HLTH301 Exam 2: Study Design and Statistical Analysis in Epidemiology - Prof. Debra Lee Hi, Study notes of Health sciences

Download HLTH301 Exam 2: Study Design and Statistical Analysis in Epidemiology - Prof. Debra Lee Hi and more Study notes Health sciences in PDF only on Docsity!

HLTH301 EXAM 2

Study Design Information Types Bias Measures Advantages Disadvantages Example Important Confounding Criteria -Confounding factor (CF) must be related to both exposure and disease -Must be association between CF and disease among those unexposed -CF cannot be part of causal chain between exposure and disease Ways to reduce confounding Randomization Restriction Matching Best way to deal with confounding = randomized control trial Ice cream and murder rates Large positive correlation does not equal causality Case report -Very common in medical journals -Describe experience of a single patient or a group of patients with similar diagnosis -Descriptive Useful for recognizing new diseases and formulating hypotheses for risk factors -Small sample size (experience of a few individuals may be a coincidence) -No comparison group Associatio n between diseases or symptoms Case-Control When Appropriate: -Before a cohort study or an experimental study to identify potential etiology -Investigate multiple exposures -Good for rare diseases, because cases can be identified and included Types of controls

• Hospital Controls: have similar quality of info and are easy to select, but they may have characteris tics or diseases that led to hospitalizat ion -Dead controls: if cases are dead, surrogates, such as spouse or children, will give informatio n of past Selection Bias: inclusion of cases or controls depends on exposure status ex) association between Pap smears and cervical cancer ( households for 1 control) Recall Bias: differences in ways exposure information is remembered or reported by cases ex) those who have been sick tend to think about causes for their illnesses Case selection -Should be homogenous -Selected using strict case definitions -Incident cases vs. prevalence cases - Misclassificat ion Control Selection -Most difficult part of case control study -Controls should be selected from the same population at risk of disease from which cases -Quick and inexpensive -Good for diseases with long latency periods -Good for rare diseases -Can examine multiple factors that may contribute to a single disease -No loss to follow-up -Inefficient for rare exposures -Temporal relationship between exposure and disease is difficult to establish -Prone to selection and recall bias -Unable to estimate incidence Risk factors for developme nt of TSS w/ association with a tampon brand

1. Definition Bias a. Systematic error that results in an incorrect estimate of the association between an exposure and outcome b. “Different” c. 30+ different types d. Difficult to identify and correct
2. Types a. Selection Bias i. Distortions that result from procedures used to select subjects and from factors that influence participation in the study ii. Error introduced when the study population does not represent the target population iii. Defining feature
3. Occurs at the stage of recruitment of participants and/or during the process of retaining them in the study iv. In case-control studies b. Information bias i. A flaw in measuring exposure, covariate, or outcome variables that results in different quality (accuracy) of information between comparison groups ii. Bias in an estimate arising from measurement errors iii. A distortion in the measure of effect caused by lack of accurate measurements of exposure or disease status iv. Defining feature
4. Information bias occurs at the stage of data collection
5. Misclassification of exposure and/or outcome status is the main source of error, and this, in turn, has the potential to bias the effect estimate v. Types of information bias
6. Recall bias individuals with particular health problem report their previous exposure differently than those who are not affected
7. Interviewer bias difference in the soliciting or interpreting of information from study participants
8. Misclassification: participants are erroneously categorized with respect to exposure or disease c. Ways to reduce bias i. Study design 1. Choice of study population 2. Methods of data collection should be the same for all study groups 3. Use of objective measures 4. Use of preexisting records

5. Blinding
6. Measure of Association a. Odds Ratio i. ii. Measure of how strongly the exposure is related to the disease iii. Ratio of 2 odds  odds of having exposure and disease iv. OR=ad/bc v. Odds ratio interpretation
7. OR=1.00  no association between the exposure and the disease
8. OR> 1.00  odds of disease are X (4) times higher among the exposed person than among the nonexposed vi. Need confidence interval
9. 95% CI  2.07 to 8. vii. Advantages of Odds Ratio
10. Easy to calculate
11. Easy to interpret
12. Can be used to make treatment decisions b. Relative Risk i. Likelihood of developing the disease in the exposed group relative to those that are not exposed
13. RR = [a/(a+b)]/ [c/(c+d)] ii. RR> 1= Risk Factor

c. 95% Confidence Interval i. Represents the range in which the true magnitude of effect lies with a certain degree of assurance ii. OR or RR is always reported with 95% CI iii. Why is 1 important for OR and RR

1. Null value
2. Odds or risk in one group is the same as the other
3. Internal Validity a. Is there an alternative explanation that can explain the results? i. Bias systematic error that resulted from study design or how study was conducted ii. Confounding mixing of effects between the exposure, the disease, and a third factor iii. Role of Chance Need to quantify the degree to which chance may account for results test for statistical significance
4. External Validity [Generalizability] a. Whether results are applicable to other populations
5. Confounding a. Mixing of the effect of the exposure on the disease with a third factor b. Relationship between the exposure and the disease can be attributed to the confounder c. Criteria for confounding i. Must be associated with the exposure and the disease
6. If no association between the exposure and the confounder OR no association between the disease and the confounder then no confounding by that factor
7. Age as a potential confounder a. Age is related to exposure aspiring use increases with age b. Age is related to outcome pancreatic cancer increases with age ii. Must be an association between the confounder and disease among those non-exposed iii. The potential confounder cannot be an intermediate link in the causal chain between the exposure and the disease
8. Requires knowledge of the relationship between the exposure and the disease
9. Exposure alters level of potential confounder which in turn effects disease status
10. Aspirin doesn’t cause changes in age, which then causes pancreatic cancer d. Identifying and addressing confounding i. Often difficult, requires knowledge of the disease ii. Past research e. How to address potential confounding

i. Through study design

1. Randomization equal out known and unknown confounders
2. Restriction if race and gender are potential confounders could include only white females
3. Matching  potential confounders distributed equally among groups ii. In analysis
4. Stratified analysis if potential confounding by sex, examine relationship between exposure and disease separately for males and females
5. Multivariate analysis analysis of data that takes into account multiple potential confounders at the same time f. What are we looking for? i. Factors related to the disease- pancreatic cancer ii. Factors related to the exposure- aspirin
6. Confounding vs. Effect Modification a. Confounding distortion of a true effect, want to get rid of this b. Effect modification 2 factors together amplify effect; want to describe
7. Association a. Refers to a linkage between or among variables; variables that are associated with one another can be positively or negatively related i. Positive association if the value of one variable increases, the value of the other variable increases as well ii. Negative association if the value of one variable increases, the value of the other variable decreases
8. Cause a. Antecedent event, condition, or characteristic, that was necessary for the occurrence of disease at the moment it occurred, given that other characteristics are fixed
9. Models of causality in epidemiologic studies a. Counterfactual i. Formalizes notions of cause and effect ii. Association considered causal when: the exposure was altered outcome would have been different iii. What would have happened had the exposure not occurred iv. Ideal experiment
10. Person experiences some exposure and followed over time
11. Go back in time
12. Same person not exposed (or exposed to something different) and followed over the same initial time period

v. Counterfactual framework

1. Forces clear idea of effect investigator is interested in- clearly define exposed and unexposed b. Hill’s criteria of causality An expanded list of causal criteria i. Strength: strong associations give support to causal relationship between factor and disease
2. Big odds ratios and/or Rs ii. Consistency: an association has been observed repeatedly
3. Many times
4. Many types of studies iii. Specificity: association is constrained to a particular disease-exposure relationship iv. Temporality: the cause must be observed before the effect v. Biological gradient: also known as a dose-response; shows a linear trend in the association between exposure and disease vi. Plausibility: the association must be biologically plausible from the standpoint of contemporary biological knowledge vii. Coherence: the cause-and-effect interpretation of our data should not seriously conflict with the generally known facts of the natural history and biology of the disease viii. Analogy: relates to the correspondence between known associations and one that is being evaluated for causality
5. Effect Modification a. Association between exposure and disease differs depending on the value of another variable i. Example) association between smoking and lung cancer is stronger among people exposed to asbestos
6. Multicausality a. Disease processes tend to be multifactorial i. Very few exposures cause disease entirely by themselves
7. Exposure to measles can cause measles only if somebody is susceptible
8. Development of melanoma among those with high UV light exposure who also have fair skin ii. The one-variable-at –a-time perspective has several limitation iii. Both confounding and effect modification are manifestations of multicausality (reality is multivariate!)
9. Effect modifier a. Effect of exposure on the disease is modified (altered) depending on the value of a third variable called “effect modifier”

Practice Problems  study design?

1. Smoking histories are obtained from all patients entering a hospital who have lip cancer and are compared with smoking histories of patients with cold sores who enter the same hospital.
2. The entire population of a given community is examined, and all are who are judged free of bowel cancer are questioned extensively about their diet. The people are then followed for several years to see whether their eating habits will predict their risk of developing bowel cancer.
3. To test the efficacy of vitamin C in preventing colds, army recruits are randomly assigned to two groups: one given 500mg of vitamin C daily, and one given a placebo. Both groups are followed to determine the number and severity of subsequent colds.
4. The physical examination records of the incoming first-year class of 1935 at the University of Minnesota are examined in 1980 to see whether the freshmen’s recorded height and weight at the time of admission to the university were related to their chance of developing coronary heart disease by 1981.
5. 1500 adult men who worked for Lockhead Aircraft were initially examined in 1951 and were classified by diagnostic criteria for coronary heart disease. Every three years they have been reexamined for new cases of the disease.

6. A random sample of middle-aged sedentary women was selected from four census tracts, and each woman was examined for evidence of osteoporosis. Those found to have the disease were excluded. All others were randomly assigned to either an exercise group or a control group, which had no exercise program. Both groups observed annually for incidence of osteoporosis.
7. Questionnaires were mailed to every 10th^ person listed in the city telephone directory. Each person was asked to provide his or her age, sex, and smoking habits and to describe the presence of any respiratory symptoms during the preceding 7 days Practice  Calculations
8. The death rate per 100,000 for lung cancer is 7 among nonsmokers and 71 among smokers. The death rate per 100,000 for coronary thrombosis is 422 among nonsmokers and 599 among smokers. The prevalence of smoking in the population is 55%. What is the RR of dying of lung cancer for smokers vs. nonsmokers? What is the RR of dying of coronary thrombosis for smokers vs. nonsmokers? What is the RR of dying of lung cancer for smokers vs. nonsmokers? RR= risk associated with an exposure in comparison to the risk associated with nonexposure RR=1: Risk of disease among exposure is not different from the risk of disease among the nonexposed. Risk of dying from lung cancer is more than 10 times as high in smokers as among nonsmokers
9. An investigator wanted to determine whether vitamin deficiency was associated with birth defects. By reviewing birth certificates during a single year in a large U.S. county, the researchers located 189 infants located with NTDs. A total of 600 other births were selected

at random from the certificates. Mothers were given a dietary questionnaire. Among mothers who gave birth to an infant with an NTD , 84 reported no use of supplementary vitamins; a total of 137 control mothers did not use a vitamin supplement. Calculate the OR between vitamin use and NTDs OR<1: protective effect Ex OR=. 5 exposure is associated with half of the risk of disease OR>1: risk factor Ex OR=2.1 odds of disease are about 2.1 times higher among the exposed than the nonexposed