The outcome is the dependent factor in your study — what you think your exposure causes.
The most basic form of the hypothesis is: X causes Y, where Y is the outcome.
Example: Smoking causes cancer. Cancer is the outcome.
(outcomes are in italics)
Your outcome should follow your exposure.
Example: Red spots cause chickenpox.
Red spots are a direct result of chickenpox infection.
Reformulated: Chickenpox causes red spots.
Exposure and outcome are relative.
Example: Household crowding is associated with gastroenteritis.
There is nothing wrong with this example. However, it is important to remember that one study’s outcome is another’s exposure.
Example: Gastroenteritis causes dehydration.
Your outcome must be measurable.
Example: Botulism makes babies floppy.
While this is true, and it may be legitimate to create a "floppiness scale" to measure floppiness with, if you are primarily interested in neurologic effects, reformulate.
Reformulated: Botulism causes progressive descending paralysis in babies.
Any point in the disease process can be an outcome. It is believed that some gastric cancers develop along the following pathway.
If you believe that Helicobacter pylori causes gastric cancer, and want to study this association, it is important to remember that gastric cancer takes a long time to develop, and a prospective study may be impractical (but certainly not impossible). However, using intermediate steps or biomarkers as outcomes can make a study far more tractable. Therefore, you may choose to study the relationship of H. pylori infection and intestinal metaplasia, with the goal that your conclusions will have some relevance to gastric cancer.
At the far end of the spectrum, death due to your disease of interest is another popular outcome, particularly for clinical trials. Using death as an outcome cuts to the heart of the question — it can never come before the exposure!