SurveyUSA for Firedoglake (1/16-18, likely voters):
Tim Bishop (D-inc): 47
Randy Altschuler (R): 45
First of all, there is a pretty glaring peculiarity from this poll: it contains almost no young voters in its sample. Voters between the ages of 18-34 make up exactly 1% of its sample. Look: I wouldn't be surprised if the youth vote is severely depressed in November, but that seems a little extreme. (Similarly, SUSA's AR-02 poll had a 5% sample of young voters, and 3% in OH-01.)
I have no interest in discussing the healthcare portions of this poll (and neither should you), but the top lines are still worth having a look at. Critics of a poll like this might point to an Altschuler internal poll from mid-November that had Bishop ahead by 46-26. However, I think you can defensibly argue that the environment has gotten weaker for Democrats in the interim. Remember, this is not a deep blue district -- Obama only won Bishop's CD by 52-48. Moreover, this is one of the few races that have actually been engaged: Altschuler went on the air in late December with blistering ads hitting Bishop on his TARP votes. Still, the lack of young voters here seems like a pretty glaring weakness of SUSA's likely voter screen. I would suspect that a more realistic model would give a few extra points to Bishop, but that still wouldn't be enough to say that he's out of the woods.
UPDATE: Nate Silver has noticed the demographic quirk, too, and is currently "doing some digging" on the poll.