What does a calibration slope indicate in a risk prediction model?

• A. A slope of 1 indicates perfect calibration.
• B. If the slope is less than 1, predictions are too modest.
• C. A calibration slope near 1 indicates good calibration; slope less than 1 indicates overfitting; slope greater than 1 indicates underfitting.
• D. Calibration slope relates to intercept only.

Answer: (C)

Calibration slope captures how well the predicted risk scores track the actual outcomes across the entire range of risk. The ideal slope is 1, meaning the predictions increase with observed risk at the correct rate.

If the slope is near 1, predictions align with reality across low, medium, and high risk groups, signaling good calibration. When the slope is less than 1, the relationship is too flat: high predicted risks become too extreme and low predicted risks become too low, so the model exaggerates differences—this is a sign of overfitting. If the slope is greater than 1, the relationship is too steep: the model doesn’t differentiate enough across risk levels, underestimating differences at the high end and overcompensating at the low end—this points to underfitting.

So, the statement that a slope near 1 indicates good calibration, with slope <1 indicating overfitting and slope >1 indicating underfitting, correctly describes how the calibration slope behaves.