Correlation between mortality and total chest tube output before thoracotomy. The mean (± SD) total output before thoracotomy was 1627 ± 945 mL. The mean (± SD) time to thoracotomy was 2.4 ± 5.4 hours. Risk for death is a logarithmic scale based on the assumption that risk for death at 250 mL is 1.
Correlation between mortality and chest tube output observed (excluding initial output on placement) before thoracotomy. These data are based on the 56 patients for whom the primary indication for thoracotomy was ongoing blood loss through a chest tube. The overall mean (± SD) of observed blood loss was 5425 ± 1560 mL during a mean (± SD) observation of 3.1 ± 4.7 hours. Risk for death is a logarithmic scale based on the assumption that risk for death at 250 mL is 1.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Karmy-Jones R, Jurkovich GJ, Nathens AB, et al. Timing of Urgent Thoracotomy for Hemorrhage After Trauma: A Multicenter Study. Arch Surg. 2001;136(5):513–518. doi:10.1001/archsurg.136.5.513
It is possible to quantify an amount of thoracic hemorrhage, after blunt and penetrating injury, at which delay of thoracotomy is associated with increased mortality.
A retrospective case series.
Five urban trauma centers.
Patients undergoing urgent thoracotomy (within 48 hours of injury) for hemorrhage (excluding emergency department thoracotomy).
Respective registries identified patients who underwent urgent thoracotomy. Injury characteristics, initial and subsequent chest tube outputs, time before thoracotomy, and outcomes were evaluated.
Main Outcome Measure
One hundred fifty-seven patients (36 with blunt and 121 with penetrating injuries) underwent urgent thoracotomy for hemorrhage between January 1, 1995, and December 31, 1998. Mortality correlated with mean (± SD) Injury Severity Score (38 ± 19 vs 22 ± 12.6 for survivors; P<.01) and mechanism (24 [67%] for blunt vs 21 [17%] for penetrating injuries; P<.01). Mortality increased as total chest blood loss increased, with the risk for death at blood loss of 1500 mL being 3 times greater than at 500 mL. Blunt-injured patients waited a significantly longer time to thoracotomy than penetrating-injured patients (4.4 ± 9.0 h vs 1.6 ± 3.0 h; P = .02) and also had a greater total chest tube output before thoracotomy (2220 ± 1235 mL vs 1438 ± 747 mL; P = .001).
The risk for death increases linearly with total chest hemorrhage after thoracic injury. Thoracotomy is indicated when total chest tube output exceeds 1500 mL within 24 hours, regardless of injury mechanism.
THE INDICATIONS for thoracotomy after traumatic injury typically include shock, arrest at presentation, diagnosis of specific injuries (such as blunt aortic injury), or ongoing thoracic hemorrhage. This latter indication is most commonly specified to mean greater than 1500 mL of initial chest tube output (CT-initial) or continued hourly blood loss of 250 mL or more for 3 consecutive hours after tube thoracostomy.1 This concept was largely derived from observations made in the early 1970s based on experience predominantly with penetrating injuries.2-4 It is not clear whether these guidelines should be or are being applied equally after blunt trauma.1,5,6 We therefore conducted a retrospective, multicenter review of trauma patients undergoing urgent thoracotomy specifically because of ongoing thoracic hemorrhage in an effort to better define indications for thoracotomy based on blood loss and subsequent outcomes.
A retrospective review of patients who underwent urgent thoracotomy after traumatic injury between January 1, 1995, and December 31, 1998, at 5 participating level I trauma centers was conducted. All participating institutions are members of the Western Trauma Association Multi-Center Study Group. Urgent thoracotomy was defined as thoracotomy performed within 48 hours of injury, specifically excluding resuscitative thoracotomy performed in the emergency department. This definition was chosen to exclude patients who required thoracotomy because of a late complication, such as empyema. An additional inclusion criterion was thoracotomy for hemorrhage, specifically characterized by 1 of the following: (1) large hemothorax, (2) initial output on placement of chest tube, and (3) ongoing chest tube output over a period of observation. These criteria were determined by the reviewers and based on personal knowledge of the case or results of review with the primary surgeon and of the operative records. Thus, large hemothorax was defined as one that was apparent on plain chest radiographic findings (usually greater than one third and/or opacification of the involved hemithorax) and was the primary indication for performing thoracotomy. If the records indicated that the decision to perform thoracotomy was prompted by the amount of blood released on placing the chest tube(s), this was categorized as initial output. Similarly, if thoracotomy was performed specifically because of continued bleeding through the chest tube(s), as distinct from the other 2 criteria, this was categorized as ongoing chest tube output. The categorization was based not on a specific volume, but rather on what was determined as the indication for thoracotomy.
Demographic data collected included mechanism of injury, sex, and age. Injury was characterized by Injury Severity Score (ISS) and by Abbreviated Injury Score (AIS) for each body region, and specific thoracic injuries were recorded.7 Physiological data collected included presenting systolic blood pressure (SBP), SBP on arrival to the operating room (SBP-OR), time from admission to operative intervention, and outcome. The volume of CT-initial, the volume of chest tube output during observation before thoracotomy (CT-observed), and the total amount of chest tube output before thoracotomy (CT-initial + CT-observed = CT-total) were recorded.
Results are expressed as mean ± SD. Univariate analysis was performed using the χ2 test for categorical data. When a table had at least 1 cell with an expected frequency of less than 5, the Fisher exact test was used. The independent-pairs t test, with the Levene test for equality of variances, was used for continuous variables. All variables significant at a level of P≤.10 were entered into a logistic regression model to determine predictors of outcome. Statistical significance was taken to be P≤.05. Odds ratios were used to predict the risk for a particular outcome. All statistical analyses were performed using commercially available software (SPSS 7.5 for Windows; SPSS Inc, Chicago, Ill). Approval was obtained from the respective institutional review boards before commencing the study.
During the 4-year study period, a total of 451 patients underwent urgent (nonresuscitative) thoracotomies in the 5 centers (161 after blunt injury mechanisms and 290 after penetrating trauma). Of these, 157 patients underwent thoracotomy for thoracic hemorrhage (36 [22%] with blunt and 121 [42%] with penetrating injuries) and constitute the study group of this report (Table 1). One hundred thirty-five patients underwent thoracotomy within 4 hours, 11 from 5 to 8 hours, 10 from 9 to 24 hours, and 1 patient from 25 to 48 hours after admission. Mean time to thoracotomy was 2.4 ± 5.4 hours. Operative approaches included anterolateral thoracotomy (n = 78), posterolateral thoracotomy (n = 64), sternotomy (n = 8), clamshell incision (n = 5), and rigid thoracoscopy (n = 2).
Associated thoracic injuries based on mechanism are listed in Table 2. The overall mortality was 45 (29%). Of the 24 deaths occurring after blunt injury, 13 (54%) were directly attributable to thoracic hemorrhage. Thoracic hemorrhage was the primary cause of death in 10 (48%) of the 21 patients who died after penetrating trauma.
By univariate analysis, the following factors were found to be significant to the level of P≤.10: advanced age; increased ISS; and increasing AIS in the head and neck, abdominal and pelvic, and pelvic girdle and extremity body regions, but not AIS in the chest body region. In addition, SPB-OR, CT-observed, CT-total, need for laparotomy, female sex, injury mechanism, blunt aortic injury, and chest wall bleeding were also significant (Table 3 and Table 4).
A regression model was developed incorporating the factors found to be significant to the level of P≤.10. Injury Severity Score and CT-total before thoracotomy were used in place of regional AIS and CT-observed. Total chest tube output before thoracotomy and ISS retained independent significance, as seen in the following tabulation:
Furthermore, increasing CT-total was associated with a progressive increase in the risk for death; odds ratios were calculated to be an increase in risk for death by 1.6 times (95% confidence interval, 1.3-1.9 times) for each 500 mL of CT-total before thoracotomy. The impact of CT-observed and CT-total on risk for death is represented graphically in Figure 1 and Figure 2.
The analysis was repeated using AIS for head and neck, abdominal and pelvic, and extremities and pelvic girdle regions in place of ISS, as well as replacing CT-total with CT-observed. The only factors independently associated with mortality were CT-observed (P = .08) and SBP-OR (P = .05).
Patients who had blunt trauma had significantly greater ISS than those admitted after penetrating trauma (43 ± 18.5 vs 21 ± 11.8; P = .001), although there were no differences in the severity of chest injuries (chest AIS 3.6 ± 1.3 for blunt-injured vs 3.8 ± 0.8 for penetrating-injured patients; P = .26). In addition, patients with blunt injury had a significantly longer observation compared with those with penetrating injuries (4.4 ± 9.0 hours vs 1.6 ± 3.0 hours; P = .01) and a larger CT-total (2.2 ± 1.2 L vs 1.4 ± 0.7 L; P = .001). In an attempt to determine what factors might be influencing these differences, we studied the number of diagnostic tests performed before thoracotomy and the timing of thoracotomy in relation to other operations (ie, laparotomy) in patients with multisystem injuries. Eighteen blunt-injured patients had a laparotomy, of which 15 (83%) underwent laparotomy before the thoracotomy. This differed significantly from the 34 penetrating-injured patients who required laparotomy. In this group, 19 (56%) underwent laparotomy before the thoracotomy (P = .05). Of the 121 patients who underwent thoracotomy after penetrating injuries, 100 (83%) went directly from the emergency department to the operating room with no diagnostic tests other than a portable chest x-ray. This was significantly greater than among patients with blunt trauma, of whom only 21 (58%) went straight from the emergency department to the operating room (P = .002). Blunt-injured patients underwent more complex radiological evaluations, in particular diagnostic angiography (14% vs 2%; P = .02).
It generally has been reported that thoracotomy will be required in approximately 30% of cases presenting after penetrating chest injury and in 15% after blunt chest trauma.1,3,6,8-10 Outcomes after blunt injury are significantly worse, predominately due to extrathoracic injuries.5,6,9 In addition, there is a tendency to greater delay in performing surgical intervention after blunt compared with penetrating chest trauma.8,11 There are, in general, 3 indications for thoracotomy after traumatic injury: shock or arrest with suspected correctable intrathoracic lesion; specific diagnoses (eg, penetrating cardiac or blunt aortic injury), and evidence of ongoing thoracic hemorrhage.1,5,8,10
This latter indication became recognized as experience with penetrating chest injuries increased. In 1970, based on their experiences with thoracic causalities arising from the Vietnam conflict, McNamara and associates2 described a reduction in mortality when thoracotomy was performed early after penetrating trauma. They commented that
The current report cites criteria for exploration of 1000 to 1500 mLs of blood in the thoracic cavity upon initial insertion of a tube and 500 mLs of bleeding within the first hour after insertion. This is somewhat more aggressive than the criteria suggested by the World War II experience and is commensurate with the decreased risk of thoracotomy in current surgical practice.
In the discussion, Webb,4 commenting on the experience derived from the "domestic battlefield," noted that "any patient who has an early loss of as much as 1500 mL of blood is in need of operation." In 1976, Kish and associates3 described a series of 59 patients who required thoracotomy after injury. Four (9%) of the 44 patients with penetrating and 2 (13%) of the 15 patients with blunt injuries required thoracotomy 6 to 36 hours after admission, not because of shock or identification of a specific injury, "but because the rate of bleeding either was greater than 150 mL per hour for more than 10 hours or amounted to 1500 milliliters in a shorter period." Currently, "initial chest tube output exceeding 1500 milliliters or a continued hourly output of more than 250 milliliters for 3 consecutive hours" are accepted guidelines for surgical exploration after penetrating chest trauma.1
In contrast, it has been argued that "urgent thoracotomy for blunt trauma is rarely justified on the basis of chest tube output alone."1 A significant reason is that patients who need thoracotomy after blunt injury are usually in shock or have clearly identified injuries, usually blunt aortic injuries.1,3,11 Despite this, other authors have noted that ongoing chest tube output is not an uncommon indication for thoracotomy after blunt injury, and also that increasing blood loss appears to be related to increased mortality, independent of other factors discussed previously.5,11 Our group also has been concerned that the concept of "hourly output" may lead to delays in intervention, either because of fluctuations in chest tube drainage or because of nonfunctioning tubes, which inadequately drain the hemithorax. There may be, in addition, a nihilistic attitude because of the high incidence of bleeding originating from diffuse chest wall sites as well as the recognition that outcome after blunt chest trauma is related to associated extrathoracic injuries in most cases.
In our combined experience, all but 1 patient underwent thoracotomy within 24 hours of injury. A significant component of delay in operative intervention appeared to be a period of prolonged observation of chest tube output, particularly after blunt trauma. The risk for death if CT-total exceeded 1500 mL before thoracotomy was 3.2 times greater than with CT-total of 500 mL or less. Based on this increase in mortality after a CT-total of 1500 mL, it may be worthwhile to consider thoracotomy in patients who have CT-total of greater than 1500 mL in the first 24 hours of admission, even if there is no evidence of shock. Using these criteria, it may be possible to simplify indications for thoracotomy by no longer relying on hourly output calculations. Indeed, the findings that the time to thoracotomy did not appear to be related to mortality but rather the absolute amount of blood loss emphasize the importance of not being distracted by the concept of latest hourly output. It is critically important to ensure that chest tubes are adequately placed, are not occluded, and are appropriately draining the hemithorax. In selected stable patients, thoracoscopy or limited thoracotomy may be a reasonable alternative to standard thoracotomy.2,12
All centers recognize that there are individual cases in which the presentation is delayed for several hours. In such cases, there may be initial large chest tube output, but in stable patients with no further evidence of hemorrhage, thoracotomy is usally not indicated. In addition, our study suffers not only from being retrospective in design, but also arbitrary in terms of defintions of indications for thoracotomy. Despite these limitations, our results support the concept that, in the absence of shock or specific diagnosis, increasing blood loss, coupled with prolonged delay before performing thoracotomy, is associated with increased morbidity and mortality. The data also indicate that criteria for thoracotomy should be applied equally to blunt and penetrating trauma. Our data again confirm that thoracotomy for blunt trauma is associated with a significantly worse prognosis than for penetrating injury. Admittedly, there are significant confounding factors after blunt injury, including the need for laparotomy, pelvic stabilization, and embolization of pelvic bleeding and/or ruling out blunt aortic injury, sometimes all occurring in the same patient. Nevertheless, delaying thoracotomy in the face of ongoing thoracic hemorrhage simply because the mechanism of injury is blunt is associated with a significantly increased risk for morbidity and mortality. It is significant that approximately 50% of deaths in the blunt as well as penetrating injury groups were believed by the reviewing institutions to be directly related to thoracic hemorrhage.
Ongoing hemorrhage without shock, although a more common indication for thoracotomy after penetrating injury, is not unusual after blunt trauma. Thoracic bleeding of greater than 1500 mL within the first 24 hours of admission after injury is a simpler criteria to prompt consideration of operative intervention. This should be applied equally to penetrating and blunt trauma.
Presented at the 108th Scientific Session of the Western Surgical Association, Dana Point, Calif, November 13, 2000.
Corresponding author: Riyad Karmy-Jones, MD, Department of Surgery, Campus Box 359796, 325 Ninth Ave, Harborview Medical Center, Seattle, WA 98104 (e-mail: firstname.lastname@example.org).
Jorge J. Rodriguez, MD, Minneapolis, Minn: I would like to congratulate Dr Karmy-Jones and his colleagues for an excellent presentation and more importantly for the effort to look at the question utilizing a multicenter approach. In brief, they have defined that utilizing the traditional concept of greater than 1500 cc of initial chest tube output and/or continuous loss of 250 cc or more for 3 hours following tube thoracostomy was associated with an increased blood loss, a delay in performing thoracotomy, especially in blunt trauma patients, and in increased mortality and morbidity. They urged us to utilize a much more simple criteria for operative intervention, which is greater than 1500 cc within the first 24 hours of injury, and feel that if this is applied to both penetrating and blunt trauma, we should avoid the increased risk of death.
I have 3 simple questions for the authors. First, the overall mortality of this group was 45%. You attributed approximately 50% of the deaths as directly due to thoracic hemorrhage. How did you reach this conclusion? It is not clear in the manuscript. What criteria did you utilize to define these results? Second, and more importantly, from your data regardless of the mechanism of injury, the patient group presented to the emergency room with an average SBP of 92 mm Hg. By the time they reached the OR, their average SBP was 79 mm Hg. In fact, the patients who died had a mean SBP of 69 mm Hg on presentation to the operating room, and those who survived had a mean SBP of 82 mm Hg. Furthermore, in your manuscript you state that the time from admission to OR thoracotomy in blunt trauma patients was 4.4 hours and for penetrating patients, 1.6 hours. The piece of information that is critical is the management of resuscitation during the period of observation. I think whether we are vascular surgeons, endocrine surgeons, or trauma surgeons, we are well aware that continuous shock with an SBP of less than 90 mm Hg is associated with significantly poor morbidity and mortality. Therefore, from your data one can conclude that underresuscitated patients who have blood loss from the chest don't do well, especially if they are observed. I think it is important if you would address what the management was during the resuscitation period of observation before operative intervention occurred.
Last, based on your data, is the Western Trauma Study Group, which is part of the group that you acquired this data from, are they willing to put together a prospective randomized study to validate your criteria?
Ernest E. Moore, MD, Denver, Colo: Having been identified as one of the skeptics of the conclusions, I feel compelled to make some comments. It is laudable to perform multi-institutional studies, but I think the authors would acknowledge that retrospectively collected information has serious limitations in drawing firm conclusions. One of the key factors in analyzing blood loss is the time interval. Certainly the patient who sustains a stab wound in the wee hours of the morning and arrives the next day by private car hemodynamically stable with 1500 cc of blood out of a chest tube is far different than a patient with a gunshot wound who arrives by ambulance 5 minutes after being shot with 1500 cc from the chest tube. Furthermore, a major flaw in the study design is selection bias. Only patients who underwent operative intervention were evaluated. Consequently, we have no idea of the denominator, ie, how many patients sustaining blunt trauma had 1500 cc of blood out in the first 48 hours and did not require an operation? I am particularly uneasy about your recommendation for emergent thoracotomy in blunt multisystem-injured patients with a coagulopathy. Interventional radiologists now have the capability of embolizing intercostal arteries. I noted in the abstract that over two thirds of the patients who had thoracotomy for blunt trauma had bleeding attributed to intercostal vessels.
Finally, with respect to the multiple regression analysis, I am curious if you added base deficit; I suspect that that would be a highly powerful predictor of the need for urgent thoracotomy.
James J. Tyburski, MD, Detroit, Mich: I also want to congratulate the authors for trying to answer a question that every medical student or first-year surgical resident asks, which is, when should we take these patients to the operating room? I echo Dr Rodriguez' comments about the interval time of the resuscitation, particularly that even on your multivariate analysis the SBP was a significant factor. I would like the authors to comment about the management of that and for how many of these patients did they go to the operating room for hypotension? Also, when they got to the operating room, what was done? Were these primarily vascular repairs? Were they lung resections? What was done? And, finally, kind of a philosophical question: if a patient came in and had 1400 cc of output for the first 22 hours and then put 200 cc in the last 2, would that be a patient who you would actually take for a thoracotomy?
H. Gil Cryer, MD, Los Angeles, Calif: It appears that the patients who died from hemorrhage died in the operating room, probably early, and from the presentation it appeared that there was not a statistical correlation with the delay to the operating room and mortality. I would just like to specifically ask, was the mortality occurring in any of the delayed patients? Finally I would just like to reiterate what Dr Moore said. I think that the main weakness here before you can make any real recommendations is that you don't really have the denominator in this study. We need to know how many patients were managed without operation who had greater than 1500 cc of blood loss.
Dr Jurkovich: As Dr Karmy-Jones did mention, this study was stimulated by the Western Trauma Association's desire to look at the indications, timing, and outcome of thoracotomy following traumatic injury. One subset of that population reflecting this question, as Dr Tyburski noted, was, what are the indications for ongoing bleeding?
Dr Rodriguez, I thank you very much for your kind review of the manuscript. You had 3 questions. The cause of the mortality was primarily determined by the senior reviewer at each institution. As Dr Moore mentioned, this is a subjective analysis, and that bias is evident throughout. That is something that we don't deny but do acknowledge. So the senior reviewer at each institution assigned the reason for the mortality in their best judgment based on either firsthand knowledge of the case, since these were fairly contemporary patients, or their review of the medical records.
The second question related to the mean SBP dropping from 92 to 79 mm Hg from the ER to the operating room, and in fact that is the case. Our argument is that in these patients, the team simply waited too long to get these patients to the operating room to control surgical bleeding. Notably, however, this group of patients fell into what would be considered "responders to resuscitation." That is, they weren't persistently hypotensive, but they were intermittently hypotensive, responding to fluid resuscitation boluses, and because of that response there was delay by trying to manage them nonoperatively.
The third question relates to a number of comments regarding plans for a prospective study. As Dr Moore noted, selection bias is a significant concern in any retrospective study, particularly missing the denominator of all patients who had chest tube outflow of greater than 1500 cc in this study. This is something that is going to be raised in the winter meeting of the Association, and I really don't know the answer at this point.
Dr Moore also commented upon the role of angiography for embolizations of intercostal bleeding, correctly noting that in fact 67% of the findings at blunt thoracotomy were chest wall or intercostal bleeding. I can't give you the exact number, but notably there were a number of attempts at arterial embolization in the blunt chest trauma patients which failed, and those invasive and diagnostic tests, as well as the distracting injuries associated with blunt trauma, are the cause of the 3- to 4-fold increased time delay in blunt trauma patients observed in this study.
What was done in the operating room was a question raised by Dr James Tyburski. What was done was control of hemorrhage in each of the cases. To clarify one of Dr Tyburski's concerns, the presentation of hypotension was the presence of hypotension at start of the operation, not in the emergency room.
And finally the question that Dr Gil Cryer raised concerned the cause of the late mortality. One half of the deaths were acute mortalities related to ongoing thoracic hemorrhage. The remainder of the deaths were late mortalities and primarily due to the complication of multiple organ failure, most notably ADRS [adult respiratory distress syndrome]. The relationship between ARDS and ongoing blood loss and delay to the operating room was nearly significant at the P≥.10 level, suggesting that delays to the operating room and ongoing blood loss were associated with the late occurrence of ARDS and contributed to the late mortality.