Portal de Periódicos da CAPES

Stress Markers Predict Mortality in Patients With Nonspecific Complaints Presenting to the Emergency Department and May Be a Useful Risk Stratification Tool to Support Disposition Planning

2013; Wiley; Volume: 20; Issue: 7 Linguagem: Inglês

10.1111/acem.12172

1553-2712

Christian H. Nickel, Anna S. Messmer, Nicolas Geigy, Franziska Misch, Beat Müeller, Frank Dusemund, Sabine Hertel, Oliver Hartmann, Sven Giersdorf, Roland Bingisser,

Heart Failure Treatment and Management

To the authors' knowledge, no prospectively validated, biomarker-based risk stratification tools exist for elderly patients presenting to the emergency department (ED) with nonspecific complaints (NSCs), such as generalized weakness, despite the fact that an acute serious disease often underlies nonspecific disease presentation. The primary purpose for this study was to validate the retrospectively derived model for outcome prediction using copeptin and peroxiredoxin 4 (Prx4), in a different group of patients, in a prospective fashion, in a multicenter setting. The secondary goals were to evaluate the potential contribution of the midregional portion of the precursor of adrenomedullin (MR-proADM) for outcome prediction and to investigate whether disposition decisions show promise for potential improvement by using biomarker levels in addition to a clinical assessment. The Basel Nonspecific Complaints (BANC) study is a delayed-type cross-sectional diagnostic study, carried out in three EDs in Switzerland, with a prospective 30-day follow-up. Patients presenting to the ED with NSCs, as defined previously, were included if their vital signs were within predefined limits. Measurement of biomarkers was performed in serum samples with sandwich immunoluminometric assays. To examine the disposition process, the final disposition was compared with a combination of the first clinical disposition decision and the risk assessment, which included the biomarker MR-proADM in a retrospective simulation. Patients were divided into three groups according to MR-proADM concentration, defining three risk classes with three disposition possibilities (admission to tertiary care, transfer to geriatric hospital, discharge). Thirty-three 30-day nonsurvivors were observed from among 504 study patients with NSCs. Biomarker levels were significantly greater in nonsurvivors than survivors (p < 0.0001 for all three biomarkers). Univariate Cox models reveal a C-index of 0.732 for MR-proADM, 0.719 for Prx4, and 0.723 for copeptin. The incremental added value for chi-square obtained via multivariate modeling showed that models inclusive of MR-proADM, copeptin, or Prx4 are superior to and independent of models limited to sex and age. The incrementally added chi-square for MR-proADM, beyond the chi-square of a base model consisting of age and sex, was 29.79 (p < 0.00001). In a multimarker approach, only Prx4 provided additional information to MR-proADM alone (C-index = 0.77). Applying an algorithm combining physicians' first clinical assessment plus biomarker information to derive a modified risk assessment, reassignment would lead to a potential decrease of 48 admissions to acute care, seven additional transfers to geriatric care, and 41 additional discharges (negative likelihood ratio [–LR] = 0.13). Analysis of 30-day mortality reveals that our algorithm is not inferior in terms of safety. In this study the authors confirm that these new stress biomarkers permit reliable prognostication of adverse outcomes in a heterogeneous group of patients with NSCs. A simulation showed that this prognostic information could be useful to enhance the appropriateness of disposition decisions of ED patients with NSC. The use of biomarkers for risk stratification in this patient group should be evaluated with prospective intervention studies. Según nuestro conocimiento, no existen herramientas basadas en biomarcadores, prospectivamente validadas para estratificar el riesgo de los pacientes ancianos que acuden al SU con sintomatología no específica (SNE), como por ejemplo debilidad generalizada, a pesar del hecho que una enfermedad grave aguda a menudo subyace en una presentación inespecífica. El objetivo principal de este estudio fue validar el modelo obtenido de forma retrospectiva para la predicción de resultados usando la copeptina y la peroxiredoxina 4 (Prx4), en un grupo diferente de pacientes, con un diseño prospectivo y en un escenario multicéntrico. El objetivo secundario fue evaluar la contribución potencial del precursor de la adrenomedulina (MR-ProADM) para la predicción de resultados, sí el uso de biomarcadores, en conjunción con la valoración clínica, permite mejorar las decisiones de ubicación del paciente tras su atención en el SU. El estudio Basel Nonspecific Complaints (BANC) es un estudio diagnóstico transversal tipo diferido, llevado a cabo en tres SU en Suiza, con un seguimiento prospectivo a 30 días. Los pacientes que acudieron al SU con SNE, según la definición previa, se incluyeron si sus constantes vitales estaban entre los límites predefinidos. Se realizó la medición de biomarcadores en muestras plasmáticas con una prueba inmunoluminométrica. Para examinar el proceso de ubicación, la ubicación final se comparó con una combinación de la primera decisión de ubicación clínica y el riesgo de valoración, que incluyó el biomarcador MR-proADM en una simulación retrospectiva. Los pacientes se dividieron en tres grupos según la concentración de MR-proADM, definiendo tres clases de riesgo con tres posibilidades de ubicación (ingreso en hospital terciario, traslado al hospital geriátrico, alta). Se documentaron 33 muertos a los 30 días entre los 504 pacientes con SNE. Los niveles de biomarcadores fueron significativamente más altos en los no supervivientes que en los supervivientes (p < 0,0001 para los tres biomarcadores). Los modelos univariables de Cox revelan unos índices C de 0,732 para el MR-proADM, de 0,719 para el Prx4 y 0,723 para la copeptina. El valor incremental añadido para χ² obtenido según el modelo multivariable mostró que los modelos que incluyen MR-proADM, copeptina o Prx4, son superiores a, e independientes de, los modelos limitados al sexo y la edad. El χ² incremental añadido para MR-proADM, por encima de la χ² de un modelo que consiste en edad y sexo, era 29,79 (p < 0,00001). En una aproximación con multimarcadores, solo Prx4 proporcionó una información adicional al MR-proADM aislado (índice C 0,77). Si se aplicara un algoritmo que combinase la primera valoración clínica del urgenciólogo junto con la información del biomarcador para obtener una valoración del riesgo modificada, la reasignación permitiría un descenso potencial de 48 ingresos en unidad de agudos, 7 traslados adicionales a centro geriátricos y 41 altas adicionales (razón de verosimilitud negativa 0,13). Un análisis de mortalidad a 30 días revela que el algoritmo no es inferior en términos de seguridad. En este estudio se confirma que estos nuevos biomarcadores de estrés permiten un pronóstico fiable de resultados adversos en un grupo heterogéneo de pacientes con SNE. Una simulación mostró que esta información pronóstica podría ser de utilidad para mejorar la pertinencia de las decisiones de ubicación de pacientes del SU con SNE. El uso de biomarcadores para la estratificacion del riesgo en este grupo de pacientes debería ser evaluada con estudios de intervención prospectivos. Elderly patients presenting to the emergency department (ED) are at risk of acquired disability, admission to residential care, death, and larger health care expenditures.1-3 They have special care needs with potentially coexistent medical, functional, psychological, and social problems.4, 5 This can lead to nonspecific disease presentation requiring a different approach to care. To our knowledge, to date no biomarker-based risk stratification tools exist for elderly patients with nonspecific complaints (NSCs) such as functional decline or generalized weakness, despite the fact that nonspecific disease presentation may well be due to acute serious disease.6, 7 In patients with NSCs, the evaluation is complex, which can lead to undertriage, longer diagnostic work-ups, and adverse health outcomes.8, 9 The differential diagnosis of NSCs is extremely broad, and acute morbidity is underlying in almost 60% of cases.10, 11 Alternatively, unnecessary hospitalizations of elderly patients to acute care should be avoided, because hospitalization has been shown to carry significant risk for this population.12, 13 Therefore, a timely and safe disposition (i.e., hospital admission, transfer to a geriatric hospital, or discharge to home) of patients with NSC is crucial, given the rapid workflow of the ED. In a retrospective monocenter study we have recently shown that two novel stress biomarkers, copeptin and peroxiredoxin 4 (Prx4), might be helpful to support risk stratification in patients with NSC by predicting 30-day mortality irrespective of the final diagnosis, suggesting that they are nonspecific markers for disease severity.14 The biomarker performance of both copeptin and Prx4 in patients with NSC is encouraging, given the high heterogeneity and potential confounders, such as comorbidities and medication.14 Another new and promising biomarker, adrenomedullin (ADM), is a member of the calcitonin peptide family and is widely expressed.15, 16 It is up-regulated during severe infections and has been shown to have multiple effects, including immune modulation, vasodilation, stimulation of diuresis, increase of cardiac output, and antimicrobial activity.17, 18 The midregional portion of the precursor of ADM, proadrenomedullin (MR-proADM), is a more stable fragment that is easy to measure and stoichiometrically reflects levels of ADM.19 MR-proADM appears to be a prognostic marker in sepsis and lower respiratory tract infections,20-22 and its plasma concentrations have been shown to be elevated in myocardial infarction, as well as to correlate with severity of congestive heart failure.23, 24 Furthermore, in a symptom-based study of patients presenting to the ED with dyspnea, it was shown that MR-proADM was a predictor for 30-day mortality, regardless of the underlying diagnosis.25 Plasma concentration of mature ADM is also increased in various types of glomerulonephritis and progressively increases in patients with chronic renal failure.26, 27 The aim of this study was to validate, in a multicenter setting, which biomarkers might be helpful for risk stratification in patients with NSCs by predicting 30-day mortality. Simultaneously, our secondary objective was to prospectively examine the value of a new biomarker, MR-proADM, as an alternative predictor of 30-day mortality in patients with NSCs, and compare the biomarkers` predictive performance. Finally, we aimed to investigate how the information derived from MR-proADM could be applied to an individual patient encounter by retrospectively studying whether disposition planning could be improved by biomarker level assessment. The third Basel Nonspecific Complaints (BANC III) study is a delayed-type cross-sectional diagnostic study with a prospective 30-day follow-up. The study protocol was approved by the local ethics committees of all participating sites, and is registered with ClinicalTrials.gov (NCT00920491). It is in compliance with the Helsinki Declaration. Written informed consent was obtained from each participating patient. The BANC III study was performed between May 27, 2009, and February 8, 2011, at three ED sites with annual censuses ranging from 12,000 to 45,000 visits. The study was carried out at the EDs of the University Hospital Basel, Switzerland (an urban, tertiary-care hospital); the Kantonsspital Liestal (a regional hospital); and the Kantonsspital Aarau (a tertiary care center). In the ED of the University Hospital Basel, all nontrauma patients presenting to the ED with Emergency Severity Index (ESI) scores of 2 or 3 were consecutively screened for inclusion. Elderly patients account for about 20% of all ED visits in this institution, similar to the other centers. As previously shown, admission rates are 28% for the total ED population and 55% for the elderly population.8 There are three geriatric hospitals, two hospices for palliative care, and 39 nursing homes with almost 2,900 nursing beds in the Basel community. Ambulatory nursing support is provided for 6,700 patients. In Aarau and Liestal, the enrollment was linked to presence of the study personnel at the study site on weekdays from 8 a.m. until 6 p.m. The ESI was used to exclude all patients in need of life-saving interventions (ESI level 1), as well as patients in whom full work-ups were not intended (ESI level 4) or patients who qualified for the "see-and-treat" pathway (ESI level 5).28 Many of our elderly ED patients are admitted to the university department of acute geriatrics specializing in the care of acutely ill older patients after evaluation in the ED. About 50% of elderly patients seen in our ED are admitted to tertiary care.8 Additionally, three geriatric hospitals in the urban region and smaller local hospitals admit older patients after initial ED evaluation for further treatment. They are facilities specialized in geriatric care, in which the patients receive medical care and rehabilitation to regain abilities for activities of daily living, so that they can eventually be discharged to home. From there, patients can also be transferred to long-term nursing facilities if needed. Decisions on disposition (i.e., admission, discharge, transfer to local geriatric hospital) are documented in the electronic health record. The first disposition plan is made after initial patient work-up, including diagnostics ("intention to transfer" [ITT]) in the acute assessment unit. Due to regular exit block (no inpatient beds available for transfer to geriatric hospitals; only during the daytime) and the availability of a decision unit for up to 24-hour stays, the final disposition plan ("effective transfer" [ET]) is made often after an overnight stay on the decision unit when inpatient beds, or beds in the three geriatric hospitals, become available. In the decision unit, typically no further diagnostic work-up is performed. However, disposition decisions are influenced by observation of patients. Patients were eligible if they were age 18 or older and if they gave informed consent for the study. Patients were included if they presented to the ED with NSCs such as weakness or "functional impairment" as previously described.10 Exclusion criteria were specific complaints (e.g., dyspnea), patients with clinical features suggestive of a working diagnosis (e.g., jaundice), persistent signs of shock or vital parameters out of the normal range (blood pressure < 80 mm Hg, respiration rate >20 breaths/min, tympanic body temperature > 38.5°C, SaO2 < 92%), patients referred from other hospitals, and palliative patients. Data collection was performed as previously described.14 In brief, standardized data collection forms were used to record demographic data, patients' complaints, comorbidities as assessed by the Charlson Comorbidity Index (CCI), the patients' prescribed medications, and physical examination information. Furthermore, in each individual, the Katz Index of Independence in Activities of Daily Living (Katz ADL) was recorded.29 A score of 6 indicates a highly independent patient, and a score of 0 indicates a very dependent patient. Outcome data after 30 days were obtained from the patients' primary care physicians using questionnaires and hospital discharge reports for all hospitalized patients. Information on discharge, transfer to other care facilities, and rehospitalizations were collected for all study patients and patient reports of the respective hospitalization were obtained. Blood testing was performed in all study patients. A blood specimen was frozen at –80°C for the estimation of biomarker levels. Imaging and other diagnostic tests were not required and ordering was determined by the treating emergency physicians (EPs). MR-proADM was measured using an automated sandwich chemiluminescence immunoassay on the KRYPTOR system (Thermo Scientific Biomarkers, Hennigsdorf, Germany), described elsewhere30-32; EDTA-plasma for the MR-proADM estimation was obtained by centrifuging at 3000 × g for 10 to 15 minutes. The BRAHMS MR-proADM KRYPTOR has a detection range of 0.05 to 100 nmol/L and a functional assay sensitivity of 0.25 nmol/L. Copeptin and Prx4 were measured as described previously.14 The laboratory personnel performing the assay were blinded to the patients' medical information and final diagnoses to exclude observer bias. The primary endpoint of this study was the predictive value of the biomarkers MR-proADM, copeptin, and Prx4 for all-cause 30-day mortality in patients with NSCs. For the secondary endpoint we used a simulation to retrospectively assess whether the disposition process of patients with NSC could potentially be improved by using the biomarker information. A simulation was performed to assess whether the disposition process of patients with NSC could be optimized by using the biomarker information for MR-proADM exemplarily. Patients were treated by the EPs in charge, who determined further management without the interference of the study team. After clinical assessment, patients are assigned to one of three risk classes on the basis of the preliminary disposition plan mentioned above (ITT). If acute morbidity requiring inpatient treatment is identified, admission to an inpatient ward is planned ("A" for "acute," high-risk). Patients who do not need tertiary care, but do need inpatient treatment, are transferred to a community geriatric hospital (labeled "G" for "geriatric," intermediate-risk). If patients do not need further inpatient treatment or observation, they are discharged for outpatient treatment (labeled "D" for discharge, low-risk). Because the biomarker levels were determined at a later stage, the treating physicians had no access to results of biomarker measurements. For the biomarkers, three risk classes were specified by two cutoffs. First, we evaluated a series of MR-proADM cutoff pairs starting with tertiles (0.84; 1.32 nmol/L) to find the ideal one that would keep as many events (death) as possible in the highest-risk group, which was intended for tertiary inpatient care and, on the other hand, to assign as many patients as possible to a lower level of care. We chose the cutoffs 0.75 and 1.5 nmol/L for MR-proADM for the simulation, which are not far from tertiles, because they meaningfully separated patients as low-, intermediate-, or high-risk. These cutoffs have been proposed previously for patients with lower respiratory tract infections.33 The simulated risk assessment combining clinical assessment with biomarker information was based primarily on the clinical assessment (see Figure 1 for details). The result of the biomarker was used retrospectively and as additional information to optimize the disposition decision for the individual patient. To examine whether it is of any advantage to use the biomarker in the disposition decision process, the final disposition (ET) was compared with the combination of the preliminary disposition (ITT) and the information of the biomarker MR-proADM in a simulation. Both the final disposition and the combined risk assessment were separated by survivors and nonsurvivors. Percentage survivors moving down in the disposition categories from the highest-risk class (A) to the intermediate-risk class (G) or to the low-risk class (D) were compared with the percentage of patients moving up. The percentage of nonsurvivors moving to a lower-risk class was compared with the percentage of nonsurvivors moving to a higher-risk class. Patient characteristics and biomarkers are expressed as medians and interquartile ranges (IQR) or counts and percentages where appropriate. Group comparisons of continuous variables were performed using the Wilcoxon rank sum test. Categorical data were compared using Fisher's exact test or the chi-square test, where appropriate. Due to the right-skewed distribution of the biomarkers, these variables were log10-transformed prior to analysis. To test if biomarkers are useful for prediction of death within 30 days, univariate and multivariate Cox proportional hazard regression models were used. Sex, age, CCI, and the Katz ADL were considered in addition to the biomarker. The proportional hazards assumption was tested for all variables, and all tests were nonsignificant. The predictive value of each model was assessed by the model likelihood ratio (LR) chi-square statistic. The concordance index (C-index) was given as an effect measure. It was equivalent to the concept of area under the curve adopted for binary outcome. The C-indices were bootstrap-corrected for multivariable models to account for bias. The added discriminative power offered by the addition of biomarkers to clinical variables was analyzed using the LR chi-square test for nested models. For illustrative purposes, Kaplan-Meier survival curves for patients are presented. Statistical p-values are directly reported. No adjustment for "significance" had been made to account for the multiple comparisons. Sensitivity and specificity were calculated for single biomarkers, ITT, and ET, as well as for multivariate models. For biomarkers, different cutoffs including the optimal cutoff were used. For disposition variables, we analyzed three risk groups and therefore two cutoffs. For the calculation of the sensitivity and specificity at the upper cutoff level, we merged the lower two risk groups. For calculations at the lower cutoff, the two higher-risk groups were merged, respectively. Also, for multivariate models, optimal cutoffs were determined from model scores. Further, to provide more detail on sensitivity and specificity, positive LR (+LR) and negative LR (–LR) showing the strength of association with a positive or negative test result and outcome were calculated. The statistical analyses were prespecified in a statistical analysis plan prior to constitution of the database and performed using R version 2.5.1 (http://www.r-project.org, library Design, Hmisc, ROCR). The study population consisted of 544 patients. Fifteen patients were excluded by the outcome assessors because of protocol violations, and in 25 patients an aliquot for determination of biomarker levels was not available, so that 504 patients remained for final analysis. Of these 504 patients, 497 MR-proADM levels, 495 copeptin levels, and 493 Prx4 levels could be determined. Analysis was impossible due to a lack of sufficient sample volume for six samples for MR-proADM, six samples for copeptin, and 11 samples for Prx4 and due to hemolysis for one sample for MR-proADM and four samples for copeptin. Baseline characteristics of the population are presented in Table 1. The median age of the population was 82 years (IQR = 75 to 87 years); 38.9% were men. The median CCI was 2 (IQR = 0 to 3). At 30-day follow-up, there were 471 survivors and 33 nonsurvivors. The median age did not differ between survivors and nonsurvivors, but we observed more male patients among nonsurvivors (57.6%) than among survivors (37.6%). Also, the median CCI was higher in nonsurvivors than in survivors (3 vs. 2). The Katz ADL index was lower in nonsurvivors than in survivors (3 vs. 5). Emergency department referral occurred exclusively from the community, mostly by ambulance (252 cases), but also from primary care providers (124 cases); self-referral (64 cases); or family, neighbors, and others (64 cases). Most patients (96%) lived at home before study inclusion. Twenty patients (4%) lived in nursing homes. Hospitalization rates were 86.5%. Of 504 patients, most of the study patients were admitted to tertiary care hospitals (292, 57.9%). A total of 145 patients were transferred to geriatric hospitals (28.8%), and 67 were discharged home (13.3%). At 30-day follow-up, 57 (11%) were in acute care, 233 (46%) were in geriatric hospitals, and 214 (42%) were outpatients. Of all outpatients, 40 were living at home independently, whereas before inclusion 202 lived independently. Eighteen patients (3.6%) were rehospitalized within the 30-day follow-up, mainly for new causes. Of the 33 nonsurvivors, 29 were patients admitted to tertiary care hospitals. Four were transferred and died at geriatric hospitals. Nonsurvivors were assigned to an inpatient ward or ICU by clinical decision only with a specificity of 44%, whereas survivors were assigned to the group of patients to be discharged home with a sensitivity of 100%. For detailed calculations of sensitivity, specificity, and likelihood ratios, see the Data Supplement S1 (available as supporting information in the online version of this paper). The 30-day survival probability was 93.5% (95% confidence interval [CI] = 90.9% to 95.3%). Biomarker levels were significantly higher in non-survivors than in survivors (p < 0.0001 for all three biomarkers; Table 1). Univariate Cox models show that MR-proADM has a C-index of 0.732, slightly higher than the other two biomarkers, indicating the best numerical predictive performance. Based on the results of C-index and added value, we used MR-proADM for the following evaluations. A time-dependent ROC plot for 30-day survival is presented in Data Supplement S2 (available as supporting information in the online version of this paper). Table 2 shows univariate and multivariate Cox regression results and C-index of MR-proADM and other variables. Data for Prx4 and Copeptin are provided in Data Supplement S3 (available as supporting information in the online version of this paper). The added chi-square for nested models show that MR-proADM and the other studied biomarkers are superior and independent of sex and age. The added chi-square of MR-proADM on top of a base model consisting of age and sex was 29.79 (p < 0.00001). All biomarkers were also independent and superior to Katz ADL; the added chi-square for MR-proADM was 25.94 (p < 0.00001). Further, all biomarkers were also independent and superior of the CCI; here the added chi-square for MR-proADM was 23.78 (p < 0.00001). In a multimarker approach, out of all possible combinations, only Prx4 provided additional information to MR-proADM, and the C-index of a model including age, sex, MR-proADM, and Prx4 increases to 0.77 (Table 2). To demonstrate the prognostic value of MR-proADM to predict mortality, we calculated Kaplan-Meier survival curves. Patients were divided into three groups according to the MR-proADM concentration, with cutoffs of 0.75 and 1.5 nmol/L, respectively (Figure 2). A clear decrease of survival with higher MR-proADM concentrations was observed. In the MR-proADM group with the lowest values, 98.4% (95% CI = 93.8% to 99.6%) survived after 30 days. In the group with intermediate MR-proADM levels, the survival rate was 94.6% (95% CI = 90.9% to 96.8%), and in the group with highest MR-proADM values, survival rate decreased to 86.4% (95% CI = 79.2% to 91.2%). At the cutoff 1.5 nmol/L, merging the two lower-risk groups, nonsurvivors were assigned to an inpatient ward/ICU with a specificity of 75%, whereas at the lower cutoff 0.75 nmol/L, merging the two higher-risk groups, survivors were assigned to the group of patients to be discharged home with a sensitivity of 94% (for details see Data Supplement S1). Applying the algorithm combining clinical and biomarker risk assessment to the cohort with MR-proADM values available (n = 497), patient disposition would have changed in our simulation as shown in Figure 3. A total of 169 patients (34%) would be affected by the reassignment with 49 patients moving to a higher-risk class and 120 patients to a lower-risk class. The effect of the simulation can be summarized as a potential decrease of 48 hospitalizations to acute care, seven additional geriatric hospitalizations, and 41 additional discharged patients. Comparing the number of events stratified for risk classes in the clinical disposition decision (ET) with the simulated biomarker-aided disposition (ITT + MR-proADM), three events move from the highest-risk class to the intermediate-risk class. Applying the combined algorithm, one patient would be discharged. However, in this case, the initial disposition decision of the clinician was to admit the patient. This was overruled by the patient and his family, and therefore ITT was changed to discharge. The patient had a MR-proADM level of 1.04 nmol/L (indicating intermediate risk), which would not have had any effect on the risk stratification according to the proposed algorithm. Ultimately, after observation, this patient was admitted. The effect of a biomarker-supported risk assessment for patient disposition, in comparison to the clinical-only risk assessment, can be summarized that 9% fewer patients would be admitted to inpatient wards or ICUs, with an increase in mortality of 0.7% in this group; 2% more patients would be referred to intermediate care facilities, with a mild to moderate increase in mortality of 1.9%; and 7% more patients would be discharged safely, with a decrease in mortality of 0.6%. In summary, the proposed algorithm (ITT + MR-proADM) assigns survivors with a sensitivity of 97% to the group of patients to be discharged home and nonsurvivors with a specificity of 53% to an inpatient ward or ICU (for details see Data Supplement S1). In this prospective study, the primary objective was met. We were able to prospectively validate, in a multicenter setting with a new group of patients, our prior finding that copeptin and Prx4 are useful prognostic markers for elderly patients with NSCs. In addition, the secondary objective of evaluating the utility of MR-proADM showed that this additional marker has efficacy in predicting outcomes for patients with NSCs. In fact, it appears that the addition of MR-proADM to age and sex numerically outperforms the other two biomarkers. In a simulation we found that the prognostic information provided by the biomarker could potentially be useful for disposition decisions in patients with NSC. These new findings require prospective validation before incorporation into medical practice. The management of patients with NSCs is challenging, a fact well recognized for a long time.34 Leaders in both emergency medicine and geriatric medicine have issued a call for further research in this area.35 Furthermore, geriatric experts believe that generating an age-specific differential diagnosis for patients presenting to the ED with general weakness or dizziness should be among the competencies of residents taking care of elderly patients.36 New tools are needed to support clinical decision-making. The application of biomarkers appears attractive in this patient population, where efficient work-up strategies and risk stratification tools are lacking. Adrenomedullin is an almost ubiquitous peptide. Its production is up-regulated by oxidative stress, proinflammatory cytokines such as TNF-α and IL-1, angiotensin II, and endothelin-1.37-39 Initially the effects of ADM were considered to be primarily vasodilatory, diuretic, and natriuretic.15, 40, 41 However, this peptide appears to be a multifunctional mediator with effects on normal and malignant growth, inflammation, and immunity.41, 42 ADM has an important role in the pathophysiology of a wide spectrum of clinical conditions, such as lower respiratory tract infections, septic shock, arterial hypertension, acute coronary syndrome, heart failure, and renal failure. Elevated plasma ADM levels may provide important prognostic information in these conditions. Similarly, elevated Prx4 levels, reflecting oxidative stress, have been shown to be correlated with endogenous antioxidants and markers of inflammation in sepsis patients and are associated with elevated morbidity and mortality in this patient group.43 Other studies have demonstrated that the expression or plasma levels of Prx4 change under various pathological conditions such as cardiovascular disease, infection, cancer, and diabetes.43-46 It has been speculated that Prx4 secretion into the circulation could lead to a loss of intracellular or membrane-bound Prx4, resulting in diminished capacity to reduce cellular oxidative stress, which in turn could lead to an increased probability of adverse outcome and death.47 Copeptin, the C-terminal part of the arginine vasopressin precursor peptide, has been shown to be a surrogate marker for the activity of the hypothalamic–pituitary–adrenal axis. It is released in response to a stressor that disrupts the homeostatic balance and is a marker of the individual stress level.48 It has been shown to be a predictor of mortality in a wide variety of clinical conditions such as acute exacerbated chronic obstructive pulmonary disease, lower respiratory tract infection, hemorrhagic and septic shock, acute ischemic stroke, and heart failure (for review see Nickel et al.49). Considering the heterogeneity of underlying conditions in patients with NSCs,11, 14 the predictive power of stress biomarkers may be helpful to improve the disposition process in the following typical situations: 1) safe immediate discharge (–LR = 0.13), if clinical risk assessment is combined with MR-proADM; and 2) immediate disposition to acute tertiary care (specificity of 75.2% for mortality at higher cutoff of MR-proADM). For the first situation, our simulation showed that an additional 41 patients could have been safely discharged. This accounts for a potential economic effect of biomarker-aided discharge planning. For the second situation, our simulation showed that 48 fewer patients would have been admitted to tertiary care, accounting for further sparing of resources. In reality, when comparing the preliminary (ITT) to the final decision for disposition (ET), we noted a tendency to admit more patients to the highest level of medical care after observation. The tendency to admit more patients than those at risk for adverse outcomes may be attributed to a well-funded health care system. Our algorithm may therefore assist in disposition decisions, potentially reducing admissions and increasing safe discharge. Furthermore, stress biomarkers could play a role as a "red flag," identifying patients at high risk for mortality. Therefore, inappropriate referral to a lower level of care or even discharge could be avoided using biomarker-aided decision-making. Finally, applying the proposed algorithm has the potential to reduce time spent in the ED, as it is based on the first disposition decision (ITT). A final disposition decision could be made in a more timely fashion using biomarkers, instead of observation, if prospective validation supports our retrospectively derived findings. This would allow limited resources to be allocated to patients at risk of adverse outcome. Analysis of 30-day mortality reveals that our algorithm is not inferior in terms of safety. Future intervention studies are needed to clarify whether this derived algorithm leads to practical improvements in risk stratification of patients with NSCs. The disposition process (i.e., the delay from ITT to ET) of our patients may be unique to our setting. In most hospital EDs that do not have observation units, patients are admitted and sent relatively quickly to inpatient units. The secondary phase of observation, characteristic of the main site hospital, might not exist in these places. However, this setting allows us to study the disposition process of patients with NSC in detail. We noted a tendency to admit more patients to a higher level of medical care, most likely not to discharge patients at risk of adverse health outcomes. However, no other variables were collected systematically that might have determined why this occurred. We are aware that there are many reasons why patients are admitted to the hospital, including family expectations, patients' underlying comorbidities, physicians' motivations, and fear of liability. In our retrospective simulation, these variables were not incorporated into the analysis. In addition, no geriatric assessment took place in our ED. The 40 patients who were considered low-risk (i.e., discharge to home) in the simulation may well have needed nursing support due to their social situations or due to functional decline. However, in our setting ambulatory nursing support is readily available and might therefore be an option for this low-risk group of patients. The C-statistics reported in our study indicate that the accuracy of the biomarkers (copeptin, Prx4, and MR-proADM) in predicting 30-day mortality is moderate. Combining these biomarkers did not significantly improve the accuracy. However, the broad spectrum of underlying conditions in patients with NSCs should be taken into account when evaluating accuracy.11 Our study mostly included white patients from northwestern Switzerland. It therefore lacks generalizability to other ethnic populations. One of the study centers was not able to provide consecutive enrollment. Our methods did not assess for all missed eligible patients where a stronger assessment of enrollment bias could be determined. In this prospective study we confirm in a multicenter setting that new stress markers are reliable prognosticators in the heterogeneous group of patients with nonspecific complaints. Within this group of quite similarly performing biomarkers, MR-proADM revealed a slightly better performance predicting 30-day mortality. In a simulation we found that the prognostic information provided by MR-proADM may be useful for disposition decisions of ED patients with nonspecific chief complaints. Obviously, the biomarker information must always be interpreted in the context of a careful clinical assessment and the individual patient's situation. Therefore, the use of biomarkers for risk stratification in patients with nonspecific complaints requires prospective validation in interventional studies before incorporation into medical practice. We are grateful to the emergency nurses, physicians, and patients who participated in the study. The authors thank Nicola Liversidge and Karen Delport, MD, for helpful discussions. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.