- Email: [email protected]
Predictors of mortality in pulmonary contusion
Predictors of Mortality in Pulmonary Contusion Daniel R. Kollmorgen, MD, Kathleen A. Murray, MD, John J. Sullivan, PhD, Mary C. Mone, RN, Richard G. Barton, MD Salt Lake City, Utah
BACKGROUND: Associated injuries and central nervous system (CNS) trauma are historically associated with poor outcome in patients with pulmonary contusions, but the value of specific factors reflecting shock, fluid resuscitation requirement and pulmonary parenchymal injury in predicting mortality in this population is not well established. METHODS: The medical records of 1 0 0 consecutive patients with pulmonary contusion, adndtted over a 5-year period, were retrospectively reviewed. Survivors and nonsurvivors were compared in terms of age, Injury Severity Score (ISS), Glasgow Coma Score (GCS), PaO2/FiO 2 (oxygenation ratio), the severity and adequacy of shock resuscitation reflected in plasma lactate, resuscitation volume and transfusion requirements, using oneway ANOVA. To determine the contribution of individual, interdependent variables to mortality, the data were then analyzed using multivariable analysis. RESULTS: ISS and transfusion requirement were significantly higher, and GCS and PaO2/FiO 2 at 2 4 and 4 8 hours after admission were significantly lower in nonsurvivors than in survivors. After multiple regression analysis, the factors most strongly associated with mortality included patient age, oxygenation ratio at 2 4 hours after admission, and resuscitation volume. CONCLUSIONS: Outcome in patients with pulmonary contusion is dependent upon a number of variables including the severity of pulmonary parenchymal injury as reflected in PaO2/FiO 2 ratio. horacic trauma is involved in nearly one third of acute admissions to trauma centers ~ and accounts for 25% of all trauma-related deaths in the United States. 2 Pulmonary contusion is often considered an inconsequential injury in these patients, but it is the sec"ond most c o m m o n injury in blunt thoracic trauma and
T
From the Departments of Surgery (DRK, JJS, MCM, RGB) and Radiology (KAM), University of Utah School of Medicine, Salt Lake City, Utah. Requests for reprints should be addressed to Richard G. Barton, MD, Department of Surgery, 3B313, Universityof Utah School of Medicine, 50 North Medical Drive, Salt Lake City, Utah 84132. This work was supported in part by NIH grant 1 RO HD 30455-01. Presented at the 46th Annual Meeting of the Southwestern Surgical Congress, Tucson, Arizona, April 17-20, 1994.
is associated with a mortality rate ranging from 14% to 40%. 1,3 Historically, the most consistent predictors of mortality have been patient age, severity of injury, and the presence of central nervous system (CNS) trauma, although a number of factors have been evaluated. 4-9 We reviewed our experience in the management of patients with pulmonary contusion in order to determine if other variables were associated with outcome and to examine our success in treating this injury. METHODS
Patient Population We reviewed the medical records of 100 consecutive patients admitted to the University of Utah Medical Center between January 1988 and December 1993 with a diagnosis of pulmonary contusion.
Diagnosis of Pulmonary Contusion The diagnosis of pulmonary contusion was based on the radiographic findings of nonsegmental areas of opacification resulting from a blow to the chest. Flail chest was defined clinically (paradoxical motion of the chest wall or sternum) or radiographically (the presence of 6 or more rib fractures or more than 3 segmental rib fractures in 1 hemithorax).
Patient Management All patients who had pulmonary contusions were resuscitated with intravenous fluids sufficient to support blood pressure, maintain urine output, and correct lactic acidosis. Fluid restriction and corticosteriods were not employed for pulmonary contusion but were used selectively for patients with major CNS trauma. Indications for intubation and ventilation included a respiratory rate >40 or <8, a PaCO 2 >50 mm Hg without a history of chronic obstructive disease, a PaO 2 <60 mm Hg or requirement of an FiO 2 exceeding 0.50, the need to establish or maintain a patent airway, clinical signs of respiratory distress, or hemodynamic instability. Patients with flail chest were managed selectively according to the same criteria.
Data Collection Data extracted from patient medical records included age, sex, mechanism of injury, requirement for ventilatory support, length of intensive care unit (ICU) and hospital stay, Injury Severity Score (ISS), Glasgow Coma Score (GCS) on admission, transfusion requirement in the first 48 hours after admission, resuscitation volume (total fluid volume received in the first 48 hours after admission), and admission chest roentgenogram results. In addition, plasma lactate and PaO2/FiO 2 were recorded at initial presentation as well as 24 and 48
THE AMERICANJOURNALOF SURGERY*' VOLUME 168 DECEMBER1994 659
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
The information in the database was analyzed in two phases. Initially, survivors and nonsurvivors were compared in terms of the previously described variables using one way analysis of variance. Subsequently, in order to identify factors that may be independent predictors of mortality in this patient population, information from the database was analyzed using multiple regression analysis. Plasma lactate values were excluded from this analysis due to inadequate information in many study patients. Statistical analysis was performed on an IBM PC using SPSS/PC (SPSS Corp., Boca Raton, Florida). A P value of <0.05 was considered significant. All data are expressed as mean value +_ standard deviation.
ber of units of blood transfused, and volume of resuscitation fluid received during the first 48 hours after admission. Differences in the length of ICU stay and the length of hospital stay between survivors and nonsurvivors are also included in Table II. Compared to survivors, nonsurvivors had a significantly higher ISS, lower GCS, higher plasma lactate 48 hours after admission, and a significantly lower oxygenation ratio at 24 and 48 hours after admission. Nonsurvivors also required significantly more blood transfusions than did survivors. Age, plasma lactate on admission and at 24 hours after admission, oxygenation ratio on admission, and volume of intravenous fluids in the first 48 hours of admission were not significantly different between survivors and nonsurvivors. To examine the relative importance of the interdependent variables described in Table II on mortality in patients with pulmonary contusion, the data were analyzed using multiple regression analysis. The results of this analysis are summarized in Table III. The regression equation produced a multiple R value of 0.642 and an adjusted R square of 0.303. The variables included in the multiple regression equation are listed in order of decreasing [3 value, The variables most strongly associated with mortality (the highest [3 values) in this analysis are patient age, oxygenation ratio 24 hours after admission, resuscitation volume, and GCS. The variables with the least influence on mortality are transfusion requirements, and oxygenation ratio on admission and 48 hours after admission. Due to inadequate data sets, plasma lactate values were not included in the analysis.
RESULTS One hundred patients (73 male, 27 female), with a mean age of 33 years, were admitted with a diagnosis of pulmonary contusion during the period reviewed. Eighty-six percent of the injuries were caused by motor vehicle or motorcycle accidents. The remainder were due to auto-pedestrian accidents (4%), crush injuries (3%), falls (2%), skiing accidents (2%), and miscellaneous injuries (3%). Flail chest was present in 17% of the study population, 14% of survivors and 30% of nonsurvivors. Fifty-two patients (52%) required intubation, including 49% of the survivors and all of the nonsurvivors. Ten patients died (10% mortality) and the causes of death in these patients are summarized in Table I. The mean duration of survival for patients who died was 5.5 days (range 0-17); however, 8 of the 10 nonsurvivors expired within the first 4 days following injury. Two patients survived for 12 and 17 days and died of multiple organ failure. Seven of 10 patients who died did so as a direct consequence of pulmonary failure or lung injury; 4 patients with acute respiratory failure, 2 patients with acute respiratory distress syndrome associated with multiple organ failure, and 1 of uncontrolled pulmonary hemorrhage. Table II compares survivors and nonsurvivors of pulmonary contusion in terms of factors previously associated with increased mortality. These factors include age ISS, GCS, plasma lactate and oxygenation ratio on admission, and at 24 and 48 hours after admission, num-
COMMENTS Although the adverse effects of associated injuries and CNS trauma in patients with pulmonary contusions are well documented, it is less clear how other clinical findings contribute to mortality. Advanced patient age 4.5,14 and the presence of hypotension 4'5'9't5 have been inconsistently correlated with poor outcome. Similarly, requirement for and volume of transfusions 4'5:4 or intravenous fluids 4'16'17are variably associated with mortality from pulmonary contusion. Further, attempts to define pulmonary parenchymal injury and measure its contribution to mortality in these patients 4,9'14"~82° have not established a clear, reproducible association. The present study was conducted to determine the value of factors reflecting shock, fluid resuscitation requirement, and pulmonary parenchymal injury in predicting outcome for patients of a similar age or severity of injury with pulmonary contusion. The demographic findings in this study are similar to those previously described for pulmonary contusion/'9 Young men comprise the majority of patients who suffer pulmonary contusions. Motor vehicle accidents (cars, all-terrain vehicles, and motorcycles) are the most common mechanisms of injury, accounting for nearly 90% of pulmonary contusions. Craniocerebral injuries are the most commonly associated nonthoracic injuries in blunt chest trauma, occurring in nearly 40% of cases. 9 In our study, 34% of all patients with pulmonary contusions also suffered head injuries.
TABLE I Causes of Death in Patients With Pulmonary Contusion Cause of Death
No.
Survival (d)
Acute respiratory failure Severe head injury Multiple organ failure Uncontrolled hemorrhage Multiple blunt injuries
4 2 2 1 1
1, 1, 2, 4 1, 4 12, 17 1 1
hours after admission. ISS was calculated using the method of Baker et aP ° and the 1990 Abbreviated Injury Scale. jl GCS was determined according to the criteria of Teasdale and Jennett.12 Statistical Analysis
660
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
TABLE II Comparison of Survivors and Nonsurvivors Predictor Age (y) ISS GCS Lactate (mmol/dL) Admission 24 hours 48 hours Oxygenation ratio (PaO2/Fi02)* Admission 24 hours 48 hours Transfusions (units/48 hours) Resuscitation volume (mL/48 hours) ICU length of stay (d) Hospital length of stay (d)
Survivors n = 90
Nonsurvivors n = 110
P Value
31.8 ± 15.2 19.0 ± 8.6 13.1 ± 3.1
43.9 ±25.8 34.6 ± 10.2 7.3 ± 5.4
0.18 0.001 0.01
3.7 ± 2.4 2.1 ± 1.7 1.1 ± 0.7
6.9 ± 4.1 4.0 ± 2.3 1.8 ± 0.5
0.12 0.11 0.04
215 238 241 1.9 6,361 5.1 12.8
± ± ± ± ± ± ±
119 104 116 4.1 2,990 5.9 11.0
138 121 175 9.2 9,660 4.8 4.8
± 107 _+ 64 ± 59 ± 9.1 ± 6,881 ± 5.5 ± 5.5
0.06 0.002 0.05 0.03 0.19 NS NS
*Ratio of partial pressure of oxygen in arterial blood to fraction of inspired oxygen. ISS = Injury Severity Score; GCS = GlasgowComa Score; ICU = intensivecare unit.
TABLE III MulUvariate Analysis of Mortality in Patients With Pulmonary Contusion Variable Age (y) PaO2/i=iO2* 24 hours Resuscitation volume (mL/48 hours) GSC ISS PaO2/FiO2 Admission 48 hours Transfusion requirement (units/48hours)
13
Sig T
0.42997 0.22073 0.21975 0.19779 0.07685
0.0032 0.1188 0.1076 0.1520 0.5979
0.04325 0.04048 0.02820
0.7313 0.7650 0.8396
Multiple R 0.64197. R Square 0.41213. Adjusted R Square 0.30275. *Ratio of partial pressure of oxygen in arterial blood to fraction of inspired oxygen GSC = GlasgowComa Score; ISS = Injury Severity Score.
At 10%, overall mortality with pulmonary contusion in the present series was lower than the 13% to 40% that has been reported previously. This may reflect the trend toward standardization of treatment (selective ventilatory support, aggressive pulmonary toilet, no steroid use or fluid restriction unless required by severe head or spinal injury), but it is probably due to less severe head injuries in our series. Our patients' mean ISS was 21, compared to 26 in a previous report4; our patients also had a shorter mean length of hospital stay (12 versus 22 days). Patients who suffer pulmonary contusions are often victims of blunt trauma with multiple injuries. 4"6'9 Severity of associated injuries and CNS injuries are the primary predictors of outcome in patients with pulmonary contusion. 4-7'9"~3 Previous work 5.6.15 suggested
that GCS can be used to predict mortality, requirement for ventilatory support, and hospital stay in patients with pulmonary contusions. These findings have led many clinicians to consider mortality in patients with pulmonary contusions to be primarily due to associated injuries.5.7, t4 Shock has been associated with mortality in several studies of blunt chest trauma, 5'6'9"j5 but in these studies, shock has been defined vaguely. In this study, plasma lactate was used as an indicator of shock and resuscitation. Elevated plasma lactate reflects anaerobic metabolism at the cellular level and is associated with severity of shock and the adequacy of resuscitation. 2t'22 Serial determinations of lactate levels objectively measure the adequacy of resuscitation over time. 23 Although lactate levels tended to be higher in nonsurvivors, the
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994 661
MORTALITY IN PULMONARY CONTUSIONfKOLLMORGEN ET AL
difference was not significant when compared with those of survivors. This suggests the presence of shock on admission, as measured by plasma lactate, is not independently associated with poor outcome in pulmonary contusion. Rather, it is an index of the severity of injuries incurred. Incomplete data precluded inclusion of plasma lactate in the multiple regression analysis of factors affecting mortality in patients with pulmonary contusion. Over the last 4 years of this 5-year review, it has been our practice to obtain serial plasma lactate levels in seriously injured trauma patients until the patients are hemodynamically stable and lactate levels are normal (<2 mmol/dL). In many cases, plasma lactate was within the normal range prior to 24 to 48 hours after admission, and levels were not obtained at these time intervals. The requirement for transfusion has been reported as both associated with 4 and unrelated tO 5"t4 increased mortality in patients with pulmonary contusions. We found that nonsurvivors required a significantly higher number of units of red blood cells in the first 48 hours of hospitalization than survivors (9.2 versus 1.9). When analyzed by multiple regresssion, we found transfusion requirements to be the weakest factor in the regression equation for predicting mortality (Table III). This suggests that transfusion requirements in the first 48 hours of hospitalization, like the presence of shock, reflect the severity of other injuries in patients with pulmonary contusions. They are not independently predictive of mortality. At one time, progression of pulmonary contusion was thought to be related to excessive crystalloid infusion, 1624 which led to the routine use of fluid restriction and/or diuretics in the care of patients with this injury. ]5.25 Recent studies have suggested that mortality is not related to the form ]4 or volume4 of fluid required in resuscitation. In our study, there was no significant difference between survivors and nonsurvivors in terms of the volume of intravenous (IV) fluid administered over the first 48 hours of hospitalization; but in the multivariate analysis for this injury, resuscitation volume figured prominently. This association could reflect either the requirement for more fluid in patients with more severe injuries or the pulmonary consequences of excess fluids. The correlation between resuscitation volume and ISS in our review was significant (0.44; P = O.01), whereas the correlation between resuscitation volume and oxygenation ratio at 48 hours was not significant (0.01; P <0.49). This may suggest that IV fluid volume given to patients with pulmonary contusions is related to mortality as a reflection of the severity of associated injuries and not as a contributor to lung dysfunction or hypoxia. After lung injury, blood and interstitial fluid accumulate in the damaged parenchyma resulting in decreased alveolar membrane diffusion and compliance. 24.26 This, in turn, leads to worsening atelectasis, intrapulmonary shunting, and hypoxemia. 3.14 The degree of hypoxemia roughly parallels the extent of parenchymal involvement in pulmonary contusion 18 and other forms of lung injury, 9 but is not consistently associated with mortal669.
ity in pulmonary contusions. 4'5"9'14 In our study, there was no significant difference in PaO2/FiO 2 on admission between survivors and nonsurvivors. Twenty-four hours after injury, when interstitial fluid accumulation is maximal, 27 the shunting and hypoxemia associated with the extent of lung injury should also be maximal. Multivariate analysis suggests that the oxygenation ratio 24 hours after admission, as opposed to oxygenation ratio on admission or at 48 hours after admission, is an influential and independent factor in predicting mortality in patients with pulmonary contusion. This observation supports the concept that interstitial water accumulation and subsequent hypoxemia peaks at 24 hours post-injury and then begins to decrease. 27 It is likely that the oxygenation ratio reflects the extent of lung parenchymal injury, a factor that is not measured by ISS or GCS. In summary, these results confirm that mortality in pulmonary contusion is related to a variety of factors including age, Injury Severity Score and Glasgow Coma Score. These results also suggest that mortality from pulmonary contusion is dependent upon the severity of the underlying lung parenchymal injury as reflected in the oxygenation ratio at 24 hours after admission. The fact that the majority of patients who died did so as a direct consequence of respiratory failure or primary lung injury further supports the concept that the severity of lung injury is an important determinant of outcome in patients suffering pulmonary contusion.
REFERENCES
1. Truncey SZ, Rodriguez A, Cowley RA. Preface. In: Truncey SZ, Rodriguez A, Cowley RA, eds. The Management o f Cardiothoracic Trauma. Baltimore: Williams and Wilkins; 1990:ix. 2. Lewis FR. Thoracic trauma. Surg Clin North Am. 1982;62: 97-104. 3. Shackford SR. Blunt chest trauma: the intensivist's perspective. J hzt Care Med. 1986;1:125-136. 4. Johnson, JJ, Cogbill TH, Winga ER. Determinants of outcome after pulmonary contusion. J Trauma. 1986;26:695-697. 5. Stellin G. Survival in trauma victims with pulmonary contusion. Am Surg. 1991;57:780-784. 6. Clark GC, Schecter WP, Trunkey DD. Variables affecting outcome in blunt chest trauma: flail chest vs. pulmonary contusion. J Trauma. 1988;28:298-304. 7. RodriguezA. Injuries of the chest wall, the lungs and the pleura. In: Truncey SZ, Rodriguez A, Cowley RA, eds. The Management of Cardiothoracic Trauma. Baltimore: Williams and Wilkins; 1990: 155-177. 8. Allen JE, Schwab CW. Blunt chest trauma in the elderly. Am Surg. 1985;51:697-700. 9. Pinella JC. Acute respiratory failure in severe blunt chest trauma. J Trauma. 1982;22:221-226. 10. Baker SP, O'Neill B, Haddon W Jr, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluation of emergency care. J Trauma. 1974; 14:187-196. 11. Committee on Injury Scaling. The Abbreviated hljur3., Scale, 1990 Revision. Des Plaines, Illinois: Associationfor Advancement of Automotive Medicine; 1990. 12. Teasdale G, Jennett B. Assessment of coma and impaired consciousness, a practical scale. Lancet. 1974;2:81-84.
THE AMERICANJOURNALOF SURGERY** VOLUME 168 DECEMBER1994
MORTALITY IN PULMONARY CONTUSIONfKOLLMORGEN El' AL
13. Svennevig JL, Bugge-Asperheim B, Vaage J, et al. Corticosteroids in the treatment of blunt injury of the chest. Injury. 1984; 16:80-84. 14. Bongard FS, Lewis FR. Crystalloid resuscitation of patients with pulmonary contusion. Am J Surg. 1984;148:145-15t. 15. Richardson JD, Adams L, Flint LM. Selective management of flail chest and pulmonary contusion. Ann Surg. 1982;196: 481-487. 16. Fulton RL, Peter ET. Compositional and histologic effects of fluid therapy following pulmonary contusion. J Trauma. 1974; 14:783-790. 17. Trinkle JK, Furman RW, Hinshaw MA, et al. Pulmonary contusion: pathogenesis and effect of various resuscitative measures. Ann Thorac Surg. 1973;16:568-573. 18. Erickson DR, Shinozaki T, Beekman E, Davis JH. Relationship of arterial blood gases and pulmonary radiographs to the degree of pulmonary damage in experimental pulmonary contusion. J Trauma. 1971; 11:689-694. 19. Wagner RB, Jamieson PM. Pulmonary contusion: evaluation and classification by computed tomography. Surg Clin North Am. 1989;69:31-34. 20. Wagner RB, Slivko B, Jamieson PM, et al. Effect of lung contusion of pulmonary hemodynamics. Ann Thorac Surg. 1991; 52:51-58. 21. Tuchschmidt J, Fried J, Swinney R, Sharma P. Early hemodynamic correlates of survival in patients with septic shock. Crit Care Med. 1989;17:719-723. 22. Bakker J, Coffernils M, Leon M, et al. Blood lactate levels are superior to oxygen derived variables in predicting outcome in human septic shock. Chest. 1991;99:956-962. 23. Vincent JL, Dufaye P, Berre L, et al. Serial lactate determinations during circulatory shock. Crit Care Med. 1983;11: 449-45 I. 24. Fulton RL, Peter ET. The progressive nature of pulmonary contusion. Surgery. 1970;76:499-506. 25. Taylor GA, Miller HA, Shulman HS, et al. Symposium of trauma. I. Controversies in the management of pulmonary contusion. Can J Sterg. 1982;25:167-170. 26. Oppenheimer L, Craven KD, Forkert L, Wood LD. Pathophysiology of pulmonary contusion in dogs. J App! Physiol. 1979;47:718-728. 27. Tranbaugh RF, Elings VB, Christensen JM, Lewis FR. Determinants of pulmonary interstitial fluid accumulation after trauma. J Trauma. 1982;22:820-826. DISCUSSION F r e d e r i c k A. M o o r e , M D (Denver, Colorado): This is a retrospective, 5-year review of 100 patients admitted to a university hospital trauma center with pulmonary contusion. The primary purpose was to determine w h a t factors contributed to the 10 deaths in this severely injured group of patients. Using univariate analysis the authors identified six variables that were associated with death. Not surprisingly, these include a 'high ISS, a low GCS, high lactate at 24 and 48 hours, a low PaO2/FiO 2 ratio at 24 to 48 hours, and high transfusion requirements. The authors next did a stratified risk analysis based on the fact that previous studies consistently identified advanced age, high ISS, and low GCS to be independent predictors of death. They chose to stratify their patients by three factors. Each of these factors was categorized into three groupings. For example, ISS was divided as less than 14, 14 to 25, and greater than 25. Thus, they ended up with 9 separate subsets of patients. They then used these 9 patient sub-
sets to determine whether eight additional factors were associated with death. Altogether, they did 72 univariate subset analyses. What they found was that a lower P a O J F i O 2 ratio at various time points was associated with death in 3 of the 9 age subsets, 4 of the 9 ISS subsets, and 4 of the 9 GCS subsets. Additionally, higher transfusion requirement was associated with death in the highest ISS and the lowest GCS subsets. Lastly, larger transfusion volume was associated with death in patients older than 45 years. Based on these analyses, the authors went on to make a series of conclusions that are consistent with my biases. Unfortunately, I have concerns about the database and statistical analyses used. First, the authors are studying a complex issue in a small group of patients where a large number of confounding variables may contribute to a small number of deaths. Secondly, the authors stratified the patients by three likely confounding variables, but in their analyses they only control for one of these variables at a time. What they need to do is to control all three variables simultaneously. This might be possible with the stratified risk analysis technique that they employed, but they would need much larger groups of patients. In my opinion, multivariate logistic regression analysis should have been used. The third major issue is that the authors are dealing with incomplete data sets. For example, 89 of 300 potential P a O J F i O 2 ratios are missing. This represents a real problem because it introduces clinician bias. For example, a low P a O J F i O 2 ratio may predict death because the residents choose to get arterial blood gases in patients they think are sicker. If this is true, then your analysis proves that your residents are good clinicians. In conclusion, I have several questions for the authors. First, this was a retrospective chart review. How did you confirm that the patients truly had pulmonary contusions? As you know, the radiologic diagnosis of pulmonary contusion can be confused with other entities such as aspiration or resolving lobar collapse. Why did you not use multivariate analysis? From the methods section, it appears that you do have the computer capabilities. Could you please address the issue of missing data? Do you believe that your finding for the low PaO2/FiO 2 ratio predicts death is specific to patients sustaining pulmonary contusion? I ask this because Dr. Holcroft recently published his validated trauma ICU score in which a low PaO2/FiO 2 ratio at 24 hours was predictive of death in generic trauma ICU patients. Daniel R. K o l l m o r g e n , MD: To make the diagnosis of pulmonary contusion, we looked at the timing of the x-ray changes, the location of the x-ray abnormality relative to the location of the traumatic blow, and whether the opacification was nonsegmental. Segmental opacification can represent atelectasis, aspiration, or contusion. In other words, although not 100% accurate, the timing and the anatomic distribution were the main criteria we used to differentiate contusion from atelectasis and aspiration. The second question about the statistical methods and multivariate analysis makes a good point. We have per-
THE AMERICANJOURNALOF SURGERY® VOLUME 168 DECEMBER1994 663
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
formed multivariate analysis since the manuscript was submitted and found that age, PaO2/FiO 2 ratio at 24 hours, and resuscitation volume are the three strongest, independent variables associated with mortality in the studied population. The third question about the presence or lack of data in the database also raises an important point. The patients who were doing well didn't always have lactates or blood gases checked. This lack of complete data sets is an inherent weakness of a retrospective study. Obviously, a larger, prospective study where the labs were checked at all times would be stronger.
664
Finally, Dr. Moore's last comment is about oxygenation ratio and whether it is specific for pulmonary contusion. This ratio has been used in a number of areas to measure the degree of lung injury; for example it has been used as one of the criteria for diagnosing acute respiratory distress syndrome. Oxygenation is a systemic problem in severely ill patients and can be related to gas exchange, oxygen delivery, or cellular consumption. Although this ratio is not specific for pulmonary contusion, we feel that in these patients, in this time course, the degree of pulmonary parenchymal injury is the major cause of intrapulmonary shunt and therefore poor oxygenation.
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994
BACKGROUND: Associated injuries and central nervous system (CNS) trauma are historically associated with poor outcome in patients with pulmonary contusions, but the value of specific factors reflecting shock, fluid resuscitation requirement and pulmonary parenchymal injury in predicting mortality in this population is not well established. METHODS: The medical records of 1 0 0 consecutive patients with pulmonary contusion, adndtted over a 5-year period, were retrospectively reviewed. Survivors and nonsurvivors were compared in terms of age, Injury Severity Score (ISS), Glasgow Coma Score (GCS), PaO2/FiO 2 (oxygenation ratio), the severity and adequacy of shock resuscitation reflected in plasma lactate, resuscitation volume and transfusion requirements, using oneway ANOVA. To determine the contribution of individual, interdependent variables to mortality, the data were then analyzed using multivariable analysis. RESULTS: ISS and transfusion requirement were significantly higher, and GCS and PaO2/FiO 2 at 2 4 and 4 8 hours after admission were significantly lower in nonsurvivors than in survivors. After multiple regression analysis, the factors most strongly associated with mortality included patient age, oxygenation ratio at 2 4 hours after admission, and resuscitation volume. CONCLUSIONS: Outcome in patients with pulmonary contusion is dependent upon a number of variables including the severity of pulmonary parenchymal injury as reflected in PaO2/FiO 2 ratio. horacic trauma is involved in nearly one third of acute admissions to trauma centers ~ and accounts for 25% of all trauma-related deaths in the United States. 2 Pulmonary contusion is often considered an inconsequential injury in these patients, but it is the sec"ond most c o m m o n injury in blunt thoracic trauma and
T
From the Departments of Surgery (DRK, JJS, MCM, RGB) and Radiology (KAM), University of Utah School of Medicine, Salt Lake City, Utah. Requests for reprints should be addressed to Richard G. Barton, MD, Department of Surgery, 3B313, Universityof Utah School of Medicine, 50 North Medical Drive, Salt Lake City, Utah 84132. This work was supported in part by NIH grant 1 RO HD 30455-01. Presented at the 46th Annual Meeting of the Southwestern Surgical Congress, Tucson, Arizona, April 17-20, 1994.
is associated with a mortality rate ranging from 14% to 40%. 1,3 Historically, the most consistent predictors of mortality have been patient age, severity of injury, and the presence of central nervous system (CNS) trauma, although a number of factors have been evaluated. 4-9 We reviewed our experience in the management of patients with pulmonary contusion in order to determine if other variables were associated with outcome and to examine our success in treating this injury. METHODS
Patient Population We reviewed the medical records of 100 consecutive patients admitted to the University of Utah Medical Center between January 1988 and December 1993 with a diagnosis of pulmonary contusion.
Diagnosis of Pulmonary Contusion The diagnosis of pulmonary contusion was based on the radiographic findings of nonsegmental areas of opacification resulting from a blow to the chest. Flail chest was defined clinically (paradoxical motion of the chest wall or sternum) or radiographically (the presence of 6 or more rib fractures or more than 3 segmental rib fractures in 1 hemithorax).
Patient Management All patients who had pulmonary contusions were resuscitated with intravenous fluids sufficient to support blood pressure, maintain urine output, and correct lactic acidosis. Fluid restriction and corticosteriods were not employed for pulmonary contusion but were used selectively for patients with major CNS trauma. Indications for intubation and ventilation included a respiratory rate >40 or <8, a PaCO 2 >50 mm Hg without a history of chronic obstructive disease, a PaO 2 <60 mm Hg or requirement of an FiO 2 exceeding 0.50, the need to establish or maintain a patent airway, clinical signs of respiratory distress, or hemodynamic instability. Patients with flail chest were managed selectively according to the same criteria.
Data Collection Data extracted from patient medical records included age, sex, mechanism of injury, requirement for ventilatory support, length of intensive care unit (ICU) and hospital stay, Injury Severity Score (ISS), Glasgow Coma Score (GCS) on admission, transfusion requirement in the first 48 hours after admission, resuscitation volume (total fluid volume received in the first 48 hours after admission), and admission chest roentgenogram results. In addition, plasma lactate and PaO2/FiO 2 were recorded at initial presentation as well as 24 and 48
THE AMERICANJOURNALOF SURGERY*' VOLUME 168 DECEMBER1994 659
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
The information in the database was analyzed in two phases. Initially, survivors and nonsurvivors were compared in terms of the previously described variables using one way analysis of variance. Subsequently, in order to identify factors that may be independent predictors of mortality in this patient population, information from the database was analyzed using multiple regression analysis. Plasma lactate values were excluded from this analysis due to inadequate information in many study patients. Statistical analysis was performed on an IBM PC using SPSS/PC (SPSS Corp., Boca Raton, Florida). A P value of <0.05 was considered significant. All data are expressed as mean value +_ standard deviation.
ber of units of blood transfused, and volume of resuscitation fluid received during the first 48 hours after admission. Differences in the length of ICU stay and the length of hospital stay between survivors and nonsurvivors are also included in Table II. Compared to survivors, nonsurvivors had a significantly higher ISS, lower GCS, higher plasma lactate 48 hours after admission, and a significantly lower oxygenation ratio at 24 and 48 hours after admission. Nonsurvivors also required significantly more blood transfusions than did survivors. Age, plasma lactate on admission and at 24 hours after admission, oxygenation ratio on admission, and volume of intravenous fluids in the first 48 hours of admission were not significantly different between survivors and nonsurvivors. To examine the relative importance of the interdependent variables described in Table II on mortality in patients with pulmonary contusion, the data were analyzed using multiple regression analysis. The results of this analysis are summarized in Table III. The regression equation produced a multiple R value of 0.642 and an adjusted R square of 0.303. The variables included in the multiple regression equation are listed in order of decreasing [3 value, The variables most strongly associated with mortality (the highest [3 values) in this analysis are patient age, oxygenation ratio 24 hours after admission, resuscitation volume, and GCS. The variables with the least influence on mortality are transfusion requirements, and oxygenation ratio on admission and 48 hours after admission. Due to inadequate data sets, plasma lactate values were not included in the analysis.
RESULTS One hundred patients (73 male, 27 female), with a mean age of 33 years, were admitted with a diagnosis of pulmonary contusion during the period reviewed. Eighty-six percent of the injuries were caused by motor vehicle or motorcycle accidents. The remainder were due to auto-pedestrian accidents (4%), crush injuries (3%), falls (2%), skiing accidents (2%), and miscellaneous injuries (3%). Flail chest was present in 17% of the study population, 14% of survivors and 30% of nonsurvivors. Fifty-two patients (52%) required intubation, including 49% of the survivors and all of the nonsurvivors. Ten patients died (10% mortality) and the causes of death in these patients are summarized in Table I. The mean duration of survival for patients who died was 5.5 days (range 0-17); however, 8 of the 10 nonsurvivors expired within the first 4 days following injury. Two patients survived for 12 and 17 days and died of multiple organ failure. Seven of 10 patients who died did so as a direct consequence of pulmonary failure or lung injury; 4 patients with acute respiratory failure, 2 patients with acute respiratory distress syndrome associated with multiple organ failure, and 1 of uncontrolled pulmonary hemorrhage. Table II compares survivors and nonsurvivors of pulmonary contusion in terms of factors previously associated with increased mortality. These factors include age ISS, GCS, plasma lactate and oxygenation ratio on admission, and at 24 and 48 hours after admission, num-
COMMENTS Although the adverse effects of associated injuries and CNS trauma in patients with pulmonary contusions are well documented, it is less clear how other clinical findings contribute to mortality. Advanced patient age 4.5,14 and the presence of hypotension 4'5'9't5 have been inconsistently correlated with poor outcome. Similarly, requirement for and volume of transfusions 4'5:4 or intravenous fluids 4'16'17are variably associated with mortality from pulmonary contusion. Further, attempts to define pulmonary parenchymal injury and measure its contribution to mortality in these patients 4,9'14"~82° have not established a clear, reproducible association. The present study was conducted to determine the value of factors reflecting shock, fluid resuscitation requirement, and pulmonary parenchymal injury in predicting outcome for patients of a similar age or severity of injury with pulmonary contusion. The demographic findings in this study are similar to those previously described for pulmonary contusion/'9 Young men comprise the majority of patients who suffer pulmonary contusions. Motor vehicle accidents (cars, all-terrain vehicles, and motorcycles) are the most common mechanisms of injury, accounting for nearly 90% of pulmonary contusions. Craniocerebral injuries are the most commonly associated nonthoracic injuries in blunt chest trauma, occurring in nearly 40% of cases. 9 In our study, 34% of all patients with pulmonary contusions also suffered head injuries.
TABLE I Causes of Death in Patients With Pulmonary Contusion Cause of Death
No.
Survival (d)
Acute respiratory failure Severe head injury Multiple organ failure Uncontrolled hemorrhage Multiple blunt injuries
4 2 2 1 1
1, 1, 2, 4 1, 4 12, 17 1 1
hours after admission. ISS was calculated using the method of Baker et aP ° and the 1990 Abbreviated Injury Scale. jl GCS was determined according to the criteria of Teasdale and Jennett.12 Statistical Analysis
660
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
TABLE II Comparison of Survivors and Nonsurvivors Predictor Age (y) ISS GCS Lactate (mmol/dL) Admission 24 hours 48 hours Oxygenation ratio (PaO2/Fi02)* Admission 24 hours 48 hours Transfusions (units/48 hours) Resuscitation volume (mL/48 hours) ICU length of stay (d) Hospital length of stay (d)
Survivors n = 90
Nonsurvivors n = 110
P Value
31.8 ± 15.2 19.0 ± 8.6 13.1 ± 3.1
43.9 ±25.8 34.6 ± 10.2 7.3 ± 5.4
0.18 0.001 0.01
3.7 ± 2.4 2.1 ± 1.7 1.1 ± 0.7
6.9 ± 4.1 4.0 ± 2.3 1.8 ± 0.5
0.12 0.11 0.04
215 238 241 1.9 6,361 5.1 12.8
± ± ± ± ± ± ±
119 104 116 4.1 2,990 5.9 11.0
138 121 175 9.2 9,660 4.8 4.8
± 107 _+ 64 ± 59 ± 9.1 ± 6,881 ± 5.5 ± 5.5
0.06 0.002 0.05 0.03 0.19 NS NS
*Ratio of partial pressure of oxygen in arterial blood to fraction of inspired oxygen. ISS = Injury Severity Score; GCS = GlasgowComa Score; ICU = intensivecare unit.
TABLE III MulUvariate Analysis of Mortality in Patients With Pulmonary Contusion Variable Age (y) PaO2/i=iO2* 24 hours Resuscitation volume (mL/48 hours) GSC ISS PaO2/FiO2 Admission 48 hours Transfusion requirement (units/48hours)
13
Sig T
0.42997 0.22073 0.21975 0.19779 0.07685
0.0032 0.1188 0.1076 0.1520 0.5979
0.04325 0.04048 0.02820
0.7313 0.7650 0.8396
Multiple R 0.64197. R Square 0.41213. Adjusted R Square 0.30275. *Ratio of partial pressure of oxygen in arterial blood to fraction of inspired oxygen GSC = GlasgowComa Score; ISS = Injury Severity Score.
At 10%, overall mortality with pulmonary contusion in the present series was lower than the 13% to 40% that has been reported previously. This may reflect the trend toward standardization of treatment (selective ventilatory support, aggressive pulmonary toilet, no steroid use or fluid restriction unless required by severe head or spinal injury), but it is probably due to less severe head injuries in our series. Our patients' mean ISS was 21, compared to 26 in a previous report4; our patients also had a shorter mean length of hospital stay (12 versus 22 days). Patients who suffer pulmonary contusions are often victims of blunt trauma with multiple injuries. 4"6'9 Severity of associated injuries and CNS injuries are the primary predictors of outcome in patients with pulmonary contusion. 4-7'9"~3 Previous work 5.6.15 suggested
that GCS can be used to predict mortality, requirement for ventilatory support, and hospital stay in patients with pulmonary contusions. These findings have led many clinicians to consider mortality in patients with pulmonary contusions to be primarily due to associated injuries.5.7, t4 Shock has been associated with mortality in several studies of blunt chest trauma, 5'6'9"j5 but in these studies, shock has been defined vaguely. In this study, plasma lactate was used as an indicator of shock and resuscitation. Elevated plasma lactate reflects anaerobic metabolism at the cellular level and is associated with severity of shock and the adequacy of resuscitation. 2t'22 Serial determinations of lactate levels objectively measure the adequacy of resuscitation over time. 23 Although lactate levels tended to be higher in nonsurvivors, the
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994 661
MORTALITY IN PULMONARY CONTUSIONfKOLLMORGEN ET AL
difference was not significant when compared with those of survivors. This suggests the presence of shock on admission, as measured by plasma lactate, is not independently associated with poor outcome in pulmonary contusion. Rather, it is an index of the severity of injuries incurred. Incomplete data precluded inclusion of plasma lactate in the multiple regression analysis of factors affecting mortality in patients with pulmonary contusion. Over the last 4 years of this 5-year review, it has been our practice to obtain serial plasma lactate levels in seriously injured trauma patients until the patients are hemodynamically stable and lactate levels are normal (<2 mmol/dL). In many cases, plasma lactate was within the normal range prior to 24 to 48 hours after admission, and levels were not obtained at these time intervals. The requirement for transfusion has been reported as both associated with 4 and unrelated tO 5"t4 increased mortality in patients with pulmonary contusions. We found that nonsurvivors required a significantly higher number of units of red blood cells in the first 48 hours of hospitalization than survivors (9.2 versus 1.9). When analyzed by multiple regresssion, we found transfusion requirements to be the weakest factor in the regression equation for predicting mortality (Table III). This suggests that transfusion requirements in the first 48 hours of hospitalization, like the presence of shock, reflect the severity of other injuries in patients with pulmonary contusions. They are not independently predictive of mortality. At one time, progression of pulmonary contusion was thought to be related to excessive crystalloid infusion, 1624 which led to the routine use of fluid restriction and/or diuretics in the care of patients with this injury. ]5.25 Recent studies have suggested that mortality is not related to the form ]4 or volume4 of fluid required in resuscitation. In our study, there was no significant difference between survivors and nonsurvivors in terms of the volume of intravenous (IV) fluid administered over the first 48 hours of hospitalization; but in the multivariate analysis for this injury, resuscitation volume figured prominently. This association could reflect either the requirement for more fluid in patients with more severe injuries or the pulmonary consequences of excess fluids. The correlation between resuscitation volume and ISS in our review was significant (0.44; P = O.01), whereas the correlation between resuscitation volume and oxygenation ratio at 48 hours was not significant (0.01; P <0.49). This may suggest that IV fluid volume given to patients with pulmonary contusions is related to mortality as a reflection of the severity of associated injuries and not as a contributor to lung dysfunction or hypoxia. After lung injury, blood and interstitial fluid accumulate in the damaged parenchyma resulting in decreased alveolar membrane diffusion and compliance. 24.26 This, in turn, leads to worsening atelectasis, intrapulmonary shunting, and hypoxemia. 3.14 The degree of hypoxemia roughly parallels the extent of parenchymal involvement in pulmonary contusion 18 and other forms of lung injury, 9 but is not consistently associated with mortal669.
ity in pulmonary contusions. 4'5"9'14 In our study, there was no significant difference in PaO2/FiO 2 on admission between survivors and nonsurvivors. Twenty-four hours after injury, when interstitial fluid accumulation is maximal, 27 the shunting and hypoxemia associated with the extent of lung injury should also be maximal. Multivariate analysis suggests that the oxygenation ratio 24 hours after admission, as opposed to oxygenation ratio on admission or at 48 hours after admission, is an influential and independent factor in predicting mortality in patients with pulmonary contusion. This observation supports the concept that interstitial water accumulation and subsequent hypoxemia peaks at 24 hours post-injury and then begins to decrease. 27 It is likely that the oxygenation ratio reflects the extent of lung parenchymal injury, a factor that is not measured by ISS or GCS. In summary, these results confirm that mortality in pulmonary contusion is related to a variety of factors including age, Injury Severity Score and Glasgow Coma Score. These results also suggest that mortality from pulmonary contusion is dependent upon the severity of the underlying lung parenchymal injury as reflected in the oxygenation ratio at 24 hours after admission. The fact that the majority of patients who died did so as a direct consequence of respiratory failure or primary lung injury further supports the concept that the severity of lung injury is an important determinant of outcome in patients suffering pulmonary contusion.
REFERENCES
1. Truncey SZ, Rodriguez A, Cowley RA. Preface. In: Truncey SZ, Rodriguez A, Cowley RA, eds. The Management o f Cardiothoracic Trauma. Baltimore: Williams and Wilkins; 1990:ix. 2. Lewis FR. Thoracic trauma. Surg Clin North Am. 1982;62: 97-104. 3. Shackford SR. Blunt chest trauma: the intensivist's perspective. J hzt Care Med. 1986;1:125-136. 4. Johnson, JJ, Cogbill TH, Winga ER. Determinants of outcome after pulmonary contusion. J Trauma. 1986;26:695-697. 5. Stellin G. Survival in trauma victims with pulmonary contusion. Am Surg. 1991;57:780-784. 6. Clark GC, Schecter WP, Trunkey DD. Variables affecting outcome in blunt chest trauma: flail chest vs. pulmonary contusion. J Trauma. 1988;28:298-304. 7. RodriguezA. Injuries of the chest wall, the lungs and the pleura. In: Truncey SZ, Rodriguez A, Cowley RA, eds. The Management of Cardiothoracic Trauma. Baltimore: Williams and Wilkins; 1990: 155-177. 8. Allen JE, Schwab CW. Blunt chest trauma in the elderly. Am Surg. 1985;51:697-700. 9. Pinella JC. Acute respiratory failure in severe blunt chest trauma. J Trauma. 1982;22:221-226. 10. Baker SP, O'Neill B, Haddon W Jr, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluation of emergency care. J Trauma. 1974; 14:187-196. 11. Committee on Injury Scaling. The Abbreviated hljur3., Scale, 1990 Revision. Des Plaines, Illinois: Associationfor Advancement of Automotive Medicine; 1990. 12. Teasdale G, Jennett B. Assessment of coma and impaired consciousness, a practical scale. Lancet. 1974;2:81-84.
THE AMERICANJOURNALOF SURGERY** VOLUME 168 DECEMBER1994
MORTALITY IN PULMONARY CONTUSIONfKOLLMORGEN El' AL
13. Svennevig JL, Bugge-Asperheim B, Vaage J, et al. Corticosteroids in the treatment of blunt injury of the chest. Injury. 1984; 16:80-84. 14. Bongard FS, Lewis FR. Crystalloid resuscitation of patients with pulmonary contusion. Am J Surg. 1984;148:145-15t. 15. Richardson JD, Adams L, Flint LM. Selective management of flail chest and pulmonary contusion. Ann Surg. 1982;196: 481-487. 16. Fulton RL, Peter ET. Compositional and histologic effects of fluid therapy following pulmonary contusion. J Trauma. 1974; 14:783-790. 17. Trinkle JK, Furman RW, Hinshaw MA, et al. Pulmonary contusion: pathogenesis and effect of various resuscitative measures. Ann Thorac Surg. 1973;16:568-573. 18. Erickson DR, Shinozaki T, Beekman E, Davis JH. Relationship of arterial blood gases and pulmonary radiographs to the degree of pulmonary damage in experimental pulmonary contusion. J Trauma. 1971; 11:689-694. 19. Wagner RB, Jamieson PM. Pulmonary contusion: evaluation and classification by computed tomography. Surg Clin North Am. 1989;69:31-34. 20. Wagner RB, Slivko B, Jamieson PM, et al. Effect of lung contusion of pulmonary hemodynamics. Ann Thorac Surg. 1991; 52:51-58. 21. Tuchschmidt J, Fried J, Swinney R, Sharma P. Early hemodynamic correlates of survival in patients with septic shock. Crit Care Med. 1989;17:719-723. 22. Bakker J, Coffernils M, Leon M, et al. Blood lactate levels are superior to oxygen derived variables in predicting outcome in human septic shock. Chest. 1991;99:956-962. 23. Vincent JL, Dufaye P, Berre L, et al. Serial lactate determinations during circulatory shock. Crit Care Med. 1983;11: 449-45 I. 24. Fulton RL, Peter ET. The progressive nature of pulmonary contusion. Surgery. 1970;76:499-506. 25. Taylor GA, Miller HA, Shulman HS, et al. Symposium of trauma. I. Controversies in the management of pulmonary contusion. Can J Sterg. 1982;25:167-170. 26. Oppenheimer L, Craven KD, Forkert L, Wood LD. Pathophysiology of pulmonary contusion in dogs. J App! Physiol. 1979;47:718-728. 27. Tranbaugh RF, Elings VB, Christensen JM, Lewis FR. Determinants of pulmonary interstitial fluid accumulation after trauma. J Trauma. 1982;22:820-826. DISCUSSION F r e d e r i c k A. M o o r e , M D (Denver, Colorado): This is a retrospective, 5-year review of 100 patients admitted to a university hospital trauma center with pulmonary contusion. The primary purpose was to determine w h a t factors contributed to the 10 deaths in this severely injured group of patients. Using univariate analysis the authors identified six variables that were associated with death. Not surprisingly, these include a 'high ISS, a low GCS, high lactate at 24 and 48 hours, a low PaO2/FiO 2 ratio at 24 to 48 hours, and high transfusion requirements. The authors next did a stratified risk analysis based on the fact that previous studies consistently identified advanced age, high ISS, and low GCS to be independent predictors of death. They chose to stratify their patients by three factors. Each of these factors was categorized into three groupings. For example, ISS was divided as less than 14, 14 to 25, and greater than 25. Thus, they ended up with 9 separate subsets of patients. They then used these 9 patient sub-
sets to determine whether eight additional factors were associated with death. Altogether, they did 72 univariate subset analyses. What they found was that a lower P a O J F i O 2 ratio at various time points was associated with death in 3 of the 9 age subsets, 4 of the 9 ISS subsets, and 4 of the 9 GCS subsets. Additionally, higher transfusion requirement was associated with death in the highest ISS and the lowest GCS subsets. Lastly, larger transfusion volume was associated with death in patients older than 45 years. Based on these analyses, the authors went on to make a series of conclusions that are consistent with my biases. Unfortunately, I have concerns about the database and statistical analyses used. First, the authors are studying a complex issue in a small group of patients where a large number of confounding variables may contribute to a small number of deaths. Secondly, the authors stratified the patients by three likely confounding variables, but in their analyses they only control for one of these variables at a time. What they need to do is to control all three variables simultaneously. This might be possible with the stratified risk analysis technique that they employed, but they would need much larger groups of patients. In my opinion, multivariate logistic regression analysis should have been used. The third major issue is that the authors are dealing with incomplete data sets. For example, 89 of 300 potential P a O J F i O 2 ratios are missing. This represents a real problem because it introduces clinician bias. For example, a low P a O J F i O 2 ratio may predict death because the residents choose to get arterial blood gases in patients they think are sicker. If this is true, then your analysis proves that your residents are good clinicians. In conclusion, I have several questions for the authors. First, this was a retrospective chart review. How did you confirm that the patients truly had pulmonary contusions? As you know, the radiologic diagnosis of pulmonary contusion can be confused with other entities such as aspiration or resolving lobar collapse. Why did you not use multivariate analysis? From the methods section, it appears that you do have the computer capabilities. Could you please address the issue of missing data? Do you believe that your finding for the low PaO2/FiO 2 ratio predicts death is specific to patients sustaining pulmonary contusion? I ask this because Dr. Holcroft recently published his validated trauma ICU score in which a low PaO2/FiO 2 ratio at 24 hours was predictive of death in generic trauma ICU patients. Daniel R. K o l l m o r g e n , MD: To make the diagnosis of pulmonary contusion, we looked at the timing of the x-ray changes, the location of the x-ray abnormality relative to the location of the traumatic blow, and whether the opacification was nonsegmental. Segmental opacification can represent atelectasis, aspiration, or contusion. In other words, although not 100% accurate, the timing and the anatomic distribution were the main criteria we used to differentiate contusion from atelectasis and aspiration. The second question about the statistical methods and multivariate analysis makes a good point. We have per-
THE AMERICANJOURNALOF SURGERY® VOLUME 168 DECEMBER1994 663
MORTALITY IN PULMONARY CONTUSION/KOLLMORGEN ET AL
formed multivariate analysis since the manuscript was submitted and found that age, PaO2/FiO 2 ratio at 24 hours, and resuscitation volume are the three strongest, independent variables associated with mortality in the studied population. The third question about the presence or lack of data in the database also raises an important point. The patients who were doing well didn't always have lactates or blood gases checked. This lack of complete data sets is an inherent weakness of a retrospective study. Obviously, a larger, prospective study where the labs were checked at all times would be stronger.
664
Finally, Dr. Moore's last comment is about oxygenation ratio and whether it is specific for pulmonary contusion. This ratio has been used in a number of areas to measure the degree of lung injury; for example it has been used as one of the criteria for diagnosing acute respiratory distress syndrome. Oxygenation is a systemic problem in severely ill patients and can be related to gas exchange, oxygen delivery, or cellular consumption. Although this ratio is not specific for pulmonary contusion, we feel that in these patients, in this time course, the degree of pulmonary parenchymal injury is the major cause of intrapulmonary shunt and therefore poor oxygenation.
THE AMERICANJOURNALOF SURGERY® VOLUME168 DECEMBER1994