If a risk model result does not make sense to you, it may actually not make sense, but it is still what that model produces in that situation. In other words, the model result may not be intuitive or may actually be incorrect clinically in certain situations. You need to look at the models in the context of what you want to accomplish.
This screenshot from Hughes RiskApps (Figure 1) show an extreme example as to how the models calculate breast cancer risk differently for the same patient. In most cases, the models are more consistent.
There is a reason for these discrepancies, and the example above is a patient that accentuates these differences. It does not mean the models are useless. It means they are only useful in the context of what you are trying to accomplish.
Figure 2 below helps to show what each model should be used for, and also shows what inputs are used for each. For example, as BRCAPRO does not recognize Atypical Hyperplasia, it would be the wrong model to determine breast cancer risk for a patient with only ADH and no family history.
If you think the calculations shown by Hughes RiskApps are incorrect, you should download the original model and compare. You will usually find the results from Hughes RiskApps are the same as the results when the model is run directly(Within a few percentage points, accounting for varying conventions for missing or incomplete data).
If you try this comparison, be sure you enter the EXACT data you entered into Hughes RiskApps.
If the numbers come out differently, there are several possibilities:
- Rounding error (Hughes RiskApps says 13 but the model directly says 14)
- The data you entered into the model directly differed from what you entered into your system (Note: This happens a lot. Patient was 45 in your system but 35 was entered into the model directly)
- Hughes RiskApps has a bug. (Note: We strive to make a high fidelity product but rely on the user community to constantly make improvements. With release 3.1, there are no known bugs in our risk calculations)