What is an Appropriate Value for C-for-Benefit in Medical Classification Models

I am currently studying this AI4M course and recently came across the concept of “C-for-Benefit”. The example in the lecture showed a model with a C-for-Benefit value of 0.6. This has led me to wonder: what are typical values for C-for-Benefit in models that are actually deployed in a clinical setting? My curiosity stems from the fact that general image classification models for diagnosing diseases usually have an accuracy above 0.9, which seems quite high compared to a C-for-Benefit value of 0.6. Would anyone with experience in this area be willing to share their insights? I would greatly appreciate your expertise. Thank you.

Dear @boy.skier ,
Welcome to the discourse community. I am a AI4M Mentor and I will do my best to give you a reply. Thanks a lot for asking this question. In my reply I will clarify concepts such as “C-For-Benefit” and “accuracy” and give answers to your additional questions.

C-for-Benefit, also known as the concordance statistic for benefit, is a metric used to evaluate the discriminative ability of a treatment benefit predictor. It represents the probability that, from two randomly chosen matched patient pairs with unequal observed benefit, the pair with greater observed benefit also has a higher predicted benefit. In other words, it measures how well a model can predict which patients will benefit more from a treatment compared to others.

It’s important to note that C-for-Benefit and accuracy are different metrics. While accuracy measures the proportion of correct predictions in a classification problem, C-for-Benefit focuses on the model’s ability to predict treatment benefits. Therefore, comparing these two metrics directly may not be appropriate.

In the example you mentioned, the C-for-Benefit value was 0.6. This value indicates that the model has a moderate ability to discriminate treatment benefits. However, it’s challenging to provide a typical range for C-for-Benefit values in clinical settings, as it may vary depending on the specific context, treatment, and patient population.

In summary, C-for-Benefit is a metric used to evaluate the discriminative ability of a treatment benefit predictor, and it’s not directly comparable to accuracy. The typical values for C-for-Benefit in clinical settings may vary depending on various factors, and it’s essential to consider the specific context when interpreting these values.

Please feel free to ask a followup if my reply was unclear to you. And feel free to ask more questions on Discourse. I will be happy to help as much as I can.
Can Koz

1 Like