Chapter II | The Language of Decision Making 6. Relevant Prevalence Title: The Use of "Overall Accuracy" to Evaluate the Validity of Screening or Diagnostic Tests Authors: A. Alberg, J. Park, B. Hager,, M. Brock, and M. Diener-West Source: J Gen Intern Med 2004: 19:460-465. Overall Accuracy is defined as Overall Accuracy = (Prevalence) x Sensitivity + (1 - Prevalence) x Specificity. The prevalence of a condition is just the probability that the condition will be present in the population studied. The authors of this study discuss the pitfalls of using overall accuracy as a summary measure of how well a test is doing. The following table regarding a screening test for liver cancer was recreated from summary numbers presented in this article. a. What is the sensitivity of the test based on these screening results (rounded to three decimal places)? TABLE 8.19 Cancer Test Results i. 0.949 li. 0.050 Truth Regarding Cancer Totals ill. 0.400 Predicted by No Liver Liver iv. 0.922 Test Cancer Cancer b. What is the specificity for the test based No Liver 543 18 561 Cancer on these screening results (rounded to three decimal places)? Liver Cancer 12 41 i. 0.949 Totals 572 602 ii. 0.050 iii. 0.400 iv. 0.922 c. What is the overall accuracy for this cancer screening test based on the screening results above (answer rounded to three decimal places)? 1. 0.949 11. 0.050 ili. 0.400 iv. 0.922 d. Overall accuracy is not a completely reasonable concept in this application? Explain why it is not. What is causing the problem? Be specific