Delirium is a clinical syndrome, not a diagnosis: the precipitant has to be found. The causes that kill quickly (hypoglycaemia, sepsis, hypoxia, intracranial bleeding, meningitis, toxic states and alcohol withdrawal) can all be screened for in the first few minutes with a glucose, observations and a quick neurological look, which is why they come before any cognitive test. The sepsis guidance for people aged 16 or over (NICE NG253, November 2025) now replaces NG51 for adults; a new altered mental state still counts as a high-risk feature, so a confused older person with a suspected infection and any other high-risk sign needs emergency transfer rather than a community antibiotic. Subdural haematoma is the classic trap: the fall may have been days ago, may not have been witnessed, and the only sign may be fluctuating drowsiness in someone on an anticoagulant. Wernicke's encephalopathy rarely shows the full triad, so the threshold for parenteral thiamine in a heavy drinker who is confused should be low.

The 4AT is weighted so that a patient who is too drowsy or agitated to be tested still scores 4 or more, which prevents the most unwell patients being falsely reassured. The acute-change item carries 4 points on its own, which is why the collateral history matters so much: without it you cannot score the test properly and you cannot distinguish a new delirium from a long-standing dementia. In the Tieges meta-analysis (Age and Ageing, 2021) the 4AT performed consistently across emergency, acute medical, surgical and care-home settings. Hypoactive delirium is underdiagnosed because the patient causes no trouble; a carer who says "she's just not herself, sleeping all day" is describing it. Always record the 4AT score in the notes so that the trajectory can be tracked at review.

A structured sweep stops premature closure on the first abnormality found — a positive urine or a low sodium may be one contributor among several. Medicines are among the most modifiable contributors, and total anticholinergic load (added up with an anticholinergic burden scale) matters more than any single drug. Wernicke's encephalopathy is easily missed outside obvious alcohol dependence; malnutrition from any cause can produce it. In older people, bacteria in the urine without urinary symptoms are common and are not by themselves an explanation for confusion (UKHSA diagnostic tool for suspected UTI in adults over 65, updated December 2024), so keep looking for another source if the picture does not fit.

Urinary retention and faecal impaction are common, painless-looking causes in older men and people on opioids or anticholinergics, and they are found only by examining the abdomen. Low oxygen saturation in someone with COPD or pneumonia may explain the whole picture. Asterixis and myoclonus point towards a metabolic or toxic cause rather than a structural one, which changes the investigation plan. Unexplained bruising should also prompt thought about falls on anticoagulation and, where the story does not fit, about safeguarding. Checking the medicines physically in the home often reveals double-dosing from old and new boxes or a recently started sedative that is not on the repeat template.

Blood tests find the silent causes: hyponatraemia, hypercalcaemia, AKI, hepatic dysfunction and hypothyroidism rarely announce themselves on examination. Severity of hyponatraemia correlates loosely with symptoms, and a rapid fall is more dangerous than a chronic low level, so the trend matters as much as the number. Dipstick testing in older people is unreliable because asymptomatic bacteriuria becomes common with age; treating a positive dipstick often means the true cause is missed and an unnecessary antibiotic adds to the problem. CT head has a low yield when there is a clear medical cause and no focal signs, head injury, anticoagulation or reduced consciousness, but it becomes essential when any of those are present. Record a baseline QTc before haloperidol, because both the drug and the electrolyte disturbances of delirium prolong it.

Delirium is associated with poor outcomes whatever the setting: in a meta-analysis of older patients, delirium was linked to higher mortality (hazard ratio 1.95) and to more institutionalisation and dementia, independent of age and illness severity (Witlox et al., JAMA, 2010). That argues against under-triage. At the same time, unfamiliar surroundings, noise, broken sleep and immobility can prolong delirium, so the decision should weigh what admission adds. Capacity is decision-specific and fluctuates in delirium; if the person cannot understand, retain and weigh information about going to hospital, a best-interests decision with the family is lawful and appropriate, and should be documented clearly.

No drug treats delirium; drugs only reduce dangerous distress while the cause is treated. The 2023 update of NICE CG103 narrowed the drug option to short-term haloperidol and added a reminder to take account of the MHRA's advice on its risks in older people. The MHRA's 2021 review found no new safety signal but noted variable practice and neurological adverse effects, so the lowest dose for the shortest time, with monitoring, is the standard. In dementia with Lewy bodies and Parkinson's disease, antipsychotics can cause severe sensitivity reactions and worsen motor function, so check the dementia subtype before prescribing. Benzodiazepines are the right choice only when the delirium is due to alcohol or sedative withdrawal. Any antipsychotic started during delirium should have a stop date written on the prescription and be reviewed at every contact.

The landmark trial of multicomponent prevention (Inouye et al., NEJM, 1999) targeted six risk factors — cognitive impairment, sleep deprivation, immobility, visual impairment, hearing impairment and dehydration — and reduced the number and duration of delirium episodes. The 2021 Cochrane review confirmed that multicomponent programmes probably reduce delirium incidence in hospital patients outside intensive care. These interventions are cheap and are largely things families and care homes can do if they know why they matter. NICE CG103 asks that people at risk and their families receive information about delirium; many relatives find episodes frightening and are reassured to hear that it is a recognised, usually temporary, response to illness.

Tracking the 4AT over time turns a subjective impression ("seems a bit better") into a documented trajectory, which makes escalation decisions clearer and helps out-of-hours colleagues. Delirium that persists or recurs is a warning sign: it can indicate an unresolved cause, a second precipitant, or underlying dementia that the episode has revealed. Cognitive testing during delirium is unreliable, so the formal assessment for dementia should wait until attention has recovered. For the SCA, examiners look for explicit safety-netting given to the carer in plain words, a clear plan for who reviews and when, and recognition of carer strain.