In a linkage study, individuals participating in a population health survey (the Manitoba Heart Health Survey of 1990-1991) were linked with the Manitoba database of health care utilization, consisting of claims submitted by physicians and hospitals for reimbursement. Each claim contained personal data as well as the diagnoses of the conditions treated. It was found that of 2,731 survey participants, 205 had a self-report of diabetes, 186 had a physician/hospital claim for diabetes, and 145 had both. If we take the physician and hospital claims data as the “gold standard”: (4 marks)
a) What is the sensitivity of survey self-report as a means of detecting diabetes? (1 mark)
b) What is the specificity, and how would you define “specificity” in this situation? (2 marks)
If someone in the survey reports having diabetes, what is the likelihood that he or she actually has a physician or hospital claim for diabetes? What do you call this measure?