ABSTRACT
This final chapter of the book will present some additional advanced topics in statistical techniques that are employed for the analysis of data encountered in diagnostic medicine. For example, verification bias is an important topic that occurs when not all of the test results are subject to a gold standard. Screening for breast cancer is a good example of this. When the mammogram is positive, the patient is usually sent for biopsy, which serves as the gold standard. On the other hand, if the test result is negative, a biopsy is usually not performed. In such as situation, follow-up for those patients who test negative serves as the gold standard, however, this requires time and years can pass before the results are confirmed. This is a case of extreme verification bias and it is not possible to directly estimate the usual measures (true and false positive fractions) of test accuracy. Less extreme forms of verification bias occur when a certain percentage of negative test results are subject to the gold standard. For example, negative test results may be subject to the gold standard if there are other patient characteristics that put them at high risk.
