Reference no: EM13858367

Part 1:

A good proposal will be organized around three paragraphs.

The first paragraph should describe the public health problem to which the epidemiologic research topic pertains. How many cases of the disease occur per year? How are these cases distributed in the population (e.g., by gender, age, race/ethnicity, geography, socioeconimic status, etc)? What is the research paper or research topic and how does it fit into the overall context of the public health problem? For example, if the paper or topic investigates an association between an exposure and the disease of interest, how prevalent is the exposure? how strong might the association be? what is the potential for an intervention to reduce the exposure prevalence? After reading this paragaph, I should understand the scope of the public health problem and how the research paper or topic fits into that problem.

The second paragraph should describe the particular epidemiologic research paper or topic that will be the focus of the project. Summarize the design and results (if available) before bias analysis. Describe the data available for bias analysis (record-level data if you have the dataset, crude or stratified 2x2 table if using a publication).

The third paragraph should summarize one bias that threatens the validity of the result (unmeasured confounding, selection bias, measurement error (preferably misclassification)). If you will use a publication, what have the authors said about this bias? Conclude with a statement of why it is important to quantify the bias before relying on the research paper or topic in the context of a public health program.

Part 2:

The second part of the semester project builds on Part 1. In Part 1, you identified a study or dataset susceptible to at least one bias, and assured that the primary data or applicable 2x2 table was available to assess the bias. In the second part, you will describe the structure of the bias, describe the bias model to evaluate the bias, and describe the sources you will use to assign values to the bias parameters of the bias model. The assignment will be due by 11:59 PM ET on Sunday October 4, 2015. The document should be no more than one page long, excluding references if any, and use one inch margins and 12-point font. The assignment should be submitted via Blackboard.

A good assignment will include at least these three paragraphs.

•Paragraph one: describe the structure of the bias. If selection bias, describe how initial or longitudinal participation in the study may have been affected by both the exposure and the outcome. If information bias, describe how one of the variables has been mismeasured or misclassified, and whether the measurement error is likely to be differential and/or dependent. If unmeasured confounding, describe the variable that has not been measured, and why it is likely to be associated with both the exposure and the outcome, without being on the causal path from the exposure to the outcome.

•Paragraph two: describe the bias model that you will use to evaluate the bias. You can use equations, or you provide a description of the bias model. Equations for each type of bias can be found in the textbook, and examples in the textbook provide an idea for how to describe the bias model without using equations. By the end of this paragraph, I should understand what bias parameters you will use in the bias model to evaluate the bias. For selection bias, this may include selection proportions for each combination of exposure and outcome. For misclassification, this may include sensitivity and specificity of classification, or positive and negative predictive values of classification. For an unmeasured confounder, this may include the strength of the association between the exposure and the confounder and between the outcome and the confounder.

•Paragraph three: describe the sources you will use to assign values to the bias parameters of the bias model. For example, has there been an internal validation substudy that measured selection proportions (selection bias), sensitivity or specificity (information bias), predictive values (information bias), or the strength of association between the confounder and exposure and between the confounder and outcome (unmeasured confounder)? If there has been no internal validation study, will you use instead external information to obtain the values to assign to the bias parameters and from where will that external information come (presumably published information from another population). Defend the applicability of the external information to your study population. If there is no internal or external information, you will have to make an educated guess. What will be the basis for that educated guess? Will it in any way incorporate published data to support the guess (hopefully so)?

Part 3:

The third part of the semester project will implement bias analysis methods. The assignment should require no more than two pages to present (1 inch margins, 12-point font). Note that you do not have to interpret the results of the bias analysis for this assignment. The assignment will be due by 11:59 PM ET on Sunday October 25, 2015. The assignment should be submitted via Blackboard.

A high quality assigment should contain the following (or a closely related adaption, depending on the project topic and design):

•Show a crude table (most often 2x2) that shows the data used to estimate the crude association. For example, a case-control study with dichotomous exposure would show. Also show the crude estimate of association (for example, the case-control odds ratio).
•Show the bias model equation, first with the bias parameters, and second with values assigned to the bias parameters. The bias model and values assigned to the bias parameters were part of the second assignment.
•Show a revised table showing the data after bias modeling. Also show the bias adjusted estimate of association.
•Vary at least one, and no more than two, of the values assigned to one or two of the bias parameters of the bias model. For each combination of bias parameters, calculated the bias adjusted estimate of association. Summarize the results of this sensitivity analysis of the bias analysis in a multidimensional bias analysis table.
Do not use probabilistic bias analysis methods for this assignment.

Part 4:

The objective of the fourth, and final, part of the semester project is to present and interpret the results of your bias analysis. The assignment should be no longer than one page of text (1 inch margins, 11-point font). A second page can optionally include one-figure or table and literature citations. The assignment will be due by 11:59 PM on Sunday November 15, 2015 and should be submitted via Blackboard. A high quality submission will have three paragraphs.

Paragraph 1: Summarize the objective of the research, the conventional result, and the bias that may affect the summary result. Avoid duplicating information provided in assignment 1; summary description is the key.

Paragraph 2: Summarize the bias analysis method that was applied, including summary descriptions of the bias model, its parameters, how values were selected for the parameters, and the results of the bias analysis. Equations are not necessary or encouraged. Avoid duplicating information provided in assignments 2 and 3; summary description is the key.

Paragraph 3: Summarize the inference that you and/or the original authors made based on the conventional result, and how that inference changed as a result of the bias analysis. Focus on the direction, magnitude, and uncertainty in the estimate of association that was yielded by the bias analysis. It is possible for the original inference to be strengthened (e.g., the bias did not seem so large, or the bias adjusted result was further from the null than conventional estimate). It is also possible for the original inference to be weakened (e.g., the bias adjusted result or the additional uncertainty calls to question whether the original result may have been wholly or largely due to the bias). Conclude with one sentence about what, if any, further research may be useful.