Sepsis is a major healthcare concern (Adhikari et al, 2010), resulting in an estimated 30 million cases globally each year and around 6 million deaths annually related to sepsis (Fleischmann et al, 2016). According to the UK Sepsis Trust (2019), there are at least 250 000 cases of sepsis and 52 000 sepsis-related deaths in the UK every year. One in four patients with sepsis die during hospitalisation, and sepsis survivors frequently experience long-term functional and cognitive impairment, resulting in significant social and economic impacts (Iwashyna et al, 2010; Mostel et al, 2020).
Acute changes in physiological parameters (respiratory rate, heart rate, oxygen saturations, blood pressure, temperature and level of consciousness) can be warning signs of deterioration in a patient. Early warning scores (EWSs) were developed to identify patients at risk of deterioration. The National Early Warning Score (NEWS) is an EWS developed by the Royal College of Physicians (RCP) for use in acute hospital settings, with the aim of standardising EWS use. It was initially introduced into the NHS in 2012, although there were systems of routine monitoring of patients being used prior to this as well (RCP, 2017).
The use of EWS in the pre-hospital setting remains controversial due to the lack of available evidence relating to its effectiveness in these settings (Roland and Jahn, 2012).
Despite increased awareness of sepsis, this condition is frequently missed in the community until it has progressed to severe sepsis or septic shock. EWSs are used in hospital settings to identify and respond to deteriorating patients. Use of these tools in pre-hospital settings could facilitate identification of sepsis earlier and provide clinicians with a common language to provide more seamless and quicker patient care. There have been numerous studies and reviews conducted in acute settings (Alam et al, 2014; Corfield et al, 2014; Smith et al, 2014; Nannan Panday et al, 2017), but there is less known about the effectiveness of EWSs in pre-hospital settings. This systematic review aimed to assess the difference in patient outcomes if an EWS is used pre-hospitalisation in patients with suspected sepsis.
Methods
Search strategy
This narrative review was conducted in accordance with the PRISMA statement (Moher et al, 2009). Two bibliographic databases-Ovid Medline and CINAHL-were searched for studies evaluating the pre-hospital use of EWS in patients with suspected sepsis published in English and with no date restrictions. The Ovid Medline search strategy is detailed in Table 1, and additional search strategies are available in Appendix 1. Reference lists of relevant studies and systematic reviews were examined for sources of further relevant studies.
Table 1. Eligibility criteria
| Type of participants | The target of the intervention was an adult (16 years and over) of any gender with suspected sepsis in a primary care setting. Only studies that were conducted in the UK were included. Studies that only included children with suspected sepsis were excluded. Studies that were solely hospital-based were excluded. Studies that included A&E were included but only if outcomes of the pre-hospital use of EWS and A&E use were reported separately. |
| Type of intervention | The type of interventions were those use an early warning score (EWS), defined as a score used to identify patients at risk of clinical deterioration, as an assessment tool in patients with suspected sepsis. Interventions that use National Early Warning Score (NEWS), NEWS2, Modified Early Obstetric Warning Score (MEOWS), Modified Early Warning Score (MEWS) and quick Sequential Organ Failure Assessment scores (qSOFA) were included. Interventions that use Paediatric Early Warning Score (PEWS) were excluded. |
| Type of outcome | This systematic review was concerned with the ability to predict adverse outcomes in a pre-hospital population. Outcomes were time to diagnosis, time to initiation of treatment, intensive care unit (ICU) admission within 48 hours and 30 day mortality. Outcomes, time to diagnosis and time to treatment could be measured by time to initiation of Sepsis 6 Bundle. |
| Language | All studies not in English were excluded due to a lack of translation resources |
| Type of study | Randomised controlled trials (RCTs), controlled non-randomised studies, case control studies, cross-sectional studies and cohort studies were included in the systematic review |
Predefined eligibility criteria (Table 2) were used to determine whether a study could be included in this systematic review. Studies were screened initially by titles and abstracts. All potentially relevant articles were obtained in full. If the information contained in the title or abstract was inconclusive, then the full article was obtained. Full articles were assessed against the eligibility criteria. Studies were included if they examined the effect of the pre-hospital use of EWS on identification of a patient with suspected sepsis or infection and whether it was predictive of their prognosis or outcome. The outcomes of interest were ICU admission within 48 hours and 30-day mortality. Only studies conducted in the UK were included. This systematic review only included randomised controlled trials (RCTs), controlled non-randomised studies, case-control studies, cross-sectional studies and cohort studies. Studies that only included children or were solely hospital-based were excluded. Only studies written in English were included.
Table 2. Study characteristics
| Study/country | Population | EWS | Comparator | Outcome |
|---|---|---|---|---|
| Booth and Bloch (2013) Aberdeen Royal Infirmary, Scotland Mixed urban/rural | Prospective study, 7 weeks 104 patients transported to ED resus by ambulance while investigator was on shift | Pre-alert tool. Pre-alert requirement determined by consultant physician blinded to outcome | Clinician judgement | Pre-alert sensitivity and specificity |
| Abbott et al (2018) Royal London Hospital, England Urban | Retrospective cohort study, 25th March–13 April 2013 180 adult patients admitted to acute assessment unit brought to hospital by ambulance | NEWS—pre-hospital, ED, admission to ward | Clinician judgement | Critical care unit escalation and death within 48 hours, length of hospital stay |
| Hargreaves et al (2019) Two non-specialist UK hospitals South East Coast Ambulance Trust, England Urban | Retrospective cohort study, 2015–2017 1233 adult patients transported to ED by emergency ambulance | Pre-hospital NEWS and ED lactate | Clinician judgement | 30-day mortality ICU admission |
| Shaw et al (2017) London Ambulance Trust, West London, England Urban | Retrospective cohort study, April 2012–January 2013 287 adult patients treated by ambulance clinicians and transferred to ED | Pre-hospital NEWS | Clinician judgement | Hospital discharge disposition |
| Silcock et al (2019) Royal Alexandria Hospital, Paisley, Scotland Mixed urban/rural | Retrospective cohort study, 1 October to 30 November 2012 1713 adult patients transported to ED by emergency ambulance | NEWS or qSOFA | Clinician judgement | 48-hour and 30-day mortality, ICU admission combined 30-day mortality ICU admission within 48 hours |
Data extraction
Methodological quality was assessed using the Critical Appraisal Skills Programme (CASP) checklist for cohort studies (CASP, 2018). Data on cohort characteristics, cohort outcomes, study aims, risk of bias and study methods and results were extracted from the eligible studies. A data-extraction form was adapted from the Cochrane Collaboration's data collection form. The Cochrane risk of bias tool was used to assess study bias as part of the data extraction process (Brown et al, 2017).
The following information was collected for each study included in this systematic review:
- Study setting, to ensure eligibility, that is, that the study was related to the pre-hospital setting
- The use of an EWS as well as details of any comparators, such as clinical judgement
- Outcome definition, type of measurements used in the study, time points measured and time points reported
- Outcomes, results, unit of analysis and statistical methods used
The results of the data extraction forms and the Higgins I2 test determined whether a meta-analysis was viable. There was significant heterogeneity, and, in line with the Cochrane Handbook for Systematic Reviews (Higgins et al, 2019), a meta-analysis was ruled out as, although most studies had similar primary outcomes (ICU admission or 30-day mortality), they used different study designs and selection criteria, and there were important differences in the key components of the intervention.
Results
Figure 1 shows the flow of studies through the review process and reasons for exclusion from this systematic review. A screening form (Appendix 2) was developed to determine whether the studies met the primary inclusion criteria. Finally, five studies met the selection criteria and were included in the review.
Figure 2. Flow chart of the study selection process
Study characteristics
Of the five studies that met the primary inclusion criteria, four studies were retrospective cohort studies and one was a prospective study. The number of patients totalled 3517 (range, 104–1713). Some 49% of the patients were from a single study. Studies were summarised based on the ability of EWS to identify a patient with critical illness; the ability of EWS to predict the risk of adverse outcome (ICU or critical care admission and 30-day mortality); and whether the use of EWS in pre-hospital settings improved patient outcomes.
All the studies had different numbers of study participants, and the duration of each study varied. Table 2 describes the key characteristics of each study. Four studies used NEWS as the EWS (Shaw et al, 2017; Abbott et al, 2018; Silcock et al, 2019; Hargreaves et al, 2020), and one study used a pre-alert tool (Booth and Bloch, 2013). Ambulance crews were the primary source of NEWS scores in all five studies. Four studies used patient observations documented by ambulance crews to calculate a retrospective NEWS score (Shaw et al, 2017; Abbott et al, 2018; Silcock et al, 2019; Hargreaves et al, 2020). The remaining study looked at pre-alert requirement determined by a consultant physician blinded to outcome compared with ambulance crew alerts (Booth and Bloch, 2013). No studies reported modifying or adapting the interventions.
The primary outcome in three studies was ICU or critical care unit escalation within 48 hours (Abbott et al, 2018; Silcock et al, 2019; Hargreaves et al, 2020); two of these studies also looked at 30-day mortality (Silcock et al, 2019; Hargreaves et al, 2020), and one looked at length of hospital stay (Abbott et al, 2018). One study examined hospital discharge disposition (Shaw et al, 2018), and the final study determined pre-alert sensitivity and specificity (Booth and Bloch, 2013).
The primary target group across all studies was adult patients. One study did not clearly identify its target age group, but the author of this review found no evidence of data involving paediatric patients from the study (Booth and Bloch, 2013). Two studies had a solely urban population (Shaw et al, 2017; Abbott et al, 2018), and three studies had a mix of urban and rural populations (Booth and Bloch, 2013; Silcock et al, 2019; Hargreaves et al, 2020).
All five studies used clinician judgement as a comparator. Clinician judgement is considered usual care in patient assessment and was often not explained in the studies. Only Booth and Bloch (2013) explicitly referenced clinician judgement as a comparator and detailed the process of comparison.
Methodological quality
The level of evidence was generally low, and none of the studies included were RCTs. However, no study was excluded due to poor-quality study methodology. Table 3 summarises the results of the CASP cohort study checklist. The results of all five studies were determined to be valid. It was determined that the cohorts in each of the five studies had been recruited in an acceptable way so as to ensure the generalisability of the results.
Table 3. Quality appraisal summary
| Study | Are the results valid? | Was the cohort recruited in an acceptable way? | Confounding factors | Implications for clinical practice |
|---|---|---|---|---|
| Booth and Bloch (2013) | Yes | Yes | Age: mean age was 65 yearsGender: 55 men and 49 women |
|
| Abbott et al (2018) | Yes | Yes | Age: mean age of all patients was 61 years, mean age of ambulance patients was 67 years Gender |
|
| Hargreaves et al (2019) | Yes | Yes | Age: mean age was 79 years |
|
| Shaw et al (2017) | Yes | Ambulance and ED records were electronically linked for a total of 47 964 patients: 28 757 (60%) were discharged, 18 957 (39.5%) admitted to a ward, 164 (0.3%) were admitted to ITU and 86 (0.2%) diedUsing systematic sampling, the authors aimed to include 100 patients in each disposition category. In total, 287 patients were included | Age: mean age was 63 yearsGender |
|
| Silcock et al (2019) | Yes | Yes | Age: mean age was 58 years |
|
Risk of bias
Table 4 presents the results of the risk of bias appraisal. Detection bias was either high risk or unclear in all studies. Booth and Bloch (2013) and Hargreaves et al (2020) were determined to have a high risk of selection bias, whereas the other three studies were considered to have a low risk. Reporting and non-reporting bias were assessed to be low risk in all studies. The risk of other bias was determined to be low in three studies and high in two.
Table 4. Risk of bias
| Study | Selection bias | Detection bias | Reporting bias | Nonreporting bias | Other bias |
|---|---|---|---|---|---|
| Booth and Bloch (2013) | High risk | High risk | Low risk | Low risk | Unclear |
| Abbott et al (2018) | Low risk | Unclear | Low risk | Low risk | Unclear |
| Hargreaves et al (2019) | High risk | Unclear | Low risk | Low risk | Low risk |
| Shaw et al (2017) | Low risk | High risk | Low risk | Low risk | Low risk |
| Silcock et al (2019) | Low risk | Unclear | Low risk | Low risk | Low risk |
EWS for rapid identification and diagnosis of critically ill patients
This review identified two studies that examined the identification of critical illness, both including patients with suspected infection (Booth and Bloch, 2013; Hargreaves et al, 2020). Both studies are considered low-quality studies and used different EWS tools to identify sepsis and measured different outcomes. Booth and Bloch (2013) used a pre-alert guidance tool to prompt an alert to the emergency department before the arrival of critically ill patients. The pre-alert tool proved to have superior sensitivity and specificity to that of ambulance crew decisions. The tool alerts had a high sensitivity and a moderate specificity compared with ambulance crew decisions to alert. Although the pre-alert tool was not used to specifically identify sepsis, it does contribute towards developing an understanding as to whether a well-formulated EWS improves the identification and diagnosis of patients with suspected sepsis. The study by Hargreaves et al (2020) used NEWS to screen for suspected sepsis. The diagnostic value for NEWS≥5 measured pre-hospital for 30-day mortality in patients with suspected sepsis had a high sensitivity but a low specificity. A NEWS≥5 pre-hospital and at admission to ED also had a high specificity and a low specificity. Lastly, a NEWS≥5 pre-hospital, at admission to ED and at admission to a ward had a modest sensitivity and specificity.
EWS for predicting the risk of an adverse outcome
Four cohort studies determined the prognostic effects of NEWS on adverse outcomes (Shaw et al, 2017; Abbott et al, 2018; Silcock et al, 2019; Hargreaves et al, 2020). ICU or critical care unit admission and 30-day mortality were the outcomes of primary interest. The single-centre study by Abbott et al (2018) determined the association between NEWS and critical care unit admission or death within 48 hours. Pre-hospital NEWS was assessed as being associated with the primary outcome. NEWS on admission to hospital was also determined to be associated with the primary outcome. Patients with a pre-hospital NEWS≥7 had a four times higher risk of being admitted to the critical care unit or dying compared with patients with a pre-hospital NEWS≤4.
Hargreaves et al (2020) analysed the effect of NEWS on ICU admission and 30-day mortality. NEWS was measured at three points: pre-hospital, on arrival at ED and on admission to ward. Pre-hospital NEWS≥5 was associated with a 19.6% mortality rate within 30 days compared with the 11.9% mortality rate associated with a pre-hospital NEWS<5. If NEWS≥5 persisted after arrival at ED, then it was associated with a 22.1% mortality rate compared with the 10.2% rate associated with NEWS<5. A NEWS≥5 pre-hospital, at arrival to ED and on admission to ward was associated with a mortality rate of 32.1%, whereas NEWS<5 had a mortality rate of 14.3%. In this study, ICU admission was 6%. In the group with NEWS≥5 pre-hospital and at arrival in ED, ICU admission increased to 8.3% versus 1.8% for pre-hospital NEWS≥5.
Shaw et al (2017) examined the association between pre-hospital NEWS and hospital disposition. All NEWS scores were calculated from the initial set of observations taken by the ambulance crew. Of patients discharged from the ED, 81% were classed as low risk (NEWS=0–4), 16% as medium risk (NEWS=5–6) and 3% as high risk (NEWS≥7). Of patients admitted to a ward, 65% were classed as low risk, 25% as medium risk and 10% as high risk. Some 14% of those patients admitted to the ICU were classed as low risk, 33% as medium risk and 53% as high risk. Of the patients who died, 13% were classed as low risk, 24% as medium risk and 63% as high risk. Patients discharged from ED were 128 times more likely to be in the low-risk category than those who died in the ED. This study showed a clear pattern—an increasing severity of risk level is associated with a poorer outcome.
Silcock et al (2019) compared the ability of NEWS and quick Sepsis-related Organ Failure Assessment (qSOFA) score to predict adverse outcomes in a pre-hospital population. The NEWS risk category was found to be significantly associated with 30-day mortality, which was 2% for patients in the medium-risk group (NEWS=5–6) versus 13% for those in the high-risk group (NEWS≥7). The NEWS risk category was also associated with ICU admission within 48 hours, which was 3% for the medium-risk group and 5% for the high-risk group. The qSOFA score was significantly associated with 30-day mortality and ICU admission within 48 hours. For both EWSs examined in this study, the risk of mortality and ICU admission was higher as the score increased.
Impact on patient outcomes
This systematic review was unable to determine if the use of an EWS in the pre-hospital setting definitively improved patient outcomes, as none of the included studies directly compared the effectiveness of an EWS with clinical judgement in improving patient outcomes.
Discussion
Numerous studies have been conducted in relation to the use of EWSs in an in-hospital setting, but very few studies have examined their use in the pre-hospital setting. This systematic review is a comprehensive appraisal of the evidence of the use of an EWS in a pre-hospital setting. All studies included in this review were conducted in the ambulance service; no studies were found in GP, community or mental health settings. The five studies included in this review had substantial heterogeneity in terms of the EWS tool used, population, how the studies were constructed and their outcomes, which prevented a reliable meta-analysis from being conducted. Nevertheless, the narrative synthesis of the evidence enabled a number of observations.
The narrative synthesis suggested that the use of an EWS was helpful in identification of patients with critical illness. Sepsis is a time-critical illness, and, in the pre-hospital setting, identifying sepsis early may be beneficial to patient outcomes, since it allows therapeutic interventions to be administered earlier than they would otherwise have been (Gaieski et al, 2010; Ferrer et al, 2014). An EWS tool can also be used to guide pre-hospital clinicians in their decision to admit to hospital or to pre-alert before arrival in ED. Pre-alerting before arrival in the ED has been found to almost halve the time of in-hospital treatment for patients (Studnek et al, 2012). Of the two studies in this review that examined the efficacy of EWS in the identification of patients with critical illness (Booth and Bloch, 2013; Hargreaves et al, 2020), only one study included a pre-alert to ED. Booth and Bloch (2013) showed that the pre-alert tool showed superior sensitivity and specificity to advise a pre-alert compared to ambulance crew decisions. EWSs are used to trigger ED pre-alerting (Williams et al, 2016).
The use of an EWS appears to predict the likelihood of an adverse outcome. In the four studies that examined patient outcomes (Shaw et al, 2017; Abbott et al, 2018; Silcock et al, 2019; Hargreaves et al, 2020), a high NEWS score corresponded with increased ICU or critical-care admission and mortality. The RCP (2017) highlighted that a NEWS score ≥5 is a key trigger point for an urgent clinical alert and response, and, in patients with a known or suspected infection, sepsis should be considered. The UK Sepsis Trust (2019) advocated using a NEWS of ≥3 to trigger a screening for sepsis. NEWS has been demonstrated to be effective at determining the risk of serious clinical deterioration and acute mortality (Smith et al, 2013). A moderate (5 or 6) to high (≥7) pre-hospital NEWS was predictive of adverse outcomes in all included studies. The studies suggested that patients with a NEWS of 0 are unlikely to deteriorate and patients with a high score (NEWS ≥7) are more likely to deteriorate and require ICU or critical care unit admission or die. A study conducted in Finland, which was not included in this systematic review due to not having been conducted in the UK, determined that the high-risk category for pre-hospital NEWS had a much higher association with 1-day mortality than the medium- or low-risk NEWS category (Hoikka et al, 2018). However, this was not as noticeable an effect for 30-day mortality (Hoikka et al, 2018). This suggests that pre-hospital NEWS would be a beneficial tool in patient assessment, as it would allow for earlier recognition of those patients most at risk of deterioration or death and will allow for earlier treatment and escalation of care for these at-risk patients.
Limitations
A major limitation of this systematic review was the substantial heterogeneity among the included studies, and the low number of studies included in the review prevented a meta-analysis of study findings being conducted. The studies themselves were also a limitation, since no RCTs were included. RCTs are considered the gold standard for research evidence, and, hence, their absence creates a gap in the quality of the research evidence. None of the studies included data from GPs, advanced nurse practitioners (ANPs) or community nurses. Therefore, the findings of this review may have limited generalisability beyond ambulance staff. Evidence from large multi-centre studies is required before a pre-hospital EWS can be implemented.
Comparison with previous literature
The author is aware of two systematic reviews that examined the use of EWS in the pre-hospital setting (Williams et al 2016; Patel et al, 2018). These studies differed from this review, as they restricted the pre-hospital setting to studies conducted in the ambulance settings, and one review (Williams et al, 2016) included a wider range of EWSs that were excluded from this review. One study from this review (Booth and Bloch, 2013) was included in the review by Williams et al (2016), and one other study (Shaw et al, 2017) was included in the review by Patel et al (2018). Therefore, there is minimal overlap between these previous reviews and the present systematic review.
Conclusion
This systematic review demonstrated the successful use of EWSs in the pre-hospital setting in predicting patients most at risk of deterioration and as a useful tool in assisting decision-making. However, four of the five studies included used NEWS as the EWS, and, as such, the results cannot confidently be said to be generalisable. The pre-hospital use of EWS is an emerging theme in community practice but is a controversial topic, as the evidence base for the benefits of EWS is almost entirely from an acute setting. Very little research evidence is available for the pre-hospital use of EWS. Therefore, it is premature to conclude that the use of EWS in the pre-hospital setting improves the outcomes of patients with suspected sepsis.
KEY POINTS
- Sepsis is a life-threatening complication that is often missed in the pre-hospital setting
- Early detection of sepsis may improve patient outcomes
- The use of early warning scores (EWSs) in the pre-hospital setting assist in identifying patients with sepsis
- Pre-hospital EWSs are predictive of patient outcomes
- Although this review shows that the use of EWSs pre-hospital may be beneficial, there is not enough evidence to recommend their implementation in pre-hospital settings
CPD REFLECTIVE QUESTIONS
- What are the main signs of deterioration in those with sepsis?
- What are the challenges to using early warning scores in pre-hospital settings?
- Why is it easy to miss the signs of sepsis in pre-hospital settings?