Odds Ratios—Current Best Practice and Use With Dr Norton
Nov 2, 2023
auto_awesome
Dr. Edward C. Norton discusses the best practices and applications of odds ratios in different fields, from gambling to medical research. They also explore the impact of controlling for factors in regression and demonstrate how odds ratios can be altered by various models. Additionally, they discuss the significance of controlling for specialty when examining the effect of gender on industry payments.
Odds ratios are commonly used to measure the effects of treatments or risk factors on the probability of a patient experiencing a specific outcome.
Adding clinical risk factors in logistic regression models can alter the magnitude of association and necessitate alternative measures for effective interpretation.
Deep dives
Understanding Odds and Odds Ratios
Odds are a ratio of the chance of something happening divided by the chance of it not happening. They are useful for gamblers to determine fair bets. Odds ratios are commonly used to measure the effects of treatments or risk factors on the probability of a patient experiencing a specific outcome. The odds ratio compares the odds of the outcome for one group to the odds of the outcome for another. An odds ratio of 1 indicates that the outcome is equally likely for both groups, while an odds ratio below 1 suggests a lower likelihood of the outcome in the treated group.
Interpreting Odds Ratios
The value of 1 is a significant reference point for odds ratios. An odds ratio of 1 signifies that the odds for one group divided by the odds for the other group are equal. When assessing the effectiveness of a treatment, an odds ratio close to 1 suggests minimal impact. For beneficial outcomes, a higher odds ratio implies a greater likelihood in the treated group. Conversely, for negative outcomes like mortality, a lower odds ratio indicates a reduced likelihood of the outcome in the treated group.
The Impact of Statistical Models on Odds Ratios
As more clinical risk factors are considered in statistical models like logistic regression, the odds ratio can either remain consistent or become stronger. Adding clinical risk factors in linear regression enhances the precision of the treatment effect estimate, while in logistic regression, it can alter the apparent magnitude of association. An article by Trungali et al. analyzing physicians' industry payments revealed that controlling for specialty, a confounder related to gender and payment likelihood, resulted in a smaller odds ratio. Effective interpretation of the association between gender and industry payments may necessitate alternative measures, such as relative risk or risk difference, in addition to explaining the overall probability of receiving an industry payment.