The classical definition of fever of unknown origin required a temperature above 38.3 °C on several occasions, an illness lasting more than three weeks and no diagnosis after a week of inpatient investigation (Petersdorf and Beeson, 1961). A later revision shortened the required investigation and separated classical, nosocomial, neutropenic and HIV-associated categories (Durack and Street, 1991). Infection, malignancy and non-infectious inflammatory disease are the three large diagnostic groups, and Still's disease is among the more frequent inflammatory causes in adults.

The criteria of Yamaguchi et al. (1992) are the most widely used. Their major items are fever of 39 °C or more lasting at least a week, arthralgia for at least two weeks, the typical rash, and leucocytosis of at least 10,000 per cubic millimetre with 80 per cent or more granulocytes. The minor items are sore throat, lymphadenopathy or splenomegaly, liver dysfunction, and negative rheumatoid factor and antinuclear antibody. Classification needs five items including two major ones, after exclusion of infection, malignancy and other rheumatic disease. The criteria of Fautrel et al. (2002) dispense with exclusions and add a glycosylated ferritin fraction of 20 per cent or less as a major item, following earlier work showing that a low glycosylated fraction combined with a high total ferritin is more specific than ferritin alone (Fautrel et al., 2001).

Table 1 compares the instruments available. The HLH-2004 criteria were written for a paediatric treatment protocol and require five of eight findings, including a ferritin of at least 500 micrograms per litre, cytopenia in two lineages and tissue haemophagocytosis (Henter et al., 2007). They are specific but late: by the time five are met the syndrome is advanced. The HScore assigns weighted points to nine clinical, laboratory and cytological variables and returns a probability of reactive haemophagocytic syndrome (Fardet et al., 2014). The 2016 classification criteria were derived for systemic juvenile arthritis and require a ferritin above 684 nanograms per millilitre with two of four laboratory changes (Ravelli et al., 2016).

A feature shared by both diseases complicates all three instruments. Active Still's disease raises the white cell count, the platelet count and fibrinogen. When macrophage activation supervenes, those values fall, and they may pass through the normal range on the way down. A platelet count that would be unremarkable in a healthy person can therefore be an alarm signal in someone whose count was high three days earlier, and the erythrocyte sedimentation rate may fall with fibrinogen while C-reactive protein keeps rising (Carter et al., 2019).

The first step is descriptive. Temperature is charted at least four times daily so that a quotidian or double-quotidian pattern can be seen, and the skin is examined at the fever peak, when the salmon-coloured rash is most likely to be visible. Sore throat at onset, joint pain or swelling, lymph nodes, spleen and liver size, and serositis are recorded as present or absent. Baseline tests comprise a full blood count with differential, C-reactive protein, sedimentation rate, liver enzymes, lactate dehydrogenase, ferritin, triglycerides and fibrinogen. The record left by this step is a dated table of these values, which later steps depend on.

Still's disease may be diagnosed only when the alternatives have been pursued. Blood cultures, serology and molecular tests directed by exposure and geography, cross-sectional imaging, and echocardiography where endocarditis is possible form the infectious part of the search. Lymphoma is the chief malignant mimic and is also the commonest malignant trigger of haemophagocytic syndrome in adults (Ramos-Casals et al., 2014), so enlarged nodes should be biopsied and not merely observed. Autoantibody testing covers lupus and vasculitis. The pathway asks the clinician to document which alternatives were considered and on what evidence each was set aside.

With the exclusions documented, the Yamaguchi items are counted, and the Fautrel set is applied as well where glycosylated ferritin can be measured. Classification criteria were built to assemble comparable groups for research and were not intended as diagnostic rules (Giacomelli et al., 2018). A patient who falls one item short because leucocytosis is missing should not be dismissed if steps one and four suggest that macrophage activation has already lowered the count. The pathway therefore asks for the item count to be recorded twice: on the values observed, and on the highest values documented earlier in the illness.

Screening is repeated at every review and uses change over time. The dated table from step one is extended daily during the acute phase. Features that should prompt concern are a ferritin that is rising steeply, a fall in platelets, neutrophils or haemoglobin from the previous measurement, rising liver enzymes and triglycerides, a falling fibrinogen or sedimentation rate in the face of persistent fever, and a change in the fever from spiking to continuous. The HScore is calculated at each review, and the 2016 criteria are checked as a second view. A bone marrow aspirate is requested when the screen is positive, both to look for haemophagocytosis and to exclude marrow infiltration, but its result is not a precondition for escalation, since haemophagocytosis may be absent early.

A positive screen triggers same-day review by rheumatology and haematology together, with critical care informed. Published guidance for adults with haemophagocytic lymphohistiocytosis stresses that treatment should be directed at both the hyperinflammation and its trigger, and that it should not be delayed until every criterion is fulfilled (La Rosée et al., 2019). The choice of agents is outside the scope of this pathway and should follow those recommendations and the European guidance for Still's disease (Fautrel et al., 2024). What the pathway specifies is the handover: the dated table, the criteria counts, the HScore series and the list of excluded alternatives travel with the patient.

After escalation the same variables are followed until fever has settled and ferritin, counts and fibrinogen are moving steadily in the right direction. Macrophage activation can recur, and it can be precipitated by intercurrent infection, so the screen in step four is resumed whenever fever returns. Markers such as interleukin-18 and soluble CD25 reflect the biology of the disease (Feist et al., 2018) and may be added where a laboratory offers them, but the pathway is designed to work without them.