Introduction

During the last twenty years, despite research on prevention and treatments, the mortality and the morbidity rates of sepsis remain stable [1, 2]. To date, whatever the initial stage of sepsis, this latter one is responsible of 20% of all deaths worldwide, approximately 11 million deaths ever year [3]. Sepsis is also a major cause of death among hospitalized patients, especially in intensive care unit (ICU) [4, 5] where mortality and morbidity rates are estimated of 37.3% and 10.4% [6].

On 2017 the World Health Assembly and the World Health Organization defined that prevention, diagnosis, and sepsis management are the main priorities to decrease morbidity and mortality [7]. It was underpinned that sepsis outcome depends on early identification and treatment implementation with hemodynamic optimization and antibiotic therapy (ABT) administration [8, 9] both included in a bundle of care [10].

In and, out-of-hospital sepsis diagnosis and severity evaluation are the bundle of care first steps aiming to determine the appropriate level of care. Because 70% of sepsis occurs outside a hospital environment, a special attention is needed to help physician for sepsis diagnosis and severity assessment to optimize triaging between ambulatory treatment, ward, emergency department (ED) or ICU admission [11]. Outside hospital setting, severity assessment is mainly based on clinical signs due to the absence of biomarker availability [12,13,14]. Different sepsis severity assessment scores have been developed, to enhance clinical diagnostic performance, despite to date no score was prospectively validated for pre-hospital use [15, 16].

In 2012, the Royal College of Physicians launched the National Early Warning Score (NEWS) to improve the outcomes of patients suffering from acute illness [17]. The NEWS score aims to triggering a rapid and effective clinical response, in time, person and place because the triad: early detection, timeliness of response and competency of the clinical response directly impact the outcome [18,19,20,21,22,23,24,25].

This study aims to describe the association between the prehospital National Early Warning Score 2 (NEWS-2) and in-hospital, 30 and 90-day mortality of SS patients cared for in the pre-hospital setting by a mobile ICU (MICU).

Methods

Population

As previously reported, in France, pre-hospital emergency system (PEMS) relies on the Service d’Aide Médicale d’Urgence (SAMU) [12, 13, 15, 26, 27]. Briefly, SAMU is composed of dispatch operators and emergency physicians [28] with a unique national phone number, the “15”. After a telephone discussion with the patient, or a relative, or a witness, the physician, based on patient’s medical history and reported symptoms, in case of life-threatening emergencies, may decide to dispatch a mobile intensive care unit (MICU) team to the scene. The MICU team, a driver, a nurse and an emergency physician, is equipped to face initial management of major organ failures [28].

All adults patients aged ≥ 18 years, cared for by a MICU from SAMU 75 Necker - Enfants Malades Hospital, SAMU 75 Lariboisière Hospital, SAMU 75 Pitié Salpêtrière Hospital, SAMU 75 Hôtel Dieu Hospital, Paris, Assistance Publique - Hôpitaux de Paris, Paris– France; Paris Fire Brigade Emergency Medical Service, Paris– France; SAMU 972 Fort de France University Hospital, La Martinique– France; SAMU 31 University Toulouse Hospital, Toulouse– France and SAMU 31 Castres Hospital, Castres– France, between 2016, April 6th, and 2021, December 31st, were included in this retrospective study based on 2012 sepsis-2 conference septic shock definition [29]. Eligible patients were identified using electronic research using septic shock or severe sepsis keywords. No exclusion criteria were used in this study.

Ethical considerations

The French Society of Anaesthesiology and Intensive Care ethics committee (Reference: IRB00010254-2017-026, 2017/12/12), the Ethics Committee for Behavioural and Health Research (CERCES Reference 2018-04, 2018/01/16) and the National Heart Agency (2017-A02335-48–2017/07/30) approved the trial protocol waiving patient consent for this retrospective study.

Data collection

A standardized data collection template was used in order to minimize data abstraction bias [30].

Patients’ demographic characteristics (age, weight, height, calculated body mass index (BMI) and gender), prehospital supposed origin of sepsis, initial prehospital vital signs values (systolic (SBP), diastolic (DBP) and mean blood (MBP) pressure, heart rate (HR), pulse oximetry (SpO2), respiratory rate (RR), body core temperature and Glasgow coma scale (GCS)), skin mottling score (SMS from 0 to 5), capillary refill time (CRT) (seconds), duration of prehospital care, prehospital treatments delivered (antibiotic therapy (ABT) type and dose, fluid volume expansion type and dose, catecholamine type and dose) were collected from MICU pre-hospital medical records. Comorbidities: hypertension, coronary heart disease (CHD), chronic cardiac failure (CCF), chronic renal failure (CRF), chronic obstructive pulmonary disease (COPD), diabetes mellitus and history of cancer, were also collected from pre and in-hospital medical reports. Length of stay in the ICU, length of stay in the hospital, 30 and 90-day mortality were retrieved from in-hospital medical records. In France, the hospital patient monitoring software enables to know the vital status even if the patient is no longer hospitalised. Thus, the vital status “alive” or “dead” on day-90 was available for each analysed patient.

Simplified acute physiology score (SAPS2) [31] was calculated 24 h after hospital admission.

Statistical analysis

A mean with standard deviation was used to express quantitative parameters with a gaussian distribution, median with interquartile range [Q1-Q3] for quantitative parameters with a non-normal distribution and absolute values and percentages for qualitative parameters.

The main outcomes were in-hospital, 30 and 90-day mortality of septic shock patients initially cared for by a MICU in the pre-hospital setting.

The NEWS-2 was calculated based on the sum of the worst value of the 6 following physiological variables: blood pressure, heart rate, respiratory rate, temperature, oxygen saturation prior oxygen supplementation, and level of consciousness according to the 2017 Royal College of Physicians.

guidelines [17]. The NEWS-2 ranges from 0 to 20.

A NEWS-2 ≥ 7 threshold was chosen for increased clinical deterioration risk definition and usefulness in clinical practice based on previous report [17].

The relationship between each covariate and in-hospital, 30-day and 90-day mortality rates were assessed by bivariate and multivariate analyses. The NEWS-2 was analysed, as a continuous variable and as a binary variable using a threshold of NEWS ≥ 7 because its association with increased clinical deterioration risk. Results are expressed by an Odd Ratio (OR) and adjusted Odd Ratio (aOR) with a 95% confidence interval [95 CI].

A log binomial regression weighted with the inverse probability of treatment (IPTW) propensity score method was computed taking into potential cofounders. The propensity score aims to decrease bias due to non-randomized treatment allocation [32]. Cofounders included in the IPTW propensity analysis were: age, cancer history, CRF, COPD, CHD, diabetes mellitus, CCF, SAPS2, prehospital ABT administration, prehospital fluid volume expansion and prehospital catecholamine infusion. The selection of variables included in the multivariable analysis was done a priori guided by previous knowledge of factors known to influence septic shock survival. Results were expressed as adjusted risk ratio (RR) [95 CI].

All tests were 2-sided with a statistically significant p-value considered as < 0 0.05.

All analyses were performed using R 3.4.2 (http://www.R-project.org; the R Foundation for Statistical Computing, Vienna, Austria).

Results

Population characteristics

Between 2021, April 6th, and 2021, December 31st, 530 patients requiring pre-hospital MICU intervention for septic shock were analysed. The missing data rate was lower than 1%; these data were deleted for the statistical analysis.

Of the 7 participating centres, 165 patients (31%) were included by the Paris Fire Brigade Emergency Medical Service, 104 patients (20%) by SAMU 31 Toulouse, 77 patients (15%) by SAMU 75 Necker, 71 patients (13%) by SAMU Castres, 51 patients (10%) by SAMU 972 La Martinique, 31 patients (6%) by SAMU 75 Lariboisière, 14 patients (3%) by SAMU 75 Hôtel Dieu and 17 patients (3%) by SAMU 75 Pitié Salpétrière.

The overall population mean age was 69 ± 15 years and 341 patients (64%) were male gender.

One hundred eighty (34%) patients died during hospital stay, 164 (31%) patients had died by day-30, and 184 (35%) had died by day-90.

Table 1 summarise the populations’ demographic and clinical characteristics (Table 1).

Table 1 Population demographic and first prehospital clinical characteristics. Results are expressed as mean with standard deviation for quantitative parameters with gaussian distribution, as median with interquartile range for quantitative parameters with non-normal distribution and, as absolute value and percentage for qualitative parameters. p-value corresponds to the comparison between alive and deceased patients. p-value in bold indicates a significant difference between alive and deceased patients
Full size table

Presumed origin of septic shock was mainly pulmonary (43%), digestive (25%) or urinary (17%) (Table 2).

Table 2 Presumed origin of septic shock. Data are expressed as an absolute value with a percentage of all cases (due to rounding, total overpasses 100%)
Full size table

Prehospital setting

The mean duration of prehospital care was 71 ± 34 min, pre-hospital fluid expansion consisted on crystalloids (100%) with a mean volume of 932 ± 573 ml and 155 (29%) patients received norepinephrine infusion with a median dose of 1.0 [0.5–2.0] mg.h− 1.

Prehospital ABT was administered in 132 patients (25%) and no significant difference was observed between patients surviving or dying, at hospital, at 30 or 90 day (Table 1). The antibiotics were principally 3rd generation cephalosporins (n = 98, 75%).

Hospital stays

The median length of stay in the ICU was 4 [2–8] days and the median in-hospital length of stay was 10 [5–18] days.

The mean SAPS2 score was 60 ± 21 with significant difference between patients surviving or dying in hospital, at day-30 and day-90 (Table 1).

Prehospital NEWS-2

The mean overall prehospital NEWS-2 was 9 ± 3, with 406 patients (77%) patients having a prehospital NEWS-2 ≥ 7.

Bivariate logistic analysis revealed a significant association between prehospital NEWS (continuous variable) and in-hospital mortality (OR = 2.14 [1.19–3.83], p = 0.01), 28-day mortality (OR = 2.41 [1.33–4.36], p = 0.003) and 90-day mortality (OR = 1.88 [1.05–3.35], p = 0.03).

Using a threshold of a prehospital NEWS-2 ≥ 7, the association remains significant for in-hospital mortality (OR = 1.62 [1.04–2.58], p = 0.04), 28-day mortality (OR = 1.94 [1.22–3.18], p = 0.01) but not for 90-day mortality (OR = 1.43 [0.93–2.24], p = 0.11).

Propensity IPTW analyses

The log binomial regression weighted with the IPTW observed an association between a prehospital NEWS-2 ≥ 7 and in-hospital mortality: RRa = 2.34 [1.39–3.95], 30-day mortality: RR = 2.08 [1.33–3.25] and 90-day mortality: RRa = 2.22 [1.38–3.59].

Initial lactate was missing for 218 patients (41%) and initial skin mottling score missing for 220 patients (42%); both covariables were not included in the model.

The calibration statistic values were: 0.54; 0.55 and 0.53 respectively for in-hospital mortality, 30-day mortality and 90-day mortality.

Discussion

Here we report a positive association between in-hospital, 30 and 90-day mortality and (i) the prehospital NEWS-2, (ii) prehospital NEWS-2 ≥ 7 and among septic shock patients cared for in the pre-hospital setting by a mobile intensive care unit.

Previous reports underpinned that around one-third of potentially preventable deaths in the United Kingdom were related to poor clinical monitoring and/or inadequate response to clinical deterioration supporting that patients should be addressed to the most appropriate setting for clinical care [33], justifying NEWS-2 widespread deployment [17]. It is all the most true for sepsis for which diagnosis, severity assessment and treatment initiation does not suffer from any delay [14, 34, 35]. Beyond out- and in-hospital care aiming to improve sepsis outcome, early diagnosis and severity are cornerstones to decrease sepsis related mortality by initiation of the sepsis survival chain [36]. Early diagnosis is also recognised as an essential leverage arm to prevent potentially preventable deaths as it allows early initiation of treatment. Daily, PEMS are faced to a challenge aiming to, as quickly as possible, establish a right diagnosis and assess severity to adequately guide the patient towards the optimal care pathway for his or her disease. To establish a diagnosis and a clinical monitoring, simple and objective clinical tools usable at any time and reproduced by different caregivers are necessary. This is all the more important when diagnostic certainty is uncertain and/or urgent, even when combining both clinical and paraclinical, for example biological, variables, in the grey zone decision making [37]. Some clinical parameters, e.g., SMS, CRT [12, 13], are subjective and may be caught out in some situation, for example hypothermia.

Because of the lack of specificity of a single clinical sign [14], scoring system was developed to improve sensitivity and sensibility. For sepsis, the most known scores are which SOFA [38], Mortality in Emergency Department Sepsis (MEDS) [39], Predisposition, Infection, Response and Organ dysfunction (PIRO) [40] and q-SOFA since 2016 [14]. Although the latter does not require biological results and is recommended because of its simplicity outside ICU [41], q-SOFA validity remains under debate [13, 15, 16, 40, 42,43,44]. To date, in the prehospital setting, no score is validated, thus, sepsis severity assessment and prognostication still remain on clinical evaluation [45].

The score has several advantages, including: its ease of establishment since all the variables are accessible in the pre-hospital setting, its inter-observer reproducibility, and the possibility of being repeated over time in order to evaluate the treatment effect. However, one of the weakness is that NEWS-2 does not include age and major comorbidities, both reflecting frailty [46, 47] and associated with poor sepsis outcome [48,49,50].

Limitations

The current study suffers from limitations. Because of the retrospective study design, we cannot exclude that a potential selection bias affects the results validity. In addition, we are unable to conclude on a causal link between the NEWS-2 and mortality related to septic shock and sepsis. We cannot exclude unknown or missed confounders during the analysis. The NEWS-2 performance and external validation need to be confirmed by prospective studies, although the inclusion of centres of varying size and geography (a large city - Paris, one medium-sized city - Toulouse and one rural city– Castres) seems promising and represents a study strength. The study population was only adults; consequently, results extrapolation to sepsis and to paediatric population is not possible. We should keep in mind that some NEWS-2 variables may be influenced by patient previous medications, e.g., beta-blocker therapy, restricting their contribution to the NEWS-2.

Beyond all these limitations, the NEWS-2 seems to be an adequate tool for pre-hospital sepsis screening of a high risk of poor evolution and should be considered as an aid to clinical decision making, not a barrier or alternative to skilled clinical judgement.

Conclusion

Among sepsis patient requiring prehospital mobile intensive care unit intervention, a prehospital NEWS-2 ≥ 7 is associated with an increase in in-hospital, 30 and 90-day mortality. However, prospective studies are needed to confirm the usefulness of NEWS-2 to improve the prehospital triage, patient orientation to the optimal pathway and sepsis related mortality.