A hospital wants to estimate the prevalence of diabetic nephropathy in the surrounding population of adults with type 2 diabetes. Kidney biopsy samples are obtained from 500 adult patients with diabetes who receive care at the hospital. The samples are then interpreted by 10 different pathologists, 5 of whom work at the hospital and 5 of whom work at nearby institutions. A preliminary analysis shows that the pathologists who work for the hospital are 3 times more likely to interpret the biopsy samples as diabetic nephropathy compared to those who do not work for the hospital. Which of the following most likely explains this difference in interpretation?

Question

A hospital wants to estimate the prevalence of diabetic nephropathy in the surrounding population of adults with type 2 diabetes.  Kidney biopsy samples are obtained from 500 adult patients with diabetes who receive care at the hospital.  The samples are then interpreted by 10 different pathologists, 5 of whom work at the hospital and 5 of whom work at nearby institutions.  A preliminary analysis shows that the pathologists who work for the hospital are 3 times more likely to interpret the biopsy samples as diabetic nephropathy compared to those who do not work for the hospital.  Which of the following most likely explains this difference in interpretation?

Quizplus · Accepted Answer

Observer bias (also known as expectancy bias) occurs when investigators misclassify data due to preconceived expectations or prior knowledge concerning the study or its participants. This type of bias is particularly important when outcomes are subjective (ie, involve personal interpretation of clinical, microscopic, or radiographic findings). In this case, the pathologists who work at the hospital where the study is being conducted may know that the study is investigating diabetic nephropathy, or they may have access to additional medical records that indicate diabetes status. Conversely, the pathologists at other institutions are more likely to be blinded to the study's objectives and the patients' medical history. (Choice A) A confounder is an extraneous variable that is related to the exposure under investigation but that is also an independent risk factor for a disease. Confounding distorts the relationship between the exposure and the disease. For example, fatty food intake is a potential confounder in a study evaluating the association between physical activity and obesity, as people who do not exercise regularly may be more likely to consume a high-fat diet. (Choice B) Lead-time bias refers to the apparent prolongation of survival in patients who undergo a screening test that allows for earlier diagnosis but does not actually improve prognosis. For example, a patient with pancreatic cancer presents with metastases at age 58 for which he receives treatment, but he dies at age 60. Had the same patient been screened and diagnosed with pancreatic cancer at age 55, he would have received earlier treatment, but the disease could have progressed such that he would have still died at age 60 (ie, age at death would not have been affected by screening). (Choice D) Recall bias results from inaccurate recall of past exposure by people in a study; it applies most often to retrospective designs such as case-control studies. People who have experienced an adverse outcome are more likely to recall previous risk factors than people who have not experienced an adverse outcome. (Choice E) Selection bias can occur with inappropriate (ie, nonrandom) selection methods or through selective attrition of the study participants. For example, many patients drop from a clinical trial due to the severity of side effects associated with the treatment. This type of attrition is selective (ie, different attrition rates between groups) and reduces the generalizability of the study. Educational objective: Observer bias occurs when the investigator's evaluation is affected by preconceived expectations or prior knowledge, typically leading to overestimation of the disease association or treatment effects. This type of bias can be reduced by conducting a blinded study in which observers are unaware of study details and patient characteristics that could unduly influence them.

A Hospital Wants to Estimate the Prevalence of Diabetic Nephropathy