From an article in The Hospitalist: "The Top 10 Things ID Specialists Wish Every Hospitalist Knew" & from a blogpost "6 high-yield ID clinical pearls" (blogpost has lots of references to peer-reviewed journals): 
1. Beta-lactam/b-lactamase (ie unasyn, zosyn) have excellent anaerobic coverage, no need to add flagyl or clinda. Unasyn's gram neg coverage is not as good as zosyn.
2. Staph Aureus bacteremia = 2 weeks of IV antibiotics, even if its assoc with removable line. Less may lead to relapse with infections of bone/valves. 20-25% cases of S.aureus bacteremia are complicated by metastatic dx or endocarditis. Consider TE echo to r/o valve involvement. If there is no source, or there is cardiac/bone/joint involvement = 4-6 weeks of IV antibiotics.
-Staph Aureus in urine, look for another source of infection, like heart valves/bone. Exception: history of foley or GU procedure
3. Sensitivity of C.Diff stool toxin = 80-90%, which means 10-20% false negative rate-- negative test does not r/o c.diff. Stool toxin test that is positive remains positive during and after cure, do not use it to confirm treatment, follow the clinical course instead.
-Diarrhea that develops more than 1-2 days after admission = c.diff, almost never is other bacteria (salmonella, shigella, yersenia, campy) or parasite unless someone is immunocomp or recently traveled abroad.
-Absence of diarrhea does NOT rule out C.Diff.
-C.diff causes leukocytosis (WBC 30-50). A hospitalized pt with a history of abx and a high white count = c.diff
4. Uncomplicated cellulitis is most commonly strep (often group A), including in diabetics.
-Group A strep can take 3-4 days to respond to antibiotics, and worsen for the first 1-2 days because of the secretion of tissue-damaging toxins. If systemic signs (fever, WBC) are improving but the wound looks worse, it doesn't need different/more potent antibiotics, it needs time and elevation.
-Cellulitis in the context of venous stasis/edema: elevate and diurese aggressively, as the infection won't improve until the edema is gone.
5. Quinolone monotherapy should not be used for serious GN infections because of expanding GNR quinolone resistance (incl >50% of pseudomonas in some hospitals).
6. VRE in the stool does not need to be treated, as most people will clear it on their own and not become asymptomatic. Unless you are a liver transplant recipient, in which case VRE in the stool needs to be aggressively treated.
7. Candida in the blood is always real (urine or sputum it might be a contaminant). Take it seriously, treat with mica (more effective than azoles), removing the line is not enough, get an optho evaluation. If there is candida in the urine, treat if there are symptoms.
8.  Coagulase-neg staph is a contaminant in blood the vast majority of the time; odds of getting a + culture before abx: ~40%, after abx: <20%
9. New guidelines for UTI tx: 
-Nitrofurantonin, 500mg BID for 5 days
-Bactrim 160/800mg BID for 3 days if local resistance rates are low (<20%)
-Fosfomycin trometamol 3g single injection: resistance rates are low worldwide, possibly less effective than the other treatments
-Pivmecillinam 400mg BID for 3-7 days
-Fluroquinolones are bigger antibiotics, consider if the others fail
-Do not use amp or amox, resistance is too high.
10. With normal vital signs, pneumonia is extremely unlikely. With one abnormal VS (temp, RR, pulse, O2), risk is ~12%, with 4 abnormal, it is 70%. (chart review, 2007 Am J Emer Med n>4000)