proportions are measures of public health impact expressed as proportions.
- This is analogous to risk difference (RD) and population risk difference PRD.
Attributable Proportion among exposed
APexp describes the proportion of disease among the exposed population that is associated with the exposure.
This measure converts absolute risk to a proportion.
APexp = Rexp - Runexp = RD
Rexp Rexp
Attributable Proportions among total population
APtot describes the proportion of disease among the total population that is associated with the exposure.
- It answers the question: What percent of disease in total population is due to exposure?
- APtot is useful for setting priorities for public health action.
Rtot = risk (IR, CI) among total population Runexp = risk (IR, CI) among unexposed population
Measures of Frequency
Goals
- To measure the distribution of the disease
- For assessing burden
- For comparing between groups to ID disease determinants
Key Components
- # of people affected by disease/health state
- Size of source population
- Time
Source Population
Populations: groups of people with common characteristic that give rise to cases
Ex. of defining characteristics
- Geographical location
- Age, sex, ethnicity/race
- Occupation
- Life event
Can be FIXED (membership is permanent) or DYNAMIC (membership is transient defined by being in and out of our state)
Incidence
Measure of NEW disease developed in a population AT RISK over a particular time interval.
3 Key components of “at risk”
- At risk people must be disease free at the start of the study
- For those transitioning from health to diesase, there must be an element of time
- Two main types: Cumulative Incidence (CI) and Incidence Rate (IR)
Cumulative Incidence (CI)
Estimates the probability (risk) that someone will develop the disease during a specified period of time
Assumes that the entire population was followed the entire time
Is expressed as a proportion
Easier to interpret and calculate than IR but requires certain assumptions for accuracy
Incidence Rate (IR)
Occurrence of new cases of disease that arise during person time of observation
Time is the denominator (therefore units) meaning that it is not a proportion but a true rate
Does not assume that everyone was followed for the whole time
- people can contribute individual person time
- good for dynamic populations or those where a lot of people where lost to follow up
- Only person time under observation contributes to person time
Harder to calculate and less intuitive than CI. However the accuracy does not depend on the same assumptions. Most often reflects the real world in which we work with dynamic populations
Mortality Rate
= total # of deaths/ total person time of follow up
Not a true rate because …
Prevalence
Measure of current disease in a population (these are EXISTING cases) Prevalence estimates are influenced by duration which is affected by treatment and morality.
It is a proportion because the denominator is a subset of the numerator. Estimates the probability that an individual in a given population has the disease.
Answers: What is the proportion of the population affected by the disease at that specific time?
Point and Point prevalence
Time can be a single point or an interval
Point prevalence: those with the disease at a single point in time.
- includes people with different durations of disease
- excludes people who had the disease but were cured and people who died of the disease.
Period prevalence: those with the disease during a specified duration of time.
- Includes new and existing cases at start of the study
- includes people with disease at any point during the duration
Comparing Incidence and Prevalence
Prevalence: can be useful for administrative PH activities
- Impact of disease on community (burden)
- Project medical care needs
- Often easier to obtain this info
- affected by incidence and duration meaning it is less helpful for risk factor research.
Incidence: may be useful for causal inferences (risk)
- More useful for risk factor research and evaluation of prevention efforts
- Observing transition from health to disease
- Closer in time to events, not mixed with duration, survival, treatment effects.
- Harder to obtain
Passive Surveillance
Pros
- Nationwide
- Medical record verified immunization status
- Can be very simple (EMR-linked)
Cons:
-Reporting quality and frequency
varies
- Exact system varies state to state
Active Surveillance
Pros
-Allows more complete and
rigorous collection of
immunization status
Cons
- Very resource intensive
- Representative, but still limited
Sentinel Surveillance
Surveillance by a selected set of health professionals in a geographic area or specific reporting group
Examples - Gonococcal Isolate Surveillance Project, EIP
Syndromic Surveillance
Surveillance based on a set of symptoms rather than a diagnosis or test result
Examples - influenza like illness, COVID-19 and multi-system
inflammatory syndrome in children (MIS-C)
Surveillance System Process
- Define goals and scope
- Collect data
- Analyze and interpret results
- Disseminate (link to action)
- Evaluate
What is the burden of a disease?
Not a specific epidemiological term and does not have a specific calculation or equation associated with it.
Based on class: generally means we should comment on some measure of frequency
Crude Rates
A summary measure calculated by dividing the total # of cases in population by total # of individuals in population for specific time period
Problem with comparing crude rates between populations
-if people differ with respect to underlying characteristic that affects overall rate of disease (e.g. age)
Category Specific Rates
Rates specific to some sub population (e.g. race, gender, or age)
Comparing age specific rates
How to decide if surveillance is needed
The PH importance of the problem
- Severity and prevalence/incidence of disease
- Mortality
- Public perception of the concern
Ability to control/prevent problem
-Preventability and control measure/treatment
Capacity of health system to implement control/prevention methods
Evaluating Surveillance System’s Attributes
Usefulness : How useful is the system in accomplishing its objectives?
Data Quality: How reliable are the available data? How complete and accurate are data fields in the reports received by the system?
Timelines : How quickly were reports received?
Flexibility : How quickly can the system adapt to change?
Simplicity : How easy is the system’s operation
Stability : Does the surveillance system work well? Does it break down often?
Sensitivity : How well does it capture the intended cases?
Predictive Value Positive: How many of the reported cases meet the case definition?
Representativeness: How good is the system at representing the population under surveillance
Acceptability : How easy is the system’s operation?
Informatics : ?
Methods for active follow up
Collecting Sufficient locating information at baseline
Maintaining regular contact with participants
Making participation worth while for participants
Absolute Comparisons
- -Based on the difference between 2 measures of frequency
- -Comparing disease occurrence among exposed with the disease occurrence among the unexposed group by subtracting from one another
Interpreted as…
- -Excess risk or burden of the disease in the exposed group compared to unexposed group
- -The amount of disease associated with exposure
Relative Comparisons
Based on ratio of two measures of frequency (i.e. risk ratio or RR)
Risk difference
(on formula sheet)
– Alternate names: Cumulative Incidence Difference, Incidence Rate Difference, and prevalence difference (sometimes called attributable risk but also not preferred)
Interpretation: The excess risk of disease among the exposed compared to those who where unexposed is XX%
Excess # of Cases in Population
(on formula sheet)
Nearly X of the X number of cases of disease among those exposed is associated with the exposure.
Population Risk Difference (PRD)
(on formula sheet) –> Also called population attributable risk
measure of excess disease risk among total population that is associated with exposure
Describes the impact of exposure on total population
Can estimate the proportion of cases in the total population that are in excess due to exposure
Depends on the prevalence of exposure in population
Interpretation: the excess risk associated with exposure in the total population is X%
Attributable Proportions (AP)
(on formula sheet) Also called etiological fraction, attributable risk percent, attributable risk percent among exposed
Measure of public health impact as expressed by proportion among the exposed and total population
Interpretation: XX% of cases among the exposed is associated with the exposure.
Attributable Proportion among the total population
(On formula sheet) Also called population attributable risk percent
describes proportion of the disease among the total population associated with the exposure OR what % of the disease is associated with exposure among total population.
Useful for setting priorities for PH action
Interpretation: X% of cases in the total population is associated with exposure
Relative risk (RR)
(on formula sheet)
Tells us the magnitude or strength of the association/ Also gives information on the relative effect of the disease
Measure of strength or magnitude of association between exposure and outcome./ Tells us how many times higher or lower the disease outcome is among exposed compared to unexposed.
Dimensionless so no units that range from 0-infinity
Interpretation: People who where exposed had X times the risk of disease over some time period compared to the unexposed.
Excess Relative Risk
ERR= (RR-1) *100
Those exposed are at a XX% increased risk of disease compared to those unexposed.
Age-adjustment or standardization of rates
This answers the question of what would the death rates be in each group if the populations had identical age distributions?
Information needed: Age-specific rates (deaths and population size). Age distribution of standard population (i.e. US. Population in a given year…)
Computation of Age Adjusted Rate
Weights = to proportion of the standard population in each category.
Age adjusted rates are the mortality rates if each population had the same age distribution.