Correlation Analysis between Pulse Oxygen Saturation and Prognosis of Emergency Trauma Patients

Past studies are limited in which proposing SpO2 as mortality predictor of trauma patients. The aim of our study was to investigate the correlation between pulse oxygen saturation (SpO2) and prognosis in emergency trauma patients. We collected 1720 trauma patients admitted to the Emergency Department of the First Aliated Hospital of Soochow University from November 1, 2016 to November 30, 2019 to retrospective analysis. The mortality of trauma patients in the emergency department was dened as end-point of outcome. The patients were divided into six subgroups with 75%, 80%, 85%, 90% and 95% SpO2 as the bound values. SpO2>=95% subgroup was dened as basis reference, we calculated the crude HR of other subgroups and adjusted HR after the adjustment for confounders including age and sex by Cox regression model. The ROC curve was performed and the area under the curve (AUC) was calculated to evaluate the predictive value of SpO2 in emergency mortality. pulse oxygen saturation, SpO2; area under the curve, AUC; emergency department, ED; mean arterial pressure, MAP; pulse rate, P; respiratory rate, RR; revised trauma score, RTS; hours in emergency room, HER; inter quartile range, IQR; condence intervals, CIs; Injury Severity Score, ISS; the Trauma and Injury Severity Score, TRISS; the Revised Trauma Score, RTS; the Mechanism, Glasgow coma scale, Age, and Arterial Pressure, MGAP; the Glasgow Coma Scale, Age, and Systolic Blood Pressure, GAP; the New Trauma Score, NTS.


Background
Past studies are limited in which proposing SpO2 as mortality predictor of trauma patients. The aim of our study was to investigate the correlation between pulse oxygen saturation (SpO2) and prognosis in emergency trauma patients.

Methods
We collected 1720 trauma patients admitted to the Emergency Department of the First A liated Hospital of Soochow University from November 1, 2016 to November 30, 2019 to retrospective analysis. The mortality of trauma patients in the emergency department was de ned as end-point of outcome. The patients were divided into six subgroups with 75%, 80%, 85%, 90% and 95% SpO2 as the bound values.
SpO2>=95% subgroup was de ned as basis reference, we calculated the crude HR of other subgroups and adjusted HR after the adjustment for confounders including age and sex by Cox regression model.
The ROC curve was performed and the area under the curve (AUC) was calculated to evaluate the predictive value of SpO2 in emergency mortality.

Results
Compared to basis reference, the mortality of other subgroups was gradually increased with the decrease of SpO2. The crude HR and 95% CI of each subgroup calculated by univariate Cox regression were: SpO2 90%-95%, 7

Conclusion
Our study revealed that SpO2 was closely related to the prognosis of emergency trauma patients and had a good predictive value for emergency room death. The lower the SpO2, the higher the mortality.
This study was retrospectively registered in the First A liated Hospital of Soochow University.

Page 3/11
Trauma is a devastating medical issue with high mortality rate, imposing huge global burden [1]. In accordance to an epidemiological study [2], 52.5% of respondents in China experienced at least one lifetime traumatic event, which indicate that trauma is common and worthy of clinical attention. In emergency department (ED), it is important to quickly identify and treat patients with severe trauma. The factors that could predict trauma mortality contribute to discover severe patients to some extent. There have been some studies on the predictors of mortality in trauma patients [3][4][5]. Variables, including vital signs, trauma scoring systems, acute traumatic coagulopathy, ED presentation time and et al were investigated to evaluate the value of predicting trauma mortality.
As one of the vital signs, respiratory rate assessment obtained by medical personnel was unreliable and usually incorrect in the ED settings [6,7]. Pulse oxygen saturation (SpO2), an objective, e cient and unequivocal parameter, was quick to measure and widely used in the clinical setting [8].
However, studies are scarce in which proposing SpO2 as mortality predictor of trauma patients. Though some studies reported that SpO2 was not good predictor for mortality of trauma patients, they still recommend SpO2 as a variable in a model of survival probability due to the clinical ease of obtainment and its physiological signi cance [4,5]. Other studies indicated that SpO2 was independently predictive value of mortality in trauma patients [3].
This study was conducted to explore the relationship between SpO2 and prognosis of trauma patients in ED. The results of the present study will contribute to quickly identi cation of severe trauma and also improve the outcome of patients.

Subject
This study was retrospective and clinical data were exported from the emergency trauma registry information system of the First A liated Hospital of Soochow University.
We included 1720 trauma patients over 18 years of age enrolled in the database from November 1, 2016 to November 30, 2019. Information on patient characteristics was collected including sex, age, mean arterial pressure (MAP), pulse rate (P), respiratory rate (RR), SpO2, revised trauma score (RTS) and hours in emergency room (HER). SpO2>=95% was de ned as basis reference. Among the rest, SpO2 was divided into ve groups according to the following values, 90 to 95%; 85 to 90%; 80 to 85%; 75 to 80%; <75%.

Outcome
The mortality of trauma patients in ED was de ned as end-point of outcome in this study. Trauma patients were divided into two groups: survival and non-survival. According to SpO2 boundary, outcomes of ve subgroups were recorded to analyze the correlation between SpO2 and prognosis.
This study was approved by the Ethics Committees of the First A liated Hospital of Soochow University (Approval Number: 2021-231). Personal informed consent was not required due to the retrospective nature of the data.

Statistical analysis
The patient characteristics involved continuous and categorical variables. Continuous variables were expressed as median (inter quartile range, IQR). Categorical variables were expressed as frequencies and percentages. P values were calculated by Mann-Whitney test or Chi squared test. The mortality of trauma patients in ED were evaluated in crude and multivariable Cox regressions. Multivariable regression analysis was adjusted for confounders including age and gender. Receiving operating characteristic curve analyses were performed to determine the cutoff values of variables for distinguishing between survival and non-survival. Statistical analyses and graphics were achieved with STATA 15. Data were presented with the standard level of signi cance (P<0.05) and with 95% con dence intervals (CIs).

Results
Over the 3-year recruitment period, 1720 consecutive patients of trauma were enrolled in the emergency trauma registry information system of the First A liated Hospital of Soochow University, 97.5% of which survived in the end. Among survivals, 452 (26.95%) cases were female and 1225 (73.05%) cases were male. There was no signi cant sex difference between survivals and non-survivals (P=0.589). The survivors showed on statistically signi cant lower HER and higher MAP, SpO2 and RTS than another group (P<0.05), whereas age, P and RR were not different between two groups. (Table 1) Figure 1 demonstrated the ROC curve, each point on which corresponded to a sensitivity and speci city, and a high sensitivity meant a decrease in speci city. The ROC curve analysis showed that the AUC was 0.898, 95%CI (0.845-0.951), the optimal cut-off value was 94%, and the corresponding sensitivity and speci city were 79.07% and 85.84%, respectively. (Table 3)

Discussion
Although past studies had evaluated the potential outcome predictors of trauma patients, such studies in SpO2 as predictive parameter are limited. SpO2 with vital clinical signi cance is routinely measured and recorded in ED. The present study revealed SpO2 was independent predictor of outcome in emergency trauma patients.
It is a key element in rapid assessment and treatment of trauma patients in ED. In order to quickly identify severe trauma patients, various severity scoring systems were explored including Injury Severity Score (ISS) [9], the Trauma and Injury Severity Score (TRISS) [10], the Revised Trauma Score (RTS) [11], the Mechanism, Glasgow coma scale, Age, and Arterial Pressure (MGAP) score [12], the Glasgow Coma Scale, Age, and Systolic Blood Pressure (GAP) score [13], the New Trauma Score (NTS) [14]. Above trauma scoring systems were dedicated to perform well in predicting outcome of trauma patients, but none played an irreplaceable role and gained easy access since of di culties in real-time statistics.
Domingues et al. [4] proposed three new models which were equivalent to TRISS and indicated that SpO2 as an isolated adjustment did not increase the predictive ability of this model, but also considered that the frequent lack of SpO2 data might have underestimated its importance in the survival prediction models. It was also observed that SpO2 did not add signi cant value to other variables when predicting mortality in severe trauma patients [5]. An integrative review [15] found that an alteration to the method of age inclusion in the equation, and the insertion of gender, comorbidities and trauma mechanism, and exclusion of RTS presented a tendency towards improved performance of the TRISS, but use of SpO2 with neutralized RR in RTS did not result in improved performance.
Nevertheless, SpO2 was veri ed as a powerful mortality predictor across all ages by a recent study [16]. Moreover, Woodford et al discovered that mean SpO2 was a signi cant predictor of mortality after trauma during prehospital care [17]. According to a research from South Africa[18], SpO2 was demonstrated as an independently signi cant predictor of outcome in severe traumatic brain injury as well. Other studies [3,14] also con rmed that SpO2 had signi cant value of predicting mortality of trauma patients.
Such results show discordance regarding performance of SpO2 and indicate that it is necessary to propose further researches to investigate the physiological component in survival probability prediction for trauma patients. Furthermore, it is important in evaluating predictor valuables of early mortality in trauma patients, especially during emergency department. This study found that SpO2 is signi cantly associated with the mortality of trauma patients in ED, which is consistent with previous research results [3,14,[16][17][18]. We also discovered that variables including MAP, RTS and HER were independent prognostic factors for death in emergency trauma patients.
According to subgroup analysis, the lower SpO2, the higher mortality even if adjusting sex and age. This nding suggested that trauma patients with pulse hypoxia in ED deserved close attention. It is convenient to measure the level of SpO2, hereby rapid triage of trauma patients and subsequently perform intervention therapy in ED.
As far as we know, this study was the rst hospital registry-based research with large sample size (1720 patients) which revealed the relationship between SpO2 and early mortality of trauma patients in ED. The complete clinical data enabled us to study the effect of SpO2 on mortality, which has not yet being wellexplored in ED.
However, this study was conducted in a single-center, therefore our results may not be generalizable to the broader population of trauma patients, which may have biased our results. Moreover, it is required to carry out further subgroup studies by trauma classi cation in the future.

Conclusions
SpO2 is clearly associated with disparate outcomes following trauma. It could be applicable to both prompt triage and prognosis evaluation of trauma patients in ED. With trauma morbidity occurring at unprecedented speed and scale in China, more attention show be paid to trauma patients particularly with low SpO2, and clinical interventions that could be effective in improving SpO2 are required to reduce mortality of trauma in ED. Availability of data and materials All data are included in the manuscript and Supporting Information les. Figure 1