likelihood ratio (LR)

Change in odds of disease for a test result. An LR of > 1 makes disease more likely. An LR of < 1 makes disease less likely. defined as: probability of obtaining result in patient with specified disease / ----------------------------------------------------------------- probability of obtaining result in patient without specified disease In the case of dichotomous results: LR associated with abnormal test (positive LR): sensitivity/(1-specificity) LR associated with normal test (negative LR): (1-sensitivity)/specificity positive LR of 2, 5, & 10 increase probability of disease by ~ 15%, 30%, & 45% respectively. For example, if the pretest probablity of a condition is 50%, a likelihood ratio of 10 yields a post-test probability of 95% (50% + 45% = 95%). negative LR of 0.5, 0.2, & 0.1 decrease probability of disease by ~ 15%, 30%, & 45% respectively. For example, if the pretest probablity of a condition is 50%, a likelihood ratio of 0.1 yields a post-test probability of 5% (50% - 45% = 5%).

References

- Medical Knowledge Self Assessment Program (MKSAP) 11, 14, 17. American College of Physicians, Philadelphia 1998, 2006, 2015