Epidemiology 2 Flashcards
Define
Observational Studies
- Uncontrolled studies/natural experiments
- Viewed as less definitive than clinical trials
- By the end of experiment, cohort might flake → your sample size will be lower
What can/can’t you control in observational studies?
- Only thing you control is finding the boundary in which you will observe
- Cannot control certain factors (or confounders)
List
Types of Observational Studies
3
- Cohort studies
- Case-control studies
- Cross-sectional studies
Define
Cohort Study
- Starts with exposed & non-exposed group, both w/o disease
- Follow groups to observe disease incidence or mortality rates
- How many people develop the disease based on who are & who are not exposed?
In Cohort Studies, how are disease rates in both groups compared?
rate ratio or rate difference(??) lol
List
Types of Cohort Study
2
- Retrospective (past to present)
- Prospective (present to future)
Define
Retrospective studies
- Faster & cheaper
- Getting people now, and asking their past lifestyle
- Info on exposure & confounders from historical information
Define
Issues w/ Retrospective Studies
- Bias, will they tell you accurately? Or too much/too little/incorrect info?
- Di ka sure if your medical records are accurate
Define
Prospective Studies
- Take longer, more expensive
- Measure exposure levels & confounding variables at baseline
- Good when biological samples are required
- Unreasonable to keep historical biological samples
- To study diseases difficult to ascertain in retrospect
Define
Issues w/ Prospective Studies
- Lifestyle can change throughout the study: Takes longer, can be multifactorial
- Inconsistent statistically
- Tip: get too much at first, para kung umonti u arent screwed
List
What can cohort studies consider?
2
Cumulative incidence or risk (disease events per person)
* This person had disease event, this person didn’t
Incidence rates or mortality
rates (disease events per person per time)
* Over a month, x people had this disease
Cohort studies are good for ____ exposures & ____ diseases
rare exposures & common diseases
Define
Case Control Studies
- Start: diseased & non-diseased groups
- Look backward in time
- More subject to bias than cohort
- Case: has disease
- Control: no disease
- Compares odds of exposure in each group
Define
Recall bias
- Cases recall past exposures more often than controls
- typical of retrospective exposure assessment
- some people may recall more, some less
- no such thing as perfect recall
- Results biased towards finding association
Case Control studies are good for ____ exposures & ____ diseases.
why?
common exposures & rare diseases
- Since ure starting with a bias looking into the rare disease, might as well look into those patients
- Cohorting a rare disease needs a huuuge cohort
- Unicorn disease: hard to catch
Proportion Exposed vs Odds of Exposure
Proportion exposed = a / (a+b)
Odds of exposure = a / b
a is no. of exposed
b is no. of non-exposed
What conclusion can we draw when comparing odds of exposure and control?
odds of exposure > control → exposure is associated with disease
Describe
Cross-sectional Studies (Prevalence Studies)
- Measures exposure & disease at the same time
- Done when outcome of interest is subclinical/asymptomatic disease
- Difficult to determine whether exposure DID precede disease
- looking at 1 point in time & getting samples (bio & envi) in that point in time
- Tend to be weaker in design (compared to cohort & case-control)
Define
subclinical disease
illness that is staying below the surface of clinical detection
- “under” the clinic, Clinic doesn’t see it bc you don’t go to the hospital bc it’s not that bad
- Ex. blood has the lead due to low exposure, but won’t show disease associated with high exposure (neurotoxicity)
Relationship bet. Cross sectional studies vs. cohort or case-control studies
Cross-sectional is Often only possible design
* Can be confirmed in cohort or case-control studies
* Often first step before cohort or case-control studies
Define
bias
Distortion of the true relationship bet. exposure & disease
List
Kinds of bias
3
Selection Bias
Confounding Bias
Information Bias
Define
Selection bias
- Selecting a study population not representative of general population
- Cannot be corrected during analysis
- Easy to do, but difficult to correct
- Once you selected your sample, kahit anong sample test won’t explain why you selected your samples
Describe
Healthy Worker Effect
- Workers compared to general population
- Results based against finding adverse health effects among workers
- Typically ‘labour’ work (physical work)
- What representation means may be different based on what you’re looking at
Explain
how did the study on ethylene oxide & breast cancer obtain selection bias?
- Only 20% of individuals in target pop answered questionnaire on how breast cancer occurs
- Bc of how they designed where/how the interview will be, nagkakaselection bias agad
Describe
Confounding Bias
Distortion of exposure-disease relationship by a third variable association with both exposure & disease
* Ex. Age, smokers
- Adjustment for effect can be made during analysis
- Might show association even when there isn’t
- Influences both independent and dependent variables – will show association even if there isn’t
- Even if you select your cohort really well, there can still be confounding bias – important to always talk to your cohort
Describe
How does Confounding Bias impact results?
- Might show association even when there isn’t
- Influences both independent and dependent variables – will show association even if there isn’t
- Can change positively or negatively(???? what)
How can the error of confounding bias be adjusted during analysis?
Stratification of groups or multivariate analysis
* Can see odds ratio/incidence go down when your remove a confounder
However, you need to actively correct it
* Be aware that you have a confounding variable present
* pag di mo alam na may confounding variable, edi di mo maadjust sa stat
Define
Effect modification
Third variable modifies the effect of the exposure variable of interest
Define
Information Bias
- Occurs when information obtained about either exposure or disease is incorrect
- Mismeasurement or misclassification
How does information bias impact results?
Towards what
Usually to bias results towards no association
List
Kinds of Data Analysis Variables
2
- Categorical variable (or dichotomous variable)
- Continuous variable (on a spectrum)
How do we analyze data when both exposure & disease variables are categorical?
- Calculate rate ratio & odds ratio
- Can be stratified to control for confounding
How do we analyze data when both exposure & disease variables are continuous?
Regression analysis
What are the implications of regression (R)?
What is an acceptable R?
R coefficient significantly different from 0 = exposure is significant predictor of disease
- Acceptable R: usually +- 0.8, depends on nature of study
- Minsan nga .5 relationship is strong enough eh (wtf)
How do we analyze data when exposure & disease variables area mix of cat & cont?
Linear regression for outcome
* Can be used w/ categorized exposure in regression
Logistic regression - measure of interest is odds ratio
* Even when disease variable is categorical
* Either categorical or continuous variables can be included among disease predictors
Why does epidemiology use a lot of regression analysis?
kasi you’re always looking at correlation, and regression analysis shows the quantitative relationship bet 2 variables
____ sample sizes = ____ statistical power
Large sample sizes = greater statistical power
Define
Confidence Interval
What is the usual CI set?
Range of plausible values for measure of effect
Usually 95%, if more sure edi 99%
Idk this maica edit before long exam
Precision of estimate = reflection of random error
* Error likely to result form choosing sample of total population of interest
* Error from biases etc.
Why does epidemiology have the risk of being less precise/accurate?
How do we account for the lower precision/accuracy?
nature of data collection makes precision difficult
* So your stat has to remove/control your data
* But also don’t overly control your data
List
Other Aspects of Epidemiology
2
- Environmental Epidemiology
- Occupational Epidemiology
What does Environmental Epidemiology include and exclude?
Includes environmental agents
* Large number of people are exposed involuntarily
* Looks at explicit connection w/ environment
Excludes voluntary exposures
Describe
Environmental Epidemiology exposure
- Exposure is usually low level and homogenous
-
Difference in risk bet. more & less exposure is usually small
* Needs large sample size to see the small differences
Differentiate
Pandemic vs Epidemic vs Endemic
Pandemic: global
Epidemics: unusual outbreaks of disease significantly above normal levels
regional
* Sudden urge in incidences
* Often caused by known agents
Endemic: exist at constant, low or bg levels
* Always there
* COVID is endemic: it’ll be there for a long time
Define
Occupational Epidemiology & nature of its exposure
Studies of illness or iniury associated w/ workplace exposures
Relatively high exposures in relatively small numbers of people
* Often geographically isolated in a worksite
How were we able to discover carcinogens in Occupational Epidemiology?
Historical studies - needed time to identify long term effects
T/F
Current workplace exposures to suspected toxins lower than in past
T
What is the goal of measuring exposure?
- Detecting and quantifying dose-response relationship
- Support a causal relationship
What do we measure when measuring exposure?
- External exposures (the actual thing)
- Biomarkers (ex. Proteins produced by antigens)
What is the impact of a mismeasured exposure?
- Skews true dose-response to a weaker(?) relationship
- Which is why we need to understand which biomarker to get to definitively measure exposure & disease
What is the impact of Misclassification of an exposure?
- severely bias toward null hypothesis (no relationship)
- Part of information/selection bias
Describe
Job-Exposure Matrix (JEM)
Cross classification of jobs and exposure levels across time
How are jobs on JEM categorized?
Categorized based on presumably same exposure level
* Workers in all jobs in one category assigned same exposure level at any given point in time
Describe
Biomarkers
Some are measure of internal dose
* Can account for variation in absorption and metabolism
* Estimate of biologically relevant dose
T/F
Many Biomarkers persist long enough for use in retrospective studies
Why/why not?
F
few persist long enough for use in retrospective studies
* Bc we can metabolize hooray
* Ex. opioid addict who doesnt take opioids for a week will pass the drug test
Define
Quantitative Meta-analysis
- Combine results from different study designs
- Provides weighted average of results across studies
- Studies with the most weight = lowest variance
- Do not need access to raw data (can access from other papers)
Define
Pooled Analysis
- Raw data from several studies obtained and re-analyzed
- More time consuming
- More flexibility for analysis
Differentiate
Quantitative Meta-analysis vs Pooled Analysis
- Quanti: combine results, weighted average lang (easier, no need for raw data)
- Pooled: literally re-analyze raw data (more time-consuming, but also more flexibility for analysis)