Or none of the above. So, you can apply statistics to medicine in the form of incidence of a disease, and likelihood ratios to increase or decrease the chance o

Or none of the above. So, you can apply statistics to medicine in the form of incidence of a disease, and likelihood ratios to increase or decrease the chance of disease in diagnosis. The problem is, the data for these is so difficult to interpret. Likelihood ratios in papers often have huge confidence intervals (like anywhere from 2-15), making their use problematic. Then often, they are quoted only for very specific patient groups, and applying them out of these groups invalidates the rule. Then things like prognosis are so variable, taking quantitively becomes a bit pointless. It’s especially true in cases where treatment or investigations are changing constantly, such as cancer. For diagnosis, the history is often more useful. Humans are story tellers, and the description of the problem varies hugely between patients, as does the significance of certain symptoms or descriptions. It’s an art to work out what is the real symptom and what isn’t. To work out the real picture isn’t always easy either. Ultimately for things where there is relatively good evidence to practice statistics they are made into algorithms (see MDCalc for examples) but again it’s hard to apply them out of the correct patient population. Even these rules seem to poorly apply in many situations. Then, real hard statistics rarely change management. You think of the reasonable causes and risk reduce as far as possible. Knowing the actual probability doesn’t really change what you can do. Same kind of goes for prognosis. Most patients don’t really know what to do with exact numbers (especially when they vary so much between different patients). So, ultimately this is why doctors don’t use statistics regularly.