Health literacy is a huge limitation in making effective use of modern medicine. Depending on how you measure it, up to 90% of Americans may have problems following the instructions that doctors and nurses give them, or reading a prescription well enough to understand how to take medicine correctly. Even for those who do understand, the fine details can be easily lost as well. Two recent studies evaluated the communication of risks with interesting results.
First, from the Annals of Internal Medicine Woloshin et al report a study of 2,944 patients who were given data about risks that was presented in one of three ways: as a percentage chance (1% or 10%), as a natural frequency (ie: 10 in 1,000 or 100 in 1,000), or a variable frequency (ie: 1 in 10,000 or 1 in 1,000). Then the patients took a test of comprehension. They were more likely to pass the test if the data were presented as percentages. Unfortunately, the best pass rate was still only about two-thirds.
Second, from the Archives of Internal Medicine, Foppa et al report on 167 study participants (including patients, medical students, and attending/resident physicians) who were asked to give a percentage to a variety of qualitative descriptors (ie: certainly, likely, possibly, rarely, and so on). The figure from this study tells the entire story (if you can access it) but basically and not suprisingly, the range of interpretation is very wide. For example, if I told a patient that a certain complication was "unlikely", this data suggests that my patient could interpret that as being somewhere between a 0% and 40% chance of happening! Doctors are sometimes reluctant to give out precise estimates because people sometimes perseverate on them. For example, no doctor I know has even "given" someone 3 months to live, but they might say that an average life expectancy for a given patient is 3 months. Subtle, but different. However, for instances where we do have strong data, this would suggest that the more precise the communication, the better.