EPI 3901 1st Edition Lecture 4Outline of Last Lecture I. Central Axiom of EpidemiologyII. Descriptive Epidemiologya. John SnowIII. Hygiene and HealthIV. Incidence vs. PrevalenceV. The “Magical Triad” of EpidemiologyVI. Reading a Scientific JournalOutline of Current Lecture I. Modeling Epidemiologya. Covariance vs. ConfoundingII. Framing a Studya. Determining the Controlsb. Triagec. EfficiencyIII. Terminology in the Studya. Diseaseb. Syndromec. Illnessd. SicknessIV. Descriptive vs. Analytic EpidemiologyCurrent LectureI. Modeling Epidemiologya. Covariance vs. Confounding - Covariance is not the same thing as Confoundingi. Covariance – attempt to build best model possible (no perfect model since you don’t know all variables) explaining association.ii. Confounding is one step farther than covariance – confounding is when a variable has a relationship to the outcome and to the input.1. Confounding tends to be decreased by randomization. This is very hard to do though; a massive testing population is requiredb. Efficient model – explains model (showing exposure/onset) by significantly relating all variables.These notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.c. Point estimate – Using data from a sample to calculate a single value, a statistic, which best represents an unknown value.II. Framing a Studya. Determining the Controlsi. Need to determine controls. For everyone that has outcome, have to havesome as control (control can’t have had exposure to any of the variables; very difficult to find controls due to large number of variables)ii. Stats allows us to assign error, and quantifies datab. Triagei. When a submitted paper is rejected, and may not even be scored by the judging committee; you can’t resubmit as is. If you don’t take confounding factors into account, the paper will always get triaged by peersii. Occurs when intervention is associated with an extraneous cofactor; variables can either relate to burden of disease or don’t interact. Are the variables interacting at the same time? Or acting one at a time?1. It is possible to tell if they act one at a time by removing variables one at a time. If it is one at a time, onset of illness will occur less and less with each removal.c. Efficiency – the ability to remove a variable to show correlation; you won’t be able to account for all variables but an efficient model will relate to many.d. Longitudinal study: collect data over time. e. The data is only as good as the honesty of your input; there are certain questionspeople always lie on (weight [underestimated], height [guys overestimate, women underestimate], drinking, promiscuity) i. So have to come up with ways to soften the impact of inaccurate information.III. Terminology in the Studya. i.e. Gender vs. sex, race vs. ethnicityi. “Ethnicity” is more accurate, but survey designers tend to use “race” because survey subjects get confused by ethnicity and tend to identify by race insteadb. Disease – A disorder of health, impairing structure or function, often expressing specific symptoms. (Does not include infirmities directly from physical injury)c. Syndrome – a set of symptoms that frequently occur together, or a certain condition associated with these symptoms.d. Illness – a disease or period of sickness affecting the mind or body.e. Sickness – The state of being ill.IV. Descriptive vs. Analytic Epidemiologya. Descriptive – collect data, come up with hypotheses, observe distribution (TPP = time, place, and person)i. Example: early descriptive work of recording birth, death, marriage ii. Typical design is a community health surveyiii. Synonyms – cross-sectional study, descriptive study (monitor, evaluate, plan health program)b. Analytic – testing a specific hypothesis.i. Need to understand descriptive epidemiology before you can do analyticii. Typical study design: cohort, case-control (retrospective
View Full Document