Introduction

The COVID-19 pandemic has had a profound impact on healthcare systems across various levels. In Brazil, the Emergency Medical Service (EMS), known as Serviço de Atendimento Móvel de Urgência (SAMU), plays a crucial role. SAMU is a public service, universally accessible and free for all residents, and serves approximately 85.6% of the Brazilian population1. In Manaus, the capital of the Amazonas state, Western Brazilian Amazon, the EMS coverage reaches over two million people in urban, rural, and riverside areas2. The SAMU corresponds to the mobile pre-hospital component of the public emergency care system1.

SAMU and emergency medical services in other countries were challenged by a sudden change in the profile of their patients. There was an impact on the severity of patients, refusal of care, and assistance provided at the scene. This adaptation of the pre-hospital service during the pandemic must be understood, to ensure greater resilience and a more orderly response3.

Accurate prehospital judgment is vital to reduce undertriage and overtriage, both of which are detrimental to patient outcomes and health systems4. In limited resource settings and situations of overwhelmed health services, such as in disease outbreaks, investigating the inherent capability of the prehospital EMS to timely diagnose and predict prognosis could lead to improvement in decision-making and consequent outcomes. The innovative aspect of this study lies in its focus on the untapped potential of EMS data. In many health systems, EMS data is underutilized, primarily serving immediate care needs rather than contributing to broader clinical decision-making frameworks. By demonstrating the predictive value of prehospital data, this research advocates for a more integrated approach, where EMS plays a pivotal role not only in immediate patient care but also in ongoing clinical management and strategic health system planning.

Ultimately, the findings of this study could inform protocols and guidelines for EMS operations, ensuring that prehospital evaluations contribute to improved patient outcomes during health crises. Therefore, this study aimed to analyze the data obtained in the prehospital evaluation of patients with SARS during the initial response to the COVID-19 pandemic and clinical outcomes, such as mechanical ventilation, hospital discharge, and death.

Methods

Study design and databases

We performed a retrospective analysis of individuals with SARS throughout the first peak of the COVID-19 pandemic in Manaus, Brazil, from January to June 2020. We analyzed data from two different databases: the EMS database and the Brazilian Health Surveillance System database (SIVEP-Gripe) from Manaus, which consists of data from patients admitted to hospital with SARS or whose death was due to SARS, regardless of hospital admission.

The local EMS (SAMU Manaus) database consists of electronically stored medical records from pre-hospital emergency care. All emergency calls reach a medical coordination center and are entirely recorded according to Brazilian legislation5. Times (patient time of call, time from dispatch to site arrival, time to hospital arrival, etc.) are registered in detail. The data obtained include address information, a first medical interview, radio dispatch order, return call after local assessment by the ambulance team, final medical guidance/prescription, and decision on hospital destination. When necessary, paper-based medical records filled by the ambulance teams are also available, as a complement to the electronic medical records.

The SIVEP-Gripe, which was developed by the Ministry of Health of Brazil, is a system designed for the collection, storage, and analysis of data related to the epidemiological surveillance of influenza and other respiratory viruses. Its main objective is to monitor and analyze flu cases, identifying patterns and trends to enhance the response to acute respiratory diseases. Access to the SIVEP-Gripe database is typically restricted and controlled to ensure the confidentiality of information and compliance with ethical and safety standards.

To access both databases, specific procedures established by health authorities must be followed, such as requesting authorization, meeting ethical and safety requirements, and, in some cases, entering into confidentiality agreements. The availability of access may vary depending on the purpose of the research and the access control policies established by the responsible institutions.

SAMU Database: Occurrence; name; sex; age; date_time; dispatch_time; Action; destination; destination_name; reason; type; nature; neighborhood; arrival_time; dispatch_time; transport; cancel_type; municipal_zone; service_delta; service_month; attended; epi_date. Sivep—Influenza Database: notification_date; symptom_start_date Unit; name; sex; age_in_years; neighborhood_name; fever; cough; sore_throat; dyspnea; respiratory_discomfort; saturation; diarrhea; vomiting; risk factor; puerpera; cardiopathy; other disease; asthma; diabetes; pneumopathy; renal; obesity; other morbidity; vaccine; vaccine dose date; mother vaccine; mother vaccine date; breastfeeding; single dose date; first dose date; second dose date; antiviral; antiviral type; other antiviral; antiviral date; hospitalized; hospitalization date; hospital unit name; ICU; ICU entry date; ICU exit date; ventilatory support; sample; collection date; PCR result; PCR date; final classification; other classification; criterion; evolution; evolution date; closure date; observation; analysis result date; analysis result; COVID vaccine; first COVID dose; second COVID dose; booster dose.

The study was approved by the local IRB Committee, which is named CEP (Comitê de Ética em Pesquisa) da Fundação de Medicina Tropical Dr Heitor Vieira Dourado, with Certificate of Presentation of Ethical Appreciation No. 5 60491122.3.0000.0005, which waived the need for informed consent due to the retrospective nature of the study. All stages of the research were performed following the Declaration of Helsinki.

Study location and participants

All EMS calls occurring during the period of the study in Manaus, Brazil, were analyzed. Trained EMS physicians assess every call and classify the request into different syndromic diagnoses before dispatching emergency vehicles when indicated. All patients whose chief complaints were compatible with SARS in the period of the study were included. SARS was assigned to patients presenting with acute respiratory syndrome (at least two of the following: fever, chills, sore throat, headache, cough, coryza, taste/smell disorder plus at least one of the following: dyspnea, chest pressure, cyanosis, or pulse oximetry < 95% in room air), as defined by the SAMU triage system.

Those in need of medical assistance were transported by the prehospital EMS to a hospital emergency department and were eventually included in the second database, with their in-hospital data. Name, gender, age, home address, and date of system input were used to match patients’ details from the two different databases.

The patient outcome was evaluated in two moments, prehospital and in-hospital. Data obtained in the prehospital setting were age, gender, address, hospital destination, and clinical variables such as blood pressure, body temperature, pulse oximetry (first reading, while breathing room air), heart rate, and respiratory rate. Blood pressure and heart rate values obtained during the initial assessment were used to calculate the shock index (heart rate / systolic blood pressure) and the modified shock index (heart rate / mean arterial pressure5.

We also examined the association between the prehospital clinical variables and the following in-hospital outcomes: length of hospital stay, length of stay in the intensive care unit (ICU), need for mechanical ventilation, and death. Data for these outcomes were extracted from the second database, which included in-hospital care information.

Data processing and statistical analysis

The original SAMU database contained 45,780 records. Following the application of filters, 352 were excluded due to a lack of name, 8251 for missing information, 1546 due to a lack of age, 157 owing to time discrepancies, and 1675 were canceled records, resulting in a filtered SAMU dataset with 33,799 records. The original SIVEP-Gripe dataset comprised 28,112 records. After filtering, 15,718 were removed due to data outside the study period, resulting in a filtered SIVEP dataset with 12,394 records.

The data linkage was performed based on variables that were present in both databases, and the "Levenshtein" method was employed for string comparison. Linked record pairs were obtained by applying a pre-specified minimum threshold, in this case, a 100% match. The time difference between linked records was calculated using the linkage date for EMS and the notification date for SIVEP. Records with missing values were excluded. Those records with a time difference of less than 30 days between the EMS record and SIVEP were retained. After merging both databases, only variables with full completeness were used for outcome analysis.

For data presentation, demographic and clinical data from the EMS database were presented in terms of percentages and means with respective deviations or medians accompanied by their interquartile ranges, depending on data distribution, which was calculated using the Shapiro–Wilk test. For the comparison of proportions, chi-square or Fisher's exact tests were used. For comparison of means, the T and median tests were performed and, when applicable, the Wilcoxon Mann–Whitney test or the Kruskal Wallis test. The linear relationship between two quantitative variables was calculated with the Pearson correlation coefficient. A multivariate model was carried out to analyze the relationship between the variables in addition to the correlations between clinical data such as mechanical ventilation, admission to the ICU, and those who were discharged/deceased. All analyses were performed using the Stata 16.0 software. The R software (version 4.3.0) was used to merge the EMS and the SIVEP-Gripe databases.

Results

The combination of the two databases yielded a total of 1.190 patients, who received a first EMS response and were later admitted to a hospital with SARS and had data on clinical outcomes of interest available.

Table 1 displays the main characteristics of the population of the study. Patients were predominantly male (754, 63.4%), with a median age of 66 (IQR: 54–78) years. Death was more common among patients in need of respiratory support, especially in the invasive ventilation group (262/287; 91.3%) (p < 0.001). In addition, mechanical ventilation was more common among elderly individuals (p < 0.001) and in patients with a longer course of disease (p < 0.001). Prehospital pulse oximetry at the EMS transport was not associated with the need for mechanical ventilation in this population (Table 2).

Table 1 Data from patients with SARS during prehospital EMS assistance in Manaus, January to June 2020 in median (IQR).
Full size table
Table 2 Variables associated with mechanical ventilation.
Full size table

Patients admitted to ICU had a greater chance of dying when compared to non-ICU admitted patients (p < 0.001), closely related to invasive mechanical ventilation, which was more common in ICU (p < 0.001). Patients in ICU were also older (p = 0.003) and had longer hospital stay (p < 0.001) (Table 3).

Table 3 Variables associated with ICU admission.
Full size table

Mortality was associated with mechanical ventilation (p < 0.001), ICU admission (p < 0.001) and older age (p < 0.001). In addition, death was associated with lower pulse oximetry obtained during prehospital assessment (p = 0.025). Patients who died had shorter length of both ICU and total hospital stay (p < 0.001) (Table 4). We applied a regression with the main outcome death and included age (years), initial and final saturation, systolic blood pressure (SBP) and diastolic blood pressure (DBP), shock and SAMU response time. We found that the DBP (p < 0.05) and SAMU response time are significantly associated with a reduction in the probability of death. For each additional minute, the probability of death decreases by approximately 0.98% (p < 0.05) (Table 5).

Table 4 Variables associated with hospital discharge and death.
Full size table
Table 5 Multivariate regression to identify determining factors in patient deaths.
Full size table

Discussion

Our results show that clinical data from prehospital care might be useful in anticipating certain clinical outcomes and could be detrimental to decision-making in times of pandemics. Recently, few attempts have been made to pursue relevant variables that could potentially predict such outcomes. The surge of numerous respiratory emergencies related to COVID-19 worldwide has cleared the way for the investigation of pulse oximetry as a strong candidate. In this study, prehospital lower pulse oximetry values were associated with higher in-hospital mortality, which reinforces its use for such moments. Similar results have been described while investigating estimated hypoxemia through pulse oximetry and hospital outcomes during the recent pandemic6,7. The variance in the number of total patients is explained by the unavailability of specific data for each type of analysis.

Even though oximetry has its inherent limitations, it is important to reckon its low cost, and non-invasiveness, making it very simple for bedside and home situations. This is especially important for COVID-19, in which silent hypoxemia has been shown to occur, potentially obscuring severe patients during clinical examination8. Besides, research has shown that pulse oximetry might also be effective in prehospital non-emergency scenarios. A program remotely monitoring people over 65 years with confirmed COVID-19 through pulse oximetry found a 2% mortality reduction for every 10% increase in coverage of the program9.

Besides pulse oximetry, additional examination through capillary refill time and the shock index might add up to clinical data obtained in the prehospital setting. However, in our study, neither the shock index nor its modified version have yielded significance to support their use as predictors of hospital outcomes in such groups of individuals. This association has been sought before, incorporating laboratory assessment (lactatemia) to the shock index, yet results were not associated with ICU admission or 30-day mortality10.

At the time, there was an insufficient availability of ICU beds and mechanical ventilators, and patients were in extreme situations, intubated and mechanically ventilated in hospital wards or even, maintained with oxygen room air despite the need for pressurized invasive ventilation. In-hospital mortality was associated with increasing age, the need for invasive mechanical ventilation, and ICU admission. Patients admitted to ICU had also longer hospital stay when compared to patients in non-ICU. Aside from mortality, this could have led to more disability and a greater impact on quality of life after hospital discharge11. This should be further investigated, especially with increasing recognition of individuals with long COVID.

The relevance of prompt assistance for patients with SARS was brought out when we analyzed the association between the duration of symptoms and the need for mechanical ventilation. Patients who received earlier prehospital assistance—before day five of the onset of symptoms—were less likely to need assisted ventilation. This might be explained by the fact that patients with acute respiratory failure require in-hospital treatment, and earlier prehospital assistance could contribute to timely diagnosis and a more favorable prognosis12,13. Therefore, opportune EMS assistance could potentially benefit patients with acute respiratory syndrome by suitably identifying risk factors—such as hypoxemia, older age, and the presence of comorbidities—and quickly transferring to a compatible health center.

In addition, non-invasive ventilation during prehospital assistance has been studied and proven to reduce intubation and mortality rates14. High-flow nasal cannula oxygen therapy might be useful in selected COVID-19 cases15. In any case, supplemental oxygen, commonly considered the mainstay therapy for hypoxemia, is universally available in prehospital EMS16. The ability to promptly identify respiratory emergencies, initiate oxygen therapy in the field, and expedite transportation to a suitable health center probably turns prehospital EMS into a game changer in patients with acute respiratory failure.

The main limitations of the study involve original database quality and completeness. Both the EMS and the health surveillance databases presented consistency issues, especially during the transition period when the number of respiratory emergencies started to rise. The inability to test patients for COVID-19 in the EMS at the time was also a pitfall, making it difficult to associate clinical characteristics with etiology and even possible different SARS-CoV-2 variants. Immunization impact could not be assessed either, as no vaccines were available up until the end of the study period and merit further investigation. This study used secondary data from the first wave of COVID-19 in Manaus, focusing on cases with complete data to avoid imputation, due to overwhelmed health services and frequent missing data. Standardized data extraction and statistical adjustments were applied to control information and confounding biases. It is possible that with the advancement of the COVID-19 pandemic from the first epidemic peak to peaks in the following years, the pre-hospital EMS evolved in data collection and further screening of individuals. Nonetheless, the collection of data regarding previous medical history, as comorbidities, may help strengthen inferences on health outcomes in the future, which could be done by unifying health databases.

Nonetheless, the number of follow-up patients allowed a few important analyses regarding important aspects of COVID-19 and possibly other acute respiratory diseases. Knowledge of the role of EMS and their chief capabilities could contribute to outbreak response in the future, improving emergency preparedness, as recommended by world authorities17.

Conclusion

In-hospital mortality of patients with SARS transported by the EMS in Manaus during the initial phase of the COVID-19 pandemic was higher in patients with increasing age and in those presenting lower pulse oximetry values during prehospital assessment. Considering that prehospital EMS may provide feasible and early recognition of critical patients with SARS, it is of the utmost importance to regard data obtained during prehospital care to anticipate certain clinical outcomes. This is especially valuable in resource limited settings and/or during disease outbreaks, when information scarcity indisputably impacts quality of care and patient safety and outcomes. Further research is needed to determine if these findings could be extrapolated to SARS patients in different scenarios, other than pandemics.