Baldirà et al. Ann. Intensive Care (2020) 10:7 https://doi.org/10.1186/s13613-020-0625-5 RESEARCH Open Access Biomarkers and clinical scores to aid the identification of disease severity and intensive care requirement following activation of an in‑hospital sepsis code Jaume Baldirà1,2, Juan Carlos Ruiz‑Rodríguez2,3,4*, Darius Cameron Wilson4, Adolf Ruiz‑Sanmartin3,4, Alejandro Cortes3,4, Luis Chiscano3,4, Roser Ferrer‑Costa5, Inma Comas5, Nieves Larrosa6,7, Anna Fàbrega6, Juan José González‑López6,7 and Ricard Ferrer2,3,4 Abstract Background: Few validated biomarker or clinical score combinations exist which can discriminate between cases of infection and other non-infectious conditions following activation of an in-hospital sepsis code, as well as provide an accurate severity assessment of the corresponding host response. This study aimed to identify suitable blood bio‑ marker (MR-proADM, PCT, CRP and lactate) or clinical score (SOFA and APACHE II) combinations to address this unmet clinical need. Methods: A prospective, observational study of patients activating the Vall d’Hebron University Hospital sepsis code (ISC) within the emergency department (ED), hospital wards and intensive care unit (ICU). Area under the receiver operating characteristic (AUROC) curves, logistic and Cox regression analysis were used to assess performance. Results: 148 patients fulfilled the Vall d’Hebron ISC criteria, of which 130 (87.8%) were retrospectively found to have a confirmed diagnosis of infection. Both PCT and MR-proADM had a moderate-to-high performance in discriminat‑ ing between infected and non-infected patients following ISC activation, although the optimal PCT cut-off varied significantly across departments. Similarly, MR-proADM and SOFA performed well in predicting 28- and 90-day mor‑ tality within the total infected patient population, as well as within patients presenting with a community-acquired infection or following a medical emergency or prior surgical procedure. Importantly, MR-proADM also showed a high association with the requirement for ICU admission after ED presentation [OR (95% CI) 8.18 (1.75–28.33)] or during treatment on the ward [OR (95% CI) 3.64 (1.43–9.29)], although the predictive performance of all biomarkers and clini‑ cal scores diminished between both settings. Conclusions: Results suggest that the individual use of PCT and MR-proADM might help to accurately identify patients with infection and assess the overall severity of the host response, respectively. In addition, the use of MR-proADM could accurately identify patients requiring admission onto the ICU, irrespective of whether patients *Correspondence: jcruiz@vhebron.net 3 Intensive Care Department, Vall d’Hebron University Hospital, Barcelona, Spain Full list of author information is available at the end of the article © The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​iveco​ mmons​.org/licen​ses/by/4.0/. Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 2 of 11 presented to the ED or were undergoing treatment on the ward. Initial measurement of both biomarkers might there‑ fore facilitate early treatment strategies following activation of an in-hospital sepsis code. Introduction progression in the emergency department (ED) [16], An early diagnosis of sepsis, irrespective of hospitalisa- after severe burn injury on a medical ward [17], and in tion setting, is crucial in order to provide a rapid and critically ill sepsis patients with decreasing PCT con- appropriate therapeutic response. This, however, may centrations [1] who may require renal replacement be complicated by the complex and heterogeneous host therapy (RRT) [18], or are likely to progress towards response to infection, as well as the high level of patient multiple organ failure [19]. heterogeneity between departments [1]. This study therefore aimed to investigate the perfor- A standardised set of defined criteria to aid in the mance of each biomarker (MR-proADM, PCT, CRP and early identification of sepsis therefore remains elusive, lactate) and clinical score (SOFA and APACHE II) in despite the established use of specific parameters in patients fulfilling the criteria for the Vall d’Hebron Uni- conditions such as myocardial infarction and pulmo- versity in-hospital sepsis code (ISC) in order to: (1) dis- nary embolism [2]. Furthermore, the poor acceptance criminate between infected and non-infected patients; of previous [3] and current [4, 5] definitions has led (2) assess infection severity according to 28- and 90-day to significant debate concerning the most appropriate mortality prediction, and (3) identify patients presenting clinical criteria required to make an accurate diagnosis. to the ED or undergoing treatment on a medical/surgical Differential presenting signs, symptoms and severities ward who require subsequent ICU admission. at each stage of hospitalisation, as well as prior surgical history and the setting of the original infectious insult, Methods significantly exacerbate this already complex diagnosis. Study design and ethical approval The construction and validation of new parameters This single-centre, observational study prospectively to aid in the early diagnosis and severity assessment of enrolled consecutive patients across all hospital depart- sepsis patients is therefore crucial. The recent develop- ments (ED, ward and ICU) who met the criteria for the ment of tools such as the quick Sequential Organ Fail- activation of the Vall d’Hebron University Hospital in- ure Assessment (qSOFA) score to screen patients at risk hospital sepsis code (ISC) [20] between November 2016 of a poor outcome has resulted in additional param- and August 2017. The study protocol was approved by eters becoming available to the treating clinician [6]; the Clinical Research Ethics Committee of Vall d’Hebron however, significant inherent and kinetical limitations University Hospital (PR(AG)333/2016) without the need make its implementation problematic [7–9]. The use of for informed consent, and was performed in accordance a panel of blood biomarkers, however, may provide a with the ethical standards laid down in the 1964 Declara- more elegant solution. Whilst measurement of individ- tion of Helsinki and its later amendments. ual biomarkers to concurrently discriminate between infectious and non-infectious patients and provide an Inclusion and exclusion criteria accurate assessment of disease severity is highly desir- Inclusion criteria comprised adult patients ≥ 18 years of able [10], no such biomarker appears to have a high age presenting with either a suspected or documented enough accuracy to fulfil both clinical requirements. infection with the presence of at least one of the two fol- Accordingly, both requirements should be considered lowing sets of variables, as outlined by the Vall d’Hebron separately. University Hospital in-hospital sepsis code (ISC) [20]: The use of procalcitonin (PCT) and C-reactive pro- (1) an acute alteration in the level of consciousness not tein (CRP) to discriminate between infectious and non- explained by other clinical conditions, or (2) the presence infectious cases is well documented [11–15], despite of hyperthermia (axillary temperature > 38.3 °C) or hypo- neither biomarker being well-established in the field of thermia (axillary temperature < 36.0 °C), and/or tachy- emergency medicine or on the medical ward. Further- cardia (> 110 beats per minute), tachypnea (> 30 breaths more, neither biomarkers exhibit a robust performance per minute) or desaturation (­ SpO2 < 90%), as well as arte- in terms of mortality prediction or in the identification rial hypotension (systolic arterial pressure < 90 mmHg of patients requiring intensive care (ICU) admission. or mean arterial pressure < 65 mmHg, or a decrease of Conversely, the novel blood biomarker, mid-regional > 40 mmHg of baseline systolic arterial pressure). Exclu- proadrenomedullin (MR-proADM) has been shown sion criteria included non-adult patients, pregnancy, or to be elevated in the early stages of infectious disease patients where no blood sample could be obtained. Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 3 of 11 Study endpoints and analytical aims selected based on a univariate analysis (p-value < 0.005 Study endpoints and analytical aims were defined as fol- after applying a Bonferroni correction), and subse- lows. Infection identification: either proven infection quently included in a multivariate analysis. Results were resulting from a positive blood culture or microbial iden- either presented as the hazard (HR) or odds ratio (OR) tification, or a final clinical diagnosis of infection. 28- and per 1 interquartile-range increase for mortality predic- 90-day mortality: all-cause mortality within either 28 or tion and ICU admission, respectively, with correspond- 90 days following enrolment. Community-acquired infec- ing 95% confidence intervals. Previously established tion: infectious insult developed outside the hospital [16] biomarker cut-offs for 28-day mortality prediction within the community. Hospital-acquired infection: infec- were further analysed within this dataset, and mortality tious insult developed whilst a patient was hospitalised rates in patient subgroups with high biomarker values for a previously non-infectious complaint. ICU admis- compared to those with low values. All data were ana- sion: admission onto either a medical or surgical inten- lysed using the statistics software R (version 3.1.2). sive care unit within 28 days of sepsis code activation. Data collection and biomarker measurements Results Patient comorbidities and demographics were subse- Patient characteristics quently noted upon sepsis code activation, as well as data A total of 150 patients were identified and enrolled concerning triage, routine laboratory values, microbiol- after activation of the sepsis code (sepsis code group), ogy testing and final clinical diagnosis. Clinical scores comprising 43 (29.1%) presenting to the emergency (APACHE II and SOFA) were retrospectively calculated department (ED), 70 (47.3%) already undergoing treat- whenever possible upon enrolment. The following bio- ment on a medical or surgical ward, and 35 (23.6%) marker measurements were performed as standard: already present on the intensive care unit (ICU). 2 PCT using a chemiluminescent immunoassay (CLIA), patients were removed from the analysis due to a lack CRP using an immuno-turbidimetric test, and L-lactate of blood sample. using an enzymatic colour test. MR-proADM was retro- spectively analysed following blood sampling through a central catheter, and stored at − 80 °C until batch analy- Infection identification following sepsis code activation sis using TRACE technology (time-resolved amplified Based on final clinical diagnosis, a total of 130 (87.8%) cryptate emission, KRYPTOR platform, Thermo Fisher). infected (sepsis code infected group) and 18 (12.2%) non- Results were unavailable to the treatment physician infected patients (sepsis code non-infected group) could throughout patient enrolment and hospitalisation. be retrospectively identified following initial sepsis code activation. Patient characteristics of the infected popula- tion are presented in Table 1. PCT was found to have a Statistical analysis high discriminatory performance (Fig. 1a), with an opti- Data were either reported using the mean (standard mal cut-off of 2.02 ng/mL (sensitivity and specificity: deviation) for symmetrically distributed variables, or 0.73 and 0.78). This cut-off, however, varied considerably the median [first quartile − third quartile] for vari- depending on the location of sepsis code activation, with ables showing a skewed distribution. Demographic and an optimal cut-off of 0.37 ng/mL (sensitivity and specific- clinical characteristics between surviving and non- ity: 0.86 and 0.75) in the ED, 1.91 ng/mL (sensitivity and surviving patients up to 28 days following sepsis code specificity: 0.77 and 0.86) on the ward and 2.02 mg/mL activation were assessed using the Chi-square (χ2) test (sensitivity and specificity: 0.79 and 0.71) on the ICU. for categorical variables, Student’s t test for age, and the In the sepsis code infected group, a positive blood cul- Mann–Whitney U test for all other continuous vari- ture rate was obtained in 54 (42.2%) cases, with a greater ables. Area under the receiver operating characteristic prevalence of Gram-positive as opposed to Gram-nega- (AUROC) curves were used to identify the biomarker tive infections (N = 34; 26.2% vs. N = 17; 13.1%, p < 0.01). or clinical score with the greatest predictive value for A comparison of biomarker and clinical score perfor- each endpoint, with 95% confidence intervals (95% CI) mance found that PCT had a moderate performance in compared to determine significance. Youden’s criterion predicting the likelihood of a positive culture (Fig. 1b). established optimal cut-off values with corresponding The addition of further biomarkers or clinical scores to sensitivity and specificity values. Univariate and multi- initial PCT measurements resulted in no significant variate Cox regression models assessed the association increase in discriminatory (infected vs. non-infected with survival time (censored at 28 days following sepsis patients) or predictive (positive vs. negative blood cul- code activation), with potential confounding variables ture) performance. Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 4 of 11 Table 1 Clinical patient characteristics upon activation of the sepsis code with respect to the total infected patient population and subsequent 28-day mortality Patient population (N = 130) Survivors (N = 102) Non-survivors (N = 28) p-value Age (years) (mean, S.D.) 63.1 (15.2) 62.3 (15.2) 66.1 (15.0) 0.205 Male gender (N, %) 83 (63.8%) 61 (59.8%) 22 (78.6%) 0.348 Infection characterisation Infection no SIRS (N, %) 7 (5.4%) 6 (5.9%) 1 (3.6%) 0.623 Severe sepsis (N, %) 41 (31.8%) 37 (36.3%) 4 (14.3%) 0.028 Septic shock (N, %) 82 (63.1%) 59 (57.8%) 23 (82.1%) 0.031 Sepsis-2 (N, %) 122 (93.8%) 95 (93.1%) 27 (96.4%) 0.814 Sepsis-3 (N, %) 116 (89.2%) 88 (86.3%) 28 (100.0%) 0.157 Location of sepsis code activation Emergency department (N, %) 39 (30.0%) 33 (32.4%) 6 (21.4%) 0.568 Ward (N, %) 63 (48.5%) 47 (46.1%) 16 (57.1%) 0.729 ICU (N, %) 28 (21.5%) 22 (21.6%) 6 (21.4) 0.839 Surgical admissions (N, %) 47 (36.2%) 37 (36.3%) 10 (35.7%) 0.487 Medical admissions (N, %) 83 (63.8%) 65 (63.7%) 18 (64.3%) 0.487 ICU length of stay (days) (median, IQR) 6 [2–15] 5 [2–11] 12 [3–24] 0.093 Hospital length of stay (days) (median, IQR) 16.5 [8–31] 17 [8–33] 16 [5.5–27] 0.298 Life supporting and intensive care therapies Vasopressors (N, %) 67 (51.5%) 52 (40.0%) 15 (53.6%) 0.106 Renal replacement therapy (N, %) 15 (11.5%) 8 (6.2%) 7 (25.0%) 0.009 Mechanical ventilation (N, %) 44 (33.8%) 30 (23.1%) 14 (50.0%) 0.011 Mechanical ventilation duration (days) 7 [3–16.25] 5 [3–17.5] 10 [4.5–14.75] 0.081 High-flow nasal cannula use (N, %) 19 (14.6%) 16 (15.7%) 3 (10.7%) 0.386 Pre-existing comorbidities Cardiopathy (N, %) 24 (18.5%) 16 (15.7%) 8 (28.6%) 0.120 Chronic kidney disease (N, %) 19 (14.6%) 14 (13.7%) 5 (17.6%) 0.584 COPD (N, %) 19 (14.6%) 10 (9.8%) 9 (32.1%) 0.003 Immunosuppression (N, %) 63 (48.5%) 44 (43.1%) 19 (67.9%) 0.020 Liver cirrhosis (N, %) 3 (2.3%) 1 (1.0%) 2 (7.1%) 0.054 Microbiology Positive blood culture (N, %) 54 (41.5%) 40 (39.2%) 14 (50.0%) 0.238 Gram positive (N, %) 34 (26.2%) 24 (23.5%) 10 (35.7%) 0.895 Gram negative (N, %) 17 (13.1%) 13 (12.7%) 4 (14.3%) 0.797 Fungal (N, %) 2 (1.5%) 1 (1.0%) 2 (7.1%) 0.529 Origin of infection Abdominal (N, %) 41 (31.5%) 35 (34.3%) 6 (21.4%) 0.194 Bacteria—primary (N, %) 4 (3.1%) 3 (2.9%) 1 (3.6%) 0.864 Catheter-related (N, %) 8 (6.2%) 5 (4.9%) 3 (10.7%) 0.257 Central nervous system (N, %) 2 (1.5%) 1 (1.0%) 1 (3.6%) 0.324 Respiratory (N, %) 34 (26.2%) 25 (24.5%) 9 (32.1%) 0.416 Soft-tissue (N, %) 4 (3.1%) 4 (3.9%) 0 (0.0%) 0.287 Urinary (N, %) 30 (23.1%) 25 (24.5%) 5 (17.9%) 0.459 Unknown (N, %) 3 (2.3%) 1 (1.0%) 2 (7.1%) 0.054 Other (N, %) 4 (3.1%) 3 (2.9%) 1 (3.6%) 0.864 Source control Focus cleaning (N, %) 34 (26.2%) 28 (27.5%) 6 (21.4%) 0.521 Drainage (N, %) 8 (6.2%) 6 (5.9%) 2 (7.1%) 0.806 Surgery (N, %) 23 (17.7%) 20 (19.6%) 3 (10.7%) 0.275 Biomarker and severity scores MR-proADM (nmol/L) (median, IQR) 3.54 [1.89–6.69] 3.18 [1.73–5.64] 5.69 [3.98–13.43] < 0.001 Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 5 of 11 Table 1 (continued) Patient population (N = 130) Survivors (N = 102) Non-survivors (N = 28) p-value PCT (ng/mL) (median, IQR) 7.05 [1.46–28.8] 7.85 [1.61–29.13] 4.94 [1.25–29.86] 0.654 Lactate (mmol/L) (median, IQR) 2.65 [1.70–4.57] 2.50 [1.60–3.75] 4.1 [2.0–8.10] 0.005 CRP (mg/L) (median, IQR) 17.78 [11.03–27.88] 17.48 [10.83–27.60] 23.14 [12.86–30.80] 0.322 SOFA (points) (mean, S.D.) 6.39 (3.46) 5.74 (3.11) 8.65 (3.70) < 0.001 APACHE II (points) (mean, S.D.) 21.98 (7.22) 21.63 (7.58) 23.13 (6.02) 0.783 Data are presented as absolute numbers with percentages in brackets, indicating the proportion of surviving and non-surviving patients at 28 days. APACHE II Acute Physiological and Chronic Health Evaluation II score, COPD chronic obstructive pulmonary disease, CRP C-reactive protein, ICU intensive care unit, IQR interquartile range, MR-proADM mid-regional proadrenomedullin, N number, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score Fig. 1 AUROC and univariate analysis for a discriminating between patients with a final diagnosis of infection from those where no infection was found, after activation of an in-hospital sepsis code, and b prediction of a positive blood culture in patients with a final diagnosis of infection. APACHE II Acute Physiological and Chronic Health Evaluation II score, AUROC area under the receiver operating characteristic curve, CI confidence interval, CRP C-reactive protein, HR hazard ratio, IQR interquartile range, LR likelihood ratio, MR-proADM: mid-regional proadrenomedullin, N number, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score 28‑ and 90‑day mortality prediction mortality rate was 24.6% (N = 32). In both cases, MR- The 28-day all-cause mortality rate within the sepsis proADM, lactate and SOFA were all significantly ele- code infected group was 21.5% (N = 28), with a higher vated in non-surviving as opposed to surviving patients mortality rate in patients enrolled whilst on the ward (p < 0.01), whereas no significant differences were found (N = 16; 25.4%) and ICU (N = 6; 21.4%) as opposed to in PCT, CRP or APACHE II values. in the ED (N = 6; 15.4%; Table 1). The 90-day all-cause Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 6 of 11 AUROC analysis found that MR-proADM performed performance found using MR-proADM, followed by well in predicting 28-day mortality, followed by SOFA SOFA and lactate. The influence of existing COPD had and lactate, with an optimal cut-off of 4.28 nmol/L (sensi- little impact on predictive performance, whilst the addi- tivity and specificity: 0.76 and 0.65). No significant asso- tion of PCT to MR-proADM resulted in the greatest ciation was observed with either PCT, CRP or APACHE increase in predictive performance [HR (95% CI) 5.7 II (Fig. 2). Similar findings were observed in both the (3.0–10.9)]. Interestingly, no increase in performance was univariate Cox regression analysis and after adjusting for found for the combination of PCT with either SOFA or the influence of existing chronic obstructive pulmonary lactate measurements [HR (95% CI) 2.9 (1.9–4.6) and 2.1 disease (COPD), which was the only pre-existing vari- (1.3–3.2)], respectively). able found to significantly influence mortality prediction. Using an MR-proADM cut-off of 1.54 nmol/L within The addition of PCT to MR-proADM measurements fur- the infected patient population, 108 (83.1%) patients ther improved predictive performance [HR (95% CI) 7.5 were found to have values equal to or exceeding this (3.7–15.2)], greater than any other biomarker or clinical cut-off, with a corresponding mortality rate of 24.1% score combination. Similar results were also obtained (N = 26). In comparison, patients with MR-proADM for 90-day mortality prediction, with a moderate values < 1.54 nmol/L (N = 22, 16.9%) had a significantly Fig. 2 AUROC, univariate and bivariate analysis for the prediction of 28-day a and 90-day b mortality for each biomarker and clinical score. APACHE II Acute Physiological and Chronic Health Evaluation II score, AUROC area under the receiver operating characteristic curve, CI confidence interval, CRP C-reactive protein, HR hazard ratio, IQR interquartile range, LR likelihood ratio, MR-proADM mid-regional proadrenomedullin, N number, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 7 of 11 (p < 0.01) lower mortality rate of 13.6% (N = 3). No patient small overall additive effect. Similarly, AUROC and Cox died with an MR-proADN value of < 0.88 nmol/L. Only 6 regression analysis found that MR-proADM had a mod- (4.6%) patients were found to have PCT concentrations erate-to-high predictive value within both medical and < 0.25 μg/mL, of which one died within 28 days. surgical patient subgroups, followed by SOFA and lac- tate (Table 3). The combination of SOFA and APACHE II Mortality prediction based on location of infection scores resulted in the greatest predictive performance in development and reason for patient admission the surgical cohort [HR (95% CI) 9.36 (2.13–4.17)], com- Subgroup analysis found that a total of 67 (51.5%) pared to the combination of MR-proADM and PCT in patients developed an infection in the community and the medical cohort [HR (95% CI) 9.30 (3.89–22.25)]. were subsequently hospitalised, compared to 48 (36.9%) patients where the infection was acquired during hos- Intensive care unit admission after sepsis code activation pitalisation despite being admitted for a non-infectious Finally, biomarker and clinical score performance was related condition. The 28-day mortality rates of both compared in order to predict admission to both the sur- groups were, however, similar (N = 15; 22.4% vs. N = 12; gical and medical intensive care units. Activation of the 25.0%). Likewise, 47 (36.2%) patients were identified sepsis code in both the ED and wards resulted in a high using the sepsis code following a surgical procedure, as rate of admission on the same or subsequent day, with opposed to 83 (63.8%) patients presenting with a medi- a total of 16 (41.0%) patients transferred from the ED cal complaint, again with similar 28-day mortality rates and 28 (45.9%) patients transferred from the ward. MR- between groups (N = 10; 21.3% vs. N = 18; 21.7%). proADM showed a moderate-to-high predictive value in AUROC and Cox regression analysis found that MR- both settings, with similar results also found after univar- proADM had a high predictive value in patients with iate logistic regression (Table 4). community-acquired infections, followed by SOFA and lactate (Table 2), which could be further increased after Discussion the addition of PCT [HR (95% CI) 10.0 (3.6–27.5)]. The This study compared the performance of a number of optimal MR-proADM cut-off was 3.67 nmol/L, with a clinical severity scores, conventional and novel biomark- sensitivity and specificity of 0.93 and 0.52, respectively. ers following activation of an in-hospital sepsis code, in Conversely, the SOFA score had high predictive per- order to identify cases of infection and assess the severity formance in patients with hospital-acquired infections of corresponding host response. Results found that the (Table 2). Interestingly, the combination of MR-proADM use of PCT to discriminate between sepsis code infected with APACHE II resulted in the largest increase in pre- group and sepsis code non-infected group, and MR- dictive value [HR (95% CI) 7.9 (1.5–41.1)], whereas proADM to identify disease severity irrespective of the the addition of PCT to MR-proADM only resulted in a location of infectious insult, had the greatest accuracy in Table 2 28-day mortality prediction in patients with either community or hospital-acquired infections Biomarker Patients (N) Events (N) AUROC Univariate Bivariate or clinical score 2 LR χ HR IQR [95% CI] LR χ2 HR IQR [95% CI] Community MR-proADM 67 15 0.76 [0.64–0.89] 13.08 4.60 [1.95–10.83] 14.52 4.88 [1.96–12.14] PCT 65 15 0.54 [0.36–0.71] 0.09 0.89 [0.43–1.85] 1.64 0.89 [0.43–1.85] CRP 66 15 0.57 [0.40–0.74] 0.48 1.28 [0.62–2.63] 2.30 1.32 [0.62–2.80] Lactate 60 14 0.71 [0.53–0.89] 10.03 3.44 [1.59–7.45] 10.61 3.09 [1.40–6.83] APACHE II 33 8 0.70 [0.50–0.90] 2.10 2.07 [0.77–5.55] 2.79 2.14 [0.76–6.03] SOFA 67 15 0.73 [0.57–0.88] 8.93 2.85 [1.39–5.85] 11.16 3.01 [1.44–6.28] Hospital MR-proADM 48 11 0.66 [0.45–0.88] 5.59 3.15 [1.19–8.37] 5.70 3.32 [1.19–9.25] PCT 48 11 0.46 [0.27–0.65] 0.03 0.91 [0.30–2.77] 0.19 0.88 [0.29–2.70] CRP 48 11 0.53 [0.32–0.73] 0.52 1.20 [0.70–2.07] 0.61 1.19 [0.69–2.06] Lactate 43 10 0.58 [0.35–0.81] 0.87 1.40 [0.70–2.77] 0.87 1.40 [0.70–2.77] APACHE II 25 6 0.63 [0.37–0.89] 0.85 0.61 [0.21–1.74] 1.18 0.58 [0.19–1.75] SOFA 48 11 0.78 [0.61–0.95] 12.12 3.60 [1.70–7.59] 12.18 3.58 [1.70–7.55] APACHE II Acute Physiological and Chronic Health Evaluation II score, AUROC area under the receiver operating characteristic curve, CI confidence interval, CRP C-reactive protein, HR hazard ratio, IQR interquartile range, LR likelihood ratio of the χ2 test, MR-proADM mid-regional proadrenomedullin, N number, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 8 of 11 Table 3 28-day mortality prediction in patients enrolled following either surgical or medical complaints Biomarker Patients (N) Events (N) AUROC Univariate Bivariate or clinical score LR χ2 HR IQR [95% CI] LR χ2 HR IQR [95% CI] Surgical MR-proADM 47 10 0.80 [0.66–0.94] 9.03 6.53 [1.65–25.88] 10.40 6.34 [1.84–41.04] PCT 47 10 0.47 [0.26–0.68] 0.34 1.39 [0.46–4.21] 0.68 1.56 [0.48–5.10] CRP 47 10 0.59 [0.38–0.79] 0.37 1.24 [0.60–2.58] 0.55 1.26 [0.60–2.63] Lactate 44 8 0.73 [0.49–0.97] 3.97 2.72 [1.00–7.39] 4.06 2.65 [0.98–7.20] APACHE II 29 7 0.53 [0.29–0.77] 0.01 1.05 [0.39–2.80] 0.78 0.86 [0.29–2.53] SOFA 47 10 0.74 [0.56–0.92] 10.86 5.05 [1.82–14.02] 11.12 5.55 [1.87–16.46] Medical MR-proADM 83 18 0.72 [0.57–0.87] 15.05 3.84 [1.96–7.54] 16.32 3.87 [1.90–7.86] PCT 80 18 0.55 [0.40–0.71] 0.10 0.89 [0.44–1.79[ 2.89 0.90 [0.45–1.80] CRP 82 18 0.46 [0.30–0.61] 0.63 1.23 [0.71–2.15] 3.49 1.27 [0.70–2.29] Lactate 72 18 0.66 [0.51–0.82] 6.43 2.06 [1.21–3.51] 7.84 1.99 [1.14–3.46] APACHE II 35 7 0.59 [0.34–0.82] 0.20 1.28 [0.44–1.79] 0.43 1.17 [0.38–3.59] SOFA 83 18 0.71 [0.57–0.86] 16.32 3.87 [1.90–7.86] 13.33 2.50 [1.43–4.36] APACHE II Acute Physiological and Chronic Health Evaluation II score, AUROC area under the receiver operating characteristic curve, CI confidence interval, CRP C-reactive protein, HR hazard ratio, IQR interquartile range, LR likelihood ratio of the χ2 test, MR-proADM mid-regional proadrenomedullin, N number, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score Table 4 Intensive (ICU) and surgical intensive (ICU) care unit admission from the emergency department and ward Biomarker Patients (N) Events (N) AUROC Univariate or clinical score LR χ2 OR IQR [95% CI] Emergency department MR-proADM 39 16 0.80 [0.66–0.94] 10.55 8.18 [1.75–28.33] PCT 37 16 0.74 [0.57–0.90] 7.18 3.59 [1.27–10.12] CRP 38 16 0.64 [0.46–0.82] 2.83 1.98 [0.85–4.60] Lactate 33 14 0.45 [0.24–0.65] 0.72 0.60 [0.18–1.99] APACHE II 16 14 0.63 [0.33–0.92] 0.21 1.63 [0.20–13.47] SOFA 39 16 0.78 [0.64–0.93] 10.21 5.04 [1.59–15.99] Ward admission MR-proADM 63 28 0.72 [0.64–0.79] 10.1 3.64 [1.43–9.29] PCT 63 28 0.59 [0.46–0.72] 3.5 1.85 [0.93–3.69] CRP 63 28 0.57 [0.41–0.65] 0.8 0.76 [0.41–1.41] Lactate 63 28 0.57 [0.44–0.70] 1.7 1.49 [0.81–2.74] APACHE II 20 16 0.69 [0.40–0.77] 1.3 0.38 [0.06–2.27] SOFA 63 28 0.64 [0.49–0.75] 3.8 2.03 [0.97–4.24] APACHE II Acute Physiological and Chronic Health Evaluation II score, AUROC area under the receiver operating characteristic curve, CI confidence interval, CRP C-reactive protein, IQR interquartile range, LR likelihood ratio of the χ2 test, MR-proADM mid-regional proadrenomedullin, N number, OR odds ratio, PCT procalcitonin, SOFA Sequential Organ Failure Assessment score fulfilling both criteria. Interestingly, MR-proADM was biomarkers may provide an easy and rapid source of clin- also the most accurate parameter in identifying patients ical information with which to inform and help develop presenting to the ED or receiving treatment on a medi- clinical strategies. Of these biomarkers, PCT has been cal ward who subsequently required immediate ICU studied extensively in both the ED and ICU [12–15, 22– admission. 24], with routine use in the ICU especially widespread. The use of clinical or laboratory parameters which are Conversely, novel biomarkers such as MR-proADM significantly increased in the early stages of the infectious remain poorly understood, despite a number of recent cascade may allow more appropriate treatment strate- studies [16, 18, 19, 25]. Furthermore, biomarker perfor- gies to be initiated upon initial suspicion of infection, mance may vary extensively depending on whether the thus minimising the risk of patients being either under- infection had developed in the community or during hos- or over-treated—both of which may result in undesir- pitalisation. Hence a key aim of this study focused on the able consequences [21]. To this extent, the use of blood comparison of biomarker and clinical score performance Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 9 of 11 according to the location of initial infection development, appears to have been formed across all hospital depart- following activation of an in-hospital sepsis code. ments, highlighting the use of MR-proADM in identify- Based on recent publications, the use of MR-proADM ing the patients at risk of mid- and long-term mortality. was of especial interest within our investigation, with In contrast, few studies have assessed MR-proADM previous studies showing elevated concentrations performance in identifying patients requiring ICU resulting from increased levels of microcirculatory and admission. An early recognition of such patients is cru- endothelial damage, resulting in the early stages of subse- cial, since it has previously been shown that inappropri- quent organ dysfunction [26–29]. The use of a biomarker ate ED discharge can result in higher mortality rates in to identify developing microcirculatory damage may con- patients initially classified as suitable for outpatient treat- tribute significant information following the activation of ment, but later directly admitted onto the ICU. Simi- a local, in-hospital sepsis code. lar findings were also reported after initial triage onto Accordingly, results from this investigation potentially a medical ward [31]. The relatively high performance of indicate two clinically important uses of MR-proADM MR-proADM in identifying these patients after presenta- following sepsis code activation. Firstly, MR-proADM tion to the ED, with a modest performance during treat- may be used as a tool to aid in the early identification of ment on a medical ward, may provide clinicians with an disease severity, irrespective of medical setting or prior additional tool in assessing suitability for ICU admission. surgical history. Secondly, it may also be used to aid clini- Nevertheless, further observational and interventional cal decision-making in assessing ICU requirement within studies are required to confirm these initial findings, and the ED or on a medical ward. determine the actual clinical value compared to standard This study found that MR-proADM had a high accu- routine practice. racy in identifying both 28-day and 90-day mortality, We note several limitations and strengths of this study compared to all other biomarkers and clinical scores, that deserve greater discussion. This was a single-cen- although many differences were not found to be sig- tre pilot study with a relatively low number of enrolled nificant. Performance was maintained irrespective of patients, resulting in many subgroups being underpow- whether the infection was developed in the community, ered for their respective endpoints. Results must there- culminating in a visit to the ED, or following a surgical fore be treated with caution. Nevertheless, similar results procedure. Conversely, the SOFA score appeared to more were found in comparison to those previously discussed accurately predict mortality in patients following a hos- in the literature, despite the patient population enrolled pital-acquired infection, although similarly, differences in this study being preselected on the basis of fulfilling were not found to be significant. In each case, however, an in-house sepsis criteria, and can therefore be viewed the small sample size and relatively low event rate result as additional collaborative evidence. Similarly, the num- in few significant differences between biomarkers and ber of patients where the in-hospital sepsis code was scores, as highlighted when comparing 95% confidence activated, yet no infection was subsequently found, only intervals. Nevertheless, similar findings have been shown amounted to 18 (12.2%) patients. A greater degree of elsewhere. Elke et al. [25] previously found differences in confidence in the discriminatory results between infected performance depending on whether the infecting agent and non-infected patients could be achieved with a larger consisted solely of Gram positive vs. negative bacteria, sample size. Finally, biomarker incorporation within a as well as highlighting the influence of prior surgical sepsis code requires a short turnaround time to result, procedures in patients directly admitted onto the ICU. which may be achieved using point-of-care (POC) tech- A further ICU study by Andaluz-Ojeda et al. [2] in 326 nology. At this point in time, however, no such platform patients with severe sepsis or septic shock, as well as that exists for the measurement of MR-proADM, making rou- of Enguix-Armada [30] in 388 patients with septic shock, tine incorporation of this biomarker in a clinical setting also found MR-proADM to have the highest predic- problematic. tive value compared to conventional biomarkers such as PCT, CRP and lactate, although interestingly, the greatest Conclusions performance was found in patients with a lower degree This study found that the use of PCT to discriminate of organ failure. Nevertheless, no subgroup analysis was between infected and non-infected cases following activa- performed in either study relating to location of infec- tion of an in-hospital sepsis code, followed by measure- tion development or prior surgical history. It is, however, ment of MR-proADM to determine infection severity and interesting to note that similar results to these high sever- assess the requirement for ICU admission, may help guide ity, critical care investigations could be observed in our initial treatment decisions and intensify therapy in high- less severe cohort across three different hospital depart- risk patients. Interventional studies, specifically focussing ments. Accordingly, a significant body of evidence now on MR-proADM, to potentially optimise early patient Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 10 of 11 treatment strategies are now crucial before implementation References 1. Elke G, Bloos F, Wilson DC, Brunkhorst FM, Briegel J, Reinhart K, et al. in routine clinical use. The use of mid-regional proadrenomedullin to identify disease severity and treatment response to sepsis—a secondary analysis of a large ran‑ domised controlled trial. Crit Care. 2018;22(1):79. Abbreviations 2. Andaluz-Ojeda D, Nguyen HB, Meunier-Beillard N. Superior accuracy of AUROC: area under the receiver operating characteristic curve; CI: confidence mid-regional proadrenomedullin for mortality prediction in sepsis with interval; CRB-65: severity scores for community-acquired pneumonia; CRP: varying levels of illness severity. Ann Intensive Care. 2017;7(1):15. C-reactive protein; ED: emergency department; HR: hazard ratio; IQR: inter‑ 3. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, et al. 2001 quartile range; MR-proADM: mid-regional proadrenomedullin; N: number; OR: SCCM/ESICM/ACCP/ATS/SIS international sepsis definitions conference. odds ratio; PCT: procalcitonin; qSOFA: quick Sequential Organ Failure Assess‑ Crit Care Med. 2003;31(4):1250–6. ment; RRT​: renal replacement therapy; SIRS: Systemic Inflammatory Response 4. Singer M, Deutschman CS, Seymour CW. The third international Syndrome; SOFA: Sequential Organ Failure Assessment. consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801–10. Acknowledgements 5. Seymour CW, Liu VX, Iwashyna TJ, Brunkhorst FM, Rea TD, Scherag The authors are grateful to the staff at the Vall d’Hebron University Hospital A. Assessment of clinical criteria for sepsis: for the third international for their continued vigilance in identifying patients presenting with infections, consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. and for enrolling patients into this study. DCW would like to acknowledge the 2016;315(8):762–74. contributions made by Monika Wittmann concerning graphical presentation. 6. Singer AJ, Ng J, Thode HC Jr, Spiegel R, Weingart S. Quick SOFA scores predict mortality in adult emergency department patients with and with‑ Author contributions out suspected infection. Ann Emerg Med. 2017;69(4):475–9. JB, JCRR and RF conceived and designed the study. JB was the primary author, 7. Simpson SQ. New sepsis criteria: a change we should not make. Chest. editor of the manuscript and the lead clinical investigator. All authors apart 2016;149(5):1117–8. from DCW contributed to the enrolment of patients and sample collection. 8. Simpson SQ. SIRS in the time of Sepsis-3. Chest. 2017;153(1):34–8. DCW provided statistical advice and analysed the data. All authors critically 9. Simpson SQ. Diagnosing sepsis: a step forward, and possibly a step back. reviewed and approved the final manuscript. RF and JCR take responsibility for Ann Transl Med. 2017;5(3):55. the paper as a whole. All authors read and approved the final manuscript. 10. Charles PE, Peju E, Dantec A, Bruyere R, Meunier-Beillard N, Dargent A, et al. Mr-Proadm elevation upon ICU admission predicts the outcome of Funding septic patients and is correlated with upcoming fluid overload. Shock. The study was funded by a restricted grant from Thermo Fisher (Germany), 2017;48(4):418–26. consisting of free of charge kits. However, the funding organisation had no 11. Schuetz P, Affolter B, Hunziker S, Winterhalder C, Fischer M, Balestra GM, role in the collection, management, analysis, or interpretation of the data; et al. Serum procalcitonin, C-reactive protein and white blood cell levels preparation, review, or approval of the manuscript, or decision to submit the following hypothermia after cardiac arrest: a retrospective cohort study. manuscript for publication. Eur J Clin Invest. 2010;40(4):376–81. 12. Schuetz P, Albrich W, Christ-Crain M, Chastre J, Mueller B. Procalci‑ Availability of data and materials tonin for guidance of antibiotic therapy. Expert Rev Anti Infect Ther. The datasets used and/or analysed during the present study are available from 2010;8(5):575–87. the corresponding author upon reasonable request. 13. Schuetz P, Aujesky D, Muller C, Muller B. Biomarker-guided personalised emergency medicine for all—hope for another hype? Swiss Med Wkly. Ethics approval and consent to participate 2015;145:w14079. The study protocol was approved by the ethics board of the Vall d’Hebron 14. Schuetz P, Christ-Crain M, Muller B. Procalcitonin and other biomarkers to University Hospital (PR(AG)333/2016), with no written informed consent improve assessment and antibiotic stewardship in infections–hope for required from patients or legal representatives. hype? Swiss Med Wkly. 2009;139(23–24):318–26. 15. Schuetz P, Christ-Crain M, Thomann R, Falconnier C, Wolbers M, Widmer Consent for publication I, et al. Effect of procalcitonin-based guidelines vs standard guidelines No individual participant data is reported that would require consent to pub‑ on antibiotic use in lower respiratory tract infections: the ProHOSP rand‑ lish from the participant (or legal parent or guardian for children). omized controlled trial. JAMA. 2009;302(10):1059–66. 16. Saeed K, Wilson DC, Bloos F, Schuetz P, van der Does Y, Melander O, et al. Competing interests The early identification of disease progression in patients with suspected All authors have provided information on potential conflicts of interests infection presenting to the emergency department: a multi-centre directly or indirectly related to the work submitted in the journal’s disclosure derivation and validation study. Crit Care. 2019;23(1):40. forms. The authors declare that they have no competing interests. 17. Gille J, Ostermann H, Dragu A, Sablotzki A. MR-proADM: a new bio‑ marker for early diagnosis of sepsis in burned patients. J Burn Care Res. Author details 2017;38(5):290–8. 1 Intensive Care Department, Hospital de la Santa Creu i Sant Pau, Barcelona, 18. Nierhaus A, Bloos F, Wilson DC, Elke G, Meybohm P, SepNet Critical Care Spain. 2 Department de Medicina, Universitat Autònoma de Barcelona, Bar‑ Trials G. Predicting the requirement for renal replacement therapy in celona, Spain. 3 Intensive Care Department, Vall d’Hebron University Hospital, intensive care patients with sepsis. Crit Care. 2018;22(1):201. Barcelona, Spain. 4 Shock, Organ Dysfunction and Resuscitation Research 19. Elke G, Bloos F, Wilson DC, Meybohm P. Identification of developing multi‑ Group, Vall d’Hebron Institute of Research, Barcelona, Spain. 5 Biochemistry ple organ failure in sepsis patients with low or moderate SOFA scores. Crit Department, Clinical Laboratories, Vall d’Hebron University Hospital, Barcelona, Care. 2018;22(1):147. Spain. 6 Microbiology Department, Vall d’Hebron University Hospital, Barce‑ 20. Ferrer R, Ruiz-Rodriguez JC, Larrosa N, Llaneras J, Molas E, González-López lona, Spain. 7 Department de Genètica i Microbiologia, Universitat Autònoma JJ. Sepsis code implementation at Vall d’Hebron university hospital: rapid de Barcelona, Barcelona, Spain. diagnostics key to success. ICU Manag Pract. 2017;17(4):214–5. 21. Van der Does Y, Limper M, Jie KE, Schuit SCE, Jansen H, Pernot N, et al. Received: 15 July 2019 Accepted: 7 January 2020 Procalcitonin-guided antibiotic therapy in patients with fever in a Baldirà et al. Ann. Intensive Care (2020) 10:7 Page 11 of 11 general emergency department population: a multicenter noninfe‑ 28. Xie Z, Chen WS, Yin Y, Chan EC, Terai K, Long LM, et al. Adrenomedullin riority randomized clinical trial (HiTEMP study). Clin Microbiol Infect. surges are linked to acute episodes of the systemic capillary leak syn‑ 2018;24(12):1282–9. drome (Clarkson disease). J Leukoc Biol. 2018;103(4):749–59. 22. Schuetz P, Batschwaroff M, Dusemund F, Albrich W, Burgi U, Maurer 29. Vigue B, Leblanc PE, Moati F, Pussard E, Foufa H, Rodrigues A, et al. M, et al. Effectiveness of a procalcitonin algorithm to guide antibiotic Mid-regional pro-adrenomedullin (MR-proADM), a marker of positive therapy in respiratory tract infections outside of study conditions: a post- fluid balance in critically ill patients: results of the ENVOL study. Crit Care. study survey. Eur J Clin Microbiol Infect Dis. 2010;29(3):269–77. 2016;20(1):363. 23. Schuetz P, Chiappa V, Briel M, Greenwald JL. Procalcitonin algorithms 30. Enguix-Armada A, Escobar-Conesa R, La Torre AG, De La Torre-Prados MV. for antibiotic therapy decisions: a systematic review of randomized Usefulness of several biomarkers in the management of septic patients: controlled trials and recommendations for clinical algorithms. Arch Intern c-reactive protein, procalcitonin, presepsin and mid-regional pro-adre‑ Med. 2011;171(15):1322–31. nomedullin. Clin Chem Lab Med. 2016;54(1):163–8. 24. Schuetz P, Wirz Y, Sager R, Christ-Crain M, Stolz D, Tamm M, et al. Proc‑ 31. Fernando SM, Rochwerg B, Reardon PM, Thavorn K, Seely AJE, Perry alcitonin to initiate or discontinue antibiotics in acute respiratory tract JJ, et al. Emergency department disposition decisions and associated infections. Cochrane Database Syst Rev. 2017;10:CD007498. mortality and costs in ICU patients with suspected infection. Crit Care. 25. Elke G, Bloos F, Wilson DC, Brunkhorst FM, Briegel J, Reinhart K. The use of 2018;22(1):172. mid-regional proadrenomedullin to identify disease severity and treat‑ ment response to sepsis—a secondary analysis of a large randomised controlled trial. Crit Care. 2018;22(1):79. Publisher’s Note 26. Temmesfeld-Wollbruck B, Hocke AC, Suttorp N, Hippenstiel S. Adre‑ Springer Nature remains neutral with regard to jurisdictional claims in pub‑ nomedullin and endothelial barrier function. Thromb Haemost. lished maps and institutional affiliations. 2007;98(5):944–51. 27. Pittard AJ, Hawkins WJ, Webster NR. The role of the microcirculation in the multi-organ dysfunction syndrome. Clin Intensive Care. 1994;5(4):186–90.