812. The diagnostic efficacy for combination of nurse-assesed food consumptions and shaking chills in predicting bacteremia at the time of blood culture
Session: Poster Abstract Session: Bacteremia and Endocarditis
Friday, October 9, 2015
Room: Poster Hall
  • IDSA 2015 (final).pdf (595.6 kB)
  • Background:  Shaking chills (SCs) may be useful in predicting bacteremia. We have previously reported that the normal food consumption (FC) is helpful in ruling out of bacteremia with the sensitivity of 92.0%. In this study, we aimed to evaluate the clinical utility of combining noth observations for predicting the positivity of blood cultures (Bcx).

    Methods:  This was a prospective, multi-center observational study (UMIN ID; R000013768) in Japan enrolling consecutive patients undergoing Bcx. We evaluated 2793 consecutive Japanese patients who underwent Bcx between May 2013 and August 2014, inclusive. After the exclusion of patients with anorexia inducing conditions, such as gastrointestinal illness and malignant disease treated with chemotherapy, 1945 patients were the studied. We characterized the patients' oral intake based on FC and chills immediately prior to Bcx acquisition. We defined "positive" FC when more than 80% of the meal was ingested, and "negative" FC when less than 80% was ingested. To assess the reliability of the evaluations of FC by nurses, the kappa score was 0.79 (95%CI, 0.77–0.80) indicating a high level of concordance. Positive chills were defined when patients had the feeling of chills or the feeling extreme cold with SCs, and negative as the absence of chills or SCs. Sensitivity and specificity for criteria to predict bacteremia were analyzed. After multiple logistic regression analysis about components which were considered to be correlated with predicting bacteremia by principal components analysis, recursive partitioning analysis (RPA) was used to create a flow chart using the degrees of food consumption and chills.

    Results: Of 1945 patients, 223 patients (11.5%) had true bacteremia. Comparison of the positive FC group versus the negative FC group revealed a sensitivity of 92.8% for excluding true bacteremia. Comparison of the absence of chills group versus the chills group revealed a specificity of 95.1% for predicting true bacteremia. RPA suggested a decision tree for predicting bacteremia (see Figure).

    Conclusion: Normal FC in the absence of any form of chills may be helpful for ruling out bacteremia and aiding clinical decision making whether to provide empirical antibiotics to febrile patients.

    Takayuki Komatsu, MD, PhD1, Erika Takahashi, MD2, Kentaro Mishima, MD2, Takeo Toyoda, MD, PhD3, Fumihiro Saitoh, MD4, Manabu Sugita, MD, PhD1, Joel Branch, MD5 and Kenji Inoue, MD, PhD6, (1)Emergency and Critical Care Medicine, Juntendo University Nerima Hospital, Tokyo, Japan, (2)Emergency, Juntendo University Nerima Hospital, Tokyo, Japan, (3)Internal Medicine, Nerima General Hospital, Tokyo, Japan, (4)Ooizumi Health Cooperative Hospital, Tokyo, Japan, (5)Shonan Kamakura General Hospital, Kanagawa, Japan, (6)Cardiology, Juntendo University Nerima Hospital, Tokyo, Japan


    T. Komatsu, None

    E. Takahashi, None

    K. Mishima, None

    T. Toyoda, None

    F. Saitoh, None

    M. Sugita, None

    J. Branch, None

    K. Inoue, None

    Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 7th with the exception of research findings presented at the IDWeek press conferences.