people are stupid--and no I'm not like uh 17
So, here's a practice question for my exam. I'm posting it here so that anyone who is unfortunate enough to read this can be as thrilled as I am. Hey, and if you're reading it before July 5 at 8am EST give me a hollar and tell me what I left out! (oh yeah, because I'm timing myself--sans editing)
"Tobacco use is the single most preventable cause of premature morbidity and mortality in the United Stats. It is well documented that smokeless tobacco use is a major risk factor for a variety of oral pathoses such as gingival recession and leukoplakia (wow, how interesting).
The question of an association between smokeless tobacco use and oral cancer has important public health implications. Very few definitive studies exist to confirm this association.
You have been asked to design two studies (one longitudinal and one retrospective) to convice a scientific review committee that you deserve funding to carrry out the definitive studies. The population that you wil study is rural Appalchia in Ohio since they have a higher prevalence of smokeless tobacco use that most populations.
1. Describe the potential competing designs and analyses which you may employ in carrying out the studies. Describe the pros and cos of each design listed and indicate the appropriate statistical methods to be used in each.
A three-stage cluster sampling technique will be used in which the clusters are the Appalachian Ohio counties. At the second stage of the sampling process a the census bureaus list of addresses will be used to identify census tracks. At the third stage of the sampling process a systematic sample of every 3rd address within the census tract will be chosen. (This was based on the scarecly populated areas--better way to say this i'm sure)
Becuase the majority of smokeless tobacco users are male, both prospective and retrospective designs will restrict the sample to uninstitutionalized, male Appalachians over the age of 40. (This upper age cut-off was decided upon due to the expense involved in following persons over time and the rarity of developing oral cancer before this age--I'm making this up...but it sounds reasonable).
Within each chosen household, the male over 30 who had the most recent birthday was chosen to participate in the study. Participants were told about the study, it's benefits and potential costs and offered a small gift for participating. A consent form was signed before study participation. Each of the randomly chosen addresses was visited by an interviewer who assessed smokeless tobacco use, demographic factors such as age, race, income, education, etc. and other riks factors for oral cancer (alcohol use). Smokeless tobacco use was based on both self report and analysis of a saliva sample. A stratified random sample of these individuals (stratified by exposure status) were chosen for follow-up. Individuals chosen in this way were followed for 20 years or until development of oral cancer.
Pros of this method include the ability to establish temporal order of cause and effect, the lower risk of recall bias in regard to exposures, and the ability to capture changes in exposure status over time. Selection bias may also not be as much of an issue as it is in other study designs because of the random sampling of exposed and unexposed.
The primary disadvantage of this method is the expense associated with following individuals over time. Other disadvantages include selection bias due to selective attrition (those in the exposed group may find the follow-up more bothersome because we are always taking saliva samples to se if they still use smokeless tobacco or they may differ in some other way that makes them not want to participate). There is also the problem of diagnostic bias if exposed persons are monitored more carefully by their doctors (this could very well happen since tobacco use is a known risk factor for several cancers).
There are two retrospective methods we could use. A retrospective cohort which relies primarily on historical data. This does not seem to be a good choice since our population is not one that is monitored closely for any reason (I can't actually think of one that might be monitored and have this type of information). A second type of retrospective analysis is the case control study in which cases are identified and then surveyed or interviewed in order to find out about their exposure status. In this type of study, we must choose controls that are as much like the population that the cases came from as possible.
For the case control study, we used state cancer registry data to identify cancers occuring in Appalachian counties. We then selected a stratified random sample of these persons for study. A draw back of the case control approach is that we must rely on recall to ascertain past exposure status. There may be inherent bias to this approach since diseased persons often are better able to remember exposures (or more likely to search for a reason for their illness) than persons without the illness. Another drawback is that our controls must represent the same population as our cases. This is difficult in some situations. However, since our cases are from the community rather than hospitals, it is not such a problem. The biggest problem with case control studies is temporal order. We cannot necessarily establish that the smokeless tobacco use occurred before the development of cancer; the cancer may have already been developing due to other causes (although unlikely) before smokeless tobacco use was initiated. The biggest advantage of using the case control approach is the cost savings. We do not have to follow people over time until cancer develops.
"Tobacco use is the single most preventable cause of premature morbidity and mortality in the United Stats. It is well documented that smokeless tobacco use is a major risk factor for a variety of oral pathoses such as gingival recession and leukoplakia (wow, how interesting).
The question of an association between smokeless tobacco use and oral cancer has important public health implications. Very few definitive studies exist to confirm this association.
You have been asked to design two studies (one longitudinal and one retrospective) to convice a scientific review committee that you deserve funding to carrry out the definitive studies. The population that you wil study is rural Appalchia in Ohio since they have a higher prevalence of smokeless tobacco use that most populations.
1. Describe the potential competing designs and analyses which you may employ in carrying out the studies. Describe the pros and cos of each design listed and indicate the appropriate statistical methods to be used in each.
A three-stage cluster sampling technique will be used in which the clusters are the Appalachian Ohio counties. At the second stage of the sampling process a the census bureaus list of addresses will be used to identify census tracks. At the third stage of the sampling process a systematic sample of every 3rd address within the census tract will be chosen. (This was based on the scarecly populated areas--better way to say this i'm sure)
Becuase the majority of smokeless tobacco users are male, both prospective and retrospective designs will restrict the sample to uninstitutionalized, male Appalachians over the age of 40. (This upper age cut-off was decided upon due to the expense involved in following persons over time and the rarity of developing oral cancer before this age--I'm making this up...but it sounds reasonable).
Within each chosen household, the male over 30 who had the most recent birthday was chosen to participate in the study. Participants were told about the study, it's benefits and potential costs and offered a small gift for participating. A consent form was signed before study participation. Each of the randomly chosen addresses was visited by an interviewer who assessed smokeless tobacco use, demographic factors such as age, race, income, education, etc. and other riks factors for oral cancer (alcohol use). Smokeless tobacco use was based on both self report and analysis of a saliva sample. A stratified random sample of these individuals (stratified by exposure status) were chosen for follow-up. Individuals chosen in this way were followed for 20 years or until development of oral cancer.
Pros of this method include the ability to establish temporal order of cause and effect, the lower risk of recall bias in regard to exposures, and the ability to capture changes in exposure status over time. Selection bias may also not be as much of an issue as it is in other study designs because of the random sampling of exposed and unexposed.
The primary disadvantage of this method is the expense associated with following individuals over time. Other disadvantages include selection bias due to selective attrition (those in the exposed group may find the follow-up more bothersome because we are always taking saliva samples to se if they still use smokeless tobacco or they may differ in some other way that makes them not want to participate). There is also the problem of diagnostic bias if exposed persons are monitored more carefully by their doctors (this could very well happen since tobacco use is a known risk factor for several cancers).
There are two retrospective methods we could use. A retrospective cohort which relies primarily on historical data. This does not seem to be a good choice since our population is not one that is monitored closely for any reason (I can't actually think of one that might be monitored and have this type of information). A second type of retrospective analysis is the case control study in which cases are identified and then surveyed or interviewed in order to find out about their exposure status. In this type of study, we must choose controls that are as much like the population that the cases came from as possible.
For the case control study, we used state cancer registry data to identify cancers occuring in Appalachian counties. We then selected a stratified random sample of these persons for study. A draw back of the case control approach is that we must rely on recall to ascertain past exposure status. There may be inherent bias to this approach since diseased persons often are better able to remember exposures (or more likely to search for a reason for their illness) than persons without the illness. Another drawback is that our controls must represent the same population as our cases. This is difficult in some situations. However, since our cases are from the community rather than hospitals, it is not such a problem. The biggest problem with case control studies is temporal order. We cannot necessarily establish that the smokeless tobacco use occurred before the development of cancer; the cancer may have already been developing due to other causes (although unlikely) before smokeless tobacco use was initiated. The biggest advantage of using the case control approach is the cost savings. We do not have to follow people over time until cancer develops.