What are the two factors that impact Cohen's kappa?

• A. Prevalence and symmetry
• B. Bias and prevalence
• C. Sensitivity and specificity
• D. Bias and symmetry

Answer: (B)

Cohen's kappa measures agreement between two raters beyond what would be expected by chance, and two main factors that shape its value are how the data are distributed across categories (prevalence) and whether the two raters use the categories differently (bias).

Prevalence is about the imbalance in category frequencies. When one category is very common and the other rare, the margins become skewed, which changes the expected chance agreement. Even if the raters agree a lot in the common category, the way the marginal totals line up can push kappa down or produce counterintuitive results. This is the prevalence effect.

Bias refers to systematic disagreement between the raters, captured by differences in how often each rater assigns a given category. If one rater tends to label more cases as positive while the other tends to label more as negative, the marginals diverge and the expected agreement increases or decreases in a way that can reduce kappa, even when raw observed agreement seems decent.

Sensitivity and specificity describe accuracy against a gold standard for a single rater, not the agreement between two raters. Symmetry isn’t a standard descriptor of what drives kappa in this context, whereas bias and prevalence directly influence the calculation of the chance agreement that kappa corrects for.

So, the two factors that impact Cohen's kappa are bias between raters and the prevalence of categories.