Fever in adults — triage and diagnostic approach

The central principle of NICE NG51 is that sepsis is recognised from physiology and risk factors, not from the temperature: a patient with a fever of 39.5°C and normal observations is usually less unwell than one with a temperature of 37.2°C, a respiratory rate of 26 and new confusion. Older and frail patients, and those on steroids, may be afebrile or hypothermic while septic, and the raised respiratory rate is the earliest and most reliable warning sign. Neutropenic sepsis is singled out because it is the one febrile presentation where the clock runs in minutes: NICE CG151 requires immediate admission and broad-spectrum antibiotics within one hour, and mortality rises measurably with each hour of delay — any patient within six weeks of chemotherapy who reports a fever should be sent directly, without a face-to-face assessment first if that would cause delay. Spinal epidural abscess earns its place because fever with back pain is so often attributed to a coincidental musculoskeletal problem, and by the time neurological signs appear the deficit may be permanent.

Travel and immune status are asked first because they change the disposal of the patient rather than merely refining the differential. Falciparum malaria is the most important imported infection in the UK, with around 1,500–2,000 cases and a small number of deaths each year; almost all deaths involve a delay in diagnosis, and the classic teaching that the fever is cyclical is unhelpful because early falciparum fever is usually continuous or irregular. Three negative thick and thin films are needed to exclude it, which is a hospital task. Drug fever is worth holding in mind in the patient who is febrile but well with no source and a recently started medication — it typically appears seven to ten days after the drug is started and resolves within 72 hours of stopping it. Asking specifically about antipyretics already taken prevents false reassurance: a patient who took paracetamol two hours ago may be recorded as afebrile while remaining significantly unwell, which is why the physiological parameters matter more than the reading.

Keeping the non-infective causes visible prevents two common errors. The first is the repeated course of antibiotics for a fever that is actually inflammatory or malignant — the febrile patient with proximal girdle pain, jaw claudication and a raised inflammatory response has giant cell arteritis and needs steroids and an urgent referral, not a third antibiotic. The second is confusing hyperthermia with fever: in heatstroke, neuroleptic malignant syndrome, serotonin syndrome and thyroid storm the set-point mechanism is not driving the temperature, so antipyretics are ineffective and active cooling plus treatment of the cause is the priority. It is also worth remembering that fever in adults is far more likely to have an identifiable source than fever in young children, so a genuinely sourceless fever in an unwell adult is a reason to escalate rather than to reassure — and that pulmonary embolism, which is not an infection at all, commonly produces a low-grade fever and is missed when the clinician has committed to an infective explanation.

The two examination components most often skipped are the full skin survey and the spine, and both hide emergencies. A non-blanching rash may be limited to a handful of petechiae on the trunk, the soles or the conjunctivae, and it cannot be excluded without undressing the patient — in meningococcal disease the rash may also appear after the patient is first seen, which is why the safety-net must describe it explicitly. Percussing the spine takes seconds and is the only bedside test that reliably raises the possibility of discitis or epidural abscess in the febrile patient with back pain. Listening deliberately for a new murmur matters because endocarditis is a diagnosis made by suspicion: the typical patient has weeks of malaise, weight loss and low-grade fever, has often had one or more courses of antibiotics, and the crucial action is to send them for blood cultures before further antibiotics are given, since prior antibiotics can render cultures negative and delay diagnosis by months.

CRP lags behind clinical deterioration by hours, so a normal CRP in the first day of an illness never excludes serious bacterial infection — a well-documented cause of false reassurance in both primary care and emergency departments. The urine dipstick carries the opposite risk in older adults: asymptomatic bacteriuria is present in a substantial proportion of over-65s and in almost all catheterised patients, so a positive dip in a confused febrile older person is frequently mistaken for the source while a pneumonia or spinal infection goes unexamined. In practice the rule is that a urinary diagnosis requires urinary symptoms. The NG12 leukaemia criterion is included because it is the one cancer pathway triggered by fever itself: unexplained fever with night sweats, weight loss, bruising or lymphadenopathy earns a very urgent full blood count within 48 hours, and the same NG12 table requires immediate specialist assessment for a child or young person with fever and unexplained petechiae, hepatosplenomegaly or generalised lymphadenopathy.

Giant cell arteritis is the one febrile presentation where treatment must precede referral: visual loss can occur within hours and is irreversible, so high-dose prednisolone is started immediately on clinical suspicion and the patient is referred the same day if there are any visual symptoms. Suspected viral haemorrhagic fever is the mirror image — the patient should not be brought into a waiting room, and the correct first action is a telephone discussion with the local infection service and UKHSA. The statutory notification duty catches people out: for meningococcal disease, TB, malaria, typhoid and measles the duty arises on clinical suspicion, not on laboratory confirmation, and in practice notification is frequently left to the admitting hospital when the community clinician also has the obligation. On the PUO referral, the classical division into thirds is still a useful framework, but the important practical point is that a genuine PUO is not a reason for another empirical antibiotic course — blind antibiotics obscure cultures and delay the diagnosis of both endocarditis and lymphoma.

Antipyretics are for comfort, not for the number, and the response to them carries no diagnostic information — the belief that a fever which "breaks" with paracetamol is therefore viral is a persistent myth and a recognised route to missed sepsis. Routine alternating of paracetamol and ibuprofen is not supported and increases dosing errors. Ibuprofen carries specific cautions in the febrile patient: it worsens dehydration-related acute kidney injury, and its use in varicella and in possible streptococcal soft-tissue infection has been associated with more severe disease. Sick-day rules deserve emphasis because an ordinary febrile illness is the commonest precipitant of both diabetic ketoacidosis and adrenal crisis, and of drug-induced AKI in patients on ACE inhibitors, diuretics and SGLT2 inhibitors — a two-minute conversation about which tablets to pause prevents a substantial proportion of febrile admissions in older patients.

Explaining the expected natural history is the most effective consultation behaviour available for reducing both antibiotic prescribing and unnecessary re-attendance: patients return because nobody told them that a fever lasting four days is normal. The advice against tepid sponging and cold baths is worth stating explicitly because it remains widespread practice — peripheral cooling in a patient whose hypothalamic set-point is raised induces shivering and vasoconstriction, which increases heat production and is uncomfortable. The written safety-net is the single most important medico-legal and clinical component of discharging a febrile adult from a consultation: documenting which specific signs were described, and the fact that written information was given, both improves the patient's chance of returning at the right moment and records that the risk was addressed. For asplenic and immunosuppressed patients, the timeframe in the advice must be shortened from "if you get worse over the next day" to "within the hour", because that is the speed at which encapsulated organisms behave.

Almost all serious harm in febrile presentations occurs in the interval after the first consultation rather than during it, which makes the safety-net the clinically active part of the encounter. Documenting the observation set — including the respiratory rate — and the negative findings of a systematic examination is what makes a later reviewer able to judge whether the patient has genuinely changed, and it is precisely what is missing in the records of most delayed sepsis diagnoses. Recording the safety-net in the words used, rather than as "advised to return if worse", both improves the patient's response and demonstrates that the risk was recognised. The 48-hour review is where the second peak of diagnoses occurs, because a source that was invisible on day one is usually apparent by day three: the febrile illness that will turn out to be pneumonia, pyelonephritis, endocarditis or lymphoma nearly always declares itself to a clinician who looks again with the whole differential still open, rather than one who is confirming the original impression.