ACADEMIC EMERGENCY MEDICINE 2011; 18:851–859 © 2011 by the Society for Academic Emergency Medicine
Objectives: Patients presenting to emergency departments (ED) with nonspecific complaints (NSCs) such as “not feeling well,”“feeling weak,”“being tired,”“general deterioration,” or other similar chief complaints that do not have a readily identifiable probable etiology are a common patient group at risk for adverse outcomes. Certain biomarkers, which have not yet been tested for prognostic value when applied to ED patients with NSCs, have emerged as useful tools for predicting prognosis in patients with a variety of diseases. This study tested the hypothesis that two of these novel markers, copeptin (a C-terminal portion of provasopressin) and/or peroxiredoxin-4 (Prx4), an enzyme that degrades hydrogen peroxide, singly or together are helpful in predicting death in the near term among patients presenting to the ED with NSCs.
Methods: The Basel Non-specific Complaints (BANC) study is a delayed type cross-sectional diagnostic study with a prospective 30-day follow-up. ED patients with NSCs were consecutively enrolled. Patients with vital parameters out of the normal range were excluded. The primary endpoint of this study was the predictive value of copeptin and Prx4 for 30-day mortality in patients with NSCs. Measurement of both copeptin and Prx4 was performed in serum samples with sandwich immunoluminometric assays.
Results: On follow-up at 30 days after ED presentation, 28 of 438 patients with NSC had died. Copeptin and Prx4 concentrations were significantly higher in nonsurvivors than in survivors (Kruskal-Wallis test, p = 0.0001 and p < 0.0001, respectively). In univariate models, Prx4 (likelihood ratio [LR] χ2 = 22.24, p < 0.00001, concordance index [C-index] = 0.749) and copeptin (LR χ2 = 16.98, p = 0.00004, C-index = 0.724) were both predictive of 30-day mortality, and elevated levels were associated with an increased mortality. The bivariable model, which included both Prx4 and copeptin (LR χ2 = 28.22, p < 0.00001, C-index = 0.783), allows a significantly better prediction than the univariate Prx4 (p = 0.00025) and copeptin models (p = 0.00099), respectively. Both biomarkers provided independent and additional information to clinical risk scores (Katz Activities of Daily Living [ADL] and Charlson Comorbidity Index [CCI], all p < 0.0005).
Conclusions: Copeptin and Prx4 are new prognostic markers in patients presenting to the ED with NSCs. Copeptin and Prx4 might be valuable tools for risk stratification and decision-making in this patient group.
Nonspecific complaints (NSCs) such as “not feeling well,”“feeling weak,”“being tired,” and “general deterioration” are very common among emergency department (ED) patients, but such NSCs have been poorly studied in the past.1 Emergency physicians (EPs) are confronted with a broad differential diagnosis for ED patients with such NSCs, ranging from low-level threats to well-being, such as insufficient home care, to acute life-threatening conditions.2–4 Confirming a previous report,5 we have recently shown that 59% of individuals who presented to our ED with NSCs suffered from an acute medical problem that required early intervention.6
The evaluation and diagnostic workup of patients with NSCs, who are disproportionately elderly, can be very time-consuming and can be complicated by comorbidities, polypharmacy, or an altered mental status. Therefore, extensive diagnostic efforts are often undertaken to exclude a serious underlying condition. Such intensive workups can lead to prolonged throughput times and/or ED observation unit stays for these patients. On the other hand, the underlying condition that causes some patients’ NSC might be underestimated by EPs, and this may result in ineffective or inaccurate diagnostic efforts, and/or inappropriate disposition, as well as poor patient outcomes.5 Therefore, the ultimate goal of the Basel Non-specific Complaints (BANC) study is to identify risk stratification tools for patients with NSCs, toward the improved identification of patients with NSCs who are at high risk for short-term deterioration or death.
Different stress response systems (e.g., the hypothalamic-pituitary-adrenal [HPA] axis and the antioxidative response) can be assessed by biomarkers, such as copeptin and peroxiredoxin-4 (Prx4). Copeptin, the C-terminal portion of provasopressin, is a 39-amino-acid glycopeptide that has been found to be a “shadow” fragment reflecting vasopressin (AVP) release.7 In addition to its hemodynamic and osmoregulatory effects, AVP mirrors the individual stress level.8 Copeptin has been shown to reflect disease severity in a variety of clinical conditions and might therefore be a prognostic marker due to the positive correlation of the individual stress level with morbidity and mortality in our cohort.9
The concept of oxidative stress comprises another rather heterogeneous set of clinically relevant conditions with different dynamics and compensatory reactions. Diverse pathophysiologic states are propagated by oxidative stress as a consequence of increased reactive oxygen species.10,11 Depending on the coping mechanisms, this might lead to a broad spectrum of diseases.10,11 Prx4 counteracts the damaging effects of oxidative stress by reduction of hydrogen peroxide.12 Recently, a new assay to detect Prx4 in human serum of healthy and critically ill subjects was developed and validated.13 In a subsequent study in patients with sepsis on an intensive care unit, high Prx4 serum levels were predictive of in-hospital mortality.14 Therefore, patients presenting to the ED with NSCs represent a promising group to test the hypothesis that elevated serum levels of copeptin and/or Prx4 predict an increased likelihood of death.
The BANC study is a delayed type cross-sectional prognostic study with a prospective 30-day follow-up. The study protocol was approved by the local ethics committee (EKBB 73/07) in charge. The study is registered with ClincalTrials.gov (#NCT00920491) and is in compliance with the Helsinki Declaration.
Study Setting and Population
The study was carried out in the ED of the University Hospital Basel, Basel, Switzerland. The hospital is an urban 700-bed primary and tertiary care university hospital with an ED census of 42,000 patients per year.
All nontrauma patients presenting to the ED were consecutively screened for eligibility from May 24, 2007, until May 14, 2009, in a two-step process. The first step was triage by an ED nurse, according to the Emergency Severity Index (ESI)15 protocol, to identify all ESI 2 and ESI 3 patients. The second step was a more selective screening by a resident assigned to duty in our ED, via use of a one-page screening tool.
Study physicians enrolled all patients presenting with NSCs such as “generalized weakness” or “general condition impairment” as previously described.6 Briefly, inclusion was only performed by trained study physicians. The patients studied were the same ones that were described in a prior article.6
Study physicians excluded several groups of patients: hemodynamically unstable patients, patients with persistent signs of shock, and those with vital parameters significantly out of the normal range (blood pressure < 80 mm Hg, heart rate < 55 or > 120 beats/min, respiration rate >20 breaths/min, tympanic temperature > 38.5°C, SaO2 < 92%); patients with specific complaints (e.g., headache, dyspnea) or clinical features suggestive of a working diagnosis (e.g., jaundice); patients with known terminal conditions (e.g., cachexia in end-stage cancer); and patients referred from other hospitals.
Patient data were obtained by three trained study physicians with an average clinical experience of 6 years. Training for inclusion was performed under supervision of the head of the department in common training sessions.
All patient data were gathered by a structured interview: demographic baseline data, all presenting symptoms, a complete list of relevant comorbidities as assessed by the Charlson Comorbidity Index (CCI), the Katz Activities of Daily Living (ADL), and the patients’ prescribed medications. Blood testing was performed in all patients, and when available, an extra sample of serum was frozen at –80°C and assays for copeptin and Prx4 were performed. Patient evaluation and treatment was initiated at the discretion of the EP in charge.
Assay for Copeptin
Measurement of copeptin was performed with a commercial sandwich immunoluminometric assay (B.R.A.H.M.S. CT-proAVP LIA, BRAHMS Biomarkers, Hennigsdorf, Germany), as described in detail elsewhere.7 Since this initial publication, the assay was modified as follows: The capture antibody was replaced by a murine monoclonal antibody directed to amino acids 137 to 144 (GPAGAL) of provasopressin. This modification improved the sensitivity of the assay. The lower detection limit was 0.4 pmol/L, and the functional assay sensitivity (<20% interassay coefficient of variation) was less than 1 pmol/L.16
Assay for Prx4
Prx4 serum levels were measured in a blinded batch analysis using a newly developed two-step sandwich immunoluminometric assay.13 The assay applies two monoclonal mouse antibodies to amino acids 39–51 of human Prx4. Briefly, 100 μL of patient samples or standards was incubated in duplicate with 100 μL of sample buffer in antibody-coated tubes for 20 hours under agitation at room temperature. Tubes were washed four times with 1 mL of washing solution, and incubated with 100 μL of acridinium ester–labeled tracer antibody for 2 hours under agitation at room temperature. After four washing steps, bound chemiluminescence was measured in a luminometer (1-second detection time per tube). Prx4 concentrations were determined as arbitrary (arb)U/L from the included standard curve. The functional assay sensitivity (interassay coefficient of variation <20%) is 0.51 arbU/L. The assay showed linearity on dilution and undisturbed recovery of added standard material and pooled samples. Stability of the analyte was given for at least 72 hours at room temperature and 4°C and for repeated freezing/thawing up to four cycles. In 272 healthy controls, the median Prx4 serum levels were 0.71 arbU/L (interquartile range [IQR] = 0.48 to 1.02 arbU/L) with slightly higher values in women and no correlation with age.13
The primary endpoint of this study was the predictive value of copeptin and Prx4 for 30-day mortality in patients with NSCs. Data on death were obtained from hospital discharge reports after 30 days or from the patients’ family physicians using questionnaires.
Descriptive statistics of central tendency (and variability) are provided for continuous variables as median (IQR) and for categorical variables as n (%). Differences were tested using Kruskal-Wallis or Wilcoxon rank-sum tests, as appropriate to the number of groups being compared.
For prediction of death within 30 days, uni- and bivariable Cox proportional hazard regression models were used. The proportional hazards assumption was tested for all variables, and all tests were nonsignificant. To evaluate the incremental added value of a new biomarker on top of other variables, a likelihood ratio [LR] χ2 test for nested models was applied.
Due to the right-skewed distribution of copeptin and Prx4, these variables were log10 transformed prior to analysis. Hazard ratios (HRs) were standardized per IQR increase. Concordance indices (C-index) were bootstrap-corrected for multivariable models to account for bias.
Finally, to illustrate the ability of copeptin and Prx4 to predict mortality, Kaplan-Meier survival curves for patients are presented, stratifying patients by copeptin and Prx4 tertiles. Time-dependent receiver operating characteristic (ROC) curves and time-dependent area under the curve (AUC) values were determined from censored survival data using the Kaplan-Meier method.17 Statistical p-values are directly reported. No adjustment of the significance level has been made to account for the multiple comparisons.
The final BANC study population consisted of 686 consecutive patients with NSCs. For 438 patients, a sufficient serum aliquot was initially believed to be available for attempted measurement of copeptin and Prx4; levels could actually be determined for 421 and 423 of these 438 samples, respectively. A lack of sufficient sample volume (14 samples for copeptin and 15 samples for Prx4) or hemolysis (three remaining samples for copeptin) prevented analysis of all 438 samples.
Baseline characteristics of the study population are presented in Table 1. The median age was 80 years (IQR = 72 to 87 years); 85.6% of subjects were 65 years and older. Almost two-thirds (65%) of the study population were female. The median CCI was 2 (IQR = 1 to 4),18 and 43.4% of the study population was dependent in at least one Katz ADL.19 The majority (97.7%) of the patients was classified to ESI 3 and therefore needed more than one external resource in the ED.
|Characteristics||All Patients (n = 438)||30-day Survivors (n = 410)||Nonsurvivors (n = 28)||p-value|
|Sex, male||154 (35)||138 (34)||16 (57)||0.0207|
|Age, yr||80 (72–87)||80 (71–87)||83.5 (76–88)||0.1515|
|Age ≥ 65 yr||375 (85.6)||348 (85)||27 (96)||—|
|Discharged||54 (12.3)||54 (12.3)||0 (0)|
|Transfer to geriatric hospital||126 (28.8)||117 (28.5)||9 (32.1)|
|Admission to tertiary care hospital||258 (58.9)||239 (58.3)||19 (67.9)|
|2||10 (2.3)||9 (2)||1 (4)|
|3||428 (97.7)||401 (98)||27 (96)|
|Comorbidities||5 (3–6)||5 (3–6)||5 (3–6)||0.7721|
|CCI||2 (1–4)||2 (1–3)||4 (2–5)||0.0005|
|Number of medications taken daily||5 (3–8)||5 (3–8)||6 (4–8)||0.4266|
|Katz ADL||6 (4–6)||6 (4–6)||4.5 (1–6)||0.0126|
|Katz Index < 6||190 (43.4)||173 (42.2)||17 (60.7)||—|
|Copeptin, pmol/L||14.8 (4.5–45.6)||13.5 (4.0–41.5)||52.6 (26.8–78.6)||0.0001|
|Prx4, arbU/L||4.5 (2.8–7.3)||4.4 (2.8–6.8)||9.9 (5.8–14.4)||<0.0001|
Among the 410 survivors, 138 (34%) were male, and 348 (85%) were 65 years and older (median age = 80 years, IQR = 71 to 87 years). Only nine patients (2%) were classified to ESI 2. A total of 173 (42.2%) of the patients were dependent in at least one Katz ADL.19 The median CCI was 2.18
Regarding the 28 nonsurvivors, 16 (57%) were male, and 27 (96%) were aged 65 years and older (median age = 83.5 years, IQR = 76 to 88 years). Only one patient (4%) was classified to ESI 2. The median CCI was 4,18 and 60.7% of the nonsurvivors were dependent in at least one Katz ADL.19
The hospitalization rate was considerable. Only 54 of the 438 patients (12.3%) could be discharged home. Of these, no patient died during follow-up. A total of 258 (58.9%) patients were admitted to our tertiary care hospital, and 126 (28.8%) were transferred to a local geriatric hospital. Twenty-five patients (5.7%) were rehospitalized within the 30-day follow-up, the majority for a new cause.
No patient was lost to follow-up. On follow-up at 30 days, 28 patients had died, yielding a 30-day survival probability of 93.6% (95% confidence interval [CI] = 90.9% to 95.5%). Causes of death are shown in Table 2. In six cases, an autopsy was performed.
|Patient ID||Cause of Death||Biomarker Levels||Autopsy|
|Copeptin (pmol/L)||Prx4 (arbU/L)|
|59||Metastatic esophageal cancer||41.7||10.8||No|
|258||Heart failure, bronchial squamous-cell carcinoma||44.8||8.43||Yes|
|295||Pneumonia, CVI during hospitalization||61.3||1.97||No|
|311||Unclear: acute renal failure, aortic aneurysm||144.9||1.5||No|
|495||Renal failure with uremia||31.1||3.87||No|
|514||Septic shock, gallbladder empyema||5.5||7.33||No|
|603||Pneumonia, renal failure||63.4||11.4||No|
|722||Heart failure, metastatic breast cancer||178.8||16.2||Yes|
|745||Not known, hospitalized for pneumonia||3.4||8.29||No|
As shown in Figure 1, copeptin and Prx4 concentrations were significantly greater in nonsurvivors than in survivors (Kruskal-Wallis test, p = 0.0001 and p < 0.0001, respectively). Table 3 shows the uni- and bivariable Cox regression analyses. In univariate models, each biomarker was used to predict mortality within 30 days after presentation. Prx4 (LR chi-square = 22.2, p < 0.00001, C-index = 0.749) showed a numerically greater LR for death than copeptin (LR chi-square = 17.0, p = 0.00004, C-index = 0.724).
|Copeptin, Prx4 (both log10)||418||26||28.22||2||<0.00001||0.783|
|Prx4 (arbU/L, log10)||423||27||22.24||1||<0.00001||0.749|
|Copeptin (pmol/L, log10)||421||26||16.98||1||0.00004||0.724|
|Disposition (discharged, transfer to geriatric hospital, or admission to tertiary care hospital)||438||28||7.61||2||0.02223||0.567|
The bivariable model including Prx4 and copeptin (LR χ2 = 28.2, p < 0.00001, C-index = 0.783) allowed a numerically and statistically greater LR than either the univariate Prx4 model (p = 0.00025) or the copeptin model (p = 0.00099). Adjusting for age or sex, both copeptin and Prx4 significantly improved prediction. For copeptin, the added chi-square on top of age was 15.5 (p < 0.0001) and on top of sex was 15.3 (p < 0.0001). For Prx4, the added chi-square on top of age was 23.5 (p < 0.0001) and on top of sex was 19.8 (p < 0.0001). Both markers were also superior to and independent of the CCI and Katz ADL: copeptin and Prx4 each significantly added to a univariate model of CCI (added χ2 for copeptin was 12.91 [p = 0.0003] and added chi-square for Prx4 was 17.7 [p < 0.0001]) or Katz ADL (added χ2 for copeptin was 13.0 [p = 0.0003] and added chi-square for Prx4 was 24.4 [p < 0.0001]).
Patient disposition was weakly predictive of outcome (p = 0.02, C-index 0.567; Table 3). Again, both copeptin and Prx4 added prognostic value on top of the disposition of patients (added χ2 for copeptin was 16.1 [p < 0.0001] and added χ2 for Prx4 was 17.1 [p < 0.0001]). Therefore, both copeptin and Prx4 are independent predictors of outcome, and neither is significantly better than the other. In univariate analysis, the standardized HR of copeptin is 3.63 (95% CI = 1.83 to 7.21) and of Prx4 is 3.24 (95% CI = 2.01 to 5.20).
To illustrate the prognostic value of copeptin and Prx4 in predicting mortality, we calculated Kaplan-Meier survival curves and divided the patients into tertiles depending on their copeptin and Prx4 concentrations (Figure 2). This simplifies the biomarker measurement, but allows demonstration of survival times. Copeptin (pmol/L) tertiles are tertile 1 [0.1; 7.2], tertile 2 [7.2; 31.0], and tertile 3 [31.0; 446.70] and Prx4 (arbU/L) tertiles are tertile 1 [0.72; 3.46], tertile 2 [3.46; 6.01], and tertile 3 [6.01; 51.30]. In case of ties at the cut-points, samples were randomly assigned to one of the adjacent groups. When copeptin concentrations at ED presentation were in the third tertile, higher mortality rates were observed, compared with the first and second tertile. The estimated survival rate for the third tertile was 86.4%, compared with 97.9 and 97.1% for the first and second tertiles, respectively (Figure 2). Separation for Prx4 is comparable to that for copeptin. The estimated survival rate for the third tertile was 85.8%, compared with 97.2% and 97.9% for the first and second tertiles, respectively.
Figure 3 illustrates the predictive performance of copeptin and Prx4 via time-dependent ROC curves, where survival at 30 days is plotted. This simplifies observed survival times, but allows us to illustrate the biomarker as a continuous measurement. The 30-day survival AUC for copeptin is 0.73 (95% CI = 0.64 to 0.82) and for Prx4 is 0.76 (95% CI = 0.65 to 0.86), whereas the AUC for both markers combined is 0.80 (uncorrected for multiple variables, 95% CI = 0.72 to 0.88). Using the third tertile again as cutoff (31 pmol/L for copeptin and 6 arbU/L for Prx4) results in a sensitivity and specificity of 73 and 69% for copeptin and 74 and 69% for Prx4.
In this prospective, observational study, we found that copeptin and Prx4 are novel, strong, and independent prognostic markers for death in patients presenting to the ED with NSCs.
Copeptin, as a surrogate marker for the activity of the HPA axis is a marker of the individual stress level.8 The HPA axis is an important part of the neuroendocrine system responding to a stressor that disrupts the homeostatic balance.20 The bidirectional communication between the immune system and the HPA axis might explain the multitude of potential stressors causing mortality in our cohort, such as infection, cancer, or heart failure, triggering vasopressin release.21 This could form the basis for copeptin’s usefulness as a prognostic biomarker, especially under circumstances where a final diagnosis cannot be made in a timely manner.
Copeptin is already known to have prognostic value in a variety of diseases, such as hemorrhagic and septic shock,22 acute heart failure,23,24 and acute myocardial infarction.25 Furthermore, copeptin is useful in risk-stratifying patients with lower respiratory tract infections26 and acute exacerbation of chronic obstructive pulmonary disease requiring hospitalization.27 Copeptin is also an independent predictor of functional outcome and mortality in stroke28 and acute intracerebral hemorrhage.29
Considering that copeptin may not reflect all mechanisms involved in stress, we decided to simultaneously assess oxidative stress by the novel biomarker Prx4. A recent study found elevated Prx4 concentrations in serum of patients with sepsis compared to healthy controls.13 The authors speculate that an elevated Prx4 serum concentration may be a compensatory antioxidant reaction, because reactive oxygen species are excessively produced by immune cells during systemic inflammation in sepsis.30 In another cohort of intensive care unit patients with sepsis, severe sepsis, and septic shock, median Prx4 serum levels in nonsurviving patients exceeded those of survivors twofold.14 Median Prx4 serum levels gradually increased with disease severity ranging from systemic immune response syndrome to septic shock. Despite clear differences in patient composition and mortality rate, a similar extent of Prx4 increase was observed for nonsurvivors in the present patient cohort. In our heterogeneous patient population, Prx4 levels were found to be strong predictors of 30-day survival.
Other Prx isoforms have been identified in the circulation. Prx1, for example, was detected in serum from non–small cell lung cancer patients,31 and Prx1 and Prx2 were found in the plasma of nontreated HIV-infected patients.32 Moreover, Prx2 was shown to be increased in the plasma proteome of patients with severe acute respiratory syndrome, patients with fever, gastric cancer, or pneumonia, compared to healthy controls.33 However, the mechanism of Prx4 release into the circulation has not yet been characterized. An N-terminal hydrophobic signal sequence denotes Prx4 as a secretory enzyme.34 Okado-Matsumoto et al.35 localized Prx4 bound to the membrane of endothelial cells and suggest a release of the active enzyme in response to the surrounding oxidative stress. According to this hypothesis, the degree of oxidative stress is proportional to the Prx4 released from the cell surface, resulting in elevated Prx4 serum levels to protect cells from oxidative damage in the extracellular space. Besides this hypothesis, it is conceivable that intracellular Prx4 leaks from cells due to tissue damage, caused by oxidative stress. Patients presenting to the ED with NSCs may suffer from more or less severe oxidative stress and subsequent cell damage, depending on the underlying condition. Patients with a higher degree of oxidative imbalance could therefore be prone to adverse outcome. Consequently, by reflecting the level of oxidative stress, Prx4 might identify those patients at higher risk and at higher need for timely decisions on treatment strategies.
Due to the nonspecific presentation of our study population, both the final diagnoses at an early time point and the disease severity cannot be assessed easily. This is in part due to the fact that elderly patients often fail to show specific clinical features, such as cough and fever in pneumonia, indicative of serious disease.5,36–38
Disease severity has an effect not only on mortality, but also on the consumption of health care resources, for example, the need for intensive care admission, disposition to a geriatric ward, the suitability for patient transfer to step down community hospitals, or discharge home. In the ED, where these disposition decisions are made, novel biomarkers could be valuable tools to help predict which patients are at increased risk for mortality and thus which patients are more likely to benefit from more extensive workups and higher levels of inpatient care (e.g., ICU placement), beyond the information available from the patient history and clinical examination. “Red flags” to alert the clinician of an increased likelihood of adverse outcomes are needed in a population that is at risk for undertriage39 and has a high rate of missed diagnoses.5 By facilitating the identification of the more severely ill patients at risk of high mortality rates, novel biomarkers, such as copeptin and Prx4, could not only help to match the right patients into intensive care unit beds, but also assist in ameliorating overcrowding. This has been shown to be associated with increased mortality,40 by helping to discern which patients are unlikely to need extensive diagnostic workups of NSCs, and therefore unlikely to need inpatient care (possibly within an intensive care unit).
Early risk stratification during initial evaluation in the ED by determining levels of copeptin and Prx4 therefore might improve patient management, disposition, and resource allocation. However, the present findings warrant further prospective validation in future intervention studies to prove clinical usefulness and cost efficiency.
Our cohort was a select population of patients who were predominantly of European ancestry, presenting to a single urban tertiary care center in Switzerland. The lack of an external validation sample limits the generalizability of our results to EDs in other areas or countries. Due to the fact that not all patients of the study population had a blood sample available, a potential selection bias could be present. However, the data were obtained in a large sample of consecutive ED patients, minimizing the possibility for this bias. In addition, the patients tested did not differ in means of age, sex, comorbidities, drug intake, or mortality rate from the total population presenting with NSCs.
An additional limitation is that the population is primarily elderly, and the mortality occurred nearly exclusively in the elderly. Therefore, extrapolation to the younger population may not be valid. Due to the limited number of events, we were able to demonstrate the independence of the predictive information from known clinical variables of both copeptin and Prx4 in bivariable models only. Therefore we cannot guarantee whether this remains true in multivariable models combining all clinical variables.
Copeptin and Prx4 are new prognostic markers in a very heterogeneous group of patients presenting to the ED with nonspecific complaints. Copeptin and Prx4 might be valuable tools for risk stratification and decision-making in this patient group and they deserve further study.
We are grateful to the emergency nurses and physicians and patients who participated in the study. We are indebted to Mirjam Christ-Crain, Beat Müller, Claudine Blum, and Isabelle Suter for helpful discussions and Fausta Chiaverio from the central laboratory of the University Hospital Basel for technical assistance. This report will be part of a dissertation work (JSch) conducted at the Charité Universitätsmedizin Berlin.