Week 4: Random error, bias, confounding

Question 1

The goal of epidemiology and health research is to identify the ______ and _______ for disease

Answer 1

causes and preventions

Question 2

what is internal vs external validity?

Answer 2

internal validity = the observed results accurately reflect the true association

external validity = generalizability, whom the results can be applied to. requires internal validity

Question 3

T/F: if a study lacks internal validity, external validity is irrelevant
T/F: we can compromise internal validity in an effort to achieve external validity

Answer 3

T
F. DO NOT compromise internal validity

Question 4

External validity can be achieved by a sample that represents the ________ population

Answer 4

target

Question 5

If we observe an association and there is no error due to chance, bias, or confounding then we have a _____ association

Answer 5

true

Question 6

What are the 3 threats to validity?

Answer 6

1. chance –> random sampling error
2. bias –> systematic error in selection of participants or measurements
3. confounding

Question 7

1. How can we minimize random sampling error?
2. What is the p value NOT:
   a) false negative (B) error
   b) systematic error
   c) false positive (a) error
   d) random sampling error
3. P value is calculated to either reject or accept the ______ hypothesis to see if there is an __________

Answer 7

1. increase sample size
2. c)
3. null, association

Question 8

What are the 2 types of bias?

Answer 8

selection –> volunteer bias, non-response, loss to follow up
information –> exposure (recall, interviewer), outcome (respondent, observer)

Question 9

1. Bias refers to a systematic error in the design or conduct of a study. What is a systematic error?
2. when bias occurs in a study, the ________ association between the exposure and outcome will be different from the ____ association

Answer 9

1. your measurements of the same thing will vary in predictable ways: every measurement will differ from the true measurement in the same direction, and even by the same amount in some cases.
2. observed, true

Question 10

1. Explain information bias
2. explain selection bias

Answer 10

1. information bias = when the information collected from study participants or about the study participants during measurement is erroneous. Ex: recall or interviewer bias
2. selection bias = systematic error in the way participants are selected or retained in a study.

Question 11

What are the 4 types of selection bias? explain .

Answer 11

1. volunteer bias - volunteers could be more health conscious or from a different socio-economic group, could have different exposures
2. non-response bias - those suffering from a disease with a particular belief, have a differential outcome
3. membership bias - healthy worker effect; those with occupations have higher economic status and are more likely to participate in the study
4. loss to follow up - in clinical trials or longitudinal studies, the sickest leave the study early

Question 12

T/F: like sampling bias, selection bias too can be reduced by increasing sample size

Answer 12

F

Question 13

the simple exposure-outcome model is never a reality because there most likely exists a __________ which is a third variable that is an effect modifier on the pathway of exposure –> outcome

Answer 13

confounder

Question 14

1. what is confounding?
2. What type of error is it?

Answer 14

1. where 2 effects are not separate. distortion of the actual association due to a mixing of effects between the exposure and the confounder–> all or part of the observed outcomes are due to factors other than the primary exposure of interest
2. selection bias –> systematic error

Question 15

T/F: confounders are not causally associated with the outcome
T/F: confounders must be non-causally associated with the exposure
T/F: confounders are not an intermediate in the causal pathway between exposure and outcome

Answer 15

F. confounders must be causally associated with the outcome –> makes them a TRUE risk factor

F. confounders can be causally or non-causally associated with the exposure.

T. confounders are not intermediates between cause and effect. They are separate variables

Question 16

Explain each in relation to dealing with confounding:
At the design stage:
a) restriction
b) matching
c) randomization

At the analysis stage:
a) standardization

Answer 16

At the design stage:
a) restriction = limit study inclusion criteria with respect to confounding factors; study only men or women
b) matching = produce case and control (exposed or non-exposed) groups that have similar characteristics
c) randomization = experimental studies

At the analysis stage:
a) standardization = age standardization is adjustment for age

Question 17

what are multivariate models?

Answer 17

contain exposure variable and 1 or more confounding variables as predictors of the outcome variable

Question 18

Indicate the most likely source of error in each problem based solely on the on the information provided(random sampling error, confounding, selection bias, information bias):

1. The authors of a cross-sectional study hypothesized that lack of regular exercise is associated with obesity in children. In the study of 12 children, however, they failed to show a statistically significant association between exercise habits and obesity (RR=1.9; 95% CI=0.7-14.3).
2. A group of epidemiologists examine the association between the use of diet pills and migraine using a case-control study design. Their subjects are outpatients from a large community hospital. They find that those using pills are more likely to complain of migraine, but also discover that those who have migraine and use diet pills are more likely to be referred to the hospital than those who have migraine but are not using diet pills.
3. A large health study found that older adults with dementia were more likely to develop liver cirrhosis than those without dementia. The investigators, however, could offer no plausible biological explanations. Many of the subject reported heavy drinking.
4. Researchers wanted to study the relationship between ovarian cancer and socio-economic status using a case-control design. They used women admitted to a large teaching hospital over the course of a year as cases. They recruited controls by knocking on doors during the day to ask for participation from women without ovarian cancer.
5. To test whether persons exposed to pesticides during early childhood were more likely to develop neurological problems, a case-control study was designed. The results were significant. Both cases and controls (age range: 20-45) were asked detailed questions about their pesticide exposure during early childhood.
6. In a case-control study new mothers were asked about their smoking and coffee consumption habits. A strong association was reported between coffee consumption and low birth weight (OR=4.5). When the data were re-analyzed and adjusted for smoking, however, the association was found to be nonexistent (OR=1.0).
7. A health researcher (not trained at Western!) reported the following conclusion at a scientific meeting: “Daily use of 500 mg of vitamin C was associated with a reduced frequency of upper respiratory infections in children under 10 years of age (RR=0.8; 95%CI=0.3-3.7)”.
8. Dioxin has been recognized as an environmental endocrine disruptor, but there is no reported biological link to diabetes. However, in a study of serum dioxin level and the risk of diabetes, a RR of 1.56 was reported. It was found that those with high levels of dioxin also had high levels of serum triglyceride.
9. Researchers were interested in factors that influence length of stay in a geriatric rehabilitation unit. They set up a study that involved collection of plausible variables that might influence length of stay in order to test the relationship of these variables statistically. There were two research assistants hired to collect information from medical records for statistical testing. After data collection and preliminary analyses are run, the researchers find an association between length of stay and family support is positive for data collected by one of the research assistants but not this other.
10. A case-control study was conducted to investigate the association between artificial sweetener and bladder cancer. Controls were selected from a hospital sample of patients diagnosed with obesity-related conditions.

Answer 18

1. random sampling error
2. selection bias
3. confounding
4. selection bias
5. information bias
6. confounding
7. random sampling error
8. confounding
9. information bias
10. selection bias

Question 19

State whether the variable in question is a confounder in each of the scenarios:

1. In a study of the relationship between contact lens use and the risk of eye ulcers the crude relative risk is 3.0, and the age-adjusted relative risk is 1.5. Is age a confounder in this study?
2. A study of the relationship between exercise and heart attacks that is conducted among participants who do not smoke. Is smoking a confounder in this study?
3. A case-control study examined the relationship between cigarette smoking and pancreatic cancer. In this study, coffee drinking is associated with smoking and is a risk factor for pancreatic cancer among both smokers and nonsmokers. Assume that coffee drinking is not on the causal pathway between smoking and pancreatic cancer. Is coffee drinking a confounder in this study?
4. A cohort study of the risk of liver cirrhosis among female alcoholics. Incidence rates of cirrhosis among alcoholic women are compared to those among nonalcoholic women. Nonalcoholics are individually matched to alcoholics on month and year of birth. Is age a potential confounder in this study?

Answer 19

1. Yes
2. No
3. Yes
4. No

Question 20

Which method of control is used to control for smoking in the studies below?

1. A cohort study of the effect of a new lipid-lowering medication on the incidence of Coronary Heart disease (CHD) that is limited to non-smokers ≥50 years of age.
2. A case-control study of the effect of a new lipid-lowering medication on the incidence of CHD that determines the relative odds (OR) of CHD separately in smokers and non-smokers.
3. A randomized controlled trial of the effect of a new lipid-lowering medication on the incidence of CHD conducted in a population of smokers and non-smokers ≥50 years of age.
4. A case-control study of the effect of a new lipid-lowering medication on the incidence of CHD where smoking controls are selected in the same proportion as smoking cases.

Answer 20

1. Restriction - limits participants
2. Stratification - separates groups
3. Randomization - randomized controlled trial
4. Matching - have 2 equal variable amounts