The overall feel of the image is good and I love how you integrated the model into the image. The way you modeled her, it looks as if she was a doll, as there is absolutely no blood at the edges. Therefore the blood on her skin, where it is complete, looks as if it came from somewhere else, giving us a reason to think on why she is in the water (she's trying to wash it off?).
There are a few points I'd change if I was to redo the image:
*the whole background is a bit to dark, remember that it doesn't need to be hard to see to have a spooky, depressing atmosphere.
*the lighting situation is a bit different on the model than on the background, there is a definite lightsource on the left of the model, but the background has a more diffuse lightsource. I'd try to lighten the darker side of the model more, as well as darken the lighter side a bit.
*I'd try to continue the model into the water, keeping in mind that water is transparent, even though it changes the lights direction. It would look less cut off and placed on top of the water if you'd continue it a bit under the water level, using a gradient shortly after and lowering the saturation of the under-water part.
Oh, and I think it's a shame you can't really see any details on her hair!
Vision it is a great concept, even though the concept of keeping a better place in yourself than you could be able to find outside yourself isn't the newest one, your combination of dark/light, enclosure/freedom, desaturation/colour and what else is pretty unique and well choosen.
Originality as said before, it isn't the newest idea, but where do you really find new ideas? Your choice of places is original, though.
Technique you mastered the basics of your program, as well as those of composition and manage to place your 'better place' in a way that draws the viewer in.
Impact the impact is strong, it leaves the viewer thinking about it. Thinking why she doesn't like the place she is in, where the better place is and why does it break out of her? I like it when pictures ask questions.
Critique courtesy of *devCRIT