The following questions are about a study that is interested in the relationship between obesity (exposure) as defined by a BMI cutoff of >30, and incident hypertension (outcome).

The following questions are about a study that is interested in the relationship between obesity (exposure) as defined by a BMI cutoff of >30, and incident hypertension (outcome). Assume that the BMI gathered for that study are subject to information bias because investigators relied on self-reported height and weight to calculate BMI in the study. The following questions consider the impact of this measurement error of the exposure on study findings.

The following 2×2 table includes the data from your study on the relationship between obesity (exposure) and incident hypertension (outcome); this table represents the misclassified table; that is, BMIs reported with information bias are used. Calculate the risk ratio for incident hypertension among individuals with obesity as compared to individuals without obesity.

  YES incident HT NO incident HT Total

Obese 40 14 54

Non-obese 16 42 58

Total 56 56 112

Construct a new 2×2 table under the scenario that 20% of participants with true obesity under-estimated their BMI which lead them to be categorized (incorrectly) as non-obese in the study data; this table represents the true data. Assume that persons who were classified as obese correctly reported their BMI to study investigators. Based upon this “truth” table, calculate the risk ratio for incident hypertension among individuals with obesity as compared to individuals without obesity. 

In your table for question #6, draw arrows between the cells to reflect how people are migrating between the cells of your 2×2 table (from truth to misclassified) due to this exposure measurement error. Is this non-differential or differential exposure measurement error? 

Given the movement between the cells, explain the differences in the observed risk ratio (from question #5) vs. the true risk ratio (from question #6). 

The outcome data on hypertension represents the truth regarding incident hypertension given the great lengths investigators went through to collect this data (measured blood pressure, questionnaire data, and medication information). Consider the scenario whereby investigators instead, due to lack of funding, had to eliminate in-person follow-up visits and relied upon telephone contacts to determine hypertension status through only questionnaire information. As a result, new hypertension cases were missed; among individuals with obesity, 30% of true incident hypertension cases were not reported and among individuals without obesity, 50% of incident hypertension cases were not reported.

How would your 2×2 table look under this scenario? Calculate the risk ratio with the misclassification under this scenario. 

Draw arrows between the cells to reflect how people are migrating between the cells of your table (from truth to misclassified) due to this outcome measurement error? Is this differential or non-differential misclassification? 

Given the movement between the cells, explain the differences in the misclassified risk ratio (from question #9) vs. the true risk ratio (from question #5). For the purposes of this question, ignore the exposure misclassification from above and consider the table from question #5 as your true table.