Emergency. 2017; 5 (1): e27 OR I G I N A L RE S E A RC H Thoracic Injury Rule out Criteria in Prediction of Trau- matic Intra-thoracic Injuries; a Validation Study Setareh Asgarzadeh1, Bahareh Feizi 2, Farhad Sarabandi1∗, Morteza Asgarzadeh3 1. Clinical Research Development Center, Amir-Almomenin Hospital, Islamic Azad University, Tehran Medical Sciences Branch, Tehran, Iran. 2. Emergency Department, Pasteur Hospital, Bam University of Medical Sciences, Bam, Iran. 3. Harvard T.H. Chan School of Public Health, 667 Huntington Avenue, Boston, MA, 02115. Received: August 2016; Accepted: October 2016; Published online: 10 January 2017 Abstract: Introduction: Doing Chest X Ray (CXR) for all trauma patients is not efficient and cost effective due to its low diagnostic value. The present study was designed aiming to evaluate the diagnostic accuracy of thoracic in- jury rule out criteria (TIRC) in prediction of traumatic intra-thoracic injuries and need for CXR. Methods: The present study is a prospective cross-sectional study that has been carried out to evaluate the accuracy of TIRC model in screening blunt multiple trauma patients in need of CXR for ruling out intra-thoracic injuries. Results: 1518 patients with the mean age of 33.53 ± 15.42 years were enrolled (80.4% male). The most common mech- anisms of trauma were motor car accident (78.8%) and falling (13.6%). Area under the ROC curve, sensitivity, and specificity of model in detection of traumatic thoracic injuries was 0.95 (95% CI: 0.93 – 0.97), 100 (95% CI: 87.0 – 100), and 80.1 (95% CI: 78.0 – 82.1), respectively. Brier score for TIRC was 0.02 and its scaled reliability was 0.0002. Conclusion: Findings of the present study showed that TIRC has high accuracy in prediction of traumatic intra-thoracic injuries and screening patients in need of CXR. Keywords: Thoracic injuries; decision support techniques; mass chest x-ray; diagnosis © Copyright (2017) Shahid Beheshti University of Medical Sciences Cite this article as: Asgarzadeh S, Feizi B, Sarabandi F, Asgarzadeh M. Thoracic Injury Rule out Criteria in Prediction of Traumatic Intra- thoracic Injuries; a Validation Study. Emergency. 2017; 5(1): e27. 1. Introduction T raumatic injuries, as one of the causes of morbidity and mortality, inflict a big financial and social burden on health care systems (1). Meanwhile, thoracic in- juries are responsible for 20 -50% of trauma-related mortali- ties (2). Numerous diagnostic tools exist for evaluating these injuries including computed tomography (CT) scan, chest x-ray (CXR), and ultrasonography accompanied by clinical examination. Currently, CXR is considered as the first di- agnostic test in traumatic thoracic injuries (3). However, study findings have shown that doing CXR for all patients is not efficient and cost effective due to its low diagnostic value (4, 5). Therefore, researchers are seeking ways to use this tool only for patients with a higher risk of intra-thoracic injuries. In recent years, 2 clinical decision rules, namely ∗Corresponding Author: Farhad Sarabandi; Department of Pediatrics, Amir_Almomenin Hospital, Shirmohammadi Avenue, Naziabad, Bahman Square, Tehran, Iran. Tel: 09122069521 Email: farhad.sarabandi@gmail.com Nexus chest in American population and thoracic injury rule out criteria (TIRC) in Iranian population have been intro- duced for screening patients in need of CXR following blunt trauma. Based on Nexus chest criteria, if any of the fac- tors including age >60 years, rapid deceleration mechanism (falling from a height over 20 feet or being in a car accident with more than 40 mph speed), chest pain, intoxication, al- tered level of consciousness, distracting pain, and tender- ness to chest wall palpation are present, the patient is at high risk regarding presence of injury and CXR is necessary (6). In TIRC model age >60 years, hemodynamic instability, loss of consciousness, crepitation in auscultation, decreased pulmonary sounds, thoracic skin abrasion, and shortness of breath are factors predicting intra-thoracic injuries (7). These 2 models are just starting to be studied and they need to be validated in various populations. Therefore, the present study was designed aiming to evaluate the diagnostic accu- racy of TIRC in prediction of traumatic intra-thoracic injuries and need for CXR. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: www.jemerg.com S. Asgarzadeh et al. 2 2. Methods 2.1. Study design and setting The present study is a prospective cross-sectional study that has been carried out to evaluate the accuracy of TIRC in screening patients in need of CXR in multiple trauma pa- tients presented to the emergency department (ED) of Pas- teur Hospital, Bam, Iran, during 1 year in 2014-2015. Pro- tocol of the present study was approved by hospital ethic committee. Written informed consent was obtained from all the patients and the researchers adhered to the principles of Helsinki Declaration throughout the study. This project did not cause any disruption in the routine management of pa- tients. 2.2. Participants The study participants consisted of all blunt multiple trauma patients over 15 years old who were conscious and had stable hemodynamic. Exclusion criteria included presence of pene- trating chest trauma and not giving consent for participation in the study. 2.3. Data gathering Sample selection was done using non-randomized conve- nience sampling. After obtaining informed consent from the patient or their relative, the study checklist was filled. The checklist consisted of demographic data (age, gender, trauma mechanism), history and physical examination findings (dis- tracting pain, loss of consciousness, tachypnea, chest pain, dyspnea, presence of thoracic skin abrasion due to trauma, tenderness in chest, chest deformity, tenderness in up- per abdomen, crepitation in chest auscultation, decreased pulmonary sounds, and presence of crepitation), variables needed for TIRC model, and CXR findings. An emergency medicine specialist was responsible for examining, gather- ing, and recording of data in various days and working shifts. Immediately after data gathering, CXR was done for patients in 2 standard views of anterior-posterior and lateral, and the pathological findings (hemothorax; pneumothorax; fracture of rib, sternum, scapula, and clavicle; widened mediastinum; and lung contusion) were recorded. CXRs were interpreted and recorded by an emergency medicine specialist blinded to the clinical findings of the patients as well as the in-charge physician. To evaluate the accuracy of interpretations by the emergency physician, 5% of the CXRs were randomly se- lected and given to a radiologist for interpretation (Inter-rater agreement between the radiologist and emergency physician was 100%). It should be noted that the radiologist was blind to both the emergency physician’s interpretation and clinical findings. Final diagnosis of thoracic injury was done based on CXR. At times of suspicion to presence of a hidden injury, Chest CT scan was done. 2.4. TIRC model variables Based on this model CXR is necessary for patients with un- stable hemodynamics and loss of consciousness. In addition, conscious patients with stable hemodynamics that meet any of the factors including age >60 years, crepitation in auscul- tation, decrease in pulmonary sounds, thoracic skin abra- sion, and shortness of breath, are categorized in the high risk group regarding probability of intra-thoracic traumatic in- juries and should undergo CXR. 2.5. Statistical Analysis To determine sample size, considering the 6.5% prevalence of positive findings in multiple trauma patients’ CXR (8), a 95% confidence interval (CI) (α = 0.05), 90% power (β = 0.1) and maximum error of 1.5% (d = 0.015) in estimating prevalence of injury, minimum sample size was considered 1043. Data were entered to STATA 11.0 software. CXR findings were re- ported as frequency and percentage, and were divided into 2 groups of normal and abnormal. In the present study, to assess the validity of the model, a number of methods were used (9, 10) that included calculating the area under the receiver operating characteristic (ROC) curve, sensitivity, specificity, positive and negative predictive value (PPV/NPV ), and positive and negative likelihood ratio (PLR/NLR) with 95% confidence interval (CI). To evaluate discrimination, cal- ibration curve was drawn for assessing general calibration, and finally in evaluation of overall performance, Brier score was used for assessing predictive accuracy and predictive re- liability. It should be noted that in calibration curve, the per- fect calibration is the reference line that has 0 intercept and slope of 1. The closer the slope and intercept of TIRC model are to 1 and 0, respectively, the more perfect the model is for predicting presence or absence of injury in CXR (11). 3. Results Finally, data of 1518 patients with the mean age of 33.53 ± 15.42 years were gathered (80.4% male). Table 1 shows base- line characteristics of studied patients. The most common mechanisms of trauma were motor vehicle collisions (42.1%) and falling down (28.2%). 401 (26.4%) had chest pain, 107 (7.1%) had chest wall tenderness, and 104 (6.8%) had a tho- racic skin abrasion. Based on CXR findings, 33 (2.2%) pa- tients had at least 1 traumatic intra-thoracic injury. 3.1. Discrimination Area under the curve of TIRC in detection of traumatic tho- racic injuries was calculated to be 0.95 (95% CI: 0.93 – 0.97) (figure 1). Considering the presence of at least one of the TRIC risk factors, sensitivity and specificity of model were 100 (95% CI: 87.0 – 100) and 80.1 (95% CI: 78.0 – 82.1), respec- tively. PPV of the test was 10.1 (95% CI: 7.1 – 14.0) and NPV This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: www.jemerg.com 3 Emergency. 2017; 5 (1): e27 Figure 1: Area under the receiver operating characteristic curve of thoracic injury rule out criteria (TIRC). Figure 2: The calibration plot for thoracic injury rule out criteria (TIRC). was 100 (95% CI: 99.6 – 100). PLR and NLR calculated were 5.0 (95% CI: 4.5 – 5.6) and 0 (95% CI: 0.0 – 0.0), respectively (table 2). Calibration curve of TIRC in detection of an intra- thoracic injury has been presented in figure 2. This scatter plot has an intercept of 0.1 (95% CI: 0.01 -0.19) and a slope of 1.7 (95% CI: 1.3 -1.9) which shows the moderate calibration of this model. 3.2. Overall performance Brier score for TIRC was 0.02 and its scaled reliability was 0.0002. These findings are indicative of this model’s high pre- dictive accuracy and reliability. 4. Discussion Findings of the present study showed that TIRC has high ac- curacy in prediction of traumatic intra-thoracic injuries and screening patients in need of CXR. There was no false nega- Table 1: Baseline characteristics of studied patients Variable Number (%) Age (year) < 60 1468 (96.7) ≥ 60 50 (3.3) Gender Male 1220 (80.4) Female 298 (19.6) Mechanism of trauma Motor vehicle collision 1196 (78.8) Falling down 207 (13.6) Others 115 (7.6) Vital sign (admission time) Systolic blood pressure (mmHg) 119.2±9.1 Diastolic blood pressure (mmHg) 78.2±14.9 SPO2 (%) 97.7±2.9 Respiratory rate (1/minute) 13.6±1.8 Glasgow coma scale 15 1468 (96.7) Less than 15 50 (3.3) Dyspnea Yes 42 (2.8) No 1476 (97.2) Distracting pain Yes 399 (26.3) No 1119 (73.7) Thoracic skin abrasion Yes 104 (6.8) No 1414 (93.2) Chest deformity Yes 8 (0.5) No 1510 (99.5) Chest wall tenderness Yes 107 (7.1) No 1411 (92.9) Crepitation Yes 16 (1.0) No 1502 (99.0) Abdominal tenderness Yes 25 (1.6) No 1493 (98.4) Decrease in pulmonary sounds Yes 37 (2.4) No 1481 (97.6) Chest wall pain Yes 732 (25.20) No 2173 (74.80) tive result in this model and this indicates the proper power of this instrument to rule out intra-thoracic injury following blunt trauma. Based on the findings of this study, if TIRC clinical decision rule was used, only 328 (21.6%) of the 1518 studied patients would undergo CXR. This finding shows that using TIRC will lead to a significant decrease in unnecessary CXRs. In the studied population, 1485 (97.9%) of the CXRs were without any pathologic finding and TIRC predict 1190 (80.1%) of them. This finding is in line with 2 previous stud- This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: www.jemerg.com S. Asgarzadeh et al. 4 Table 2: Screening performance characteristics of thoracic injury rule out criteria (TIRC) in detection of intra-thoracic injuries Characteristics* Value (95%CI) Area under the curve 0.95 (0.93 - 0.97) Sensitivity 100.0 (87.0 - 100.0) Specificity 80.1 (78.0 - 82.1) Positive predictive value 10.1 (7.1 - 14.0) Negative predictive value 100.0 (99.6 - 100.0) Positive likelihood ratio 5.0 (4.5 - 5.6) Negative likelihood ratio 0.0 (0.0 - 0.0) True positive 33 True negative 1190 False positive 295 False negative 0 ∗ Values given are based on presence of at least one of the following symptoms: age over 60, crepitation, loss of con- sciousness, decrease in pulmonary sounds, chest wall pain, chest wall tenderness, dyspnea, and skin abrasion; CI: con- fidence interval. ies. In a study by Frouzanfar et al., it was shown that us- ing this tool reduces unnecessary CXRs by 63.5% (7). This rate was 67.7% in Safari et al. study (11). In the Safari et al. study, which was a multi-center one, evaluation of pa- tients was done by different physicians while in the present study all evaluations were done by one emergency medicine specialist. This might be the reason for the higher screening value of TIRC in this study. In comparing TIRC with Nexus chest model, it is revealed that both models have similar and good value in screening of patients for performing CXR. A study by Rodriguez et al. aiming to validate Nexus chest, indi- cated the 98.5% sensitivity of this tool in screening traumatic intra-thoracic injuries (12) while this rate was 100% for TIRC. However, it seems that fewer factors in TIRC can be advanta- geous for using it in clinic. In addition, data such as height of falling and speed of the vehicle at the time of accident (which are required in Nexus chest) are not readily available in many cases, especially in developing countries. However, it is worth noting that validation of nexus chest has only been done in the American population and validation of TIRC has only been done in the Iranian population. Therefore, further studies are needed on both in other settings and geographi- cal areas to ensure their validity. 5. Limitation One of the limitations of the present study is being carried out in 1 center. Therefore, the results may not be easily gen- eralized. However, since the findings are in line with similar previous studies, It seems that being single centered has not affected the generalizability of the data. Additionally, con- venience sampling was used, which raises the probability of selection bias. However, unlike previous studies (11, 12), pa- tient evaluation was done by a single emergency medicine specialist and CXR interpretation was done by another single emergency medicine specialist, which eliminates the effect of difference in assessor in these areas. 6. Conclusion Findings of the present study showed that TIRC has high ac- curacy in prediction of traumatic intra-thoracic injuries and screening patients in need of CXR. There was no false nega- tive result in this model and this indicates its proper power to rule out thoracic injury. 7. Appendix 7.1. Acknowledgements Authors would like to thank all the staff of Emergency Depart- ment of Pastor Hospital, Bam, Kerman, Iran. 7.2. Author contribution All authors passed four criteria for authorship contribution based on recommendations of the International Committee of Medical Journal Editors. 7.3. Funding/Support None. 7.4. Conflict of interest None. References 1. Nseir S, Zerimech F, Fournier C, Lubret R, Ramon P, Durocher A, et al. Continuous control of tracheal cuff pressure and microaspiration of gastric contents in criti- cally ill patients. American journal of respiratory and crit- ical care medicine. 2011;184(9):1041-7. 2. Chastre J, Fagon J-Y. Ventilator-associated pneumo- nia. American journal of respiratory and critical care medicine. 2002;165(7):867-903. 3. Valles J, Mesalles E, Mariscal D, del Mar Fernandez M, Pena R, Jimenez JL, et al. A 7-year study of severe hospital-acquired pneumonia requiring ICU admission. Intensive care medicine. 2003;29(11):1981-8. 4. Sopena N, Sabria M. Multicenter study of hospital- acquired pneumonia in non-ICU patients. Chest Journal. 2005;127(1):213-9. 5. Vincent J-L, Rello J, Marshall J, Silva E, Anzueto A, Mar- tin CD, et al. International study of the prevalence and outcomes of infection in intensive care units. Jama. 2009;302(21):2323-9. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: www.jemerg.com 5 Emergency. 2017; 5 (1): e27 6. Gupta A, Agrawal A, Mehrotra S, Singh A, Malik S, Khanna A. Incidence, risk stratification, antibiogram of pathogens isolated and clinical outcome of ventilator associated pneumonia. Indian Journal of Critical Care Medicine. 2011;15(2):96. 7. Herzig SJ, Howell MD, Ngo LH, Marcantonio ER. Acid- suppressive medication use and the risk for hospital- acquired pneumonia. Jama. 2009;301(20):2120-8. 8. Safdar N, Dezfulian C, Collard HR, Saint S. Clinical and economic consequences of ventilator-associated pneu- monia: a systematic review. Critical care medicine. 2005;33(10):2184-93. 9. Nomellini V, Chen H. Murray and Nadel’s Textbook of Respiratory Medicine. Academic Press; 2012. 10. Dore P, Robert R, Grollier G, Rouffineau J, Lanquetot H, Charriere J-M, et al. Incidence of anaerobes in ventilator- associated pneumonia with use of a protected specimen brush. American journal of respiratory and critical care medicine. 1996;153(4):1292-8. 11. Resende MM, Monteiro SG, Callegari B, Figueiredo PM, Monteiro CR, Monteiro-Neto V. Epidemiology and out- comes of ventilator-associated pneumonia in northern Brazil: an analytical descriptive prospective cohort study. BMC infectious diseases. 2013;13(1):1. 12. Japoni A, Vazin A, Davarpanah MA, Ardakani MA, Alborzi A, Japoni S, et al. Ventilator-associated pneumonia in Ira- nian intensive care units. The Journal of Infection in De- veloping Countries. 2011;5(04):286-93. 13. Hamishekar H, Shadvar K, Taghizadeh M, Golzari SE, Mojtahedzadeh M, Soleimanpour H, et al. Ventilator- Associated Pneumonia in Patients Admitted to Intensive Care Units, Using Open or Closed Endotracheal Suction- ing. Anesthesiology and pain medicine. 2014;4(5). 14. Afhami S, Hadadi A, Khorami E, Seifi A, Bazaz NE. Ventilator-associated pneumonia in a teaching hospi- tal in Tehran and use of the Iranian Nosocomial In- fections Surveillance Software. Eastern Mediterranean Health Journal. 2013;19(10):883. 15. Bennett JE, Dolin R, Blaser MJ. Principles and practice of infectious diseases: Elsevier Health Sciences; 2014. 16. Nadi E, Nekoie B, Mobaien A, Moghimbeigi A, Nekoie A. Evaluation of the Etiology of Nosocomial Pneumonia in the ICUs of the Teaching Hospitals of Hamadan Univer- sity of Medical Sciences. Scientific Journal of Hamadan University of Medical Sciences. 2011;18(1):26-32. 17. Carlson KK, Louis S. Advanced critical care nursing. St Louis. 2009. 18. Sabery M, Shiri H, Moradiance v, Taghadosi M, Gilasi HR, Khamechian M. The frequency and risk factors for early- onset ventilator-associated pneumonia in intensive care units of Kashan Shahid-Beheshti hospital during 2009- 2010. KAUMS Journal ( FEYZ ). 2013;16(6):560-9. 19. Chung DR, Song J-H, Kim SH, Thamlikitkul V, Huang S-G, Wang H, et al. High prevalence of multidrug- resistant nonfermenters in hospital-acquired pneumo- nia in Asia. American journal of respiratory and critical care medicine. 2011;184(12):1409-17. 20. Klompas M, Kleinman K, Khan Y, Evans RS, Lloyd JF, Stevenson K, et al. Rapid and reproducible surveillance for ventilator-associated pneumonia. Clinical infectious diseases. 2012;54(3):370-7. 21. Zarinfar N, Sharafkhah M, Bayat B, sgharFarazi A, Ma- soomehSoofian. Epidemiological Factors of Ventilator- Associated Pneumonia (VAP) among ICU patients in Valiasr Hospital of Arak. 2012. Iranian Journal of Infec- tious Diseases and Tropical Medicine. 2014;19(64). 22. Nateghian A, Omrani A, Alipour Z, Haerinejad M. Causes of ventilator associated pneumonia in pediatrics ICU. Iranian South Medical Journal. 2016;19(1):98-105. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: www.jemerg.com Introduction Methods Results Discussion Limitation Conclusion Appendix References