Peng-Chen Chien

Prediction of Mortality in Patients with Isolated Traumatic Subarachnoid Hemorrhage Using a Decision Tree Classifier: A Retrospective Analysis Based on a Trauma Registry System

Background: In contrast to patients with traumatic subarachnoid hemorrhage (tSAH) in the presence of other types of intracranial hemorrhage, the prognosis of patients with isolated tSAH is good. The incidence of mortality in these patients ranges from 0–2.5%. However, few data or predictive models are available for the identification of patients with a high mortality risk. In this study, we aimed to construct a model for mortality prediction using a decision tree (DT) algorithm, along with data obtained from a population-based trauma registry, in a Level 1 trauma center. Methods: Five hundred and forty-five patients with isolated tSAH, including 533 patients who survived and 12 who died, between January 2009 and December 2016, were allocated to training (n = 377) or test (n = 168) sets. Using the data on demographics and injury characteristics, as well as laboratory data of the patients, classification and regression tree (CART) analysis was performed based on the Gini impurity index, using the rpart function in the rpart package in R. Results: In this established DT model, three nodes (head Abbreviated Injury Scale (AIS) score ≤4, creatinine (Cr) <1.4 mg/dL, and age <76 years) were identified as important determinative variables in the prediction of mortality. Of the patients with isolated tSAH, 60% of those with a head AIS >4 died, as did the 57% of those with an AIS score ≤4, but Cr ≥1.4 and age ≥76 years. All patients who did not meet the above-mentioned criteria survived. With all the variables in the model, the DT achieved an accuracy of 97.9% (sensitivity of 90.9% and specificity of 98.1%) and 97.7% (sensitivity of 100% and specificity of 97.7%), for the training set and test set, respectively. Conclusions: The study established a DT model with three nodes (head AIS score ≤4, Cr <1.4, and age <76 years) to predict fatal outcomes in patients with isolated tSAH. The proposed decision-making algorithm may help identify patients with a high risk of mortality.

Stress-Induced Hyperglycemia, but Not Diabetic Hyperglycemia, Is Associated with Higher Mortality in Patients with Isolated Moderate and Severe Traumatic Brain Injury: Analysis of a Propensity Score-Matched Population

Background: Admission hyperglycemia is associated with higher morbidity and mortality in patients with traumatic brain injury (TBI). Stress-induced hyperglycemia (SIH), a form of hyperglycemia induced by the stress response, is associated with increased patient mortality following TBI. However, admission hyperglycemia occurs not only in SIH but also in patients with diabetic hyperglycemia (DH). Current information regarding whether trauma patients with SIH represent a distinct group with differential outcomes compared to those with DH remains limited. Methods: Serum glucose concentration ≥200 mg/dL upon arrival at the emergency department was defined as hyperglycemia. Presence of diabetes mellitus (DM) was determined by patient history and/or admission glycated hemoglobin (HbA1c) level ≥6.5%. In the present study, the patient cohort included those with moderate and severe TBI, as defined by an Abbreviated Injury Scale (AIS) score ≥3 points in the head, and excluded those who had additional AIS scores ≥3 points in any other region of the body. A total of 1798 adult patients with isolated moderate to severe TBI were allocated into four groups: SIH (n = 140), DH (n = 187), diabetic normoglycemia (DN, n = 186), and non-diabetic normoglycemia (NDN, n = 1285). Detailed patient information was retrieved from the Trauma Registry System at a level I trauma center between 1 January 2009, and 31 December 2015. Unpaired Student’s t- and Mann–Whitney U-tests were used to analyze normally and non-normally distributed continuous data, respectively. Categorical data were compared using the Pearson chi-square or two-sided Fisher’s exact tests. Matched patient populations were allocated in a 1:1 ratio according to propensity scores calculated by NCSS software. Logistic regression was used to evaluate the effect of SIH and DH on the adjusted mortality outcome. Results: In patients with isolated moderate to severe TBI, the presence of SIH and DH led to 9.1-fold and 2.3-fold higher odds of mortality, respectively, than patients with NDN. After adjusting for confounding factors, including sex and age, pre-existing co-morbidities, existence of different kinds of intracerebral hemorrhage, and injury severity, patients with SIH still had 6.6-fold higher odds of mortality than those with NDN; however, DH did not present significantly higher adjusted mortality odds. SIH and DH presented different effects on outcomes after TBI. The results also suggested that the pathophysiological effect associated with SIH was different from that of DH. Conclusions: This study demonstrated that patients with SIH and DH had significantly higher mortality than patients with NDN. However, the adjusted mortality was significantly higher only in the selected propensity score-matched patients with SIH and not in those with DH.

Higher Mortality in Trauma Patients Is Associated with Stress-Induced Hyperglycemia, but Not Diabetic Hyperglycemia: A Cross-Sectional Analysis Based on a Propensity-Score Matching Approach

Background: Stress-induced hyperglycemia (SIH) is a form of hyperglycemia secondary to stress and commonly occurs in patients with trauma. Trauma patients with SIH have been reported to have an increased risk of mortality. However, information regarding whether these trauma patients with SIH represent a distinct group with differential outcomes when compared to those with diabetic hyperglycemia (DH) remains limited. Methods: Diabetes mellitus (DM) was determined by patient history and/or admission glycated hemoglobin (HbA1c) ≥6.5%. Non-diabetic normoglycemia (NDN) was determined by a serum glucose level <200 mg/dL in the patients without DM. Diabetic normoglycemia (DN) was determined by a serum glucose level <200 mg/dL in the patients with DM. DH and SIH was diagnosed by a serum glucose level ≥200 mg/dL in the patients with and without DM, respectively. Detailed data of these four groups of hospitalized patients, which included NDN (n = 7806), DN (n = 950), SIH (n = 493), and DH (n = 897), were retrieved from the Trauma Registry System at a level I trauma center between 1 January 2009 and 31 December 2015. Patients with incomplete registered data were excluded. Categorical data were compared with Pearson chi-square tests or two-sided Fisher exact tests. The unpaired Student’s t-test and the Mann–Whitney U-test were used to analyze normally distributed continuous data and non-normally distributed data, respectively. Propensity-score-matched cohorts in a 1:1 ratio were allocated using NCSS software with logistic regression to evaluate the effect of SIH and DH on the outcomes of patients. Results: The SIH (median [interquartile range: Q1–Q3], 13 [9–24]) demonstrated a significantly higher Injury Severity Score (ISS) than NDN (9 [4–10]), DN (9 [4–9]), and DH (9 [5–13]). SIH and DH had a 12.3-fold (95% confidence interval [CI] 9.31–16.14; p < 0.001) and 2.4-fold (95% CI 1.71–3.45; p < 0.001) higher odds of mortality, respectively, when compared to NDN. However, in the selected propensity-score-matched patient population, SIH had a 3.0-fold higher odd ratio of mortality (95% CI 1.96–4.49; p < 0.001) than NDN, but DH did not have a significantly higher mortality (odds ratio 1.2, 95% CI 0.99–1.38; p = 0.065). In addition, SIH had 2.4-fold higher odds of mortality (95% CI 1.46–4.04; p = 0.001) than DH. These results suggest that the characteristics and injury severity of the trauma patients contributed to the higher mortality of these patients with hyperglycemia upon admission, and that the pathophysiological effect of SIH was different from that of DH. Conclusions: Although there were worse mortality outcomes among trauma patients presenting with hyperglycemia, this effect was only seen in patients with SIH, but not DH when controlling for age, sex, pre-existed co-morbidities, and ISS.

Identification of Pancreatic Injury in Patients with Elevated Amylase or Lipase Level Using a Decision Tree Classifier: A Cross-Sectional Retrospective Analysis in a Level I Trauma Center

Background: In trauma patients, pancreatic injury is rare; however, if undiagnosed, it is associated with high morbidity and mortality rates. Few predictive models are available for the identification of pancreatic injury in trauma patients with elevated serum pancreatic enzymes. In this study, we aimed to construct a model for predicting pancreatic injury using a decision tree (DT) algorithm, along with data obtained from a population-based trauma registry in a Level I trauma center. Methods: A total of 991 patients with elevated serum levels of amylase (>137 U/L) or lipase (>51 U/L), including 46 patients with pancreatic injury and 865 without pancreatic injury between January 2009 and December 2016, were allocated in a ratio of 7:3 to training (n = 642) or test (n = 269) sets. Using the data on patient and injury characteristics as well as laboratory data, the DT algorithm with Classification and Regression Tree (CART) analysis was performed based on the Gini impurity index, using the rpart function in the rpart package in R. Results: Among the trauma patients with elevated amylase or lipase levels, three groups of patients were identified as having a high risk of pancreatic injury, using the DT model. These included (1) 69% of the patients with lipase level ≥306 U/L; (2) 79% of the patients with lipase level between 154 U/L and 305 U/L and shock index (SI) ≥ 0.72; and (3) 80% of the patients with lipase level <154 U/L with abdomen injury, glucose level <158 mg/dL, amylase level <90 U/L, and neutrophil percentage ≥76%; they had all sustained pancreatic injury. With all variables in the model, the DT achieved an accuracy of 97.9% (sensitivity of 91.4% and specificity of 98.3%) for the training set. In the test set, the DT achieved an accuracy of 93.3%, sensitivity of 72.7%, and specificity of 94.2%. Conclusions: We established a DT model using lipase, SI, and additional conditions (injury to the abdomen, glucose level <158 mg/dL, amylase level <90 U/L, and neutrophils ≥76%) as important nodes to predict three groups of patients with a high risk of pancreatic injury. The proposed decision-making algorithm may help in identifying pancreatic injury among trauma patients with elevated serum amylase or lipase levels.

Derivation and validation of different machine-learning models in mortality prediction of trauma in motorcycle riders: a cross-sectional retrospective study in southern Taiwan

Objectives This study aimed to build and test the models of machine learning (ML) to predict the mortality of hospitalised motorcycle riders. Setting The study was conducted in a level-1 trauma centre in southern Taiwan. Participants Motorcycle riders who were hospitalised between January 2009 and December 2015 were classified into a training set (n=6306) and test set (n=946). Using the demographic information, injury characteristics and laboratory data of patients, logistic regression (LR), support vector machine (SVM) and decision tree (DT) analyses were performed to determine the mortality of individual motorcycle riders, under different conditions, using all samples or reduced samples, as well as all variables or selected features in the algorithm. Primary and secondary outcome measures The predictive performance of the model was evaluated based on accuracy, sensitivity, specificity and geometric mean, and an analysis of the area under the receiver operating characteristic curves of the two different models was carried out. Results In the training set, both LR and SVM had a significantly higher area under the receiver operating characteristic curve (AUC) than DT. No significant difference was observed in the AUC of LR and SVM, regardless of whether all samples or reduced samples and whether all variables or selected features were used. In the test set, the performance of the SVM model for all samples with selected features was better than that of all other models, with an accuracy of 98.73%, sensitivity of 86.96%, specificity of 99.02%, geometric mean of 92.79% and AUC of 0.9517, in mortality prediction. Conclusion ML can provide a feasible level of accuracy in predicting the mortality of motorcycle riders. Integration of the ML model, particularly the SVM algorithm in the trauma system, may help identify high-risk patients and, therefore, guide appropriate interventions by the clinical staff.

Stress-Induced Hyperglycemia in Diabetes: A Cross-Sectional Analysis to Explore the Definition Based on the Trauma Registry Data

Background: The diagnosis of diabetic hyperglycemia (DH) does not preclude a diabetes patient from having a stress-induced hyperglycemic response. This study aimed to define the optimal level of elevated glucose concentration for determining the occurrence of stress-induced hyperglycemia (SIH) in patients with diabetes. Methods: This retrospective study reviewed the data of all hospitalized trauma patients, in a Level I trauma center, from 1 January 2009 to 31 December 2016. Only adult patients aged ≥20 years, with available data on serum glucose and glycated hemoglobin A1c (HbA1c) levels upon admission, were included in the study. Long-term average glucose levels, as A1c-derived average glucose (ADAG), using the equation, ADAG = ((28.7 × HbA1c) − 46.7), were calculated. Patients with high glucose levels were divided into three SIH groups with diabetes mellitus (DM), based on the following definitions: (1) same glycemic gap from ADAG; (2) same percentage of elevated glucose of ADAG, from which percentage could also be reflected by the stress hyperglycemia ratio (SHR), calculated as the admission glucose level divided by ADAG; or (3) same percentage of elevated glucose as patients with a defined SIH level, in trauma patients with and without diabetes. Patients with incomplete registered data were excluded. The primary hypothesis of this study was that SIH in patients with diabetes would present worse mortality outcomes than in those without. Detailed data of SIH in patients with diabetes were retrieved from the Trauma Registry System. Results: Among the 546 patients with DH, 332 (32.0%), 188 (18.1%), and 106 (10.2%) were assigned as diabetes patients with SIH, based on defined glucose levels, set at 250 mg/dL, 300 mg/dL, and 350 mg/dL, respectively. In patients with defined cut-off glucose levels of 250 mg/dL and 300 mg/dL, SIH was associated with a 3.5-fold (95% confidence interval (CI) 1.61–7.46; p = 0.001) and 3-fold (95% CI 1.11–8.03; p = 0.030) higher odds of mortality, adjusted by sex, age, pre-existing comorbidities, and injury severity score, than the 491 patients with diabetic normoglycemia (DN). However, in patients with a defined cut-off glucose level of 350 mg/dL, adjusted mortality in SIH in DM was insignificantly different than that in DM. According to the receiver operating characteristic (ROC) curve analysis, a blood sugar of 233 mg/dL, a glycemic gap of 79 (i.e., blood sugar of 251 mg/dL), and a SHR of 1.45 (i.e., blood sugar of 250 mg/dL) were identified as cut-offs for mortality outcomes, with AUCs of 0.622, 0.653, and 0.658, respectively. Conclusions: In this study, a cut-off glucose level of 250 mg/dL was selected to provide a better definition of SIH in DM than glucose levels of 300 mg/dL or 350 mg/dL.

Comparison of the new Exponential Injury Severity Score with the Injury Severity Score and the New Injury Severity Score in trauma patients: A cross-sectional study

Objective To compare Exponential Injury Severity Score (EISS) with Injury Severity Score (ISS) and New Injury Severity Score (NISS) in terms of their predictive capability of the outcomes and medical expenses of hospitalized adult trauma patients. Setting This study was based at a level I trauma center in Taiwan. Methods Data for 17,855 adult patients hospitalized from January 1, 2009 to December 31, 2015 were retrieved from the Trauma Registry System. The primary outcome was in-hospital mortality. Secondary outcomes were the hospital length of stay (LOS), intensive care unit (ICU) admission rate, ICU LOS, and medical expenses. Chi-square tests were used for categorical variables to determine the significance of the associations between the predictor and outcome variables. Student t-tests were applied to analyze normally distributed data for continuous variables, while Mann-Whitney U tests were used to compare non-normally distributed data. Results According to the survival rate-to-severity score relationship curve, we grouped all adult trauma patients based on EISS scores of ≥ 27, 9–26, and < 9. Significantly higher mortality rates were noted in patients with EISS ≥ 27 and those with EISS of 9–26 when compared to patients with EISS < 9; this finding concurred to the findings for groups classified by the ISS and NISS with the cut-off points set between 25 and 16. The hospital LOS, ICU admission rates, and medical expenses for patients with EISS ≥ 27 and patients with EISS of 9–26 were also significantly longer and higher than that of patients with EISS < 9. When comparing the demographics and detailed medical expenses of very severely injured adult trauma patients classified according to ISS, NISS, and EISS, patients with ISS ≥ 25 and NISS ≥ 25 both had significantly lower mortality rates, lower ICU admission rates, and shorter ICU LOS compared to patients with EISS ≥ 27. Conclusions EISS 9 and 27 can serve as two cut-off points regarding injury severity, and patients with EISS ≥ 27 have the greatest injury severity. Additionally, these patients have the highest mortality rate, the highest ICU admission rate, and the longest ICU LOS compared to those with ISS ≥ 25 and NISS ≥ 25, suggesting that patients with EISS ≥ 27 have the worst outcome.

Same Abbreviated Injury Scale Values May Be Associated with Different Risks to Mortality in Trauma Patients: A Cross-Sectional Retrospective Study Based on the Trauma Registry System in a Level I Trauma Center

The Abbreviated Injury Scale (AIS) measures injury severity of a trauma patient with a numeric method for ranking anatomy-based specific injuries. The summation of the squares of the three most severe injuries in the AIS of six predefined body regions comprises the Injury Severity Score (ISS). It assumes that the mortality of a given AIS value is similar across all body regions. However, such an assumption is less explored in the literature. In this study, we aimed to compare the mortality rates of the patients with the same AIS value in different injured body regions in a level I trauma center. Hospitalized adult trauma patients with isolated serious to critical injury (AIS of 3 to 5) between 1 January 2009, and 31 December 2016, from the Trauma Registry System in a level I trauma center were grouped according to the injured body regions (including, the head/neck, thorax, abdomen, or extremities) and were exclusively compared according to their AIS stratum. Categorical data were compared using the two-sided Fisher exact or Pearson chi-square tests. ANOVA with Games-Howell post hoc test was performed to assess the differences in continuous data of the patients with injury in different body regions. The primary outcome of the study was in-hospital mortality. The adjusted odds ratios (AORs) were estimated using a stepwise selection of a multivariable regression model adjusted by controlling the confounding variables such as sex, age, comorbidities, and ISS. Survival curves were estimated with the Kaplan–Meier approach with a corresponding log-rank test. The patients with AIS of 5 for abdomen injury and those with AIS of 3 for extremity injury had a significantly lower odds of adjusted mortality (adjusted odds ratio (AOR) 0.1, 95% confidence interval (CI) 0.01–0.39, p = 0.004 and AOR 0.3, 95% CI 0.15–0.51, p < 0.001, respectively) than that of the patients with head/neck injury. However, the patients with AIS of 4 for extremity injury demonstrated significantly higher odds of adjusted mortality (AOR 8.4, 95% CI 2.84–25.07, p < 0.001) than the patients with head/neck injury. This study found that the risks to mortality in the patients with a given AIS value of serious to critical injury in different injured body regions were not the same, even after controlling for confounding variables such as sex, age, comorbidities, and ISS.

Effect of Age on Glasgow Coma Scale in Patients with Moderate and Severe Traumatic Brain Injury: An Approach with Propensity Score-Matched Population

Background: The most widely used methods of describing traumatic brain injury (TBI) are the Glasgow Coma Scale (GCS) and the Abbreviated Injury Scale (AIS). Recent evidence suggests that presenting GCS in older patients may be higher than that in younger patients for an equivalent anatomical severity of TBI. This study aimed to assess these observations with a propensity-score matching approach using the data from Trauma Registry System in a Level I trauma center. Methods: We included all adult patients (aged ≥20 years old) with moderate to severe TBI from 1 January 2009 to 31 December 2016. Patients were categorized into elderly (aged ≥65 years) and young adults (aged 20–64 years). The severity of TBI was defined by an AIS score in the head (AIS 3‒4 and 5 indicate moderate and severe TBI, respectively). We examined the differences in the GCS scores by age at each head AIS score. Unpaired Student’s t- and Mann–Whitney U-tests were used to analyze normally and non-normally distributed continuous data, respectively. Categorical data were compared using either the Pearson chi-square or two-sided Fisher’s exact tests. Matched patient populations were allocated in a 1:1 ratio according to the propensity scores calculated using NCSS software with the following covariates: sex, pre-existing chronic obstructive pulmonary disease, systolic blood pressure, hemoglobin, sodium, glucose, and alcohol level. Logistic regression was used to evaluate the effects of age on the GCS score in each head AIS stratum. Results: The study population included 2081 adult patients with moderate to severe TBI. These patients were categorized into elderly (n = 847) and young adults (n = 1234): each was exclusively further divided into three groups of patients with head AIS of 3, 4, or 5. In the 162 well-balanced pairs of TBI patients with head AIS of 3, the elderly demonstrated a significantly higher GCS score than the young adults (14.1 ± 2.2 vs. 13.1 ± 3.3, respectively; p = 0.002). In the 362 well-balanced pairs of TBI patients with head AIS of 4, the elderly showed a significantly higher GCS score than the young adults (13.1 ± 3.3 vs. 12.2 ± 3.8, respectively; p = 0.002). In the 89 well-balance pairs of TBI patients with head AIS of 5, no significant differences were observed for the GCS scores. Conclusions: This study demonstrated that elderly patients with moderate TBI present higher GCS score than younger patients. This study underscores the importance of determining of TBI severity in this group of elderly patients based on the GCS score alone. A lower threshold of GCS cutoff should be adopted in the management of the elderly patients with TBI.

Epidemiology of Bone Fracture in Female Trauma Patients Based on Risks of Osteoporosis Assessed using the Osteoporosis Self-Assessment Tool for Asians Score

Background: Osteoporotic fractures are defined as low-impact fractures resulting from low-level trauma. However, the exclusion of high-level trauma fractures may result in underestimation of the contribution of osteoporosis to fractures. In this study, we aimed to investigate the fracture patterns of female trauma patients with various risks of osteoporosis based on the Osteoporosis Self-Assessment Tool for Asians (OSTA) score. Methods: According to the data retrieved from the Trauma Registry System of a Level I trauma center between 1 January 2009 and 31 December 2015, a total of 6707 patients aged ≥40 years and hospitalized for the treatment of traumatic bone fracture were categorized as high-risk (OSTA < −4, n = 1585), medium-risk (−1 ≥ OSTA ≥ −4, n = 1985), and low-risk (OSTA > −1, n = 3137) patients. Two-sided Pearson’s, chi-squared, or Fisher’s exact tests were used to compare categorical data. Unpaired Student’s t-test and Mann–Whitney U-test were used to analyze normally and non-normally distributed continuous data, respectively. Propensity-score matching in a 1:1 ratio was performed with injury mechanisms as adjusted variables to evaluate the effects of OSTA-related grouping on the fracture patterns. Results: High- and medium-risk patients were significantly older, had higher incidences of comorbidity, and were more frequently injured from a fall and bicycle accident than low-risk patients did. Compared to low-risk patients, high- and medium-risk patients had a higher injury severity and mortality. In the propensity-score matched population, the incidence of fractures was only different in the extremity regions between high- and low-risk patients as well as between medium- and low-risk patients. The incidences of femoral fractures were significantly higher in high-risk (odds ratio [OR], 3.4; 95% confidence interval [CI], 2.73–4.24; p < 0.001) and medium-risk patients (OR, 1.4; 95% CI, 1.24–1.54; p < 0.001) than in low-risk patients. In addition, high-risk patients had significantly lower odds of humeral, radial, patellar, and tibial fractures; however, such lower odds were not found in medium- risk than low-risk patients. Conclusions: The fracture patterns of female trauma patients with high- and medium-risk osteoporosis were different from that of low-risk patients exclusively in the extremity region.