The burden and prognostic significance of suspected sepsis in the prehospital setting: A state‐wide population‐based cohort study

Despite high in‐hospital mortality, the epidemiology of prehospital suspected sepsis presentations is not well described. This retrospective cohort study aimed to quantify the burden of such presentations, and to determine whether such a diagnosis was independently associated with longer‐term mortality.

• The burden of prehospital suspected sepsis in the Australian setting is significant with an age-adjusted incidence rate of 65 cases per 100 000 person-years, greater than that of ST-elevation myocardial infarction.• Paramedics were able to identify a group of patients at high-risk of poor longer-term outcomes, with substantially higher mortality rates than patients not suspected to have sepsis.
• Patients with prehospital suspected sepsis received more frequent intervention in the prehospital setting and also faster time-to-treatment in the ED.

Introduction
Sepsis is defined as a dysregulated host response to infection that causes life-threatening organ dysfunction, 1 and has an estimated annual incidence of up to 300 per 100 000 people globally. 2,3Patients with sepsis are at high risk of dying, with reported mortality of up to 32.2%. 4,5Moreover, 23.8% of septic patients require intensive care unit (ICU) admission, at some point during their hospital stay. 4Subsequently, the total estimated cost of sepsis in Australia is $4.7 billion (AUD) annually. 6n recognition of the significant mortality and morbidity burden from sepsis, multiple state-wide hospital-based initiatives have now been developed (Think Sepsis: Act Fast VIC, 7 Sepsis Kills NSW, 8 Sepsis Breakthrough Collaborative QLD 9 ).These are aimed at improving sepsis recognition and outcomes in hospitalised patients, through streamlined pathways and bundles of care, including early antibiotics.
5][16] Despite this, the burden of prehospital suspected sepsis in the Australian setting is not well described.
Given the paucity of contemporary epidemiological data in this area, this retrospective cohort study aimed to quantify the burden of such presentations to emergency medical services (EMS) in Victoria, Australia, and to determine its association with longer-term mortality.As there is no single definitive test for sepsis, and often limited information is available in the prehospital setting (e.g.additional investigations and/or collateral history), we have used 'prehospital suspected sepsis' to reflect scenarios where the paramedic considered sepsis as a likely significant contributor to the patients presentation, as recorded in their primary or secondary assessment (see Section 2).The primary hypothesis was that prehospital suspected sepsis would be common, and that such a diagnosis would be associated with a high risk of death.

Study design and setting
This was a retrospective populationbased cohort study examining all adult prehospital presentations in Victoria, Australia, between January 2015 and June 2019.Data were extracted from a linked dataset based on data from: (i) the Ambulance Victoria (AV) electronic medical record, which houses data pertaining to pre-hospital patient interactions, (ii) the Victorian Emergency Minimum Dataset (VEMD) which collects information detailing presentations at Victorian public hospitals with designated EDs, and (iii) the Victorian Death Index (VDI).The data linkage methodology has been described previously. 17inked data were included where patients had a matched ED presentation record within 48-h of EMS attendance.The present study was approved by the Monash University Human Research Ethics Committee (No. 11681).

Selection of participants and data extraction
All adult patients (≥18 years) attended by an emergency ambulance during the study period were eligible for inclusion.Cases were excluded if their respective hospital data could not be linked, as were interhospital transfers.Patients were categorised into the diagnostic groups of 'Prehospital Suspected Sepsis' and 'Non-sepsis' presentations, based on the primary and secondary assessment recorded by paramedics.Patients allocated to the 'prehospital suspected sepsis' group had 'sepsis*' recorded in either of these domains in the AV electronic medical record.All other patients were allocated to the 'nonsepsis' group.Paramedics derive the recorded primary and secondary assessment from the patient's clinical history, on-scene assessment and vital signs.Although no specific protocol was in place during the study period, guidance on considering prehospital sepsis included the following:

Statistical analysis
Statistical analysis was performed using Stata Statistical Software version 17.0 (StataCorp 2021, USA).Incidence rates per 100 000 personyears were calculated using age and sex-specific population data from the Australian Bureau of Statistics 18 and the total number of patients meeting the study inclusion criteria.The 95% CIs for incidence rates were calculated with the assumption that the number of observations followed a Poisson distribution, and the trend significance was determined using the Cochrane-Armitage test.
Patient characteristics for the total linked cohort are presented according to diagnostic group (prehospital suspected sepsis vs non-sepsis).Continuous data are summarised using mean AE standard deviation (SD), or median [interquartile range (IQR)], according to whether they were normally distributed.Categorical data are reported as frequency (%).
Using the entire cohort, multivariable logistic regression was used to determine the adjusted odds ratios (OR, 95% CI) for 6-month mortality with pre-hospital exposure variables.Covariate selection for the models was based on clinical relevance and comprised age, sex, initial vital signs (systolic blood pressure, heart rate, temperature, GCS, respiratory rate, peripheral oxygen saturation), hospital location, administration of IV fluids, number of co-morbidities and paramedic assessment of suspected sepsis.All statistical tests were two-tailed, and considered significant if the P value was <0.05.

Results
A total of 1 416 511 AV encounters over the study period (1642 days) could be linked to a VEMD patient record.One hundred ninety-eight thousand four hundred sixty-four records were subsequently excluded because they were either younger than 18 years of age (n = 144 740), or had undergone inter-hospital transfer (n = 62 140).A total of 1 218 047 patients were included in the final analysis, with 17 981 suspected as having prehospital sepsis (Fig. 1).This represents a crude incidence rate of 11 patients per day, which when age adjusted to the standardised population, equates to 64.7 (95% CI 63.6-65.7)cases per 100 000 person-years, with a higher incidence with increasing age, and in males (Table 1).
A summary of demographic, transport, physiological, admission and mortality data, for prehospital suspected sepsis and non-sepsis patients, are provided in Table 2.Those with prehospital suspected sepsis were older, more frequently men, with greater physiological derangement.More often an IV cannula was inserted prehospital, and fluid administered.Such patients were more frequently triaged as 'resuscitation/emergency' cases on arrival at ED, with shorter time to treatment in ED.Crude in-hospital mortality was 6.5-fold higher in the prehospital suspected sepsis group (11.8% vs 1.8%).By  The results of the multivariable logistic regression analysis are presented in Table 3.After controlling for age, sex, hospital location, physiological parameters and number of comorbidities, a diagnosis of prehospital suspected sepsis was associated with 35% higher odds of death at 6 months.

Key findings
This is the first paper to explore the burden and longer-term prognostic significance of prehospital suspected sepsis in the Australian setting.Here, we identified an incidence rate of approximately 65 cases per 100 000 person-years, which is significantly greater than that of ST elevation myocardial infarction 19 and traumatic spinal cord injury. 20Indeed, our data demonstrate that interventions are frequently provided to this patient cohort, with the majority receiving intravenous cannulation, and fluid administration.Moreover, this group often received expedited care in the ED.Despite such intervention, patients with prehospital suspected sepsis manifest a crude inhospital all-cause mortality rate 6.5 times that of non-septic patients, whereas $23% were deceased within 6 months.Critically, when adjusted for age, sex, number of comorbidities, physiological parameters and location of receiving ED, a diagnosis of prehospital suspected sepsis was independently associated with increased odds of 6-month allcause mortality.

Relationship to previous literature
To date, there are little data available on the incidence and prognostic implications of prehospital suspected sepsis.In Australia, our prior analysis of ARISE trial enrolments in Victoria, identified that 70% of patients were transported to the ED by paramedics. 21In the USA, EMS personnel also care for a large proportion of patients with severe sepsis.In Washington State over 70% of septic patients were transported to ED via ambulance, 22 and one retrospective cohort study identified a prehospital incidence of sepsis of 3.3 per 100 EMS encounters. 23As a result of the difference in metrics, this incidence cannot be compared directly to the present study, but regardless it demonstrates that sepsis is consistently managed in the prehospital setting internationally.Moreover, the higher incidence of prehospital suspected sepsis reported in males in our cohort, is also consistent with independent local data on hospital treated sepsis. 24atients with prehospital suspected sepsis had significantly higher inhospital, and 6-month all-cause mortality compared to those without this diagnosis.This appears consistent with data from recent Australian and international sepsis trials that report 90-day mortality ranging from 18% to 28%. 21,25,26More specifically, with a 6-month all-cause mortality rate greater than 20%, paramedics in Victoria were able to identify a cohort of patients at highrisk of poor longer-term outcomes, independent of existing comorbidity and physiological derangement.Moreover, it appears these patients experienced shorter time to treatment in the ED, which in the setting of sepsis, is considered critical to achieving improved outcomes. 22,27- 30Indeed, the 2021 surviving sepsis campaign advises antibiotic administration within 1 h of recognition for

Study implications
The present study implies that prehospital suspected sepsis constitutes a significant proportion of ambulance encounters and that such a diagnosis is associated with increased mortality rates over the longer-term.Moreover, this cohort has relatively high rates of prehospital intervention (such as intravenous fluid administration), which carries over to more rapid care in the ED.Independent of age, physiological parameters, and number of comorbidities, a paramedic diagnosis of prehospital suspected sepsis is associated with greater 6-month mortality, suggesting this group of healthcare providers can accurately identify a highly vulnerable patient cohort.Finally, given the high mortality observed with prehospital suspected sepsis, it would seem logical to consider the potential impact of enhanced bundles of care (such as point of care lactate, blood culture collection and IV antibiotic administration) in this patient population.

Strengths and limitations
The strengths of this analysis relate to the utilisation of a large, linked dataset.To date, this is the largest analysis of prehospital suspected sepsis presentations in Australia.As such, for the first time we were able to perform a robust statistical analysis that defines the incidence and prognostic significance of this diagnosis in the Australian prehospital population.
Given the observational nature of the present study, however, there are several limitations.This was a retrospective study that utilised diagnostic coding to identify prehospital suspected sepsis cases.As a result, there is the potential for inaccuracies secondary to coding errors or missed cases that could result in missing data, albeit at random.Moreover, the dataset likely included re-presentations to AV, and whereas these are expected to represent a low proportion, would have inflated the reported incidence.In addition, given the project utilised linked data, prehospital suspected sepsis cases may have been excluded if their respective hospital data could not be identified.Despite this, our study clearly demonstrates that prehospital clinicians are able to identify patients with a significantly elevated risk of mortality.
The present study was isolated to the Victorian public hospital system and only included patients arriving at hospital via ambulance or those that had been assessed by the Victorian ambulance service in the 48 h preceding their admission.Subsequently, the results may not reflect national trends, populations presenting to private hospitals, or patients presenting by private vehicle or on foot.Although the primary outcome was all-cause mortality, we do not report any functional outcome data and hence we are unable to define the morbidity burden associated with prehospital suspected sepsis.
It is important to note that the incidence reported in the present study was based on the prehospital diagnosis of suspected sepsis rather than hospital diagnosis.This was chosen to better reflect the burden of cases managed as suspected sepsis by paramedics, as this would likely represent the cohort of patients who would receive any future sepsistargeted prehospital intervention.
Given the differences in resources between the prehospital and hospital setting, there is the potential for diagnostic inconsistency.Therefore, in the future, the linked dataset in the present study will be utilised to explore the accuracy of prehospital sepsis diagnoses.

Conclusion
The burden of prehospital suspected sepsis in the Australian setting is significant and manifests a 6-month mortality of greater than 20%.Independent of vital signs and preexisting comorbidity, prehospital suspected sepsis is associated with a higher mortality risk.Given the poor outcomes observed in this cohort, such patients may benefit from enhanced bundles of care, including prehospital antibiotic administration.

TABLE 1 .
6 months, 22.6% of those patients suspected Figure 1.Study flowchart.Sepsis cases had either a final primary or secondary assessment of 'sepsis*' recorded in the ambulance electronic record.All other patients were allocated to the 'non-sepsis' group.CAD, Computer Aided Dispatch; EMG, Emergency; EMS, Emergency Medical Services; NE, Non-Emergency; VACIS, Victorian Ambulance Clinical Information System; VEMD, Victorian Emergency Minimum Dataset.Incidence of ambulance attendances in Victoria for suspected sepsis linked with an ED presentation over the study period

TABLE 2 .
Demographic, illness severity, treatment and outcome data for patients with prehospital suspected sepsis and a linked ED presentation †Includes patients who died in ED.IV fluids include Normal Saline, Potassium Bromide and Glucose 5%.IV, intravenous; IQR, interquartile range; ED, Emergency Department; qSOFA, Quick Sepsis Related Organ Failure Assessment.

TABLE 3 .
Multivariable logistic regression model, predictors of 6-month allcause mortality for linked cohort