A Simple Algorithm for Predicting Bacteremia Using Food Consumption and Shaking Chills: A Prospective Observational Study.

BACKGROUND: Predicting the presence of true bacteremia based on clinical examination is unreliable.

OBJECTIVE: We aimed to construct a simple algorithm for predicting true bacteremia by using food consumption and shaking chills.

DESIGN: A prospective multicenter observational study.

SETTING: Three hospital centers in a large Japanese city.

PARTICIPANTS: In total, 1,943 hospitalized patients aged 14 to 96 years who underwent blood culture acquisitions between April 2013 and August 2014 were enrolled. Patients with anorexia-inducing conditions were excluded.

INTERVENTIONS: We assessed the patients' oral food intake based on the meal immediately prior to the blood culture with definition as "normal food consumption" when >80% of a meal was consumed and "poor food consumption" when <80% was consumed. We also concurrently evaluated for a history of shaking chills.

MEASUREMENTS: We calculated the statistical characteristics of food consumption and shaking chills for the presence of true bacteremia, and subsequently built the algorithm by using recursive partitioning analysis.

RESULTS: Among 1,943 patients, 223 cases were true bacteremia. Among patients with normal food consumption, without shaking chills, the incidence of true bacteremia was 2.4% (13/552). Among patients with poor food consumption and shaking chills, the incidence of true bacteremia was 47.7% (51/107). The presence of poor food consumption had a sensitivity of 93.7% (95% confidence interval [CI], 89.4%-97.9%) for true bacteremia, and the absence of poor food consumption (ie, normal food consumption) had a negative likelihood ratio (LR) of 0.18 (95% CI, 0.17-0.19) for excluding true bacteremia, respectively. Conversely, the presence of the shaking chills had a specificity of 95.1% (95% CI, 90.7%-99.4%) and a positive LR of 4.78 (95% CI, 4.56-5.00) for true bacteremia.

CONCLUSION: A 2-item screening checklist for food consumption and shaking chills had excellent statistical properties as a brief screening instrument for predicting true bacteremia.