Problem Set 1: Populations, Sampling and Causation
Submit your answers through “Problem Set 1” in the Problem Sets folder on Blackboard. Problem Set 1 has ten (10) questions and is worth 100 points.
If working as a group, adhere to the “Rules for Collaborating” (see syllabus).
Refer to Chapter 14 in the Aschengrau textbook for guidance on critiquing epidemiologic studies or the online module “Critical Reading” in the Epidemiology section. You can also review a set of slides on critical evaluation posted in the Problem Sets folder on Blackboard.
Read the paper “Socioeconomic status in relation to the risk of ovarian cancer in African-American women: A population-based case-control study” (AJE 2016; 184 (4): 274-283) posted in the Problem Sets folder on Blackboard.
We will begin to build the important skill of critically evaluating epidemiologic literature as consumers of public health data. Our assessment will include defining the study population, exposure and health outcome; considering vulnerability and susceptibility; and weighing evidence of causation. Do not belabor the details of the study design and statistical analysis – we will be working through those concepts and methods later in the course.
1) What was the primary exposure of interest and how was it measured? Briefly describe two ways in which exposure was categorized. (10 points)
The primary exposure of interest was socioeconomic status. This was measured by sampling the population’s education level and annual family income. For the purposes of sampling, annual family income was categorized ordinally as <$10,000, $10,000-24,999, $25000-$49000, $50000-$74999, or _>75000. Education level was also categorized ordinally, but in the following way: high school diploma or less, some education after high school but less than a college degree, or a college degree or more (275). For the purpose of sub-group analysis researchers additionally dichotomized these variables. For income level, the variable was dichotomized as less than $25000, or greater than or equal to $25000. Education level was dichotomized as high school, or at least some college education.
The study also required controlling for previously known risk factors for ovarian cancer. To isolate the exposure of interest researchers also collected data on “sociodemographic and lifestyle characteristics,” (275) including family history, BMI, age, reproductive history, contraceptive use and tubal litigation. Using this additional information, researchers were able to account for the fact that ovarian cancer is influenced by many different factors that are biological, social, economic and environmental in nature. In collecting this data, researchers adjusted for confounding variables to isolate the impact of socioeconomic status on risk of developing ovarian cancer. Researchers also included marital, employment, and health insurance status, and well as region of residence as additional indicators of socioeconomic status that were used in predicting risk.
2) What was the primary outcome of interest and how was it measured? (10 points)
The outcome of interest was ovarian cancer. Because this study was retrospective in nature, the outcome of developing ovarian cancer had already occurred, as a positive diagnosis of ovarian cancer was an eligibility requirement for the case population. The outcome of ovarian cancer was measured by identifying patient cases where there was a histological confirmation of the cancer by physicians. This information was primarily ascertained through cancer registries, including cancer registries in each state, Surveillance, Epidemiology and End Results (SEER) registries, or individual hospital registries that were in the geographic study area.
3) What were the eligibility criteria for the study population? Briefly describe these characteristics. (10 points)
The study population included both the control and the case population, for which there were two different sets of eligibility criteria. However, for the entire study population, participants were required to be women with enough English language proficiency to participate in the interviews. There were four pieces of eligibility criteria for the case population, specifically. The first was that all individuals had to be diagnosed with “histologically confirmed invasive epithelial ovarian cancer,” (Alberg et.al, p.275) meaning each participant had to have medical confirmation of ovarian cancer through biopsy or surgery. The second was that participants had to live in one of the following geographic areas: Alabama, Georgia, Louisiana, Michigan, New Jersey, North Carolina, Ohio, South Carolina, Tennessee, or Texas. The third criterion was that each individual had to self-identify as African American. Fourth, each individual had to be between 20 and 79 years old.
The eligibility criteria for the control population was the following: first, all individuals had to be female; second, participants could not have had a diagnosis of ovarian cancer at any point in their life; third, they must reside in Alabama, Georgia, Louisiana, Michigan, New Jersey, North Carolina, Ohio, South Carolina, Tennessee, or Texas. Fourth, they must self-identify as African American; fifth, they had to be between 20 and 79 yeas old; sixth, participants could not have had a bilateral oophorectomy, or the removal of both ovaries.
4) Is the source population fixed or dynamic? Is the study population fixed or dynamic? Briefly explain your reasoning. (10 points)
The source of the population would include all self-identified African American women between the ages of 20-79 who are English language proficient and reside in the geographic study area. This is a dynamic population because of the geographic study area. Participants are able to move in and out of the 10 states that define the study area. For example, an individual may meet all the eligibility requirements when living in Alabama, but if she chose to move to Rhode Island, she would no longer be eligible.
The study population would include both the control population and the case population, and could be considered fixed. The study population is fixed because once participants were found to be eligible, they were given a phone interview or questionnaire. There was no follow up required, which means that the moment participants became eligible and provided the required data to the study, it became a fixed event in their life that could not be undone. The BUSPH Measure of Disease Frequency Module suggests that participating in a study that requires no follow up, i.e., filling out a questionnaire, can be eligibility for becoming part of a fixed population. If, however, there were some follow up required, the population would be dynamic, because in the period of time between first contact and follow up, study participants could have moved out of the geographic area.
5) How was the study population vulnerable to the exposure? Briefly describe one way. (10 points)
Vulnerability means that a population is more likely to be exposed. Race is a social factor that makes African American women more vulnerable to the exposure. The study referenced well-documented disparities in the socioeconomic statuses of different races in the United States, especially between African American families and white families. African American families tend to have a lower annual family income than white families. Thus, if African American women are more likely to come from families that have lower incomes, they have increased exposure to low socioeconomic status, which is associated with higher risks of ovarian cancer. Thus, this population is more vulnerable to ovarian cancer.
6) How was the study population susceptible to the exposure? Briefly describe one way. (10 points)
Susceptibility means that an individual is more likely to suffer negative consequences from a particular exposure, than someone else exposed to the same thing. The study population was more susceptible to the exposure because of accumulated stress over the study population’s lifetime due to racial and gender disparities. If women, generally, in the United States experience accumulated physiological and psychological stress from the social impacts of gender inequality, African American women are more likely to experience even greater amounts of stress from the social impacts of gender inequalities, as well as well racial disparities. The compounded stressors of racial and gender inequality could augment the effects of low socioeconomic status, making this population more susceptible to the adverse effects.
7) Is the exposure best described as a risk factor or cause? Briefly explain your reasoning. (10 points)
The exposure is best described as a risk factor. While the study did demonstrate that higher levels of education was correlated with individuals having a lower risk of developing ovarian cancer, the study was not able to demonstrate why that association exists. A risk factor is a more appropriate way of categorizing the exposure because low socioeconomic status may increase the likelihood of developing ovarian cancer, but we cannot say for certain that if you are of low socioeconomic status you WILL get ovarian cancer. Not every single African American woman who is exposed to a low socioeconomic status will develop ovarian cancer. Ovarian cancer is multifactorial and influenced by many different factors. If low socioeconomic status were the cause of ovarian cancer, it would be both necessary and sufficient for the development of the disease. However, we know that individuals of high socioeconomic status can still develop ovarian cancer.
8) Ovarian cancer is multifactorial. What additional “component causes” were explored in this paper? Briefly describe how each variable was categorized. (Hint: See Table 4). (10 points)
Additional component causes of ovarian cancer included in this paper were marital status, employment status, medical insurance, and region of residence. Each of these are dichotomous variables that are categorized in terms of No/Yes; None or public/Private or other; and North/South. Each of these categories were indicators of higher economic status, and data consistently demonstrated there was an inverse association between risk of ovarian cancer and being married, having private insurance, and being employed.
9) Refer to the first paragraph of the Discussion on page 281. Did the study authors find an association between the exposure and the outcome? Briefly describe the two major findings discussed in this paragraph. (10 points)
Yes, the authors did find that there was an association between socioeconomic status and risk of ovarian cancer. The first major finding was that higher levels of education are associated with lower levels of risk for developing ovarian cancer. This was also the case for being married, having a job, having insurance, and the geographic area being lived in. Researchers also found that the inverse was true, where there was an association between low levels of education and higher levels of risk for developing ovarian cancer.
The second significant finding was the association between higher annual family income and decreased risk of ovarian cancer. This particular association was not as clear or as strong, but does suggest that higher income, especially in combination with other socioeconomic factors, may decrease risk. Due to the multifactorial nature of this disease, and that causes can be biological, social, and environmental, the subgroup analysis of this data did not have as strong an association.
10) Should we conclude that the exposure causes the outcome? Briefly make a case for your position, citing two quantitative methods concepts covered in Week 1 to support your argument. (10 points)
We cannot conclude that exposure causes ovarian cancer. We can, however, consider there to be an association. While the study did demonstrate that higher levels of education was correlated with individuals having a lower risk of developing ovarian cancer, the study was not able to demonstrate why that association exists. It is possible that socioeconomic status acts as a mediator for an unknown variable that does directly cause ovarian canser, lifestyle factors, for example.
The first concept that supports the argument that socioeconomic status is not a cause of ovarian cancer is direction. To state that something is a cause, it requires that the exposure happen before the outcome. However, it is possible for an individual to develop ovarian cancer, and then, for example, be laid off from their job and have a lower socioeconomic status. In this situation, the exposure did not precede the outcome. Likewise, someone of high socioeconomic status may still develop ovarian cancer without being exposed to this particular variable of interest.
A second concept that supports the argument that socioeconomic status cannot be considered a cause is statistical association. In the discussion of the study results and outcome, the authors of this research suggest that the statistical associations were not significant. The data suggests an association, but does not unequivocally suggest direct causation.