This post is derived from talks given by Dr. S. Andrew Josephson, neurohospitalist at UCSF, at the 2014 SHM national meeting (neurology pre-course). The material overlaps with a talk he also gave at UCSF's 17th annual hospital medicine conference about which I posted here. Dr. Josephson wrote the section on delirium in the latest edition of Harrison's. It is well worth the read for those who want to dive deeper on this topic.
Delirium: definitions, diagnosis
Delirium is defined, and distinguished from dementia, in terms of both its acuity (hours to days) and the specific dimension of cognition that is impaired, principally attention though just about any domain can be affected.
Delirium is thus, in a sense, an acute ADD and best tested by an attention maneuver, e.g. repeating digits forward. Less than 5 (the average person can do 7) indicates an attention deficit and therefore suggests delirium.
Demented patients who are not delirious can pass this test. Put another way, demented patients can be assessed for delirium. On the other hand, testing for dementia (via a tool such as the MMSE) may not be possible in delirious patients due to the prevailing attention deficit.
Confusion of these categories may result if the patient has Lewy body dementia since that condition is characterized by acute fluctuations in cognition, as part of the disorder itself, as well as visual hallucinations and periods of agitation. Additional references on that topic can be found here.  
The relationship between delirium and dementia
While the two disorders are defined separately there is some overlap and many patients have both. Though severe sepsis, intoxications and metabolic disturbances can produce delirium in a young person with a previously healthy brain, in some cases dementia or a previously asymptomatic early neurodegenerative process is the underlying substrate which makes the patient susceptible. In such instances delirium can be considered a failed stress test in an aging brain. Delirium triggers in these patients, who tend to be elderly, are milder and include drugs and infections. Anticholinergics top the list of offending drugs in the elderly. Infections are more often than not outside the CNS but, adds Dr. Josephson, you should have a low threshold for doing an LP if the patient's mental status does not improve promptly with treatment. (As an aside, an abnormal urinalysis without fever or localizing symptoms is a common weak explanation for altered mental status and a often a pretext for inappropriate use of antibiotics).
Treatment of delirium
You are far more likely to be called for agitated delirium than for non-agitated delirium and when you are called the first request, likely as not will be for pharmacologic treatment, or “something to calm the patient down.” While drugs are sometimes indicated that's the least favored option and belongs at the bottom of the list.
Treatment or removal of the underlying precipitant is the most important component. The second modality is management of the patient's environment by verbal reorientation, family visits, elimination of unnecessary lines and catheters, etc. (Believe it or not, those non-pharmacologic measures are effective). Drug therapy is sometimes necessary but it's a last resort and the wise clinician approaches it with hesitation. (Unique exceptions, appropriately treated with benzodiazepines, are alcohol or sedative withdrawal and stimulant toxidromes).
So what if you have to resort to drugs? There are a few caveats. First, benzos are generally to be avoided as they will convert agitated delirium into a non-agitated delirium (which may be worse) and ultimately prolong its duration. As stated above alcohol withdrawal and a few related conditions for which benzos are first choice are the exceptions. Antipsychotics (major tranquilizers) represent the only remaining alternatives in the pharmacopoeia. They are generally regarded as the drugs of first choice but there is no evidence that they improve ultimate outcomes. They are associated with cardiac risks and are contraindicated in dementia with Lewy bodies (DLB), in which case they can produce autonomic instability, a worsened alteration of mental status and a rigid state which may not be reversible. (There are a couple of antipsychotics in the atypical class, eg quetiapine, which may be permissible in very low doses).
Those are some sobering facts and represent a lot of stuff to sort through on a busy call night when you are being pressured to prescribe something to calm the patient down!
Things you might not think of
HSV-1 meningoencephalitis may have nonspecific manifestations and be associated with negative test results initially so consider early empiric treatment.
Seizure related altered mental status (either due to a prolonged post-ictal state or non-convulsive status) tends to be under recognized and can only be diagnosed with EEG.
Wernicke's encephalopathy (WE), due to its changing epidemiology (no longer primarily a disorder of alcoholics) is likely to be under recognized. Dr. Josephson recommends thiamine 100 mg IV be administered to nearly all altered patients. If WE is established or strongly suspected clinically very high doses of thiamine such as500 mg IV tid may be more appropriate. For more on thiamine deficiency and WE see this post.
What about using procholinergic drugs?
It is known that the brains of patients with dementia are deficient in acetylcholine. Also, delirium is known to be precipitated by anticholinergic agents. So why not give the delirious patient a dose of Aricept, for example? The idea has been studied. According to Dr. Josephson's chapter in Harrison's, however, clinical trials have yielded mixed results so it's not ready for prime time. I mention it here because it is an appealing idea which will undoubtedly be the subject of future research.