epi
Data types
Categorical: male vs. female, ethnicity, SES, age group, educational level Continuous: range of values (height, weight, physical activity level)
Descriptive variables are: a. Collected when epidemiologists ask questions of their study subjects. b. Most often biased and have to be taken as such. c. The outcome of interest in the context of model building. d. Measures of Relative Risk (RR). e. p-values.
Collected when epidemiologists ask questions of their study subjects
Reversibility
Does removal of cause lead to reduction in risk? -yes=strong causal relationship -no= question relationship OR seek alternate pathways/explanations ex. smokers who give up smoking have a lower risk of lung canc than those who continue to smoke
Temporality
Does the cause (exposure) precede the effect (outcome)? -Timing, cause must precede effect -post hoc fallacy: simply because one thing happens after another, the first even caused the second, accidents DO in fact exist ex. lower rate of injuries in drivers after seat belts became law
Epidemiologists typically use descriptive studies to: a. Reject null hypotheses having to do with exposure-outcome associations. b. Generate hypotheses. c. Examine potential biases having to do with data entry into computer programs. d. Refute causal models. e. Teach public health practitioners how to perform analytical studies
Generate hypotheses
The main purpose of this diagram is to: a. Help visualize the possible relationship between women who may have taken cocaine prenatally and the ability of their offspring to perform on a neuropsychological exam during the early school years. b. Demonstrate, based on analytical designs, that there is indeed a strong association between children who perform poorly on a standardized neuropsychological exam and their gestational age. c. Assess the linkage between birth weight and birth length. d. Test covariates that are not included in the model. e. Assess if the proposed model is backwards; that is, if reverse causality is really at work and the outcome of interest should really be the exposure.
Help visualize the possible relationship between women who may have taken cocaineprenatally and the ability of their offspring to perform on a neuropsychological exam duringthe early school years.
Judging the evidence
How many lines of evidence lead to conclusion? -ex. diabetes is caused by many lines, like obesity, BMI, genetics, age, ethnicity etc.
Inductive logic was discussed in the context of: a. Hypothesis generation b. Hypothesis testing c. Negative feedback d. The self-reinforcing nature of feedback perceived by the general public. e. Hypothesis refutation
Hypothesis generation
Study Design
Is the evidence based on strong study design? -sample size? enough power? sample composition? reference group/controls? anonymity?
Plausibility
Is this association consistent with existing biological/medical knowledge? -consistent with other knowledge increases plausibility ex. relationship between obesity and skeletal health? we know it is hard on joints so can "assume" it is plausible
Consistency
Persistent association between exposure and outcome from multiple studies of adequate power in different persons, place, circumstances, times. -similar results in other studies -greater emphasis should be given to best-designed studies -ex. tobacco smoking and lung cancer, many different studies on this topic backing it up
Odds Ratio Calculation
Table: a. b. a+b c. d. c+d a+c. b+d. a+b+c+d OR= (a/c)/(b/d) =axd/bxc Upper left to lower right multiply Upper right to lower left multiply divide first over second
Population and Samples
Target population: world, country, culture Study population: students, residents of place, people of occupation Study sample: active participants in the study
Strength
What is the strength of the relationship between the potential cause and effect? -strong association is more likely to be causal than a weak association -strong=statistically significant (p-values) ex. risk of lung cancer is 20x higher in smokers than non-smokers -RR: probability of lung canc for smokers/prob of lung canc for non smokers
Incidence
"Risk" # of new health-related events in a defined population within a specific period of time -measure via frequency count, rate, cumulative ex. percentage of HS freshman boys who develop acne over the course of their HS years
Inductive Logic/Reasoning (descriptive)
"instinct" -used to generate hypothesis & discover relationships a conclusion is drawn from particular cases or specific instances, based on facts or observations -premise may be true with a false conclusion
What are the 8 causal criteria
1. Temporality/Temporal relationship 2. Plausibility 3. Consistency 4. Strength of Association 5. Dose-Response relationship 6. Reversibility 7. Study Design 8. Judging the evidence
What are the steps of descriptive and analytical epi?
1. use descriptive analysis to generate a hypothesis/model 2. use analytics to find the strength of association along main or secondary pathway via hypotheses testing (parameter estimates, Odds Ratio, Relative Risks, Hazard Ratios) 3.Interpret results
Dose-Response Relationship
Are varying amounts of exposure to the cause associated with varying magnitude of the effect? -increase dosage=increase effect? ex. greater nose level and longer exposure=higher prevalence of hearing loss -risk vs. benefit relative to possible side effects
Main Causal Pathway
arrow from exposure to outcome -main exposure is not necessarily directly causal to the outcome, but the possibility can be determined through causal criteria
Contingency Table
cross-classification, subcategories of one characteristics are horizontal and other are vertical
Incidence & Prevalence comparison
do you currently have otitis media? point prevalence have you had OM during the last 3 years? period prevalence have you ever had OM cumulative incidence
Covariates placed along one of the secondary pathways included :a. IQ and memory learning b. language, fetal growth and neurological function c. birth length and neurological function d. gestational age, IQ and neurological function e. fetal growth and IQ
e. fetal growth and IQ
Gating
example: those who respond "no" are not asked further questions
Indirect/Secondary Causal Pathways
indirect associations of the exposure relative to the outcome
Stratification
layering/grouping variables by other covariates ex. grouping those exposed to cocaine by marital status and age group
Corollaries
logical extensions from the assumption that an idea is considered to be true without controversy (axiom) Corollary 1: nonrandom clusters of disease can be observed, understood, measured based on individual and community characteristics -"golden triangle": agent, host, environment Corollary 2: variates in disease frequency occur d/t variations in exposure to etiologic agents or remote causes, or variation in susceptibility
Prevalence
measure of disease occurrence -# of cases of disease present in population at a specified time/# of persons in population at a specified time -does not take into account when disease developed/duration -therefore do not have a measure of disease risk, just prevalence -point prevalence: at specific point in time -period prevalence: cases of disease at any point during a certain time period -proportion not a rate "Burden of disease" measurement
The model, as proposed, is a _____ model. a. univariate b. reversed c. temporally incorrect d. multivariate e. mis-specified
multivariate
Univariate Model
no secondary pathways, only one exposure of interest and its relation to the outcome
Axiom
something that cannot be proven to be true, it is assumed true without controversy ex. disease is not randomly distributed in human population groups -epidemiology studies difference in distribution -disease pattern can be identified, measured, and this is the most effect way to modify and prevent disease
Epidemiology
study of distribution of disease (states of health) and determinants of disease (deviations from health) in human populations Goals: -identify etiology of disease -Provide knowledge to prevent.control disease via interventions -Provide info to maximize timing/effectiveness of interventions
Hazard Ratio
theoretical measure of probability of occurrence of an event per unit time at risk "instantaneous incidence rate" ex.death at time "t"
Epi Model
tool to help epidemiologists study the potential association between an exposure and an outcome of interest
Deductive Logic/Reasoning (analytical)
used to test, refute, falsify a hypothesis -conclusions drawn from general principles or premises -conclusions have no uncertainty; they are certain provided the premises are true
Model efficiency
using fewest number of variables needed to explain an association "most bang for your buck"
Covariates
variables in indirect pathways, may impact the outcome & help better predict the outcome of interest
Multivariate Model
when there are multiple variables of interest that may help explain the relationship of interest