HS 300 Exam 2 study guide PDF

Preview

Summary

Download HS 300 Exam 2 study guide PDF

Description

STUDY DESIGN ● Ecological Design: correlation of rates of exposure and rates of disease among different groups or populations ○ How to classify: ■ Unit of observation: group ■ Disease or exposure status: Both exposure and disease status ■ Prospective, retrospective: or concurrent: concurrent (retrospective with ecological case control) ○ Advantages ■ Quick, simple, inexpensive ■ Good for generating future hypotheses ■ Use of secondary data ○ Disadvantages ■ Ecological fallacy: observations on group level may not represent people within the population correctly ■ Little individual data is accounted for ■ Little adjustment for confounds ○ Measures ■ Compare proportions ● Cross-sectional Design: exposure and disease history collected at the same time; slice of a moment in time ○ Generally descriptive; a prevalence study ○ How to classify: ■ Unit of observation: individual ■ Disease or exposure status: both ■ Prospective, retrospective, or concurrent: concurrent (maybe retrospective) ○ Advantages ■ Quick, easy, cheap ■ No loss to follow up ■ Estimates magnitude and distribution of a health problem ■ Results are generalizable ■ Useful for intervention planning ■ Can use to create hypotheses for other studies ○ Disadvantages ■ No info about causality or temporality ■ Cannot study disease etiology ■ Not a measure of risk ■ Not useful for low prevalence diseases ○ Measures: Not really a specific measure ○ Interpretation ● Case-Control Design: Two groups: one with disease, and one without; study both groups’ past exposure

○ ○ ○

○ ○

○

○

○

○

Cases need a strict, restrictive definition ■ In order to avoid misclassification Measure of incidence Controls ■ Must be from population that produced the cases ■ Ideally, would be the same in every aspect except disease status ■ If sampled from hospital, must have illness unrelated to exposure Nested case-control ■ Cases sampled from cohort study How to classify: ■ Unit of observation: individual ■ Disease or exposure status: Disease status ■ Prospective, retrospective, or concurrent: retrospective Advantages ■ Useful when little is known about disease ■ Quick, cheap ■ No loss to follow up ■ Better for rare diseases ■ Better for diseases with long induction period ■ Can evaluate multiple exposures Disadvantages ■ Indirect estimate of risk ■ Timing of exposure-disease relationship harder to obtain ■ Possibilities for misclassification ■ Greater chance of bias than cohort ■ Not useful for rare exposures ■ Survivor factor ● People with more severe form of disease may not survive Measures ■ Odds ratio ● (AD)/(BC) ● (odds of exposure among disease group)/(odds of exposure among control group) ● OR=1: odds of exposure are equal between both groups ● OR=2: odds are twice as likely ● OR=0.5: odds are half as likely Interpretation ■ Do not discuss risk

■ (Subjects) with (disease) have ___ the odds of being exposed to (exposure) compared to controls, those without (disease) ●

Cohort Design: recruit people for study, interview for exposure, then separate into two groups; followed over a period of time ○ Start with people who do not have the disease

○

○

○

○

○

How to classify: ■ Unit of observation: Individual ■ Disease or exposure status: Exposure status ■ Prospective, retrospective, or concurrent: Prospective or retrospective Advantages ■ Only study that gives direct estimation of risk ■ Multiple outcomes can be evaluated ■ Useful for rare exposures ● Occupational exposures ■ Demonstrates temporal sequence between exposure and disease ■ Can see length of latency or incubation period Disadvantages ■ Expensive ■ Need large sample size ■ Loss to follow up ■ Long time ■ Exposure misclassification ● Exposures change over time ■ Have to wait for cases to accrue ● Solution: retrospective cohort studies Measures ■ Relative Risk: ● (A/(A+B))/(C/(C+D)) ● (incidence rate exposed/incidence rate not exposed) ■ Attributable Risk ● (A/A+B)-(C/C+D) ■ Attributable Risk Percentage ● ((A/A+B) - (C/C+D))/(A/A+B) x 100 ● (IR exposed - IR not exposed)/IR exposed x100 ■ Population Attributable Risk (Percentage) ● (IR population - IR not exposed)/IR population x100 Interpretation ■ RR: (Subjects) who had (risk factor) had ___ times the risk of having (disease) than (control group), those without (risk factor) ■ AR%: Assuming association is causal, _____ percent of (subjects) who

(risk factor) and had (disease) experienced this outcome due to having (risk factor) ●

Experimental Design ○ The only non-observational study design ○ Advantages ■ Ability to control and manipulate variables in the study ● Confounding less likely

○

○

○

■ Most rigorous ■ Highest validity Disadvantages ■ Expensive ■ Time consuming ■ Not everyone will complete the treatment as expected (noncompliance) Measures ■ Rate Ratio ● (a/a+b)/(c/c+d) ● = to relative risk formula (RR = RR) ■ Percent Relative Effect ● (((a/a+b)/(c/c+d)) - 1) x 100 Interpretation ■ For reduction: Overall, those in the experimental group, who received

(treatment) had a (percentage) reduction in the incidence of (disease) compared to those who received (placebo/standard of care) ■ For cures: Overall, those in the experimental group, who received (treatment) had a (percentage) increase in cures for (disease) compared to those who received (placebo/standard of care) ○ ○ ○

Intervention studies: used to test efficacy of preventative or therapeutic measures Clinical trials: testing a drug, prophylactic, or treatment Intent to treat: everyone who you intended to treat are in the denominator ■ Even if they are non compliant or drop out ■ Emulates every day conditions ○ Efficacy: only include those who completed treatment in the denominator ■ Overly optimistic for real life use BIAS AND CONFOUNDING ● Validity ○ The degree to which inferences drawn from a study are warranted ○ Internal: proper selection, appropriate measurement ○ External: can be generalized to population ● Error ○ Random error: false association due to chance ■ Sampling error: sample collected is not representative ■ Poor precision: study factor is not measured sharply ○ Systematic error: introduced by study process ■ Selection bias ■ Information bias ■ Confounding ● Direction of effect ○ POSITIVE: Exaggerates a true association ○ NEGATIVE: Hides a true association

● ●

●

●

●

Bias cannot be adjusted for once results are collected Selection Bias ○ Relation between exposure and disease is different for those who participate in a study and those who are eligible but do not participate ○ Most likely in case control and retrospective cohort studies ○ Control selection bias ■ If you were to select controls for a case control from a hospital, those controls wouldn’t be representative of the population that produced the cases ○ Self selection bias ■ Refusal or agreement to participate based on disease or exposure status ○ Differential surveillance ■ Different diagnosis or referral based on exposure history ○ Loss to follow up ○ Healthy worker effect ■ The working population tends to be healthier than the general population ○ Selective survival - the cancer shitshow in review Information Bias ○ A result of measurement error for both exposure and disease ○ Recall bias ■ Better recall of exposure among cases than controls ○ Interviewer/Abstractor bias ■ Interviewers probe more thoroughly for an exposure in a case than a control ○ Prevarication (lying) bias ■ Participants have ulterior motive for answering a question ● Ex. patient may lie about sexual assault due to stigma Misclassification: error in classification of exposure or disease ○ Ex. exposed may be classified as non-exposed, and vice versa ○ Differential: misclassification does not go both ways ■ Influences results a lot ○ Non differential: misclassification goes both ways ■ Not a large influence on results Confounding ○ Natural mixing of effects between exposure, disease, and a third variable called a confounder, distorting the relationship between exposure and disease ○ 3 criteria to be a confounding variable (need all 3) ■ Be a risk factor for disease ■ Be associated with exposure ■ Not an intermediate step between exposure and disease ○ Appreciable difference: usually 10-20% ○ How to control for confounding? ■ Randomization

■ ■ ■ ■ ■

Restriction Matching Standardization Stratified analysis Multivariate analysis...