Blunt Abdominal Trauma in Adults: A Score to Predict the Absence of Organ Injury

Transcription

1 Blunt Abdominal Trauma in Adults: A Score to Predict the Absence of Organ Injury M.E.R. Bongers s Supervisor: J.B.F. Hulscher, MD/PhD Department of Surgery University Medical Centre Groningen University of Groningen The Netherlands

2 Abstract Objective The aim of this study is to generate a scoring system to identify and/or rule out the presence of intra-abdominal organ injury in trauma patients with blunt abdominal trauma, without the use of CT scan. The scoring system will be designed in such a manner that it can be performed in the Emergency room (ER) using the same data collected during initial trauma care, so it will not interfere with the ATLS protocol. Methods This retrospective study included all patients aged >16 years that were categorized as a socalled A-trauma by the paramedical team and subsequently admitted to the UMCG. All parameters needed for this study were extracted from the EPD. A multivariate logistic regression was performed with the covariates: Free fluid in the abdomen on FAST, Sex, Age, ALAT, Hemoglobin, Creatinine, Base Excess, Amylase, Fibrinogen, INR, LDH and Shock Index to identify predictors for the presence of abdominal injury. Results Free fluid on FAST, Shock Index, Creatinine, INR and LDH are the best predictors for the presence of abdominal injury. These predictors were combined to generate the BATiA-score. Free fluid on FAST is highly suggestive for the presence of abdominal injury with an odds ratio of: 416,297. The ABAT score was also generated and after a multiple logistic regression: Creatinine, LDH, Shock Index, Sex, ALAT and Hemoglobin seem to be the best predictors for the presence of abdominal injury in a setting where no sonographic scanner is available. Conclusions A score to predict the presence of abdominal injury (BATiA) was generated, free fluid on FAST proved to be highly predictive for the presence of abdominal injury and comprised the major part of the score. Therefore, we suggest that the FAST procedure is performed in every trauma patient admitted to the ER, only with a positive result of the FAST further abdominal CT-scanning should be performed. The ABAT proved to be a promising score for institutions where no sonographic scanners are available. Further research has to be done to improve and validate this score to be a helpful aid in third world countries. 2

3 Index Introduction Abdominal Injury Physiologic and Immunologic response to trauma Advanced Trauma Life Support Physical examination in Trauma care Laboratory findings in Trauma care Focused Assessment Sonography in Trauma (FAST) Computed Tomography in Trauma Radiology in Trauma care, United States versus Europe The current diagnostics, what can improve? Improvement of the ATLS with a scoring system Currently existing scoring systems Hypotheses Aim of the study Patients and Methods Inclusion Exclusion Blunt Abdominal Trauma Abdominal Pain Radiology FAST Computed Tomography Patients and characteristics ISS, EMTRAS and Shock Index ISS EMTRAS Shock Index Abdominal versus non-abdominal injury Statistics Results Patients and Characteristics Logistic regression Hosmer-Lemeshow and calibration Receiver operating characteristics and Area under Curve Free Fluid on FAST A prediction model without FAST

4 3.7 Calibration and discrimination for the ABAT Discussion FAST is a strong predictor for abdominal injury Patient characteristics Abdominal pain Patients with a negative FAST but with abdominal injuries Imputation Limitations Prospects Conclusion Acknowledgments References

5 Introduction 1.1 Abdominal Injury The treatment of severely injured patients relies on efficient diagnostic intervention and rapid life-preserving therapy. Three different mortality peaks are observed in which death occurs after trauma: the trimodal distribution of deaths. In the first period death usually occurs in seconds or minutes after the injury. These patients die from apnea due to neurological damage or ruptures of the heart, aorta or other large blood vessels. This group represents half of the patients that die due to trauma (1). During the second period, which comprises the first hours after the injury, deaths are usually due to cerebral hematomas or significant blood loss elsewhere, such as hematopneumothorax, hemorrhages in the abdomen or pelvic fractures. The second period accounts for 30 percent of the deceased trauma patients (1). The third peak, responsible for the remaining 20 percent of deceased trauma patients, occurs several days or weeks after the injury. These patients die as a result of severe sepsis or multiple organ failure (1). In the Netherlands more than 1700 patients died in the hospital (second and third period) in 2011 (2). The present study focuses on preventing patients from dying during the second mortality peak, those that decease in the first hours after injury. In the Netherlands registered trauma patients were brought to an Emergency Department in the year Of these patients, 2729 had an abdominal injury of which 636 were severely injured patients (2). These data were derived from a survey representing 89 percent of the emergency departments in the Netherlands (2). In the Northern part of the Netherlands, where the Blunt Abdominal Trauma in Adults (BATiA) study takes place, all emergency departments submitted data (3). In this region there were 7479 trauma patients of which 66 suffered a severe abdominal trauma in the same year as mentioned above (3). The average age of the trauma patient in the Northern part of the Netherlands was 55 years old in 2011, with a standard deviation of 28. The largest group of trauma patients where 80 years of age or older, the rest of the patients were divided equally over the other age groups, which are groups of each 10 years (3). The Dutch Landelijke Trauma Registratie (LTR) divides two different ways of subdividing the injury type of the trauma. First one can distinguish between blunt force trauma and sharp/penetrating trauma. In the Netherlands, 96 percent of all injuries are due to blunt force trauma, 3 percent to penetrating trauma and 1 percent is undecided. One can also make a division by body region. Abdominal injury occurs in two percent of trauma patients (3). Abdominally injured patients are frequently polytrauma patients making it difficult to calculate the actual mortality for patients with blunt abdominal trauma (BAT). 1.2 Physiologic and Immunologic response to trauma When a trauma occurs, multiple defense mechanisms are put into effect to establish a normal homeostasis in the body. A correlation has been found between the severity of the trauma and the response of these defense mechanisms (4). The sympathetic system is activated by the brain and the hypothalamus due to a reaction on the stress response. Catecholamines are released from the adrenal medulla and vasopressin from the pituitary gland (5). Due to the activation of the sympathetic system the cardiac output and the peripheral vasoconstriction are increased, leading to an increase in blood flow to the vital organs, such as the brain, liver and kidneys causing a lesser perfusion of other intra-abdominal organs(6). 5

6 1.3 Advanced Trauma Life Support In 1976 a surgeon crashed with an airplane with his family in it. Both he and other members of his family sustained critical injuries. The surgeon recognized that the initial trauma care was inadequate. He developed a way to improve the initial trauma care, two years later the first version of the Advanced Trauma Life Support (ATLS) was used in the U.S.(1). Studies have shown that after the initiation of the present ATLS protocol the mortality has decreased in trauma patients (1,7,8). The goal of the ATLS protocol is to decrease the mortality in the first hours after admission. One of the main points of attention is to treat first what kills first ; other injuries are attended to later. When a patient enters the ER the primary survey is initiated. Vital functions of the patient must be evaluated quickly and efficiently. The ABCDE principle is used during the primary survey to identify and rule out life-threatening conditions (see paragraph 1.4). The ABCDE principal needs to be performed from A to E, not skipping any steps, if lifethreatening conditions appear they must be managed immediately. After the primary survey is completed the next ATLS step is the secondary survey. During the secondary survey a head-to-toe evaluation of the patient is made. The whole body, including all extremities and a reassessment of the vital signs are examined. X-rays are made, if indicated by examination (1). 1.4 Physical examination in Trauma care As mentioned in the previous section the primary survey of the ATLS protocol the physical examination is carried out by the ABCDE principle in the following sequence: 1. Airway maintenance with cervical spine protection 2. Breathing and ventilation 3. Circulation with hemorrhage control 4. Disability: Neurologic status 5. Exposure/Environmental control: Completely undress the patient, but prevent hypothermia If airway problems occur, these must be managed before any breathing and ventilation problems. These are then treated before circulation problems, like hemorrhages and shock. A neurologic evaluation, using the Glascow Coma Scale (GCS) and a clinical inspection by undressing the patient whilst preventing hypothermia, completes the primary survey (1). During the primary survey, Focused Assessment Sonography in Trauma (FAST), x-rays of the cervical spine, chest and pelvis can be made if they are not interfering in the primary survey. Also pulse oximetry, blood pressure and other monitoring can be examined in the primary survey (1). If any deterioration occurs in any stage of the trauma care, urgent reassessment of the ABCDE protocol of the primary survey must be performed (1). The secondary survey can begin after the primary survey is completely finished, resuscitative efforts are underway, and if the vital functions have been normalized (1). 6

7 As stated in the first paragraph, 80 percent of the trauma patients die in the first hours after injury, of which 30 percent die in the hospital. For these associated injuries a fast and decisive method of identification or ruling out is essential, especially for the investigation of the abdomen, because intraabdominal hemorrhages are difficult to identify. 1.5 Laboratory findings in Trauma care An accurate biomarker for the presence of intraabdominal injury has not yet been found. Although for several organs there is proof that an elevated value of an individual biomarker in the blood or urine can indicate the presence of organ damage. ALT, AST, LDH and GGT levels are significantly higher in patients with various grades of liver injury (9). Another study shows that a raised ALT level after BAT has a 100 percent sensitivity for liver injury (10). Pediatric patients who developed acute renal injury after BAT show a significant higher creatinine kinase (CK) and creatine however only rhabdomyolysis after trauma has been associated with renal impairments in adults (11). Levels of CRP, base deficit, lactate, LDH, AST and D- dimer are significantly higher if intestinal ischemia has occurred. In the same study the Intestinal Fatty Acid Binding Protein (I-FABP) was found to be a promising biomarker for detecting vascular ischemia in the intestine (12). However, these finding have not yet been studied in trauma patients. Airway OK Breathing OK Circulation OK Disability OK Environmental Secondary Survey Not OK Not OK Not OK Not OK Not OK Control Airway Control Breathing Control Circlulation Control Disability Control Environmental Often Repeat Fig.1 Flowchart of the ABCDE principal in the ATLS 1.6 Focused Assessment Sonography in Trauma (FAST) FAST is a rapid non-invasive examination tool for identifying the presence of intra-abdominal hemorrhage. For the assessment of the circulatory survey, FAST is essential for hemodynamic unstable patients, in which injuries need to be managed immediately (13). Every patient who suffered from a high-energetic trauma should be examined by FAST. The fact that FAST is a non-invasive, inexpensive, portable and easy to use procedure, determines that it is a preferred way to examine the abdomen. Also FAST does not expose ionizing radiation and can be repeated easily at the bedside of the patient (1). FAST also has limitations; the sensitivity of FAST for intra-peritoneal fluid extends from 63 to 100 percent (14,15). This makes it difficult to rule out the presence of intra-abdominal injuries. Another disadvantage of sonography is that its value is highly dependent on the experience of the person performing it (1). However, FAST has a high specificity, therefore a positive FAST (free fluid (FF) in the abdomen) is highly suggestive for a hemoperitoneum and therefore for the presence of abdominal injury (16). 1.7 Computed Tomography in Trauma Computed Tomography (CT) is a diagnostic procedure which can identify the presence of intra-abdominal injuries that are difficult to assess with other non-invasive examinations. This procedure should only be used in patients who are hemodynamically stable, and have no 7

8 indication for an emergency laparotomy (1). CT is currently the golden standard diagnostic procedure to identify intra-abdominal injuries (17). The advantages of CT include that it is non-invasive and it gives a good estimate of the severity of the injury. The CT-scan not only detects the presence of organ injury, it also defines where the focus of the hemorrhage is located (18). The CT-scan also has disadvantages. For the radiation exposure of the CT has been proven that after a single CT-scan the risk for cancer increases, especially for the abdomen and pelvis. Older patients have a reduced chance compared to younger patients to develop cancer after receiving the radiation of a single CT-scan (19). This means that a patient who underwent a CT-scan at age eighteen has more change to develop cancer than a patient where the same CT-scan was performed at age fifty-five. Another study shows that after a single abdominal CT-scan without contrast performed on 20 year old patients, 1 in 500 women and 1 in 660 men will develop cancer (20). As mentioned in the first paragraph, trauma patients are divided equally over age groups, making it desirable to avoid CT scans in especially the younger patients. The diaphragm, bowel and some pancreatic injuries can easily be missed on the CT-scan, thus the CT is not ruling out all intra-abdominal injuries (1). Also, for some CT investigations the patients have to hold their breath, which some cannot manage due to their injuries. Another disadvantage is that the therapeutic yield of the CT-scan has not yet been investigated thoroughly. The CT-scan might show new diagnoses, but the outcome of the scan rarely changes the mode of treatment (21). Furthermore patients have to be transported to and from the radiology department, where monitoring and intervention options are suboptimal. This is also a time consuming procedure, which should be avoided in the golden hour of initial trauma care. Nellensteijn writes that the intra and inter observer agreement is only moderate in grading liver damage in children and states: Only the most experienced radiologist demonstrated good intra-observer agreement which might indicate the necessity of the presence of a senior trauma radiologist at all times (22). Another disadvantage of CT scanning is the high cost of the procedure. 1.8 Radiology in Trauma care, United States versus Europe. There is a substantial difference in trauma care between the United States and Europe. In Europe the ATLS is followed by FAST and an x-ray of the thorax, pelvis and spine. These X-ray Thorax, Pelvis and Spine Positive Abdominal CT Fig. 2 The European way of Radiological diagnosis in Trauma patients Patient arrives on Emergency Room ATLS FAST Negative No further abdominal imaging radiographic investigations are almost always adjunct to the primary survey, all the other imaging should be postponed to the secondary survey (23). If there are signs of damage elsewhere, radiologic imaging will be performed to find injuries. 8

9 In the US they have the tendency to perform a total body CT-scan (TBCT) to check for injuries after the ATLS. In the REACT-trial the TBCT method is currently investigated (24). This method might not be the best way of radiological imaging, due to the high radiation pollution and high cost mentioned earlier. Patient arrives on Emergency Room ATLS Total Body CT Fig. 3 The path of radiological diagnosis for Trauma patients often performed in the United States 1.9 The current diagnostics, what can improve? At the moment there is no single diagnostic intervention to identify or rule-out intraabdominal injuries. As mentioned above, CT-scanning is used to detect these injuries frequently. However the disadvantages of the CT-scan are vast, making it a diagnostic procedure to avoid when possible. Taking this into account, there are only few methods to non-invasively examine intra-abdominal injuries without an excess in radiation exposure and time-loss, which occurs during CT. These methods have all investigated just one of the aspects of initial trauma care, i.e. physical examination or laboratory findings, and not the overall view, including laboratory findings, physical examination and FAST (25). A scoring method for the presence of abdominal injuries in children has been proposed by Holmes; this method both covers physical examination and laboratory findings (26). However, this algorithm is tested only in children and has not yet been validated Improvement of the ATLS with a scoring system At the moment CT imaging in the ER is the golden standard for diagnosing abdominal injuries. However, taking the disadvantages into account and the fact that it rarely leads to a change in treatment it can be concluded that too many CT-scans are made. A clinical score that can rule out abdominal injuries would be a helpful adjunct to the ATLS protocol. This score could identify which patients would benefit from a CT scan and which would not. In this way the amount of radiation and costs could be reduced and time could be saved in the golden hour Currently existing scoring systems A prediction method generated by Poletti in 2004, showed a low specificity and thus only ruled out a small percentage of the intra-abdominal injuries. However this method showed that findings by physical examination can be helpful in predicting the presence of abdominal injuries (27). The Blunt Abdominal Trauma in Children (BATiC) score is an existing method for children to rule out abdominal injuries (28). This method has been externally validated in the UMCG, with 100 percent specificity if the BATiC score is lower than six. This method is a good guidance for the Blunt Abdominal Trauma in Adults (BATiA) score, because it uses parameters readily available from physical examination, laboratory findings and FAST, without interfering in the ATLS sequence. A method by Holmes previously mentioned is a good method to rule out the presence of intra-abdominal injuries; however it has not yet been validated (26). The BATiA might consist of some of the features of the scoring system by Holmes and/or the BATiC score. 9

10 The Emergency Trauma Score (EMTRAS) consists of four parameters which can be easily obtained in the Emergency Room (ER); age, GCS, base excess and prothrombin time (29-31). This scoring method might be a predicting parameter in this study and will be explained later on in the methods-section Hypotheses We expect that a scoring method can be generated, by using non-invasive findings, to rule out the presence of abdominal injuries after blunt abdominal trauma Aim of the study The aim of this study is to generate a scoring system to identify and/or rule out the presence of intra-abdominal organ injury in trauma patients with blunt abdominal trauma, without the use of CT scan. The scoring system will be designed in such a manner that it can be performed in the Emergency room (ER) using the same data collected during initial trauma care, so it will not interfere with the ATLS protocol. 10

11 Patients and Methods 2.1 Inclusion The source of patient inclusion is a trauma registry that prospectively records all trauma patients who are admitted alive to the ER of the University Medical Center Groningen (UMCG). The paramedic subdivides all these patients into groups, on scene of the accident, based on the level medical urgency. All patients with a code red also called A-trauma, the highest level of urgency, were included thus creating a group consisting of the most severe injuries for analysis. 2.2 Exclusion Excluded were all patients aged <16 and those that were not admitted to the hospital. 2.3 Blunt Abdominal Trauma Blunt abdominal trauma or non-penetrating abdominal trauma refers to physical trauma caused to the abdomen either by impact, injury or physical attack. Blunt trauma differs from penetrating or sharp trauma, in which objects, i.e. knife or bullet, enter the body. 2.4 Abdominal Pain Any sign that could indicate abdominal pain was defined as abdominal pain. If a patients indicate that they have abdominal pain this was defined as such. Also if a patient has a muscular defense this was defined as abdominal pain. 2.5 Radiology The radiologic examinations where interpreted by a radiologist. If any doubt was present a senior surgeon was asked for his opinion. FAST We defined ultrasound as abnormal if there was either free fluid, whatever the quantity or location, or any abnormal finding that could suggest organ injury. We have divided three groups: 1. ultrasound abnormality, 2. free fluid and 3. organ injury, in which ultrasound abnormality is a composite endpoint, consisting of free fluid and/or organ injury. Computed Tomography The results of the CT-scans have also been re-interpreted. We extracted if any CT-scans were performed and also if any abdominal CT-scans were performed. The outcomes of the scans were collected via the electronic patient database (EPD). We collected data if there was free fluid visible on CT or if there were other signs of organ damage. The location of these findings was also collected. If organ injury was visible the organs which were injured were noted, i.e.; liver, spleen, kidneys, bowel. If any free fluid was visible the location where the fluid was collected was noted, i.e.; pouch of Douglas, pelvis, paracolic gutters, mesentery, Morison's pouch, perihepatic or perisplenic spaces (32). 2.6 Patients and characteristics The following data were manually extracted from the EPD of the UMCG: Age, gender, mechanism of trauma, trauma energy (high energy being defined as motor vehicle accident >60 km/h, fall from a height >5 meters, fall from a horse in movement, projection far away from a bicycle, or loss of consciousness at the scene for more than 15 minutes). 11

12 Also the time from trauma to blood workup, The Injury Severity Score (ISS), The Emergency Trauma Score (EMTRAS), pulse, blood pressure, Mean Arterial Pressure (MAP), GCS, presence of abrasions (cave seatbelt sign) and abdominal pain have been extracted. The following laboratory findings have been collected: Hemoglobin (Hb) (mmol/l), Hematocrit (Ht) (v/v), Leucocytes (10E9/L), Platelets/Trombocytes (10E9/L), Prothrombin time (PT) (sec.) Activated partial thromboplastin time (aptt) (sec.), Lactate dehydrogenase (LDH) (U/L), aspartate aminotransferase (AST) (U/L), alanine aminotransferase (ALT) (U/L), gamma-glutamyltransferase (γ-gt) (U/L), total bilirubin (umol/l), direct bilirubin (umol/l), Amylase (U/L), sodium (mmol/l), potassium (mmol/l), Urea (mmol/l), Creatinine (umol/l), Alkaline phosphatase (AF) (U/L), BSE (mm/hour), CRP (mg/l), Erytrocytes (10E12/L), Base excess (ABE) (mmol/l), Lactate (mmol/l), egfr (m/m 1.73), ethanol (g/l), Fibrinogen (g/l), glucose (mmol/l) and International Normalized Ratio (INR) (x.) The collected laboratory findings are from a standard blood sample obtained during the trauma-screening and therefore no extra laboratory findings were done for this research Means of treatment, i.e. surgical, medicinal or conservative treatment, was collected and interpreted. Whether any complications occurred during the stay in the hospital was also determined. 2.7 ISS, EMTRAS and Shock Index ISS The Injury Severity Score (ISS) is a method for scoring the severity of the injury in trauma patients. It ranges from 0 (no injuries) to 75 (unsurvivable). The Abbreviated Injury Scale grade (AIS) of the three most injured body regions are first squared and then added together to calculate the ISS. The AIS is an anatomically based scoring system for each part of the body, grading injuries from 1 (minor injuries) to 6 (lethal injuries). When a patient scores 6 on any of the body regions the patient will automatically have an ISS of 75(33,34). Injury AIS Score Body Region ISS score Minor 1 Head and Neck Moderate 2 Face Serious 3 Thorax Severe 4 Abdomen and visceral pelvis Critical 5 The bony pelvis and extremities Unsurvivable 6 External structures Table 1. The Injury Severity Score (ISS) EMTRAS The determination of the Emergency Trauma Score (EMTRAS) can inform practitioners of the severity of injury of trauma patients at an early stage. It has been devised by Raum in 2009 in the UMGC and has been validated by Peris and Joosse in 2013 (29-31). The EMTRAS combines four parameters, which can be collected while in the ER and predicts mortality. The four parameters; age, GCS, base excess and prothrombin time, are assigned a score from 0 to 3. All four scores are summed to obtain the EMTRAS, now ranging from 0 to

13 Age < >75 GCS Base excess >-1-5 to to -5 < -10 Prothrombin time <80% 80% - 50% 49% - 20% >20% Table 2. Emergency Trauma Scoring system Shock Index The Shock Index (SI) is generated trough dividing the Heart Rate (HR) by the Systolic Blood pressure (SBP), i.e. SI = HR / SBP. A higher SI indicates a more severe hemodynamic instability (35). 2.8 Abdominal versus non-abdominal injury In this study we divided the patients into two groups. The first group consists of only patient with abdominal injuries; the other group consists of those without abdominal injuries. A patient was identified with intra-abdominal injury when there was evidence of intraabdominal injury on CT scan or if found during surgery. 2.9 Statistics To account for missing values, multiple imputation was used. When multiple imputation is used, missing values for a parameter are predicted using existing values from other parameters. This method is performed multiple times; generating multiple imputation sets (36). In this study 10 imputation sets were created, with a maximum of 50 iterations. Next a univariate analysis is performed on the original dataset to assess which individual parameters are risk factors for abdominal injuries. For the parameters the means and standard deviations (SD) were calculated and the t-test or Mann-Whitney-U test was used to compare the continuous parameters. Proportions were calculated, and the chi-square test was used to compare categorical parameters. Twelve chosen parameters with a significant (p <0,05) relation with organ injury in the univariate analysis, were included in a multivariate logistic regression analysis with a backward selection method, using the Akaike Information Criterion (AIC) as the significant value (p <0.157) to build a prediction model for abdominal injury (37). To test the performance of the created score the following tests were done: 1. A calibration line for the goodness of fit was made using the data derived with the Hosmer-Lemeshow test. This is a test to define how well a logistic regression model performs based on observed and expected values, also defined goodness of fit. The Hosmer-Lemeshow test assesses if the observed events match expected events in subgroups of the model population. The Hosmer-Lemeshow test makes ten subgroups based on the predicted probability and hereby divides patients in each imputation dataset into subgroups. The observed and the expected number of patients in each subgroup were both divided by the total number of patients in each subgroup. The average of all ten imputation datasets was then calculated for each subgroup. These values were plotted in the calibration model. The perfect calibration line is noted as: y = x + 0, the observed values match the expected values completely (38). Thus, a calibration line closely related to the perfect calibration line is suggestive for a good prediction model. 2. The Receiver Operating Characteristic curve (ROC-curve) and the Area under the Curve (AUC) were calculated to represent discrimination of the model. The sensitivity 13

14 and specificity were calculated for all the cut-off points and plotted in a Receiver operating characteristic curve (ROC-curve). A larger AUC suggests that the score has a better ability to correctly classify those with and without the abdominal injury (39). Statistical analysis is done using Microsoft Excel 2010 and SPSS

15 Results 3.1 Patients and Characteristics Patients admitted to the UMCG s shock room between the 20 th of April 2006 until the 31 st of December 2011 were included. Patients were excluded for the following reasons: a sharp trauma mechanism (85), transfer from another hospital (1), death upon arrival to the shock room (6). Included were 1325 patients for further analysis. Patient characteristics are presented in table 4 on the next page. In this table the non-abdominally injured group is compared to the abdominally injured group, which consists of ninety-nine patients. 3.2 Logistic regression The multivariate logistic regression was performed using a backward selection method. This means that the logistic regression was repeated until all the parameters had a significance of 0,157or lower (AIC). The dependent parameter in the logistic regression was abdominal injury. Not all of the parameters used in the univariate analysis were used in the multivariate logistic regression. The first logistic regression was performed with the following covariates: Free fluid in the abdomen on FAST, Sex, Age, ALAT, Hemoglobin, Creatinine, Base Excess, Amylase, Fibrinogen, INR, LDH and Shock Index. The first parameter to be removed from the logistic regression was Fibrinogen with a significance of in the pooled imputation dataset During the second step of the regression, sex was removed with a significance of Then during the third step of the regression, amylase with a significance value of Age was then excluded at the fourth step of the logistic regression with a significance of At the fifth step ALAT was removed from the logistic regression because the significance was Hemoglobin was removed during the sixth step of this regression, with a significant value of At the last and seventh step of the logistic regression Base Excess was removed with a significance of After the removal of the insignificant parameters the following parameters remained as the best predictors for the presence of abdominal injuries: Free fluid in the abdomen on FAST, Creatinine, INR, LDH and shock index. For these parameters a regression coefficient was calculated by SPSS. These regression co-efficients are used to generate a score to predict the presence of abdominal injury, we have named this score: the Blunt Abdominal Trauma in Adults score Parameter Reg. S.E. Sig. O.R. Free Fluid on FAST 6,031,566, ,297 Creatinine,005,003,077 1,005 INR,364,182,046 1,439 LDH,003,001,004 1,003 SI 1,967,735,010 7,150 Constant value -7,980,843,000,000 Table 3. Reg., Regression Co-efficient; S.E., Standard Error; Sig., Significance; O.R., Odds Ratio (BATiA). When computing the regression co-efficients of the selected parameters (table 3) the following score is formed. 15

16 Total Available IAI No IAI P-value Characteristics (n =1325) Percentage (n =99) (n =1226) IAI/No IAI Age, y (16-100) (16-85) (16-100) Sex, n (% male) 985 (74.3) (78.8) 907 (74) Hospital LOS ( ( ) 6.92 ( ) ) Hospital LOS ICU ( ( ) ( ) ) ISS EMTRAS SI (0-2.60) ( ) ( ) Intubated, n (%) 542 (40.9) (45.3) 495 (40.4) Complications, n (%) 417 (31.5) (47.9) 364 (29.7) Mortality, n (%) 194 (14.6) (18.8) 175 (14.3) Initial HR 80 (10-170) (45-156) 80 (10-170) Initial SBP 130 (40-220) (50-190) 130 (40-220) Abnormal abdominal physical 116 (8.8) (27.4) 86 (7.0) examination (%) Free Fluid on FAST (%) 83 (6.3) (72.7) 11 (0.9) CT-scans, n (%) Any CT-scan performed CT-abdomen performed Surgical intervention, n (%) Any Laparatomy Embolisation in Abdomen Mechanism of Injury, n (%) Traffic Accident Fall from height/stairs Struck by object Violence Suicide attempt Struck by or fallen of animal Dive in shallow water Other Lab values AST (U/L) ALT (U/L) WBC-count (10E9/L) LDH (U/L) Amylase (U/L) Creatinine (µmol/l) Hb (mmol/l) Trombocytes (10E9/L) Base Excess (mmol/l) INR 1187 (89.6) 245 (18.5) 495 (37.4) 49 (3.7) 24 (1.8) 794 (59.9) 307 (23.2) 57 (4.3) 22 (1.7) 30 (2.3) 32 (2.4) 6 (0.5) 77 (5.8) 41 (8-1994) 30 (2-2219) ( ) ( ) 55 (8-1026) 73 (28-886) 8.1 ( ) 208 (8-641) -2.6 ( ) 1.1 (0.9-10) (87.2) 91 (77.8) 72 (61.5) 45 (38.5) 22 (18.8) 86 (73.5) 12 (10.3) 5 (4.2) 1 (0.9) 2 (1.7) 4 (3.4) 0 (0) 7 (6.0) 144 ( ) ( ) ( ) 432 ( ) 674 (24-397) 89 (40-190) 6.6 ( ) 184 (52-359) -5.8 ( ) 1.2 ( ) 1067 (87.0) 159 (13.0) 426 (34.7) 8 (0.7) 0 (0) 726 (59.2) 296 (24.1) 51 (4.2) 21 (1.7) 28 (2.3) 28 (2.3) 6 (0.5) 70 (5.7) 39 (8-879) 29 (2-558) 11.6 (1-39.2) 266 ( ) 55 (8-1026) 72 (28-886) 8.2 ( ) 210 (8-641) -2.5 ( ) 1.1 (0.9-10) Table 4. Patient Characteristics IAI, Intra-abdominal injury; ISS, Injury Severity Score; EMTRAS, Emergency Trauma Score; SI, Shock Index; LOS, Length of Stay; ICU, Intensive Care Unit; HR, Heart rate; SPB, Systolic Blood pressure; SD, Standard Deviation; CT, Computed Tomography. 16

17 BATiA = * (FF) * (Shock index) * (INR) * (Creatinine) * (LDH) If Free Fluid (FF) is present; 1 should be written in the FF section, if FF is not present; 0 should be written in the FF section. The other parameters are continuous. 3.3 Hosmer-Lemeshow and calibration At the final step of the logistic regression, the Hosmer-Lemeshow test was performed and the results were plotted to make a calibration line, this calibration line is shown in figure 3. Fig. 3 Calibration of the Hosmer-Lemeshow test for Goodness of Fit for the BATiA 3.4 Receiver operating characteristics and Area under Curve The ROC-curve was plotted for the original dataset and also for each of the ten imputations. In figure 4 the ROC-curve of the original dataset is shown. The Area under the Curve (AUC) was calculated for each dataset, these values are given in table 5. Imputation Number AUC Original data,989 1,975 2,982 3,978 4,950 5,975 6,979 7,974 8,980 9,954 10,955 Fig. 4 The ROC-curve of the Original dataset for the BATiA-score Table 5. Area under the Curve for the BATiA-score AUC, Area Under the Curve 17

18 3.5 Free Fluid on FAST When assessing whether the patients had free fluid or not on FAST; 83 patients had a positive FAST and 1214 patients were negative. Of the 83 patients with a positive FAST, 72 (86.7%) were abdominally injured (sensitivity) and 11 (13.3%) had no abdominal injury. From the 1214 patients with a no free fluid in the abdomen on FAST, 1203 (99.1%) had no abdominal injuries (specificity) and 11 (0.906%) patients had abdominal injuries. These findings are summarized in table 6 and 7. In 146 (12.0%) of the 1214 patients with a negative FAST an abdominal CT was also performed. Of these 146 patients, 7 (4.8%) had an abdominal injury and 139 (95.2%) were not abdominally injured. Intra Abdominal Injury + - FAST Total Total Table 6. Abdominal injury versus Free Fluid on FAST Accuracy of FAST on Abdominal Injury 95% CI Sensitivity 86.7% Specificity 99.1% PPV 86.7% NPV 99.1% LR+ 95.7% LR- 0.13% Table 7. Sensitivity, Specificity, PPV and NPV for Abdominal injury on FAST. PPV, Positive Predictive Value; NPV, Negative Predictive Value; CI, Confidence Interval; LR, Likelihood Ratio 3.6 A prediction model without FAST We decided to also perform a multivariate logistic regression without the covariate Free fluid in the abdomen on FAST. Thus, only: Sex, Age, ALAT, Hemoglobin, Creatinine, Base Excess, Amylase, Fibrinogen, INR, LDH and Shock Index were used as covariates in the second multivariate logistic regression. The logistic regression was performed in the same stepwise manner as shown in paragraph 3.2. After the stepwise removal of all the insignificant (p >0,157) parameters the following parameters remained as the best predictors for the presence of abdominal Parameter Reg. S.E. Sig. O.R. Creatinine,004,002,067 1,004 LDH,003,001,002 1,003 SI 1,242,414,003 3,462 Sex -,515,330,118,597 ALAT,006,001,000 1,006 Hb -,463,082,000,630 Constant -1,861,814,022,156 Table 8. Reg., Regression Co-efficient; S.E., Standard Error; Sig., Significance; O.R., Odds Ratio injuries: Creatinine, LDH, Shock Index, Sex, ALAT and Hemoglobin. Again using the regression co-efficients calculated by SPSS (table 8) the following score is computed, which we have named the Adult Blunt Abdominal Trauma score (ABAT) ABAT = * (Shock Index) * (ALAT) * (Creatinine) * (LDH) * (Sex) * (Hemoglobin) If the patient is a woman; 1 should be written in the Sex section, if the patient is a male: 0 should be written in the Sex section. The other variables are continuous. 18

19 3.7 Calibration and discrimination for the ABAT For the logistic regression for the ABATscore, the Hosmer-Lemeshow test was also performed. A calibration line (figure 5) was created in the same manner explained in paragraph 3.3. Fig. 5 Calibration of the Hosmer-Lemeshow test for Goodness of Fit for the ABAT-score The ROC-curve for the ABAT-score was plotted for the original dataset and for each of the ten imputation datasets for the discrimination of the regression analysis without FAST. In figure 6 the ROCcurve of the original dataset is shown. The AUC for the prediction model without FAST was also calculated for each dataset, these values are given in table 9. Imputation Number AUC Original data,840 1,846 2,853 3,856 4,845 5,857 6,846 7,852 8,861 9,850 10,849 Fig. 6 The ROC-curve of the Original dataset for the ABAT-score Table 9. Area under the Curve for the ABAT-score AUC, Area Under the Curve 19

20 Discussion This study set out to identify non-invasive parameters from the standard ATLS protocol which have the ability to predict the presence of abdominal injury in trauma patients. Free fluid on FAST, Shock Index, Creatinine, INR and LDH were the best predictors for the presence of abdominal injury. These predictors were combined to generate the BATiA-score. However, free fluid on FAST proved to be such a strong predictor for the presence of abdominal injury, with an odds ratio of 416,297, that the other predictors seem to be negligible. 4.1 FAST is a strong predictor for abdominal injury We can conclude that the calibration line created for the BATiA is closely related to the perfect calibration line (fig. 3). In the BATiA calibration line this is almost completely due to the fact that free fluid on FAST is such a strong predictor (table 3). With the Hosmer- Lemeshow test, patients are divided into ten subgroups based on the calculated chance for the presence of abdominal injury according to the BATiA. Due to the high predictive value of free fluid on FAST almost all abdominally injured patient are in the group with the highest chance for the presence of abdominal injury. The patients with a low BATiA, and therefore no free fluid on FAST, are divided over the nine other groups. For that reason only one of the ten points plotted in the calibration line is responsible for the gradient of this line and not a combination of several points. This suggests that the presence of abdominal injury is almost completely predicted by free fluid on FAST and all the other parameters are redundant. The AUC of the ROC-curve (fig.1, table 5) proved to be highly efficient. Again this shows that free fluid on FAST is an extremely strong predictor for the presence of abdominal injury. Literature also shows that free fluid on FAST would be highly suggestive for the presence of intra-abdominal injuries (16). We also generated another scoring method to predict the presence of abdominal injury without the parameter: free fluid on FAST, called the: Adult Blunt Abdominal Trauma score (ABAT-score). We chose to create this second prediction model for several reasons. First, in third world countries, sonographic scanners are not available in every hospital. A score based on simple blood work, sex, age, blood pressure and pulse, might be a helpful aid in third world countries. Second, the covariate Free fluid in the abdomen on FAST is extremely predicting in the logistic regression mentioned in paragraph 3.2. Taking this into account we would like to know how well the other covariates perform in a multivariate logistic regression without the covariate Free fluid in the abdomen on FAST. The ABAT-score consists of the following parameters: Shock Index, ALAT, Creatinine, LDH, Sex and Hemoglobin. The calibration line created for the ABAT-score is also closely related to the perfect calibration line (fig. 5). For the ABAT; Shock Index, Hb en sex are mainly responsible for the gradient of the calibration line. The AUC of the ROC-curve of the ABAT-score (fig. 6, table 9) also proved to be efficient but not as efficient as the AUC and the ROC-curve for the BATiA-score. However, because a AUC of for the original dataset is reasonably high and because the scoring method consists of several parameters this might be a helpful aid in institutions without a sonographic scanner available. 4.2 Patient characteristics In this study patients with IAI are compared to patients with no IAI, some differences in the findings between these groups are remarkable. Abdominally injured patients have a lower age than patients with no IAI in this study. This age difference might be due to the fact that a higher percentage of younger patients were 20

21 involved in traffic accidents, where the abdomen tends to strike the steering wheel during a rapid forward shift. Mechanism of injury might also be the reason why there is a higher percentage of patients which have suffered from a fall in the non IAI group, this because abdominal injury is not often the result of a fall. The patients with IAI also seemed to have a longer hospital length of stay compared to the non abdominally injured group. This can be supported by the fact that IAI patients have a higher ISS and EMTRAS and thus are more heavily injured compared to the non IAI patients. The abdominally injured patients have a higher Shock Index, and are therefore more often hemodynamic unstable than patients without abdominal injuries. That abdominal injuries are more often observed in hemodynamic unstable patients after blunt trauma is not a surprise; Farrath et al. predicted that hemodynamic instability is a predictor for the presence of abdominal injury (40). These patients are hemodynamic unstable because of possible hemorrhages in the well vascularised abdominal organs. Also the IAI patients tend to develop more complications (i.e. a unfavorable decline of the disease during the treatment, e.g. infection, psychiatric disease) than patients with no IAI. These complications might be due to the fact that abdominally injured patients have a higher rate for surgical interventions than non-abdominally injured patients, and therefore have a higher chance for surgery related complications. Outliers in the hospital length of stay are caused by the fact that psychiatric patients were admitted to the psychiatry department and these patients tend to have long durations of treatment. 4.3 Abdominal pain Due to the retrospective design of this study some obstacles had to be passed during data collection. One of them was to define the definition of abdominal pain in the trauma patient. There are several different types of abdominal pain that can be found in a trauma patient. The patient can indicate that there is pain in the abdominal region. A muscular defense can also be found in the patient, denoting an abnormal abdominal examination. Or during the physical examination the patient can experience pressure or rebound tenderness. Because of the retrospective nature of this study, all the data was collected from the Electronic Patient Database (EPD). During the reporting of trauma, sometimes too little information was reported to distinguish between the different types of abdominal pain. For that reason we chose to define a patient as having an abnormal physical examination of abdomen if any of the before mentioned types of abdominal pain was present.. In this study, 40.9% of the patients were intubated and sedated at the time they arrived on the ER. This means that these patients were not able to communicate and only hetero-anamnesis was possible. Physical examination of the abdomen was therefore not possible in a substantial part of our population. Furthermore, due to the sedation of the patients any abnormal muscle tone would be absent. We therefore deemed the physical examination of the abdomen as an unsuitable parameter for this clinical score because this could not be scored in every patient. In table 4 the abnormal abdominal physical examination scored an available percentage of 99.8% because the sedated patients did not score a missing value, but were noted as having no abdominal pain. 4.5 Patients with a negative FAST but with abdominal injuries. There were 11 patients which had abdominal injuries, but had a negative FAST. In four of these eleven patients, no CT-scan was performed and the abdominal injury was discovered in the operating room. In one of these four patients a perforated sigmoid was discovered during a laparatomy performed because the patient developed severe abdominal pains and a local peritonitis. The 21

22 second suffered from a ruptured arteria iliaca interna discovered due to an absent pulse in the left leg during an operation of the left knee. In the third patient, a ruptured rectum was found during an operation of a severely injured pelvis. The fourth patient became hemodynamically unstable three days after the trauma, an immediate laparotomy was performed and a ruptured liver and spleen were found. In the seven remaining patients with no abnormal findings on FAST but with abdominal injuries a CT-scan was performed. In two patients there was no free fluid visible on the first FAST, but on the second FAST free fluid was visible (performed because one patient became hypotensive and the other patient started to suffer from severe abdominal pain). One of these patients suffered from a ruptured liver and spleen, in the other patient a ruptured intestine was found. In a third patient a CTscan was performed but no injuries were visible. Two days later another CT scan was performed due to high infection parameters and an irritated abdomen, and a ruptured mesentery was found. The fourth patient suffered from a liver contusion and the and fifth patient had a renal contusion which were discovered on CT. These patients were treated conservatively. Thus, in only two patients in the whole study the initial CT-scan played an additional role with regard to FAST in altering the mode of treatment for abdominal injuries in trauma patients. In one patient a ruptured arteria iliaca interna was discovered and treated by embolisation on the OR, the other patient suffered from a renal contusion which was treated with bed rest. In 146 patients without free fluid on FAST an abdominal CT-scan was performed. With the current price of an abdominal CT-scan without a radiology report of 210,58- ex. tax in the UMCG, 30744,68- can be cut from costs made in the Emergency department when an abdominal CT-scan is only made when FAST showed free fluid in the study duration of five and a half years. Although several patients did have abdominal injuries when no free fluid was visible on FAST, these injuries were identified after the initial trauma care. These injuries were either found when a patient was undergoing a surgical intervention for other injuries or when a suspicion arose due to signs of clinical deterioration. 4.6 Imputation Multiple imputation has appeared as a decent way to manage missing data. However some researchers avoid imputation because of fears of making up data (41). Using multiple imputation, the missing data are predicted based on other data present in the same patient (42). De Goeij states that this method builds on and improves the idea of mean substitution, where missing values are replaced with the overall mean and imputed values are treated as real observed values (43). 4.7 Limitations The retrospective nature of this study is a limitation, because frequently data was not recorded thoroughly and occasionally data could not be retrieved from the EPD. Also, it is impossible to confirm the interpretations of the treating physicians during the physical examination. Another limitation in this study is that we only included the patients coded as an A-trauma. The patients, who were initially coded as a B-trauma (not the highest level of urgency) but did have substantial injuries similar to the injuries A-trauma patients suffer, are not included in this study and this increases the chance to bias. 4.8 Prospects With this study we found that FAST has a high specificity (99.1%), and thus has the ability to confirm the presence of abdominal injury. We suggest that a FAST is performed in every trauma patient during the initial trauma care and only with a positive outcome of the FAST a 22