It may sound far-fetched when you hear that people can predict a storm based on the pain in their arthritic joints, but there is actually some science to back this up.
A few studies (example here) have indeed found that knee pain is independently correlated (not due to other factors like age, medication use, etc.) to barometric pressures. Considering that storms often signal a change in atmospheric pressures, there is some plausibility to these claims.
One explanation of this phenomenon comes from joint specialist Dr. Javad Parvizi:
When pressure in the environment changes, we know that the amount of fluid in the joint or the pressure inside the joint fluctuates with it…Individuals with arthritic joints feel these changes much more because they have less cartilage to provide cushioning.
However, it should be noted that pressures in the atmosphere can change without a storm appearing, so the predictive powers of knee pain are probably shoddy at best. Furthermore, though a few studies have found connections between air pressure and joint pain, and even a plausible mechanism, there is to date little to no statistical evidence that the phenomena could be used for accurate predictions.
Confirmation bias (remembering evidence that confirms your beliefs and forgetting/ignoring evidence which does not) probably gives people a largely exaggerated (or false) sense of their “predictive” powers. For example, people’s knees also hurt when there is no storm around the corner, but these instances are not incorporated into the belief.