Performance of scores in the prediction of clinical outcomes in patients admitted from the emergency service

Objective: to evaluate the performance of the quickSOFA scores and Systemic Inflammatory Response Syndrome as predictors of clinical outcomes in patients admitted to an emergency service. Method: a retrospective cohort study, involving adult clinical patients admitted to the emergency service. Analysis of the ROC curve was performed to assess the prognostic indexes between scores and outcomes of interest. Multivariate analysis used Poisson regression with robust variance, evaluating the relationship between variables with biological plausibility and outcomes. Results: 122 patients were selected, 58.2% developed sepsis. Of these, 44.3% had quickSOFA ≥2 points, 87% developed sepsis, 55.6% septic shock and 38.9% died. In the evaluation of Systemic Inflammatory Response Syndrome, 78.5% obtained results >2 points; of these, 66.3% developed sepsis, 40% septic shock and 29.5% died. quickSOFA ≥2 showed greater specificity for diagnosis of sepsis in 86% of the cases, for septic shock 70% and for mortality 64%, whereas the second score showed better results for sensitivity with diagnosis of sepsis in 87.5%, septic shock in 92.7% and death in 90.3%. Conclusion: quickSOFA showed by its practicality that it can be used clinically within the emergency services, bringing clinical applicability from the risk classification of patients for the early recognition of unfavorable outcomes.


Data sources/measures
The information of the researched subjects was obtained from the hospital's assistance database, generated through a database made available in spreadsheets in the MS Excel ® program. In this database, data on patients admitted to the ES in the study period were made available. After the spreadsheet was made available, the medical records were chosen at random through a draw tool within the own software used for statistical analysis where the patients included in the sample were defined during the study period. The data were obtained exclusively through the review of electronic medical records. The patients were divided in two groups: Patients WITH sepsis and patients WITHOUT sepsis. It is worth mentioning that, after obtaining the data, they were checked and typed in the Excel ® program by two different typists (main researcher and research assistant), being subsequently compared to control possible typing errors.

Variables and Outcomes
The variables were classified in two groups: sociodemographic, referring to the risk classification, and clinical variables related to hospitalization. The institution where the study was carried out uses the Manchester Screening System (MSS) in the emergency service as a risk classification system. The MSS aims to identify the user's main complaint, and to select a specific flowchart, guided by discriminators who determine service priority.
The individual can be classified in five different priority levels: 1: Emergency; 2: Very Urgent; 3: Urgent; 4: Little Urgent; and 5: Not Urgent. Each priority level has its specific color and the recommended initial service time (10) .
The variables that were part of the database referring to the patients' risk classification were the following: flowchart used, discriminator chosen, priority of the care assigned, vital signs, and the result of the qSOFA score recorded by the nurse at the time of risk classification.
To achieve the study objectives, the values of the following scores were used: qSOFA, SIRS and Charlson's Comorbidity Index (CCI). The qSOFA score is recorded jointly with the MSS in the risk classification by the nurses, since the SIRS and the CCI are not necessarily performed at risk classification, the first one because it does not contain all the variables necessary to obtain it, such as some results of laboratory tests; and the other, for not being able to delay the risk classification in the search for the patient's previous comorbidities. Therefore, it is worth mentioning that the information for calculating them was extracted exclusively from the electronic medical records of the research participants. These calculations were performed by researchers trained to obtain the scores and were obtained as follows: In the field where the study was conducted, there is a care line for septic patients, where the application of the qSOFA score is performed in the care and initial assessment by the nurses at risk classification, as part of the protocol of this problem, in addition to the application of the MSS.
The qSOFA score is considered positive for a possible diagnosis of sepsis, when showing two or more criteria (points) of patient evaluation: respiratory rate equal to or greater than 22 respiratory movements per minute (rmpm), change in the level of consciousness, verified through the application of the Glasgow Coma Scale < 15, or systolic pressure less than or equal to 100 mmHg (2) .
In the event of an abnormality of this score, the system signals this patient in a different color from the others (purple), in order to signal to the medical team the need for early medical assistance to the patient with a probable diagnosis of sepsis, signaled by this score by the risk classification nurse. It is worth mentioning that all the nurses who carry out risk classification in the Emergency Service were trained by the Brazilian Risk Classification Brasileiro de Classificação de Risco, GBCR) to apply the MSS, in order to accurately define the priority of care for patients who seek the emergency service. In addition to this, these classifying nurses received specific training to apply the qSOFA score to all patients with symptoms for sepsis.

Group (Grupo
To obtain the SIRS score, which is defined by the presence of at least two of the following signs: central temperature > 38.3ºC or < 36ºC, heart rate > 90 bpm, respiratory rate > 20 rmpm, or PaCO 2 < 32 mmHg and total leukocytes > 12,000/mm³; or < 4,000/mm³ or presence of > 10% of young forms (left deviation) (1) ; the laboratory tests were verified after the first results came out, in order to properly assess this score. These exams were consulted in the electronic medical record of the selected patients along with the registration of vital signs at risk classification, as a way to complete the assessment of the score.
Finally, to obtain the CCI, which is a tool used to verify the prediction of in-hospital mortality, the MDCalc ® online Rev. Latino-Am. Enfermagem 2021;29:e3479. calculator was used, which analyzes the age and the list of previous comorbidities recorded in the electronic medical record of each study participant. This calculator follows the modifications for the evaluation of the updated index foreseeing 16 comorbidities that generate different scores, the result being established by the sum of all, associated with the patient's age. The higher this score, the lower the subject's life expectancy in the next 10 years (11) . The CCI had its score value categorized as <2 (without risk) and as > 3 (with risk), in order to classify the patients› risk in relation to the presence of comorbidities for the mortality outcome.
The main outcome of this study was the diagnosis of sepsis and the secondary outcomes were septic shock and the occurrence of death due to sepsis during the patient's in-hospital stay, recorded in medical records, and confirmed by reviewing the hospital discharge or death summary. The mortality rate was measured from hospital admission to death.

Sample size
Sample calculation was carried out in two stages: one for the main study objective (qSOFA and SIRS vs. Sepsis and septic shock) and one for the secondary objective (qSOFA and SIRS vs. Mortality). For the first stage, the same was done in the R/R Studio ® program, roc.test function.
Considering the prevalence of sepsis of 30% in Brazilian and international studies (12)(13) , power of 95% and significance level of 5%, a sample size of 50 patients is sufficient to detect as significant an area under the ROC curve of 0.7 considering qSOFA as a predictor of sepsis diagnosis and a clinically useful test to be used in the ES for this early identification.
For the second stage (qSOFA vs. Mortality), the sample calculation was performed with the WinPEPI program, version 11.43. Considering 80% power, 5% significance level, and the following data (13) : 75.2% of the patients with a qSOFA score below 2, 3.3% of mortality in patients with a qSOFA score below 2 and 23.9% of mortality in patients with a qSOFA score greater than or equal to 2, the size total sample of 122 subjects was defined. Thus, in an attempt to answer the two outcomes proposed, the sample size of the largest number of subjects will be used.

Quantitative variables
The continuous variables were described from their means and standard deviations and the categorical variables by using frequencies and proportions. The qualitative variables, such as gender, were compared using the chisquare and Fisher's exact tests and the continuous variables with the Student's T and Mann-Whitney tests (according to the normality of the variable). The statistical tests were defined after performing the Kolmogorov-Smirnov test to verify the normality of the numerical data. The comparison of the characteristics between groups 1 and 2 (WITH sepsis and WITHOUT sepsis) was performed.

Statistic methods
The data collected were organized and compiled in the Excel software and later submitted to the statistical

Ethical aspects
The research was submitted to and approved by

Results
The results were divided in two stages: in the first, the socio-epidemiological profile of the septic and nonseptic patients in the study was assessed and, in the second, the univariate and multivariate analyses of the other data from the database were carried out.
A total of 122 subjects were included for the study;   Table 2.

Respiratory Rate>22
Systolic Blood Pressure<100 In this study, qSOFA showed better results in relation to the SIRS score in the emergency service. qSOFA ≥2 obtained higher specificity for the diagnosis of sepsis, as well as better specificity for septic shock and specificity for mortality, whereas the SIRS showed better results for sensitivity in the diagnosis of sepsis, septic shock and in-hospital death.

SIRS
When comparing these results with previously conducted studies, a recently published systematic review with meta-analysis was found that related the diagnosis of sepsis to qSOFA and SIRS, where it described the specificities of these scores for the diagnosis of sepsis and in-hospital mortality. In this research, some articles revealed that, when comparing qSOFA with SIRS, it was shown that SIRS was more sensitive and significantly superior to qSOFA for diagnosing sepsis, but qSOFA was better for predicting hospital mortality. Current articles also show higher in-hospital mortality within 30 days for patients with qSOFA >2 (14) . When compared to the performance of these scores for the prognosis of sepsis and death, our study did not show any statistical difference (15) .
Similarly to other analyses, qSOFA ≥2 and SIRS >2 were related to a higher number of deaths in relation to patients whose scores were below two (16) . In addition to that, there was also higher sensitivity for predicting mortality for the SIRS score; however, specificity was higher in qSOFA, as already shown in previous research studies (16) .
As in other published studies, the most common infection site was the respiratory system, followed by the urinary site (15,17) . However, when related to death,  (12,23) . Other studies discuss the combination of the two scores, SIRS and qSOFA, as a method to improve the prognosis and detection of patients who come to the hospital due to some infection, but not applied by nurses at risk classification in patients admitted from emergency services (24) . There are also some articles that contribute positive and negative points of each tool, not reaching consensus on which one should be used (13,16) .
It is necessary to mention that qSOFA was not developed with the purpose of diagnosing sepsis, but rather as an alert tool so that an early assessment can be carried out on the patient who seeks the emergency service and shows possible signs of infection with a risk of early deterioration of their clinical condition. However, as it is a recently discovered score, research studies about its potentialities must be carried out in order to determine the best use of this tool in the initial clinical evaluation of these patients. To this end, this study showed that the nurse's performance at risk classification is completely feasible for the early identification of possibly septic patients.

Conclusion
In this retrospective cohort study, we found that qSOFA ≥2 had higher specificity for the diagnosis of sepsis as well as improved specificity for septic shock and specificity for mortality; on the other hand, SIRS showed Silva LMF, Diogo LP, Vieira LB, Michielin FC, Santarem MD, Machado MLP.
better statistical results for sensitivity in the diagnosis of sepsis, septic shock and hospital death. In addition to that, we were able to characterize the sample as an aged population, with previous comorbidities, and not having a great distinction as to gender. We also emphasize that qSOFA served as a better score due to its practicality and good results for clinical use in the emergency service, as it resulted in greater prognostic accuracy for in-hospital mortality; however, we warn against the need for new prospective studies that cover other tools so as to identify which would be the most accurate and with the best performance and clinical applicability in this scenario.