Advertisement

High positivity of blood cultures obtained within two hours after shaking chills

  • Tomohiro Taniguchi
    Correspondence
    Corresponding author at: Division of General Internal Medicine and Infectious Diseases, Hiroshima Prefectural Hospital, 1-5-54 Ujinakanda, Minamiku, Hiroshima 734-8530, Japan.
    Affiliations
    Division of Infectious Diseases, Department of Internal Medicine, Okinawa Chubu Hospital, 281 Miyazato, Uruma, Okinawa 904-2293, Japan

    Division of General Internal Medicine and Infectious Diseases, Hiroshima Prefectural Hospital, 1-5-54 Ujinakanda, Minamiku, Hiroshima 734-8530, Japan
    Search for articles by this author
  • Sanefumi Tsuha
    Affiliations
    Division of Infectious Diseases, Department of Internal Medicine, Okinawa Chubu Hospital, 281 Miyazato, Uruma, Okinawa 904-2293, Japan
    Search for articles by this author
  • Soichi Shiiki
    Affiliations
    Division of Infectious Diseases, Department of Internal Medicine, Okinawa Chubu Hospital, 281 Miyazato, Uruma, Okinawa 904-2293, Japan
    Search for articles by this author
  • Masashi Narita
    Affiliations
    Division of Infectious Diseases, Department of Internal Medicine, Okinawa Chubu Hospital, 281 Miyazato, Uruma, Okinawa 904-2293, Japan
    Search for articles by this author
Open AccessPublished:July 27, 2018DOI:https://doi.org/10.1016/j.ijid.2018.07.020

      Highlights

      • Blood culture positivity was higher for cultures drawn within 2 h after shaking chills than for those drawn after 2 h.
      • The adjusted odds ratio of blood culture positivity for cultures drawn within 2 h after shaking chills was 1.88 (95% confidence interval 1.01–3.51).
      • Blood cultures should be obtained without delay if patients have shaking chills.

      Abstract

      Objective

      To determine whether the time lag between blood culture draw and the start of shaking chills is associated with blood culture positivity.

      Methods

      A prospective observational study was undertaken from January 2013 to March 2015 at a referral center in Okinawa, Japan. All enrolled patients were adults with an episode of shaking chills who were newly admitted to the division of infectious diseases. The study exposure was the time lag between blood culture draw and the most recent episode of shaking chills.

      Results

      Among patients whose blood cultures were obtained within 2 h after shaking chills started, the blood culture positivity was 53.6% (52/97), whereas among patients whose blood cultures were obtained after more than 2 h, the positivity was 37.6% (44/117) (p = 0.019). The adjusted odds ratio of blood culture positivity for samples drawn within 2 h after shaking chills was 1.88 (95% confidence interval 1.01–3.51, p = 0.046). Escherichia coli were the most frequently detected bacteria (58/105).

      Conclusions

      The positivity of blood cultures obtained within 2 h after the start of the most recent shaking chills was higher than that for blood cultures obtained after 2 h.

      Keywords

      Introduction

      Although various types of testing modality have been developed for the detection of bacteria, including molecular techniques, blood culture is currently the gold standard method for identifying the etiological pathogens of bacteremia (
      • Aronson M.D.
      • Bor D.H.
      Blood cultures.
      ,
      • Liesenfeld O.
      • Lehman L.
      • Hunfeld K.P.
      • Kost G.
      Molecular diagnosis of sepsis: new aspects and recent developments.
      ). The timing of blood draw for culture, however, remains controversial (
      • Coburn B.
      • Morris A.M.
      • Tomlinson G.
      • Detsky A.S.
      Does this adult patient with suspected bacteremia require blood cultures?.
      ).
      A study reported in the 1950s recommended that blood cultures should be drawn during the hour preceding an expected chill (
      • Bennett Jr., I.L.
      • Beeson P.B.
      Bacteremia: a consideration of some experimental and clinical aspects.
      ). This was because it was believed that there was a time lag of around an hour between the abrupt influx of bacteria and the onset of chills (
      • Bennett Jr., I.L.
      • Beeson P.B.
      Bacteremia: a consideration of some experimental and clinical aspects.
      ). The most severe degree of chills, namely shaking chills or shivering, has been considered a strong predictor of bacteremia (
      • Bates D.W.
      • Cook E.F.
      • Goldman L.
      • Lee T.H.
      Predicting bacteremia in hospitalized patients. A prospectively validated model.
      ,
      • Tokuda Y.
      • Miyasato H.
      • Stein G.H.
      • Kishaba T.
      The degree of chills for risk of bacteremia in acute febrile illness.
      ). Shaking chills have been a useful indicator of bacteremia even in elderly patients with dementia, because they are easily recognized (
      • Taniguchi T.
      • Tsuha S.
      • Takayama Y.
      • Shiiki S.
      Shaking chills and high body temperature predict bacteremia especially among elderly patients.
      ).
      To date, no study has demonstrated whether the time to blood culture collection affects blood culture positivity among patients with a history of shaking chills. The objective of this research was to determine whether blood culture positivity is higher in patients whose blood cultures are drawn earlier after shaking chills begin than in patients whose blood cultures are drawn later. The results could be used to improve the treatment of patients with severe bacterial infection and could aid the medical staff who take blood cultures.

      Methods

      Study design and setting

      This was a hospital-based, prospective, observational study at Okinawa Chubu Hospital, which is located in the central region of Okinawa Island in Japan. Approximately 39 000 patients visit the emergency center and nearly 7000 patients are hospitalized through the emergency center per year. The division of infectious diseases treats adults with all kinds of infectious diseases, except for those with severe lung disease (who are admitted to the division of respiratory medicine), those with febrile neutropenia (hematology and oncology), those with chronic kidney disease who are on hemodialysis (nephrology), and those who need endoscopic or surgical management (gastrointestinal medicine or surgery). Over 500 patients per year are hospitalized in the division of infectious diseases. The study period was from January 1, 2013 to March 12, 2015.

      Participants

      All patients over 16 years of age who had a history of shaking chills and were newly admitted to the division of infectious diseases on suspicion of bacterial infection were enrolled. The suspicion of bacterial infection was based on both a clinical decision and laboratory tests including point-of-care Gram stain on admission (
      • Taniguchi T.
      • Tsuha S.
      • Shiiki S.
      • Narita M.
      Gram-stain-based antimicrobial selection reduces cost and overuse compared with Japanese guidelines.
      ). Exclusion criteria were as follows: (1) not diagnosed with a bacterial infection clinically or after laboratory tests including culture results; (2) infectious agents that were not incubated in the usual blood culture bottles: viruses, Bartonella, Chlamydia, Clostridium difficile, Legionella, Leptospira, Mycobacterium tuberculosis; and (3) the time of onset of shaking chills was not recorded, unclear, or more than 48 h before.

      Exposures

      The exposure was the time lag between the start of shaking chills and the blood culture draw. Shaking chills or shivering was defined as chills plus persistent involuntary muscle tremors. The time that blood cultures were obtained was defined as the moment that a resident physician requested two sets of blood cultures by printing identification data labels. The time of onset of shaking chills was collected from the medical charts of the emergency center. If the time of onset was not recorded, the co-authors directly asked the patient, their family members, or the healthcare workers and confirmed the time. When shaking chills occurred more than once, the time of onset of the last episode of shaking chills was used, following which blood cultures were taken.
      Previous antimicrobial use within 48 h prior to arrival was also considered an important exposure.

      Outcome measure

      The outcome measure was blood culture positivity. At least two sets of blood cultures that included aerobic and anaerobic bottles with Bactec Plus resin medium (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) were routinely drawn. These were taken mainly by resident physicians from the upper or lower limbs, not from the femoral vessels to decrease the risk of contamination, before antibiotics were started. The medical staff were encouraged to obtain at least 10 ml in each set. All culture bottles were incubated for at least 5 days in an automated blood culture system, the Bactec 9240 system (
      • Kiyoyama T.
      • Tokuda Y.
      • Shiiki S.
      • Hachiman T.
      • Shimasaki T.
      • Endo K.
      Isopropyl alcohol compared with isopropyl alcohol plus povidone-iodine as skin preparation for prevention of blood culture contamination.
      ).
      Coagulase-negative staphylococci, Bacillus species, Propionibacterium species, Micrococcus species, Clostridium species, and also α-streptococci were considered potential skin contaminants (
      • Kiyoyama T.
      • Tokuda Y.
      • Shiiki S.
      • Hachiman T.
      • Shimasaki T.
      • Endo K.
      Isopropyl alcohol compared with isopropyl alcohol plus povidone-iodine as skin preparation for prevention of blood culture contamination.
      ). With the exception of α-streptococci, if any of these was cultured from only one set of blood cultures, it was regarded as a contaminant. Otherwise, clinical information was used to judge whether contamination had occurred or not.

      Statistical analysis

      In a preliminary study under different settings, which was reported at a conference of the Japanese Association for Infectious Diseases in Kyoto, Japan in 2010, the blood culture positivity values for blood drawn within 2 h after shaking chills and for blood drawn after more than 2 h were 60% (12/20) and 31% (13/42), respectively. Another study demonstrated that the blood culture positivity with a history of shaking chills was approximately 40% in this setting (
      • Taniguchi T.
      • Tsuha S.
      • Takayama Y.
      • Shiiki S.
      Shaking chills and high body temperature predict bacteremia especially among elderly patients.
      ). It was hypothesized that blood culture positivity might be lower than that in the preliminary study because the study conditions included in-hospital patients who had catheter-related blood stream infections. It was also hypothesized that a more than 15% difference in blood culture positivity between the blood culture obtained early group and the blood culture obtained later group would be clinically meaningful. Therefore, assuming a blood culture positivity in the obtained early group of 50% and in the obtained later group of 30%, an early and later group ratio of 1:1, 80% power, and two-sided alpha level of 0.05, it was calculated that 103 patients per group would be needed. Thus, it was aimed to continue this study until at least 206 patients were included.
      After the data collection, if a cut-off of 2 h was chosen, the early and later group ratio was closer to 1:1 than it was with a cut-off of 3 h. Therefore, patients were divided into two groups: (1) those for whom blood cultures were drawn at ≤2 h after shaking chills occurred, and (2) those for whom blood cultures were drawn at >2 h after shaking chills occurred.
      Patient characteristics were analyzed using the Chi-square test or Fisher’s exact test for categorical variables. The Mann–Whitney U-test was performed for numerical variables after an analysis of the data distribution.
      A multiple logistic regression model was used to investigate the association between the risk of blood culture positivity and the time lag between the start of the most recent shaking chills and time at which blood cultures were obtained. Age and previous antimicrobial exposure were sequentially assigned to the model, because these factors were shown to be correlated with blood culture positivity in previous research (
      • Taniguchi T.
      • Tsuha S.
      • Takayama Y.
      • Shiiki S.
      Shaking chills and high body temperature predict bacteremia especially among elderly patients.
      ). Next, shaking chills that occurred more than once were added into the model to include the possibility of another correlating variable. In addition, systolic blood pressure, respiratory rate, and body temperature were entered into the model to adjust for severity. This was because the first two vital signs are included as important factors with sepsis in the quick Sequential Organ Failure Assessment (qSOFA) score (
      • Seymour C.W.
      • Liu V.X.
      • Iwashyna T.J.
      • Brunkhorst F.M.
      • Rea T.D.
      • Scherag A.
      • et al.
      Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (sepsis-3).
      ), and body temperature also differed between the two groups. The level of consciousness included in qSOFA was not used because most of the patients were too old to determine a precise Glasgow Coma Scale score. Intra-abdominal infection was finally added into the model to adjust for possible confounding, because it differed among infection sources in the two groups.
      The results were calculated using Stata software version 12.1 (StataCorp, College Station, TX, USA).

      Ethics

      This research was an observational study and, therefore, no written informed consent documents were required. The study proposal was approved by the Institutional Review Board of Okinawa Chubu Hospital.

      Results

      Two hundred and seventy-three patients were screened, of whom 59 were excluded: 13 had a non-infectious etiology (histiocytic necrotizing lymphadenitis (called Kikuchi’s disease in Japan), malignant lymphoma, drug hepatitis, pseudogout, and others), 11 had infectious agents that were not incubated in the usual blood culture bottles (Bartonella, Chlamydia, Clostridium difficile, influenza, Legionella, Leptospira, Mycobacterium tuberculosis, and mumps virus), and for 35 patients, the time of onset of shaking chills was not recorded, unclear, or more than 48 h before.
      Finally, 214 patients were enrolled. Table 1 shows a comparison between the two groups. Performance status was based on the Eastern Cooperative Oncology Group scale (
      • Sorensen J.B.
      • Klee M.
      • Palshof T.
      • Hansen H.H.
      Performance status assessment in cancer patients. An inter-observer variability study.
      ). In the early group, patients were older and performance status was lower; systolic blood pressure, respiratory rate, and body temperature were higher; and the number of abdominal infections was greater.
      Table 1Comparison between the blood culture groups: obtained early (≤2 h) vs. later (>2 h) after the start of shaking chills.
      Timing of blood culturep-Value
      ≤2 h (n = 97)>2 h (n = 117)
      Median age (IQR)79 (68–87)72 (49–83)0.002
      p<0.05.
      Male sex (%)34 (35.1%)33 (28.2%)0.28
      Performance status
      Performance status: 3=in bed or in a chair more than 50% of the time; 4=bedridden.
       0–266 (68.0%)95 (81.2%)0.026
      p<0.05.
       316 (16.5%)17 (14.5%)0.69
       415 (15.5%)5 (4.3%)0.008
      p<0.05.
      Place of residence prior to admission
       Home76 (78.4%)97 (82.9%)0.40
       Nursing facility21 (21.7%)19 (16.2%)0.31
       Hospital01 (0.9%)1.00
      Median timing of blood culture collection after recent shaking chills (h) (IQR)1 (1–2)6 (3–15)0.0000
      p<0.05.
      Shaking chills more than once8 (8.3%)19 (16.2%)0.099
      Previous antibiotic exposure within 48 h13 (13.5%)13 (11.0%)0.57
      Vital signs
       Systolic blood pressure (mmHg) (IQR)132 (114–150)120 (102–130)0.0000
      p<0.05.
       Pulse rate (/min) (IQR)104 (94–118)106 (90–118)0.89
       Respiratory rate (/min) (IQR)24 (21–28)20 (20–24)0.0000
      p<0.05.
       Body temperature (°C) (IQR)39.0 (38.1–39.6)38.3 (37.7–39.1)0.001
      p<0.05.
      Infection site
       Urinary61 (62.9%)75 (64.1%)0.85
       Skin and soft tissue16 (16.5%)22 (18.8%)0.66
       Pulmonary6 (6.2%)9 (7.7%)0.79
       Intra-abdominal10 (10.3%)3 (2.6%)0.022
      p<0.05.
       Bone and joint2 (2.1%)2 (1.7%)1.000
       Cardiovascular2 (2.1%)2 (1.7%)1.000
       Central nervous system02 (1.7%)0.50
       Other2 (2.1%)2 (1.7%)1.000
       Unknown2 (2.1%)4 (3.4%)0.69
      Cultured bacteria
      Escherichia coli31 (32.0%)27 (23.1%)0.15
      Streptococcus dysgalactiae subsp. equisimilis3 (3.1%)6 (5.3%)0.52
      Streptococcus agalactiae2 (2.1%)3 (2.6%)1.000
      Klebsiella pneumoniae3 (3.1%)1 (0.9%)0.33
      Streptococcus pneumoniae03 (2.6%)0.25
      Anaerococcus prevotii2 (2.1%)00.20
      Citrobacter koseri2 (2.1%)00.20
      Streptococcus dysgalactiae subsp. dysgalactiae1 (1.0%)1 (0.9%)1.000
      Pseudomonas aeruginosa2 (2.1%)00.20
       Others13 (13.4%)5 (4.3%)0.024
      p<0.05.
      Blood culture positivity52 (53.6%)44 (37.6%)0.019
      p<0.05.
      Death3 (3.1%)1 (0.9%)0.33
      IQR, interquartile range.
      a Performance status: 3 = in bed or in a chair more than 50% of the time; 4 = bedridden.
      * p < 0.05.
      Among the 26 antibiotics that patients had previously taken within 48 h, 18 were identified: amoxicillin, ampicillin, cefaclor, cefazolin, cefotiam, clarithromycin, levofloxacin, and piperacillin. According to the culture results, blood culture positivity was 10.0% (1/10) in those who took effective antimicrobials versus 37.5% (3/8) in those who did not (p = 0.28).
      Among the bacteria cultured (Table 1) were 58 Escherichia coli, including five that were positive for extended-spectrum beta-lactamase (ESBL), and four Klebsiella pneumoniae, including one that was ESBL-positive. The 18 bacterial species included in the ‘Others’ category in Table 1 were Acinetobacter baumannii, Actinomyces meyeri, Bacteroides fragilis, Clostridium perfringens, Dialister pneumosintes, Eggerthia catenaformis, Enterobacter cloacae, Haemophilus influenzae, Pasteurella multocida, Porphyromonas gingivalis, Salmonella enteritidis, Salmonella enterica serotype Typhi, Shewanella algae, methicillin-sensitive Staphylococcus aureus (MSSA), Streptococcus alactolyticus, Streptococcus constellatus, Streptococcus mitis, and an unidentified anaerobe.
      Blood culture positivity for samples obtained within 2 h after the most recent shaking chills began was 53.6% (52/97) and for samples obtained after 2 h was 37.6% (44/117) (p = 0.019). If the previous antimicrobial exposure within 48 h was negative, blood culture positivity for samples collected within 2 h after the most recent shaking chills occurred was 56.0% (47/84) and for samples collected after 2 h was 40.1% (43/107) (p = 0.030).
      Four patients died. All of their primary infection sites were cured or under control: cellulitis, cholangitis, septic arthritis, and severe hemorrhagic paniculitis that required amputation. All of these patients died of other complications: hospital-acquired pneumonia, empyema, multiple myeloma, and cholangiocarcinoma.
      Table 2 shows the odds ratios (OR) of blood culture positivity with their associated 95% confidence intervals (CI). After adjustment for age, frequency of shaking chills, previous antimicrobial administration within 48 h, vital signs, and intra-abdominal infection, blood cultures that were obtained within 2 h after the recent shaking chills began were more likely to be positive than those obtained more than 2 h after (adjusted OR 1.88, 95% CI 1.01–3.51, p = 0.046).
      Table 2Odds ratios of blood culture positivity.
      Unadjusted OR (95% CI)Adjusted OR (95% CI)p-Value
      Age ≥ 75 years2.19 (1.26–3.81)2.08 (1.12–3.86)0.020
      p<0.05.
      Blood cultures obtained within 2 h after recent shaking chills1.92 (1.11–3.31)1.88 (1.01–3.51)0.046
      p<0.05.
      Shaking chills more than once2.32 (1.01–5.35)3.39 (1.31–8.76)0.012
      p<0.05.
      Previous antibiotic exposure within 48 h0.40 (0.15–1.05)0.33 (0.12–0.92)0.035
      p<0.05.
      Systolic blood pressure ≤126 mmHg0.78 (0.45–1.33)0.96 (0.52–1.78)0.91
      Respiratory rate ≥22/min1.55 (0.90–2.67)1.22 (0.64–2.30)0.55
      Body temperature ≥38.5 °C1.39 (0.81–2.39)1.19 (0.66–2.15)0.56
      Intra-abdominal infection1.06 (0.34–3.26)0.92 (0.27–3.18)0.92
      OR, odds ratio; CI, confidence interval.
      * p < 0.05.
      E. coli were the most frequently detected bacteria in blood cultures (58/105). Most of these were associated with urinary tract infections (54/58). The correlation between blood culture positivity and urinary tract infection was therefore investigated. The unadjusted OR was 1.08 (95% CI 0.62–1.90, p = 0.78) (not shown in Table 2).
      With regard to the frequency of shaking chills, the median time to blood culture collection (interquartile range, IQR) was later in patients who had shaking chills more than once compared to those who had shaking chills only once (5 (IQR 2–11) versus 3 (IQR 1–6), p = 0.035). The site of infection of patients who had shaking chills more than once was urinary tract (n = 20), pulmonary (n = 2), bone and joint (n = 1), cardiovascular (n = 1), central nervous system (n = 1), skin and soft tissue (n = 1), others (n = 2), and unknown (n = 1) (two patients had two simultaneous infection sites).
      Table 3 shows the subgroup analysis of Gram-positives, Gram-negatives, and anaerobes. One patient had a Gram-positive and Gram-negative co-infection. With the exception of this co-infected patient, the median time to blood culture draw after the start of shaking chills was 2 h for Gram-negative infections versus 4 h for Gram-positive infections (p = 0.063). Six anaerobic infection sources were two urinary tract, two unknown, one cardiovascular, and one gastrointestinal system infections.
      Table 3Analysis of Gram-positives, Gram-negatives, and anaerobes.
      Gram-positives

      n = 21
      Gram-negatives

      n = 70
      Anaerobes

      n = 6
      Median age (IQR)80 (67–84)80 (68–87)77 (70–94)
      Median timing of blood culture collection after recent shaking chills (h) (IQR)4 (1–14)2 (1–4)1 (0–1)
      Shaking chills more than once2 (9.5%)12 (17.1%)3 (50.0%)
      Previous antibiotic exposure within 48 h1 (4.8%)4 (5.7%)1 (16.7%)
      IQR, interquartile range.

      Discussion

      This observational study investigated the relationship between blood culture positivity and the time lag between the latest onset of shaking chills and the time when blood cultures were collected. The results demonstrated that blood cultures obtained within 2 h after the most recent shaking chills began were associated with higher positivity than those obtained after 2 h.
      Traditionally, blood cultures have been taken prior to an expected chill or spiking fever (
      • Bennett Jr., I.L.
      • Beeson P.B.
      Bacteremia: a consideration of some experimental and clinical aspects.
      ). This recommendation was based on old and limited observational research (
      • Weiss H.
      • Ottenberg R.
      Relation between bacteria and temperature in subacute endocarditis.
      ,
      • Fox H.
      • Forrester J.S.
      Clinical blood cultures, analysis of over 5,000 cases.
      ). It was speculated that there was a lag period of approximately 1 h between the time of the sudden influx of bacteria and the onset of chills (
      • Bennett Jr., I.L.
      • Beeson P.B.
      Bacteremia: a consideration of some experimental and clinical aspects.
      ).
      As expected, the blood culture positivity was higher among those bottles that were drawn sooner after shaking chills began. This result is biologically plausible because macrophages mainly in the liver and spleen (classically known as the reticulo-endothelial system (RES)) may have removed circulating exogenous substances such as bacteria by the time fever begins (
      • Bennett Jr., I.L.
      • Beeson P.B.
      Bacteremia: a consideration of some experimental and clinical aspects.
      ,
      • Zheng M.
      • Kimura S.
      • Nio-Kobayashi J.
      • Iwanaga T.
      The selective distribution of LYVE-1-expressing endothelial cells and reticular cells in the reticulo-endothelial system (RES).
      ). Thus, when a patient has shaking chills, blood cultures should be obtained without delay.
      It is generally understood that the degree of chills is a valuable predictor of bacteremia (
      • Tokuda Y.
      • Miyasato H.
      • Stein G.H.
      • Kishaba T.
      The degree of chills for risk of bacteremia in acute febrile illness.
      ,
      • Shapiro N.I.
      • Wolfe R.E.
      • Wright S.B.
      • Moore R.
      • Bates D.W.
      Who needs a blood culture? A prospectively derived and validated prediction rule.
      ). The disadvantage of the chills grading scale, however, is that it is only useful for young and healthy patients, not for elderly patients with dementia. The infectious diseases division at the authors’ institution primarily treats the latter type of patients; the median age of the participants was 80 years and the IQR was 64–87 years (
      • Taniguchi T.
      • Tsuha S.
      • Shiiki S.
      • Narita M.
      Gram-stain-based antimicrobial selection reduces cost and overuse compared with Japanese guidelines.
      ). Therefore, this study focused solely on shaking chills and hypothesized that blood cultures obtained soon after shaking chills began would have higher blood culture positivity.
      One previous study investigated the timing of blood culture collection and fever spike, not chills or shaking chills (
      • Riedel S.
      • Bourbeau P.
      • Swartz B.
      • Brecher S.
      • Carroll K.C.
      • Stamper P.D.
      • et al.
      Timing of specimen collection for blood cultures from febrile patients with bacteremia.
      ). Although it was found that fever was not associated with blood culture positivity (
      • Riedel S.
      • Bourbeau P.
      • Swartz B.
      • Brecher S.
      • Carroll K.C.
      • Stamper P.D.
      • et al.
      Timing of specimen collection for blood cultures from febrile patients with bacteremia.
      ), the authors did not investigate prior antimicrobial use, which would have affected blood culture positivity (
      • Riedel S.
      • Bourbeau P.
      • Swartz B.
      • Brecher S.
      • Carroll K.C.
      • Stamper P.D.
      • et al.
      Timing of specimen collection for blood cultures from febrile patients with bacteremia.
      ). Fever was a very susceptible variable and was diminished easily by the use of antipyretics or corticosteroids (
      • Gleckman R.
      • Hibert D.
      Afebrile bacteremia. A phenomenon in geriatric patients.
      ).
      Although fever is an important vital sign and is easy to evaluate, high fever did not affect blood culture positivity in this study setting. It was noted that chills, especially shaking chills, are a more valuable indicator of bacteremia.
      In this study, elderly patients were also at high risk of blood culture positivity (adjusted OR 2.08, 95% CI 1.12–3.86). This result is consistent with those of previous studies (
      • Keating 3rd, H.J.
      • Klimek J.J.
      • Levine D.S.
      • Kiernan F.J.
      Effect of aging on the clinical significance of fever in ambulatory adult patients.
      ,
      • Sandberg T.
      • Skoog G.
      • Hermansson A.B.
      • Kahlmeter G.
      • Kuylenstierna N.
      • Lannergard A.
      • et al.
      Ciprofloxacin for 7 days versus 14 days in women with acute pyelonephritis: a randomised, open-label and double-blind, placebo-controlled, non-inferiority trial.
      ,
      • Taniguchi T.
      • Tsuha S.
      • Takayama Y.
      • Shiiki S.
      Shaking chills and high body temperature predict bacteremia especially among elderly patients.
      ). Advanced age might account for some of these results.
      Previous antimicrobial exposure within 48 h significantly reduced blood culture positivity (adjusted OR 0.33, 95% CI 0.12–0.92). The blood culture positivity of effective antimicrobial exposure within 48 h tended to be lower than that of ineffective antimicrobials (10.0% vs. 37.5%, p = 0.28). Although some reports have found no association between previous antibiotic usage and bacteremia (
      • Bates D.W.
      • Cook E.F.
      • Goldman L.
      • Lee T.H.
      Predicting bacteremia in hospitalized patients. A prospectively validated model.
      ,
      • Tokuda Y.
      • Miyasato H.
      • Stein G.H.
      • Kishaba T.
      The degree of chills for risk of bacteremia in acute febrile illness.
      ,
      • de Jager C.P.
      • van Wijk P.T.
      • Mathoera R.B.
      • de Jongh-Leuvenink J.
      • van der Poll T.
      • Wever P.C.
      Lymphocytopenia and neutrophil-lymphocyte count ratio predict bacteremia better than conventional infection markers in an emergency care unit.
      ), the present study confirmed that blood cultures should be taken before antibiotics are started.
      The frequency of shaking chills was also a strong indicator of bacteremia (adjusted OR 3.39, 95% CI 1.31–8.76). This result can be explained as follows: After circulating bacteria are eliminated through the RES, if the primary focus of bacteria is not under control, these pathogens invade the circulation again and cause shaking chills repeatedly. Therefore, patients were asked not only whether they had experienced shaking chills, but also when and how many times they had in order to evaluate the risk of bacteremia.
      Shaking chills may depend on the etiological agent. Although not significant, the median time to blood culture draw after the start of recent shaking chills tended to be shorter for Gram-negative infections than for Gram-positive infections (2 h vs. 4 h, p = 0.063). Former research has suggested that chills are more common in Gram-negative infections than in Gram-positive or viral infections (
      • Van Dissel J.T.
      • Schijf V.
      • Vogtlander N.
      • Hoogendoorn M.
      • van’t Wout J.
      Implications of chills.
      ,
      • Van Dissel J.T.
      • Numan S.C.
      • Van’t Wout J.W.
      Chills in ‘early sepsis’: good for you?.
      ). Endotoxin is a principal cell component that interacts with the host in Gram-negative bacterial sepsis (
      • Suffredini A.F.
      • Fromm R.E.
      • Parker M.M.
      • Brenner M.
      • Kovacs J.A.
      • Wesley R.A.
      • et al.
      The cardiovascular response of normal humans to the administration of endotoxin.
      ). It plays a central role in the pathogenesis of septic shock—the release of a wide variety of inflammatory mediators (
      • Martich G.D.
      • Boujoukos A.J.
      • Suffredini A.F.
      Response of man to endotoxin.
      ). Therefore, Gram-negative rods may cause marked signs of infection, namely shaking chills, and result in patients visiting a medical center earlier. With regard to anaerobe infections, the sample size was too small for analysis.
      Regarding the site of infection, an unknown source accounted for only 2.8% (6/214) of all cases. The blood cultures of five of these cases were positive, but their source was unclear. The blood culture of the remaining case was negative, but the patient had a history of splenectomy and he recovered after cefotaxime administration. The main reason for the low rate of unknown sources was the exclusion of patients with obscure diagnoses. Since the study objective was to clarify the association between bacterial infection and blood culture positivity after shaking chills, focus was placed only on definitely confirmed bacterial infection cases.
      Overall mortality was only 1.9% (4/214) despite the inclusion of elderly patients. This low rate may have been due to selection bias: only patients who had shaking chills were included, while those with severe lung disease, febrile neutropenia, and chronic kidney disease, as well as those who needed endoscopic or surgical management were not admitted to the infectious diseases division. The Japanese medical system might differ from that in other countries: Japan has a national public insurance system, patients have free access to any hospital, and ambulances are free to use. Therefore, sick patients are brought promptly to the hospital.
      This study has several limitations. First, there is the potential for selection bias. As the enrolled patients and the primary outcome were based mainly on culture results, the study population would be representative of a population with only culture-confirmed bacterial infection. Second, there is the potential for allocation bias. There is a possibility that the time lag between the recent shaking chills and the blood culture request was shorter in sick patients and longer in non-sick patients, because practitioners were urged to obtain blood cultures earlier in sick patients. Third, there is the potential for misclassification. The precise onset of shaking chills was difficult to recognize, although it was attempted to determine the timing not only directly through patients, but also from their family or healthcare workers. Fourth, there was uncertainty regarding the volume of the blood cultures. The volume of blood per culture has been shown to be an important variable in recovering microorganisms from patients with sepsis (
      • Bouza E.
      • Sousa D.
      • Rodriguez-Creixems M.
      • Lechuz J.G.
      • Munoz P.
      Is the volume of blood cultured still a significant factor in the diagnosis of bloodstream infections?.
      ). Although the medical staff were encouraged to obtain at least 10 ml in each set, it was not possible to determine the exact volume. Fifth, the results may not be applicable in tropical areas. Malaria or relapsing fever is famous for causing shaking chills and high body temperature (
      • Poulsen L.W.
      • Iversen G.
      Relapsing fever: a differential diagnosis to malaria.
      ).
      In conclusion, this analysis demonstrated that blood culture bottles that were drawn within 2 h after the start of the most recent shaking chills were more likely to become positive compared with those drawn after 2 h. Medical staff should be urged to obtain blood cultures as soon as possible if patients are suspected of having a bacterial infection and have a history of shaking chills. Further studies are needed to confirm whether these findings are applicable in other populations.

      Acknowledgements

      We thank all resident physicians who drew blood cultures for this study. We are grateful to Dr Tomokazu Kishaba, the former chief of the Division of Infectious Diseases of Okinawa Chubu Hospital, who retired in 2006. He was aware of the importance of shaking chills and educated all of the young trainee doctors on it, including all of the co-authors.

      Funding

      This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

      Conflict of interest

      None.

      References

        • Aronson M.D.
        • Bor D.H.
        Blood cultures.
        Ann Intern Med. 1987; 106: 246-253
        • Bates D.W.
        • Cook E.F.
        • Goldman L.
        • Lee T.H.
        Predicting bacteremia in hospitalized patients. A prospectively validated model.
        Ann Intern Med. 1990; 113: 495-500
        • Bennett Jr., I.L.
        • Beeson P.B.
        Bacteremia: a consideration of some experimental and clinical aspects.
        Yale J Biol Med. 1954; 26: 241-262
        • Bouza E.
        • Sousa D.
        • Rodriguez-Creixems M.
        • Lechuz J.G.
        • Munoz P.
        Is the volume of blood cultured still a significant factor in the diagnosis of bloodstream infections?.
        J Clin Microbiol. 2007; 45: 2765-2769
        • Coburn B.
        • Morris A.M.
        • Tomlinson G.
        • Detsky A.S.
        Does this adult patient with suspected bacteremia require blood cultures?.
        JAMA. 2012; 308: 502-511
        • de Jager C.P.
        • van Wijk P.T.
        • Mathoera R.B.
        • de Jongh-Leuvenink J.
        • van der Poll T.
        • Wever P.C.
        Lymphocytopenia and neutrophil-lymphocyte count ratio predict bacteremia better than conventional infection markers in an emergency care unit.
        Crit Care. 2010; 14: R192
        • Fox H.
        • Forrester J.S.
        Clinical blood cultures, analysis of over 5,000 cases.
        Am J Clin Pathol. 1940; 10: 493-504
        • Gleckman R.
        • Hibert D.
        Afebrile bacteremia. A phenomenon in geriatric patients.
        JAMA. 1982; 248: 1478-1481
        • Keating 3rd, H.J.
        • Klimek J.J.
        • Levine D.S.
        • Kiernan F.J.
        Effect of aging on the clinical significance of fever in ambulatory adult patients.
        J Am Geriatr Soc. 1984; 32: 282-287
        • Kiyoyama T.
        • Tokuda Y.
        • Shiiki S.
        • Hachiman T.
        • Shimasaki T.
        • Endo K.
        Isopropyl alcohol compared with isopropyl alcohol plus povidone-iodine as skin preparation for prevention of blood culture contamination.
        J Clin Microbiol. 2009; 47: 54-58
        • Liesenfeld O.
        • Lehman L.
        • Hunfeld K.P.
        • Kost G.
        Molecular diagnosis of sepsis: new aspects and recent developments.
        Eur J Microbiol Immunol (Bp). 2014; 4: 1-25
        • Martich G.D.
        • Boujoukos A.J.
        • Suffredini A.F.
        Response of man to endotoxin.
        Immunobiology. 1993; 187: 403-416
        • Poulsen L.W.
        • Iversen G.
        Relapsing fever: a differential diagnosis to malaria.
        Scand J Infect Dis. 1996; 28: 419-420
        • Riedel S.
        • Bourbeau P.
        • Swartz B.
        • Brecher S.
        • Carroll K.C.
        • Stamper P.D.
        • et al.
        Timing of specimen collection for blood cultures from febrile patients with bacteremia.
        J Clin Microbiol. 2008; 46: 1381-1385
        • Sandberg T.
        • Skoog G.
        • Hermansson A.B.
        • Kahlmeter G.
        • Kuylenstierna N.
        • Lannergard A.
        • et al.
        Ciprofloxacin for 7 days versus 14 days in women with acute pyelonephritis: a randomised, open-label and double-blind, placebo-controlled, non-inferiority trial.
        Lancet. 2012; 380: 484-490
        • Seymour C.W.
        • Liu V.X.
        • Iwashyna T.J.
        • Brunkhorst F.M.
        • Rea T.D.
        • Scherag A.
        • et al.
        Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (sepsis-3).
        JAMA. 2016; 315: 762-774
        • Shapiro N.I.
        • Wolfe R.E.
        • Wright S.B.
        • Moore R.
        • Bates D.W.
        Who needs a blood culture? A prospectively derived and validated prediction rule.
        J Emerg Med. 2008; 35: 255-264
        • Sorensen J.B.
        • Klee M.
        • Palshof T.
        • Hansen H.H.
        Performance status assessment in cancer patients. An inter-observer variability study.
        Br J Cancer. 1993; 67: 773-775
        • Suffredini A.F.
        • Fromm R.E.
        • Parker M.M.
        • Brenner M.
        • Kovacs J.A.
        • Wesley R.A.
        • et al.
        The cardiovascular response of normal humans to the administration of endotoxin.
        N Engl J Med. 1989; 321: 280-287
        • Taniguchi T.
        • Tsuha S.
        • Shiiki S.
        • Narita M.
        Gram-stain-based antimicrobial selection reduces cost and overuse compared with Japanese guidelines.
        BMC Infect Dis. 2015; 15: 458
        • Taniguchi T.
        • Tsuha S.
        • Takayama Y.
        • Shiiki S.
        Shaking chills and high body temperature predict bacteremia especially among elderly patients.
        Springerplus. 2013; 2: 624
        • Tokuda Y.
        • Miyasato H.
        • Stein G.H.
        • Kishaba T.
        The degree of chills for risk of bacteremia in acute febrile illness.
        Am J Med. 2005; 118: 1417
        • Van Dissel J.T.
        • Numan S.C.
        • Van’t Wout J.W.
        Chills in ‘early sepsis’: good for you?.
        J Intern Med. 2005; 257: 469-472
        • Van Dissel J.T.
        • Schijf V.
        • Vogtlander N.
        • Hoogendoorn M.
        • van’t Wout J.
        Implications of chills.
        Lancet. 1998; 352: 374
        • Weiss H.
        • Ottenberg R.
        Relation between bacteria and temperature in subacute endocarditis.
        J Infect Dis. 1932; 50: 61-68
        • Zheng M.
        • Kimura S.
        • Nio-Kobayashi J.
        • Iwanaga T.
        The selective distribution of LYVE-1-expressing endothelial cells and reticular cells in the reticulo-endothelial system (RES).
        Biomed Res. 2016; 37: 187-198