Background In many endemic regions, serological tests for dengue are unavailable and diagnosis relies solely on clinical signs and basic laboratory workup.

Objective To identify clinical signs and basic haematological laboratory results potentially useful to distinguish dengue from other febrile illnesses.

Methods Prospective study in an outpatient setting in Rio de Janeiro from 2005 to 2008. Subjects included patients >12 years of age referring fever with duration ≥7 days (acute fever) and without evident focuses of infection. Logistic regression analysis was used to identify symptoms, physical signs and haematological features accurate for the dengue diagnosis in patients evaluated between days 0–3 and days 4–7 from fever outset.

Results Based on serological tests and virus genome detection by PCR, 202 patients were classified as dengue and 103 as non-dengue. For patients evaluated on days 0–3, a model including conjunctival hyperaemia and leucopenia achieved a sensitivity of 83% and specificity of 71% for dengue diagnosis, with an area under the receiver-operating characteristic curve of 0.82. In patients evaluated on days 4–7, a model including rash and thrombocytopaenia achieved an area under the receiver-operating characteristic curve of 83%, with a sensitivity of 71% and specificity of 87.5% for dengue diagnosis.

Conclusions Predictive models including clinical and simple laboratory tests data achieved moderate accuracy for diagnosing dengue in ambulatory febrile patients. In our population, these models were more accurate than currently used WHO dengue case definition (1997) and could be potentially useful for surveillance.