| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Medical Intensive Care Unit of the Department of Internal Medicine (Drs. Bossink, Groeneveld, and Thijs), Academisch Ziekenhuis Vrije Universiteit; the Central Laboratory of the Netherlands Red Cross Blood Transfusion Service (Dr. Hack); and the Institute for Cardiovascular Research at the Vrije Universiteit, Amsterdam, the Netherlands.
Correspondence to: A.B.J. Groeneveld, MD, PhD, Medical Intensive Care Unit, Academisch Ziekenhuis Vrije Universiteit, De Boelelaan 1117, 1081 HV Amsterdam, the Netherlands; e-mail: johangroeneveld@compuserve.com or johan.groeneveld{at}azvu.nl
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
38.0°C axillary or
38.3°C rectal) were analyzed and compared with the criteria for
the systemic inflammatory response syndrome (SIRS), including an
abnormal body temperature and WBC count, tachypnea and tachycardia, and
sepsis, defined as SIRS and the presence of a clinical infection. Design: A prospective cohort study.
Setting: Department of internal medicine at a university hospital.
Patients: In 300 hospitalized medical patients with new onset of fever, demographic, clinical, and laboratory variables were obtained during the first 2 days after inclusion, and peak and nadir values, when appropriate, were taken. Microbiologic results for 7 days were collected. Clinical information was used to decide on the presence of a clinical infection.
Measurements and results: One hundred thirty-three of 300 patients (44%) had a microbial infection: 26% suffered from local microbial infection only, 9% from bacteremia only, and 9% had bloodstream plus local microbial infections. Patients with a microbial infection had a higher World Health Organization performance score at home (p < 0.05), higher peak body temperature (p < 0.001), higher nadir and peak WBC counts (p < 0.05), lower nadir platelet count (p < 0.01), higher peak alanine and aspartate aminotransferases (p < 0.01), and lower nadir albumin (p < 0.001) levels in blood during the first 2 days after inclusion than those without infection. Using multivariate techniques, predictors for microbial infection or bacteremia alone, independent of age, sex, underlying disease, and clinical infection, were peak temperature, peak WBC count, and nadir platelet count and albumin level. In contrast, conventional SIRS/sepsis definitions and criteria predicted microbial infection less well, mainly because tachypnea and tachycardia were of no predictive value.
Conclusions: In febrile medical patients, microbial infection can be predicted with use of easily obtained clinical and laboratory variables, including peak temperature, peak WBC count, and nadir platelet count and albumin level within the first 2 days. The new model predicted microbial infection better than conventional SIRS/sepsis criteria. This may help to improve the clinical recognition of the systemic host response to microbial infection and to refine SIRS/sepsis definitions.
Key Words: bacteremia • fever • infection • prediction • sepsis • systemic inflammatory response syndrome
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Moreover, even though clinicians empirically use multiple clinical signs and symptoms to define the systemic host response to microbial infection, no one knows whether this picture is correct.2 15 Thus, many other predictors of bloodstream infection and its mortality have been described in a variety of "sepsis" patients, including those with (nosocomial) fever.3 13 16 17 18 19 20 21 22 23 24 25 26 27 The character and value of these predictors differed between studies, and this may relate in part to differences in entry criteria. The studies more often looked at risk factors and localizing features than at the clinical and laboratory signs and symptoms that may be involved in the systemic host response to infection; thus, they contributed little to the understanding of the predictive value of the SIRS criteria, particularly in the elderly, who, to complicate matters, might have infection with only a few signs and symptoms.9 16 17 19 20 21 22 23 25 27 28 Nevertheless, bloodstream infection has been associated with advanced age,17 20 22 28 a low premorbid performance status,21 22 prior chronic disease and procedures,17 19 20 22 27 the presence of a possible focus,18 19 20 21 27 an elevated temperature,17 19 20 22 25 27 28 shaking chills,19 21 27 a rise in respiratory rate (sometimes blunted),25 a fall in BP,19 22 23 25 and an altered mental status.16 22 23 27 Also, an elevated erythrocyte sedimentation rate (ESR),17 19 20 28 altered WBC counts,16 17 19 20 22 23 25 27 28 thrombocytopenia,9 25 an elevated plasma bilirubin, alkaline phosphatase (AF), and alanine and aspartate aminotransferase (ALAT, ASAT) levels,22 29 decreased plasma albumin levels,21 22 30 and renal dysfunction16 20 21 22 may predict bacteremia. Only a few studies tried to predict local and bloodstream microbial infections combined.3 17 20 Finally, sepsis scoring systems, often based on an arbitrary set of criteria, have been developed that may be representative for the severity of infection and predictive for mortality, but none of them addressed the predictive value for microbial infection itself.8 13 26
Early and accurate prediction of a microbial infection as the cause of fever at the bedside would help the physician, particularly in hospitalized patients, to decide on antimicrobial therapy and supportive care before culture results would become available.3 If the likelihood for a severe local or bloodstream infection is high,6 11 13 14 21 22 23 an early start of empiric antibiotic medication before the results of cultures are available would improve chances of survival. Moreover, accurate prediction of microbial infection would help to select patients for future studies on compounds that bind or neutralize circulating microbial products or inflammatory mediators: such studies so far have largely failed, mainly because of the poor specificity of the present inclusion criteria and definitions.6 24 26 31
It was the aim of the current study to evaluate the predictive value of easily obtainable demographic, clinical, and laboratory variables for microbial infection, ie, local and bloodstream microbial infection, in hospitalized medical patients with fever, with emphasis on the clinical features that could be associated with the systemic host response, independent of age, underlying diseases, risk factors, and focus. We also compared the predictive value of our models with that of the SIRS and sepsis definitions and their criteria.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
38.0°C axillary or 38.3°C rectal) admitted to the
Department of Internal Medicine were included into the study, which was
approved by the local committee on ethics. All patients or their
closest relatives had given informed consent orally before inclusion.
Exclusion criteria were the presence of pregnancy, shock, a life
expectancy of < 24 h, and current therapy with cytokines,
eg, interferon-
or interleukin 2, or cytostatic drugs for
solid tumors or malignant hematologic disease, which may grossly alter
a normal host response to infection. Patients were taken care of by
physicians not involved in the study. They ordered blood tests not in
the protocol, and prescribed antibiotics and supportive measures. The
clinical judgment by the treating physician was supplemented, if
necessary, by imaging techniques for assessing a possible focus. If a
focus was clinically evident, the patient was considered to suffer from
a clinical infection. At inclusion, an estimation of the interval between hospital admission and development of fever was made. Patients were considered to have nosocomial fever if fever developed at least 72 h after admission into the hospital. Eleven of 15 febrile patients transferred from other hospitals had community-acquired fever, either at home or within 72 h after admission into the referring hospital. Eight of nine febrile patients from nursing homes were also considered to have community-acquired fever, whereas one patient had nosocomial fever in the hospital (> 72 h after admission). All other patients were considered to have community-acquired fever. In the presence of a microbial infection, the patients were considered to have a nosocomial or a community-acquired infection. At inclusion, demographic variables were recorded, such as age, sex, and devices or procedures done within 3 days before inclusion, including insertion of indwelling central venous catheters, bladder catheters, start of artificial ventilation, and bronchoscopy. Also recorded were underlying diseases and other factors that may predispose to infection, such as use of immunocompromising drugs, ie, prior cytostatic or current corticosteroid treatment, prior use of antibiotics, surgery within 2 months before inclusion, neurologic disease acquired within 2 months before inclusion, GI disease, endocrine disease, malignancies, infectious disease (HIV, AIDS, active tuberculosis), skin disease, cardiovascular disease, respiratory disease, autoimmune disease, and urogenital disease. In 20 patients, information about prior antibiotic use was inconclusive. The International Classification of Disease (ICD-9) definitions were used to describe the disease states. The demographic variables recorded also included the patient performance status expressed as World Health Organization (WHO) and Karnofsky scores. The first score ranges from 0 to 4, in which 4 indicates most severe disability. In contrast, the Karnofsky performance score ranges from 0 to 100, in which 100 represents no disability.
At inclusion and each morning of the 2 days thereafter, clinical variables such as body temperature, shaking chills, respiratory rate, heart rate (HR), mean arterial BP, Glasgow Coma Score, and WHO performance scores were assessed. At the time points of the study, blood samples were obtained in standard aliquots (Becton and Dickinson; Erembodegem, Belgium) and processed for biochemical variables using a Hitachi 747 automatic analyzer (Tokyo, Japan) and for hematologic variables using a Sysmec SE 9000 analyzer (Kolbe, Japan). Hematologic variables included ESR, hemoglobin concentration, hematocrit, and WBC and platelet counts. Biochemical variables included AF, ASAT, ALAT, bilirubin, lactate dehydrogenase, creatinine, lactate, and albumin levels. Patients were classified as having SIRS or sepsis if meeting criteria at or within 2 days after inclusion. According to the American College of Chest Physicians/Society of Critical Care Medicine consensus conference,5 SIRS was considered present if two or more of the following criteria were met: a body temperature > 38°C or < 36°C; a HR > 90 beats/min; a respiratory rate > 20 breath/min; or a WBC count > 12 x 109/L, < 4 x 109/L, or > 10% immature (band) forms. Patients with SIRS and a clinical infection were considered to suffer from sepsis.5 All local and blood culture results during a follow-up period of 7 days after inclusion were taken for analysis. At admission, local specimens for microbiologic evaluation were collected, depending on the focus of clinical infection, as judged by the treating physician on the basis of physical examination and supplementary imaging techniques. Local specimens were processed using standardized procedures. Two blood samples for culture were obtained by venipuncture at inclusion as part of the protocol; blood was taken for culture from arterial catheters in ICU patients. Supplementary blood for culture was collected whenever the treating physician considered this necessary. Blood for culture was processed using "delayed vial entry" bottles for aerobic and anaerobic cultures and the Bactec 9120/9240 automatic analyzers (Becton Dickinson). Bottles were incubated for a maximum of 7 days. If the analyzers showed growth, Gram's stains were prepared and identification and sensitivity cultures were processed. Cultures for viruses and chlamydiae were not done. Specific stains for tuberculous (n = 1) or fungal (n = 13) infections were performed when indicated on clinical grounds. One patient had malaria (Plasmodium vivax), as shown by a thick smear of blood. Local positive microbiologic results were only taken into account if considered to reflect microbial infection rather than colonization, and the former was defined as the clinical need, as judged by the treating physician, to continue or start antibiotic treatment. Blood cultures containing Staphylococcus epidermidis were considered contaminated if only one bottle revealed growth and there were no intravascular catheters. A clinical infection was considered microbiologically proven in case of a local microbial infection. Clinical intravascular catheter infections were considered microbiologically proven in case of positive tip and blood cultures, including those with S epidermidis. Clinical intracardiac infection was considered microbiologically proven in case of echocardiographic evidence for vegetations and positive blood culture results.
Statistical Analysis
Patients were divided into four groups, so that group 1 had
negative cultures, group 2 had a local microbial infection only
(ie, positive local cultures including positive specific
stains for tuberculous or fungal infections), group 3 had a microbial
bloodstream infection (positive blood cultures) only, and group 4 had
both microbial infection locally and in the bloodstream (ie,
positive local and blood cultures), including one patient with P
vivax parasitemia. Patients with microbiologically proven
intracardiac or intravascular infections were classified into group 4.
For each clinical and laboratory variable, the peak (ESR, hemoglobin
concentration, hematocrit, ASAT, ALAT, HR) or, where appropriate, the
nadir value (mean arterial BP, albumin), or both (WBC and platelet
counts) among the three determinations at inclusion and daily at the
following 2 days was used for analysis. Comparison of groups was
performed using the Kruskal-Wallis test or the Mann-Whitney
U test for continuous variables (group 1 vs groups 2, 3, and
4), and the 
2 test with Yates's correction or
Fisher's Exact Test for categorical variables. To find the smallest
set of demographic, clinical, and laboratory variables significantly
contributing to the final model for prediction of microbial infection,
forward stepwise multiple logistic regression was performed, beginning
with all variables. We evaluated the predictive value for microbial
infection of models without (model 1) or with (model 2) inclusion of
the presence or absence of a clinical infection, for comparison with
the value of the SIRS and sepsis criteria, and prediction of bacteremia
on the basis of a model (3) with inclusion of the types of foci of
clinical infection. The models are of the type P (probability for
infection) = 1 / (1 + e-q), where
q = constant + (coefficient x variable). Forced-entry multiple
logistic regression was used, by treating SIRS/sepsis-defining
parameters as continuous variables, to find the predictive value for
microbial infection of the body temperature, heart and respiratory
rates, and WBC counts. The models were evaluated by comparing predicted
with observed infection/bacteremia rates, using the Hosmer Lemeshow
statistic and Spearman rank correlation coefficient, in deciles
of risk. The discriminating power of the categorical SIRS/sepsis
definitions and of the logistic regression models was evaluated by
specificity, sensitivity, and positive (PPV) and negative predictive
(NPV) values. The likelihood ratio (LHR) is
sensitivity/(1 - specificity). Receiver-operating characteristics
curves plotting sensitivity vs 1 - specificity were constructed to
examine the diagnostic value of the regression models, with an area
under the curve (AUC) approaching 1 indicating greater diagnostic
power. Values are given as median and range to obviate the influence of
potential outliers. A p value < 0.05 was considered statistically
significant, and exact probabilities are reported for the logistic
regression models.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Microbial Infection
One hundred thirty-three patients (44%) had a microbial
infection, so that 106 of 133 patients (80%) with a microbial
infection had clinical infections (vs 106 of 167 patients in group 1;
p < 0.05). Con-versely, clinical infection was microbiologically
proven in 84 of 212 patients (40%). Fifty-three patients (18%) had
bacteremia. Thirty-four patients (42%) in group 2 and 3 patients
(11%) in group 4 had microbial infections with a clinical respiratory
foci (p < 0.005), and microbial infections with clinical skin foci
in 15 patients (19%) and 12 patients (44%), respectively
(p < 0.05). Conversely, clinical infections differed (p < 0.001)
in the frequency of a microbial infection, with the highest rate of
microbial infections in 82% of patients with clinical urinary tract
infections and the lowest rate of microbial infections in 19% of
patients with clinical GI infections. Groups did not differ with
respect to the microbial species involved either (Table 2
). However, infectious foci differed in the microorganisms involved, so
that Gram-negative bacteria predominated in microbiologically proven
genitourinary infections and Gram-positive bacteria in skin infections
(p < 0.005). The prevalence of a microbial infection was the same in
patients with or without prior antibiotic usage (29 of 69 patients
[43%] and 92 of 211 patients [44%], respectively). There were no
differences in the prevalence of local microbial or bloodstream
infections between patients with community-acquired and nosocomial
fever: 97 patients with local microbial infections (44%) and 35
patients with bacterema (16%) of 217 patients with community-acquired
fever, vs 36 patients with a local microbial infection (42%) and 18
patients with bacteremia (22%) of 83 patients with nosocomial fever.
The interval between inclusion and reports of the first
microbiologic results was 40 h (range, 20 to 72 h).
|
Predictive Value of Study Variables for Microbial Infection
Age and the distribution of underlying disease did not differ
among infection groups (Table 1)
. However, 36 women (25%) vs 17 men
(11%) had positive blood cultures (p < 0.05). Women outnumbered men
in urinary tract and skin infections (p < 0.05). Four of 5 ICU
patients receiving mechanical ventilation (n = 16) included in this
study had a microbial infection vs 80 of the 240 patients without any
device or procedure (p < 0.05). Six of 15 patients with indwelling
central venous catheters had bacteremia vs 38 of 240 patients without
any device or procedure (p < 0.05). All other possibly predisposing
devices or procedures, ie, bladder and urinary stoma
catheters (n = 8), thoracic drains (n = 7), biliary drains and
stents (n = 6), bronchoscopy (n = 2), digestive tract procedures
(n = 9), and peritoneal dialysis (n = 4) or hemodialysis (n = 4),
did not significantly predispose to local or bloodstream infections or
both. Table 3
shows that patients with a microbial infection had a relatively high
peak body temperature. Groups differed with respect to the WHO score at
home and in the hospital, with a lower performance status in
microbially infected vs noninfected patients. Several laboratory
variables also significantly differed between groups (Table 4
).
|
|
|
|
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In 44% of our febrile patients, results of cultures either from a local site or from the bloodstream were positive. The frequency of microbial foci of infection in our febrile patients generally agrees with the literature.4 13 14 17 24 The reported bacteremia rate of 21% in febrile medical patients17 21 also conforms with our results (18%). Our febrile culture-negative patients may have suffered from nonbacterial, nonfungal, or noninfectious diseases.1 3 We limited our cultures to bacterial and fungal infections, however, because of their risks and therapeutic implications, and did not routinely perform viral and chlamydial cultures, so that the frequency of microbial infection may have been underestimated somewhat in our study. The low frequency of polymicrobial infections, usually associated with intraabdominal, intravascular catheter, or wound infections, agrees with the literature,1 4 16 28 as does the frequency and type of the infectious foci, the predominance of community-acquired over nosocomial infections, and the microorganisms involved.1 4 10 13 14 20 21 23 27 28
We cannot explain the somewhat higher bacteremia prevalence in women than in men. Women, however, more often had infections of the skin and urinary tract. Also, bacteremia occurred more frequently in skin and urinary tract infections, the latter in accordance with results of others.21 22 Age and underlying disease did not affect the infection frequency. Conversely, the multivariate prediction of microbial infection was independent of age, and this may disagree with studies suggesting a higher likelihood for bacteremia and relative paucity of clinical signs in elderly as opposed to young febrile patients.3 17 20 21 22 23 28 Otherwise, our risk factors for infection largely agree with those in the literature describing mechanical ventilation, indwelling catheters, and malignancies as predisposing factors to serious local and bloodstream infections.1 12 13 14 16 18 19 21 22 27 The prior use of antibiotics did not influence our results, in contrast to the reported association with fungemia.18 27 Finally, prior surgery was a negative predictive factor for infection in our febrile medical patients, both in univariate and multivariate analyses.
The peak temperature differed between groups and was predictive of a microbial infection in multivariate analysis, as others have found.17 19 21 22 25 28 Shaking chills were not predictive in our study and this is in accordance with some but not other studies.19 21 22 23 An elevated HR was, in contrast to common belief, not predictive of a microbial infection, confirming the results of several other studies,9 23 25 which argues against inclusion of this criterion in the SIRS/sepsis definition.
The literature describes a relatively low BP19 22 23 25 and, as our study, a blunted rise in respiratory rate24 in septic/bacteremic patients, but this is not confirmed by others.9 23 24 27 Mental changes can be part of the clinical picture of severe infections,1 16 22 23 27 28 32 but there was no difference among groups in the GSC score. However, this scoring system may not be sensitive enough to detect the subtle mental changes associated with infections.2 27 32 As far as laboratory variables are concerned, groups differed in WBC and platelet counts. In fact, leukocytosis is a sign that, like fever, often suggests development of sepsis. The leukocytosis associated in our study with microbial infection, particularly if local, supports the predictive value of leukocytosis for microbial infection found before and the use of WBC abnormalities as a criterion for the SIRS/sepsis definition.2 3 16 17 19 20 21 22 23 27 Bacterial infection may induce low-grade disseminated intravascular coagulation,2 partly explaining the low platelet counts in patients with a microbial infection.9 25 The albumin plasma level is low in patients with inflammation- or infection-causing fever, and the decrease contributed to prediction of infection and bacteremia in previous studies.21 22 30 In our study, the albumin levels were lowest in the presence of bloodstream infection. AF and aminotransferase plasma values were elevated in patients with infection, particularly in those with bacteremia, which may indicate a deleterious effect of microbes and released cytokines on hepatocytes.2 21 22 29 The aminotransferase levels did not contribute to infection prediction in multivariate analysis, however, possibly because of interdependency with other factors. Finally, a high creatinine plasma level or renal dysfunction may also predict microbial infection.16 20 21 22 Apart from temperature, multivariate analyses confirmed univariate analyses for WBC and platelet counts and the albumin plasma level.
We mainly focused on the clinical features of the host response to infection in medical patients with fever, and previous studies predominantly paid attention to the prediction of bacteremia by risk factors in other patient groups. Nevertheless, some comparisons can be made with studies using multivariate techniques.16 17 18 19 20 21 22 23 27 For instance, Bates et al19 found that elevated temperature, rapidly and ultimately fatal underlying disease, presence of chills, IV drug abuse, acute abdominal symptoms, and major comorbidity predicted bacteremia in hospitalized patients. Their model did less well when prospectively validated at another site.33 In a later study, these observations were extended in patients with sepsis.27 Leibovici et al21 22 found in febrile medical patients that a low premorbid performance status, prior diabetes mellitus, chills, leukocytosis, a low albumin level, renal dysfunction, and clinical urinary tract infection were predictive of bacteremia. Fontanarosa et al23 found mental changes, vomiting, and increased WBC band forms to be predictive for bacteremia in elderly emergency patients. Peduzzi et al25 predicted bacteremia and Gram-negative bacteremia in hospitalized patients with sepsis. They found an elevated temperature, low platelet count, and low systolic BP to be independent predictors for bacteremia. Our data partly confirm and extend these observations, even though the diagnostic performance of the multiple logistic regression models was relatively poor in some studies.23 25 In contrast, the diagnostic performance of our model 3 was superior to that of models 1 and 2, suggesting greater precision for prediction of bacteremia than of local microbial infection. In a previous paper on febrile medical patients,34 we reported that hospital-acquired fever, peak respiratory rate, the nadir score on the Glasgow Coma Score, and the nadir plasma albumin level were of predictive value for death. Apparently, the clinical variables predictive for infection may not fully correspond with those predictive for mortality in febrile patients.
We found that 95% of patients with fever fulfilled two or more SIRS criteria, indicating that these criteria are closely related to fever and that the SIRS definition is too sensitive and aspecific for bedside prediction of microbial infection. Indeed, 56 to 93% of hospitalized patients may have SIRS.4 5 6 7 11 12 14 In agreement with our results, the frequency of microbial infection may increase somewhat with increasing SIRS criteria,14 but the predictive value of SIRS for infection and bacteremia is poor. Nevertheless, patients with SIRS and bacteremia may be more likely to have a demonstrable local infection, ie, sepsis, than bacteremic patients without SIRS.14 Of SIRS patients, only about 25 to 49% may have sepsis, and more than half of the sepsis patients may have a proven microbial infection,4 6 11 12 17 agreeing with our results. Twenty-two percent of our sepsis patients had bacteremia, and this is also in accordance with observations by others.4 6 9 10 12 17 24 The poor predictive value of continuous SIRS criteria for infection in our study was indeed improved by including a clinical infection, so that sepsis was predictive for a microbial infection. Hence, sepsis may be a better predictor of infection than the SIRS definition.4 5 11 12 Nevertheless, our new models predicted microbial infection better than models based on conventional criteria. This suggests that the present SIRS and sepsis definitions poorly indicate the host response to microbial infection.
Our results may not apply to patients admitted to other hospital services after trauma and surgery, and further study into prediction of bacteremia, for instance by applying our models prospectively, is warranted.33 Because the study aimed to identify the systemic host response rather than risk factors for microbial infection, we did not split the cohort into two populations for derivation and validation of models predictive for microbial infection, even though a model usually predicts less if applied prospectively than in the population it is derived from.21 22 33 Another limitation of our study may be the choice of fever as entry criterion, since patients with serious, often hospital-acquired infections can present with hypothermia.15 16 28 We did not study the latter, because fever is more common and more likely to be associated with bloodstream infection in hospitalized patients than hypothermia.1 15 16 If the multivariate models of this study are to be used in practice, empiric antibiotic treatment in patients with predicted infection (model 2), which is not proven microbiologically or vice versa, will result in overtreatment in 13% but undertreatment in 21% of all patients. Our results do not eliminate the need for taking cultures and starting empiric treatment in febrile patients.23 Nevertheless, they may help to objectify the systemic host response to microbial infection and thereby the criteria for defining sepsis. Our results specifically suggest that tachycardia and tachypnea in the SIRS/sepsis definition hardly predict microbial infection and should be replaced, perhaps, by lowered platelet counts and albumin levels. Thus, our study shows the limitations of SIRS/sepsis definitions in predicting microbial infection in febrile medical patients and suggests that other clinical and laboratory variables may be more helpful.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Abbreviations: AF = alkaline phosphatase; ALAT = alanine aminotransferase; ASAT = aspartate aminotransferase; AUC = area under the curve; ESR = erythrocyte sedimentation rate; HR = heart rate; LHR = likelihood ratio; NPV = negative predictive value; PPV = positive predictive value; SIRS = systemic inflammatory response syndrome; WHO = World Health Organization
Received for publication October 26, 1998. Accepted for publication March 2, 1999.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. B. J. Groeneveld, A. W. J. Bossink, G. J. van Mierlo, and C. E. Hack Circulating Inflammatory Mediators in Patients with Fever: Predicting Bloodstream Infection Clin. Vaccine Immunol., November 1, 2001; 8(6): 1189 - 1195. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
