Factors related to early and rapid assessment of in-hospital mortality among older adult trauma patients in an earthquake

Hai Hu , Ni Yao, Xiao-qin Lai

1 Emergency Management Office of West China Hospital, Sichuan University, Chengdu 610041, China

2 China International Emergency Medical Team, Chengdu 610041, China

3 Sichuan University’s Emergency Medical Rescue Base, Chengdu 610041, China

4 Department of Critical Care Medicine, West China Hospital, Sichuan University, Chengdu 610041, China

5 Day Surgery Center, West China Hospital, Sichuan University, Chengdu 610041, China

6 West China School of Nursing, Sichuan University, Chengdu 610041, China

KEYWORDS: Trauma; Mortality; Earthquakes; Elderly patients

INTRODUCTION

Older adult earthquake trauma patients (OAETPs)have a higher risk of death than younger patients, and their care is complicated by comorbidities and increased social service needs.[1-3]In order to manage this increased complexity, care coordination and marshalling of personnel and other resources is required. During an earthquake, care coordination can be difficult for mass casualties.[1]

After an earthquake, health workers can screen older adults with higher risk and pay more attention to them to reduce mortality among these patients by understanding rapid and effective related factors.[2]

Most previous studies on older adult trauma death factors have mainly focused on trauma in daily life,[4-6]rather than trauma in disasters such as an earthquake. There are only a few studies that investigate factors related to the death of patients with earthquake trauma, and the evidence related to elderly trauma death in earthquakes is limited.Most factors in previous studies relied on laboratory or imaging examinations, which were not easily applied to rapid screening earthquake trauma victims in emergency departments.[2,7,8]Furthermore, the multicollinearity between age and preexisting diseases might contribute to statistical challenges. Therefore, simple and feasible factors related to the early screening of OAETPs needed to be explored. Further research on risk factors associated with OAETP in-hospital mortality is neccessary.

Therefore, this study aimed to investigate the impact of the simple and rapidly factors on the in-hospital mortality of OAETPs and create a screening tool for assessing such patients.

METHODS

Study design

This retrospective study analyzed the patient characteristics and outcomes from the West China Earthquake Patients Database (WCEPD) to screen potential in-hospital mortality related factors of hospitalized patients. The WCEPD held the pre-hospital,emergency, inpatient, and discharge information of inpatients from four earthquakes (supplementary files),which occurred over the past 14 years in West China. The WCEPD is managed by Sichuan University’s Emergency Medical Rescue Base and had 36,604 patient records from 701 hospitals by November 14, 2021.

The local institutional review committee approved the study and waived the requirement for informed consent due to the study design. The protocol number of the ethics approval by the local ethics committee was 2020-477. The study complied with the international ethical guidelines for human research, such as theDeclaration of Helsinki. The accessed data were anonymized.

Selection of participants

Data were obtained from the WCEPD. The inclusion criteria for this study were older adults (aged ≥65 years old) with trauma transferred from the earthquake field to the hospital rather than patients transferred from other hospitals. Of the total trauma cases due to earthquakes,7,799 older adult cases were selected. We then excluded 326 cases transferred from other hospitals, and 7,473 older adult trauma cases were enrolled. After excluding 165 cases with missing data, we analyzed the records of the remaining 7,308 patients in our study (Figure 1).Among them, 114 patients died during hospitalization.

Figure 1. The inclusion and exclusion flow chart of older adult earthquake trauma patients.

Key variables

From the WCEPD, the inpatient deaths were included as the dependent variable,Y, and coded as 0 for survivors and 1 for non-survivors. Additionally, we collected data on 35 available variables based on previous reports.[2,7,9-13]The variables can be classified into six categories:demographic characteristics, pre-hospital treatment,traumatic condition, existing comorbidities, vital signs,and rapid trauma score. These variables required no laboratory or imaging tests, and they could be obtained quickly even in the earthquake disaster setting. During calculation, all the variable scores would be cumulative with the scores of other variables. We defined these variables as follows.

(1) Demographic characteristics in this study included sex and age. Sex (variable code:X1) was either male or female and coded as 0 for male and 1 for female;while age (X2) was measured in years from the patient’s birth to earthquake onset.

(2) Pre-hospital treatment, which described whether the earthquake patient received primary medical treatment from healthcare providers before being transferred to any hospital, included four variables.[2]Code 1 was used when the treatment had been provided,while 0 was used when it was not.

The four variables were as follows: arresting bleeding (X3), where the patient’s open wound received pre-hospital hemostatic treatment, including wrapping with a bandage, compression bandage, or similar material; assisted ventilation (X4), defined as the work of assisting the patient’s respiration with a mask-balloon or portable ventilator during the pre-hospital period;fixation (X5), defined as pre-hospital immobilization for patients with suspected fractures; and fluid therapy (X6),referring to the administration of pre-hospital infusion to expand the intravascular fluid volume of patients.

(3) Traumatic condition only refered to the situation of the patient’s injury that could be quickly screened and included. Eight variables were coded 1 as yes, while were coded 0 as no.

Crush injury (X7) is a kind of trauma resulting from extreme pressure applied by heavy objects (such as collapsed buildings), interrupting blood flow and damaging of the muscle fibers or the soft tissue. Crush injury leads to crush syndrome, which is a major cause of mortality in earthquake trauma patients.[14]However,the diagnosis of crush syndrome requires laboratory indicators, which might be difficult to achieve in the rapid screening of earthquake trauma patients. Therefore,in this study, crush injury was selected as a variable instead of crush syndrome. An open wound (X8) was used to define an earthquake trauma patient with at least one of their wounds left open. The remaining six variables of the traumatic condition represented different injured parts, including the head/neck (X9), face(X10), chest (X11), abdomen/pelvis (X12), extremities(X13), and external injuries (X14). The classification of the injured part was consistent with the six parts of the injury severity score (ISS). If a patient had trauma in a particular part, regardless of whether they had other trauma in other parts, this variable was coded 1;otherwise, it was coded 0. For example, if a patient had trauma in the head and chest, bothX9 andX11 would be assigned a value of 1.

(4) Existing comorbidities in our study referred to chronic diseases and comorbidities from the patients’medical history at hospital arrival. These variables referred to the patient’s history instead of the new-onset disease in the earthquake.

We extracted 13 variables from the cases’ initial diagnosis, which were dementia (X15), hypertension(X16), diabetes (X17), chronic obstructive pulmonary disease (COPD;X18), coronary heart disease (CHD;X19), malignant tumor (X20), stroke (X21), deep vein thrombosis (DVT;X22), chronic kidney disease (CKD;X23), liver cirrhosis (X24), lung infection (X25), urinary tract infection (X26), and bedsores (X27).[10-17]

(5) Vital signs referred to the initial value of vital signs when the patients arrived at the hospital and were composed of six variables.

Axillary temperature (T,X28) was measured in degrees Celsius (°C). The variablesX29 andX30 were respiratory rate (RR) and pulse rate (PR), measured as counts per minute. The variablesX31,X32, andX33 were systolic blood pressure (SBP), diastolic blood pressure(DBP), and mean artery pressure (MAP), respectively.

(6) The rapid trauma score included two variables:Glasgow Coma Scale (GCS) and Triage Revised Trauma Score (T-RTS).

The GCS score (X34), a widely known tool for mental status assessment, is subdivided into severity categories and is ubiquitous in the trauma setting and trauma treatment guidelines.[18]The GCS score ranged from 3 to 15, and a low score indicated severe trauma. The T-RTS (X35) was a tool for disaster triage. Its superior efficiency was due to both speed and objectivity.[19]The T-RTS was composed of a combination of results from three categories: GCS, SBP,and RR. All of these results can be quickly assessed with minimal equipment. The T-RTS ranged from 0 to 12,and a low score indicated severe trauma (supplementary Tables 1 and 2).

The outcome variable was inpatient death, including death in the emergency department of a hospital and death during hospitalization. The mortality rate was defined as the number of deaths to the total number of patients in each of the variables mentioned above.

Statistical analysis

Statistical analyses were performed using R software version 4.1.1 (R Foundation for Statistical Computing,Austria) and SPSS version 20.0 (IBM Corporation,USA). Continuous variables were expressed as mean with standard deviation or medians with interquartile ranges. Categorical variables were expressed as absolute values and percentages. The continuous variables were compared using independent groupt-tests for normally distributed data and the Mann-Whitney test for non-normally distributed data. The Chi-square or Fisher’s exact test was used to compare proportions as appropriate.

Traditional regression techniques were limited in the analysis and synthesis of large numbers of covariates,including multicollinear variables.[20]Most data on the related factors of either earthquake victims or daily older adult trauma have utilized traditional statistical methods. Least absolute shrinkage and selection operator(LASSO) regression is a methodology that permits many covariates in the model. It helps to automatically remove unnecessary/uninfluential covariates by penalizing the absolute value of a regression coefficient.[21-23]

LASSO regression, which is suitable for analyzing high-dimensional data, was used to select the most significant predictive features.[24,25]We utilized the“glmnet” package (version 2.0-16) of R software to fit the LASSO regression. We utilized ten-fold crossvalidation to select the penalty term, lambda. The binomial deviance was computed for the test data to measure the predictive performance of the fitted models. The built-in function in R produces two lambda values: one that minimizes the binomial deviance and the other that represents the largest lambda that is still within one standard error of the minimum binomial deviance. We choose the stricter value to reduce the number of covariates. The standard errors of the LASSO coefficients were obtained via bootstrapping within the primary sampling unit and strata.[26]Features with nonzero coefficients in the LASSO regression model were selected in the forward stepwise logistic regression model.[25]The features are presented as odds ratios (ORs)with 95% confidence intervals (95%CIs) and two-tailedP-values. Two-tailedP-values ＜0.05 were considered significant. Finally, the nomogram was formulated using the independent related factors selected by LASSO regression.

RESULTS

The demographic and clinical characteristics of patients

In total, 7,308 OAETPs in the WCEPD were enrolled in this study. The age was 74.83±6.95 years for the survivors and 77.99±8.95 years for the non-survivors.The median (interquartile range) of ISS were 9 (4-16) for survival, and 10 (9-18) for death (P=0.002). There were 22 variables with significant differences (P＜0.05): sex,age, crush injury, head/neck trauma, extremity trauma,dementia, diabetes, COPD, CHD, malignant tumor, DVT,CKD, lung infection, urinary tract infection, bedsores,RR, PR, SBP, DBP, MAP, GCS, and T-RTS (Table 1).

Selection of independent factors

From the results of the LASSO regression, 10 variables with non-zero coefficients were selected.The details of LASSO regression can be found in Supplementary Figures 1 and 2. The 10 variables,including age (OR=1.061, 95%CI: 1.031-1.090),dementia (OR=5.146, 95%CI: 1.169-17.856), CHD(OR=23.441, 95%CI: 4.799-83.927), malignant tumor (OR=8.497, 95%CI: 3.583-17.967), DVT(OR=7.110, 95%CI: 1.369-27.168), CKD (OR=11.783,95%CI: 5.419-24.407), PR (OR=1.036, 95%CI:1.022-1.048), MAP (OR=0.960, 95%CI: 0.945-0.975),GCS (OR=0.864, 95%CI: 0.760-0.972), and T-RTS(OR=0.485, 95%CI: 0.351-0.696), were the independent related factors contributing to the in-hospital mortality of the OAETPs (Model 1, Table 2).

Although LASSO regression is currently a better statistical method for resolving multicollinear variables,for prudence, some variables that are likely to have multicollinearity, including the GCS, T-RTS, and dementia,were re-analyzed. For both T-RTS and dementia, the GCS might be a confounding factor. Then, model 2 (excluded the GCS), model 3 (excluded the T-RTS), and model 4(excluded dementia) were operated on (supplementary Table 4). Model 1 included all 10 variables. Model 2 excluded the GCS, and the maximum absolute value of the change ratio ofORwas that of T-RTS (6.39%). Model 3 deleted the T-RTS, and the maximum absolute value of the change ratio ofORwas that of GCS (9.26%). Model 4 excluded dementia, and the maximum absolute value of the change ratio ofORwas that of a malignant tumor (3.50%).All the absolute values of the change ratio ofORwere less than 10%.[2,27]Although dementia versus GCS and GCS versus T-RTS were likely to interact, all of them were the related factors for in-hospital mortality among OAETPs.Therefore, we retained all 10 variables in our LASSO regression model.

Table 2. The results of the regression analysis for predicting in-hospital mortality of older trauma patients in an earthquake

Construction of the prognostic nomogram

The nomogram used for assessing in-hospital death of OAETPs was established using the 10 variables selected by LASSO regression (Figure 2). Each variable was assigned a score according to the related factors of each case, and the total score was computed by summing the individual scores. The probabilities of inhospital death of OAETPs could be calculated using this nomogram. Both the calibration curve and the receiveroperating characteristic (ROC) curve of the nomogram are shown in supplementary Figure 3.

Figure 2. The nomogram for assessing in-hospital mortality of older adult trauma patients in an earthquake. The nomogram included ten variables,which were age (X2), dementia (X15), coronary heart disease (X19), malignant tumor (X20), deep vein thrombosis (X22), chronic kidney disease(X23), pulse rate (X30), mean artery pressure (X33), Glasgow Coma Scale (X34), and Triage Revised Trauma Score (X35). The nomogram summed the points from the scale of each variable. The total points indicated the probability of inpatient death of older adult trauma patients in an earthquake.

DISCUSSION

Early and rapid identification of fatal earthquakerelated trauma in older adults is vital for providing optimal care; however, understanding the related factors of this group is challenging. Exploring factors related to OAETPs may be valuable for reducing the mortality of this group during an earthquake.[28]Based on LASSO regression, 10 factors associated with the inpatient death risk of OAETPs were found. The factors were age,dementia, CHD, malignant tumor, DVT, CKD, PR, MAP,GCS, and T-RTS. We then developed and validated a nomogram based on the selected factors to identify OAETPs with a higher risk of in-hospital mortality.The nomogram included the above 10 parameters and showed good discrimination and calibration. Application of the model might allow emergency physicians to assess OAETPs more effectively and allow medical decisionmakers to utilize limited medical and transportation resources properly after a disaster.

Among the 10 rapid and simple factors selected by LASSO regression, age was still one of the high-risk factors in older patients. In general, older patients have weaker bodies. In a study of the Nepal earthquake, Pant et al[29]found that the earthquake did more than just cause personal injury to older adults; drug shortages and irregular inspections in post-earthquake care facilities contributed to chronic disease consequences. A previous study on the Wenchuan earthquake also reported asupply shortage of medicines and equipment due to the destruction of infrastructure such as road systems after the earthquake.[28]

Elderly adult trauma patients often had comorbidities, and the lack of medical treatment in the earthquake setting was likely to lead to poor outcomes in these patients. Our study selected five key comorbidities, including dementia, CHD, malignant tumor, DVT, and CKD. Some previous studies[30-34]on older adults with trauma illuminated similar results. In Bai’s meta-analysis of hip fractures in older adults, dementia significantly increased the 30-day mortality (relative risk: 1.57).[31]Kirshenbom’s study[32]showed a significant association between inhospital mortality of OAETPs and the existence of certain comorbidities, such as dementia, CHD, and renal insufficiency caused by CKD. Grossman et al[33]pointed out that pre-existing conditions including malignant tumors (OR=1.8) and CKD (OR=3.1)had a substantial effect on mortality in older adult trauma patients. Moreover, a high risk of DVT and venous thromboembolism were an important cause of death for patients with trauma. The prevention of DVT might reduce death related to venous thromboembolism. A systemic review by Barrera et al[34]showed that pharmacological prophylaxis was more effective than mechanical methods in reducing DVT (relative risk: 0.48). However, obtaining such medicines in an earthquake-hit area might be difficult.Thus, we should pay more attention to OAETPs with these comorbidities rather than just assess their trauma severity in disaster management. Furthermore, we should consider prioritizing medicine supply for these comorbidities in such conditions.

In our study, upon arrival at the hospital, the PR and MAP of the vital signs and the GCS and T-RTS of the rapid trauma score were also related factors to the inpatient death of the OAETPs. The PR and MAP indicated blood circulation status, and the GCS represented the level of consciousness.[35,36]The T-RTS,including the GCS, SBP, and RR, was considered a valuable scoring system for the overall triaging of OAETPs. Several previous studies on earthquakes had similar results and considered these factors related to trauma patients.[19,37]

Our study had some limitations. First, we did not account for seismic features of earthquakes and building structural factors.[38]We also did not consider differences in search and rescue time because we could not gain enough pertinent information from the database. All of these factors could be related to earthquake-based inpatient death. Second, our definition of crush injury is too broad, which may lead to misjudgments. Further studies are needed to screen early and rapid variables for identifying crush syndrome. Third, due to a lack of data, several issues were not discussed in our study, including the analysis of critically ill patients transferred from other hospitals, the comparison of mortality between different hospitals, and the exploration of oxygen saturation as an early indicator. In addition, all the four earthquakes included in our database occurred in rural areas. Compared with urban areas, rural areas are characterized by low population density and low variety and volumes of medical services. Thus,casualties and factors affecting inpatient deaths may differ. It is necessary to investigate these potential factors associated with the in-hospital mortality of OAETPs in the future.

CONCLUSION

To reduce the in-hospital mortality of OAETPs, the rapid and simple related factors identified and analyzed by LASSO regression in this study included age,dementia, CHD, malignant tumor, DVT, CKD, PR, MAP,GCS, and T-RTS. These 10 related factors that could be quickly obtained at hospital arrival should be the focal point of future earthquake response strategies regarding hospitalized OAETPs. Furthermore, we established a nomogram for such patients incorporating the above 10 related factors. This nomogram could be conveniently used, at an early stage, to facilitate the prediction of the individual risk of in-hospital mortality of OAETPs and assist emergency physicians and emergency trauma surgeons in identifying hospitalized OAETPs who might develop fatal conditions in the disaster management following an earthquake.

Funding:This work was supported by the Strategic Priority Research Program of the Chinese Academy of Science(XDA23090502) and Science and Technology Department of Sichuan Province (21KJPX0207).

Ethical approval:The local institutional review committee approved the study and waived the requirement for informed consent due to the study design. The protocol number of the ethics approval by the local ethics committee was 2020-477. The study complied with the international ethical guidelines for human research, such as theDeclaration of Helsinki. The accessed data were anonymized.

Conflicts of interest:The authors declared there are no conflicts of interest.

Contributors:HH conceived the study, designed the trial, and obtained research funding. HH and NY provided statistical advice on study design and analyzed the data. HH and XQL supervised the conduct of the data collection from the database, and managed the data. All the authors drafted the manuscript, and contributed substantially to its revision.All the supplementary files in this paper are available at http://wjem.com.cn.