[Fever in patients of a Service of Internal Medicine: prospective study in 204 cases].

OBJECTIVES Fever in hospitalized patients (FHP) is a difficult problem. We study clinical characteristics and predictive parameters of infection, sepsis and outcome. PATIENTS AND METHODS Prospective study of 204 patients with fever > 38 degrees C admitted in an Internal Medicine ward. In each patient clinical evaluation, complete blood count (CBC) urinalysis (UA), C-reactive protein (CRP), chest X- ray (CXR), blood and urine cultures were performed. RESULTS 115 patients suffered infection (75.9% nosocomial, 7.4% sepsis), 35 had a non-infectious etiology and 54 an unknown cause. CBC and CRP did not distinguish infection, bacteremia or sepsis. In UA, positive nitrites, leukocytes and bacteriuria were predictive of infection. 18.6% of the patients were diagnosed by CXR. 18% of urine cultures and 13% of blood cultures were positive. 71.6% received antibiotics (deemed unnecessary in 18%). Microbiological results modified 25% of initial treatments. Average hospital stay was 17 days; 16.7% died. Chronic lung disease, fever duration, dysphagia and urinary tract alterations predict infection. Obesity, diabetes mellitus, liver failure, immunocompromised host, central vascular access, fever duration and nosocomial infection predict bacteremia. Sepsis is more prevalent in males, with malignancy and vascular or nosocomial infection. Chronic liver disease, nosocomial infection, severe lung infection and sepsis increase mortality. CONCLUSIONS Clinical diagnosis of FHP is inaccurate. Infection is it s most frequent cause; UA, CXR, and blood and urine cultures are useful. FHP increases mortality and hospital stay. We have established predictable models of infection, bacteremia, sepsis and mortality. However, its sensibility and specificity are low.