PSYC 1010 Midterm: Bias and confounding
EPI 511 Introduction to Epidemiology
PS 5—Bias, Confounding, and Effect Modification
**Please show your calculation(s) and/or logic for each question**
1. In each of the following hypothetical study descriptions, there may be potential bias which could
impact the validity of the results. For each study:
• Indicate whether confounding or bias is present. For the latter, identify the type of bias
(i.e., selection bias or information bias).
• State what impact you would expect to see on the calculated measure of association (i.e.,
overestimate of the true association; underestimate of the true association; or
unknown).
• Provide your rationale.
Confounding is when an exposure is associated with other factors which may influence the outcome.
There are three criteria that are required for a confounding factor including (1) the confounder is
associated with the outcome (2) the confounder is associated with the exposure and (3) the confounder
is not a result of the exposure. All three criteria are required.
Bias is a systematic error in the design, conduct or analysis of a study that results in a mistaken estimate
of an exposure’s effect on the risk of disease. The two main types are selection bias, which is a
systematic error in how the cases vs controls or exposed vs non-exposed were selected, which leads to a
false association even if no association truly exists. In information bias, the information gathered
regarding exposures and outcomes are incorrect. This may be due to inadequate measurements,
sources of information, or methods.
The types of bias include
a. In a case-control study of the association between intravenous drug use and HIV infection, the
investigators discovered after the study that cases (persons with HIV) had tended to
underreport their drug use due to fears arising from previous convictions for drug-related
violations. Underreporting was not a problem among controls (persons without HIV), however,
who were not under medical supervision at the time.
Answer:
b. A large epidemiologic study found that elderly adults with dementia were more likely to develop
liver cancer than those without dementia. The investigators, however, could offer no plausible
biological mechanism for the association between dementia and liver cancer. Many of the
subjects with dementia reported heavy drinking, however, which can cause liver cancer. Most
subjects without dementia did not report heavy drinking.
c. A prospective cohort study followed 8,542 women over a 10-year period to determine if alcohol
consumption increased the risk of breast cancer. No statistically significant association was
observed. At study initiation, 12,000 women had been enrolled in the study: 6,000 heavy
drinkers of alcohol and 6,000 non-drinkers of alcohol. 3,458 had withdrawn during the 10-year
follow-up period. Subsequent interviews with the spouses, friends and relatives of the women
who withdrew revealed that 60% of the withdrawals had developed breast cancer and that
nearly 84% of the women who withdrew were regular drinkers of alcohol. Assume that
ascertainment of the exposure was accurate in both groups.
EPI 511 Introduction to Epidemiology
PS 5—Bias, Confounding, and Effect Modification
**Please show your calculation(s) and/or logic for each question**
2. You have conducted a case-control study to investigate the association between use of oral
contraceptives (OC) and the risk of stroke among women. You are considering the possibility that
age might be a confounding factor. You look at the data relative to age, which you have
operationally defined in your study as < 40 years or > 40 years.
Stroke
Subjects Exposure Cases Controls Totals
All subjects OC user 39 24 63
Non-user 114 154 268
> 40 years OC user 18 7 25
Non-user 88 95 183
< 40 years OC user 21 17 38
Non-user 26 59 85
Determine if age is a confounder in the association between oral contraceptive use and stroke, using
the following steps:
a. Calculate the crude OR and provide an interpretation for the value.
Odds ratio is the probability of occurrence of an event to that of non-occurrence. Specifically, the odds
of disease in the exposed divided by the odds of disease in non-exposed.
If >1, risk in exposed is greater than non-exposed. If <1, the risk in exposed is less than the risk in the
non-exposed.
b. Calculate the stratum-specific ORs and compare to each other and to the crude OR. What does
this comparison tell you about age as a possible confounder?
63
PSYC 1010 Full Course Notes
Verified Note
63 documents
Document Summary
**please show your calculation(s) and/or logic for each question** In each of the following hypothetical study descriptions, there may be potential bias which could impact the validity of the results. Confounding is when an exposure is associated with other factors which may influence the outcome. There are three criteria that are required for a confounding factor including (1) the confounder is associated with the outcome (2) the confounder is associated with the exposure and (3) the confounder is not a result of the exposure. Bias is a systematic error in the design, conduct or analysis of a study that results in a mistaken estimate of an exposure"s effect on the risk of disease. The two main types are selection bias, which is a systematic error in how the cases vs controls or exposed vs non-exposed were selected, which leads to a false association even if no association truly exists.
