Time to failure of late (<48 hours) failures of nonoperatively managed patients.
McIntyre LK, Schiff M, Jurkovich GJ. Failure of Nonoperative Management of Splenic InjuriesCauses and Consequences. Arch Surg. 2005;140(6):563-569. doi:10.1001/archsurg.140.6.563
Presenting patient characteristics can predict which patients will fail nonoperative therapy of blunt splenic injuries.
Retrospective descriptive population study.
All patients admitted with blunt splenic trauma were identified from a statewide trauma registry between January 1, 1995, and December 31, 2001.
Patients and Methods
Patients were characterized as requiring immediate intervention or successful or failed nonoperative management based on time from emergency department arrival to intervention (surgery or angioembolectomy). Injury and patient characteristics included age, the presence of hypotension, Injury Severity Score, and the Glasgow Coma Scale score. Risk factors for the failure of nonoperative management were evaluated using the χ2 statistic. The failure of nonoperative management associated with the admitting hospital’s trauma designation level was evaluated using logistic regression.
Main Outcome Measures
Determine factors associated with failure of nonoperative management of blunt splenic injuries.
Two thousand two hundred forty-three patients met criteria for inclusion in the study. Six hundred ten patients (27%) underwent immediate splenectomy, splenorrhaphy, or splenic embolization (within 4 hours). Of the remaining 1633 patients who were admitted with planned nonoperative management, 252 patients (15%) failed. Of the injury and patient characteristics reviewed, being older than 55 years and having an ISS higher than 25 were significantly associated with failure. Risk of failure also increased with admission to a level III or IV trauma hospital compared with a level I trauma hospital.
Being older than 55 years and having an ISS higher than 25 along with admission to a level III or IV trauma hospital were associated with a significant risk of failure of nonoperative management of splenic injuries. The Glasgow Coma Scale score, associated injuries, and presenting hemodynamics were not predictive of failure in this large retrospective review.
For more than 2 decades, the nonoperative management of splenic injuries has been a topic of considerable debate. The benefits of splenic conservation are well recognized and include the elimination of the risk of overwhelming postsplenectomy sepsis as well as avoiding potentially unnecessary surgery and the complications of a laparotomy. Failure of nonoperative management, however, must be recognized promptly to provide expeditious management of ongoing, life-threatening hemorrhage. Therefore, patients managed nonoperatively are felt to be best served by being under the care of a surgical team capable of rapid operative intervention.
Despite the development of a clear role for this management strategy in selected patients, no clear risk factors have been identified to stratify patients based on their likelihood for failure. Computed tomographic scan grading of splenic injuries was initially thought to be a useful tool in stratifying high- and low-risk patients, but it has subsequently been shown to have only a modest ability to predict the likelihood of persistent or recurrent bleeding.1,2 Moreover, there is evidence suggesting that intrarater and interrater reliability in grading splenic injuries is poor.3 Further, the time to failure has been poorly defined, resulting in variability in the intensity and duration of in-hospital observation. The goal of our study was to evaluate factors associated with failure of nonoperative management.
All patients admitted to a Washington State trauma hospital, Seattle, with splenic injury via blunt trauma during the period of January 1, 1995, to December 31, 2001, were identified from our statewide trauma registry. Splenic injury was identified using the International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code 865. The Washington State Trauma Registry enrollment criteria include patients who died at a trauma hospital, all pediatric patients admitted to a trauma hospital, and all adult patients (>14 years) who were admitted to a trauma hospital with a length of stay of more than 48 hours.
To capture patients who might have been discharged from one hospital with a diagnosis of splenic injury but then were readmitted to another hospital for eventual splenectomy, we linked the patients identified in the trauma registry with the Washington State Comprehensive Hospital Abstract Reporting System (CHARS), the statewide hospitalization database. First, we linked patients identified in the trauma registry to the CHARS database for the 30 days prior to the trauma registry admission. This linkage was performed to see if any patient had had a previous admission at any hospital with the diagnosis of blunt trauma, whether or not a splenic injury was identified. Second, we linked the patients identified in the trauma registry with the CHARS database for the 30 days after the trauma registry discharge to identify if any of these patients were subsequently readmitted elsewhere for splenectomy (or splenic embolization), if a splenic procedure was not performed at the initial admission.
Patients were then classified according to the type and timing of the outcome of their splenic injury. One group of patients required immediate intervention, which was defined as splenectomy, splenorrhaphy, or embolization within 4 hours of admission from the emergency department (ED). Another group was admitted with a presumption of planned nonoperative management. Planned nonoperative management was presumed if a patient did not have a splenectomy within 4 hours of presentation to the ED. Nonoperatively managed patients were then stratified into those who failed nonoperative management (required splenectomy, splenorrhaphy, or embolization after 4 hours following admission from the ED) and those who had successful nonoperative management (no splenic procedure during initial or subsequent admissions). To evaluate specific characteristics of patients with splenic injuries by time from ED presentation to operative intervention, patients who failed nonoperative management were categorized by the time from admission from the ED to the splenic intervention: more than 4 to 8 hours, more than 8 to 24 hours, more than 24 to 48 hours, and more than 48 hours. Patients were excluded from the analysis if the time from ED admission to intervention was not known.
Patient and injury characteristics that we evaluated included age, hemodynamics (heart rate and systolic blood pressure) on presentation in the ED, Injury Severity Score (ISS), and Glasgow Coma Scale (GCS) score. Patients were also analyzed for the admitting hospital’s trauma designation level.
Initial analysis compared the immediate intervention patients, the successfully nonoperatively managed patients, and the patients who failed nonoperative management, categorized by time from ED presentation to splenic intervention. To evaluate patient and injury factors associated with failure of nonoperative management, patients who failed nonoperative management were grouped together, regardless of time from ED presentation to splenic intervention, and compared with patients who were successfully managed nonoperatively. All patient groups were compared using the χ2 statistics.
The effect of the admitting trauma hospital’s designation (level I through V) on the risk of failed nonoperative management was assessed using logistic regression. The outcome for the regression model was success or failure of nonoperative management. The risk factor of interest was the admitting hospital’s designation, with the level I trauma center being the reference category. Odd ratios (ORs) and 95% confidence intervals (CIs) were calculated, adjusting for confounders of age and ISS.
The Washington State institutional review board approved this study.
A total of 2303 patients were found to have splenic injury from blunt trauma during the designated period, of which, 2243 had complete data for analysis. Six hundred ten patients (27%) underwent immediate splenectomy or embolization. Of the remaining 1633 patients who were admitted with planned nonoperative management of their splenic injuries, 252 patients (15%) ultimately required operative or angiographic intervention. Eighteen patients failed following discharge from the index hospitalization.
Table 1 lists the characteristics of all patients, stratified into those who were successfully managed nonoperatively, those who required immediate intervention, and those who failed nonoperative management, based on time from presentation in the ED to the operating room or angiography suite. There was a significant relationship among age, GCS score, and ISS and the timing of operative (or nonoperative) management. For example, within the 3 age strata, younger patients were more likely to be managed nonoperatively. Table 2 is a similar analysis, yet all the nonoperative columns are combined, and in doing so, the GCS score is no longer statistically significantly different.
Of the 252 patients who failed nonoperative management, 100 (40%) failed between 4 and 8 hours of admission. Arguably, this cohort could represent patients who were actually admitted with a plan for “immediate” operative intervention but were delayed owing to logistical problems (eg, operating room or surgeon availability). The remainder of the patients who failed nonoperative management and underwent intervention more than 8 hours from admission represented 152 (9.3%) of all 1633 patients admitted with an initial nonoperative plan and 152 (6.8%) of all 2243 patients admitted having a diagnosis of a splenic injury.
When we then compared patients who were successfully managed nonoperatively with those who failed nonoperative management (therefore, excluding patients who underwent immediate splenectomy), patients who failed nonoperative management were more likely to be older than 55 years or to have an ISS higher than 25 (Table 2). Note that in this table, and in subsequent analyses based on these data, vital sign results are present only for adult patients, as we recognize the variability in defining hypotension or tachycardia in pediatric patients.
We investigated whether the success or failure of nonoperative management of splenic injury differed depending on the level of trauma hospital in which they were admitted. All facilities with designated trauma center levels have strict standards including the organization and direction of a multidisciplinary trauma committee that determine a center’s basic resources and capabilities. While these standards are specific to Washington State, they largely follow the criteria of the American College of Surgeons’ Resources for Optimal Care of the Injured Patient.4 A critical component of designation level is the availability of emergency physicians, general surgeons, radiologists, operating rooms, and surgical subspecialists. Level I through III trauma centers have ED physicians available within 5 minutes of a trauma patient’s arrival and can have an operating room available within 5 minutes of notification. In addition, level I centers have a general surgeon (defined as a surgeon at least at the level of a postgraduate year 4 surgical resident) in house and available within 5 minutes of trauma team activation. A general surgeon is available within 20 minutes of trauma team activation at a level II trauma center and within 30 minutes at a level III trauma center. The primary function of level IV and V trauma centers is to stabilize a trauma patient for transfer to a higher-level center for definitive care.
Table 3 summarizes the number of patients admitted with splenic injury according to the admitting hospital’s level of trauma designation. The highest percentage of patients (38.3%) was admitted to Washington State’s only level I trauma center. This group also had the highest mortality rates, implying that the most severely injured patients were appropriately triaged there. Eighty-five patients (3.7%) were admitted to a nontrauma designated hospital.
The admitting hospital’s trauma designation level was associated with risk of splenectomy. Table 4 lists the unadjusted outcomes stratified by the admission hospital’s level of trauma designation. Level II trauma centers were not more likely to perform splenectomy (immediate or delayed) compared with the level I trauma center (adjusted OR, 1.2; 95% CI, 0.9-1.5) after adjusting for age and ISS. Patients in level III (adjusted OR, 2.0; 95% CI, 1.5-2.6) and level IV (adjusted OR, 1.9; 95% CI, 1.2-2.9) trauma centers were significantly more likely to have a splenectomy compared with patients in the level I trauma center.
The admitting hospital’s trauma designation level was also associated with the risk of failure of nonoperative management of splenic injuries. Level II trauma centers were no more likely to have a failure of nonoperative management (adjusted OR, 1.0; 95% CI, 0.7-1.4) compared with the level I trauma center after adjusting for age and ISS. However, level III (adjusted OR, 2.1; 95% CI,1.4-3.1) and level IV (adjusted OR, 2.5; 95% CI, 1.5-4.1) trauma centers were more likely to have a failure of nonoperative management compared with the level I trauma center.
There were 62 patients who failed nonoperative management of their splenic injury more than 48 hours after their admission. The Figure shows the time to failure of these patients. They were evenly distributed among the different hospitals. Most patients (73%) failed within 2 to 7 days of their injury. However, 27% failed beyond the first week of their injury (range, 7-21 days), the latest failure occurring about 3 weeks after the injury; 18 of these patients failed after index hospital discharge.
The nonoperative management of splenic injuries grew out of a desire to “save” the spleen to prevent overwhelming postsplenectomy sepsis5,6 and from the initial reports from pediatric hospitals of nearly exclusive nonoperative management of splenic injuries in children.7 The push to adopt a nonoperative strategy in adults was slower to evolve, in part because of the risk of blood transfusions8 but also because of the remarkable success of rapid operative intervention and splenectomy in decreasing mortality of blunt abdominal trauma. Numerous studies on preventable mortality in the 1970s and 1980s repeatedly demonstrated unrecognized splenic bleeding as the primary cause of preventable mortality, and general surgeons developed great facility in performing rapid splenectomy, often without the need for a blood transfusion.9,10
Nevertheless, the 1990s saw widespread adoption of a nonoperative management strategy for all abdominal solid organs, namely, the spleen, liver, and kidney. The initial concerns with nonoperative management centered on defining indications for abandoning nonoperative therapy in favor of an intervention (usually splenectomy, occasionally splenorrhaphy or angioembolization), as well as concerns about missed associated injuries and blood transfusion risks.11- 14 However, with widespread adoption of nonoperative management of splenic trauma in adults, it became clear via isolated case reports that delayed ruptures or rebleedings presented a real and potentially disastrous complication that could not be reliably predicted.15- 19
In this article we have used a statewide population database (CHARS) to examine the risk of failure of nonoperative management of splenic injuries. This study covers 7 consecutive years in 1 state (Washington) and takes advantage of both a state-mandated uniform hospital discharge diagnosis database (CHARS) as well as a state-mandated trauma registry (Collector) in each of the state’s 77 designated trauma centers. Merging these administrative and patient-specific databases resulted in the ability to identify an index hospitalization for splenic intervention (splenectomy or embolectomy) following injury (E-code identified) and to look backward and forward 30 days from this index admission for associated admissions and interventions. To our knowledge, the result is the largest population-based study of nonoperative management of splenic injuries, and importantly, it sheds some additional information on the likelihood of late failure of this management strategy.
Most notably, 610 (27%) of 2303 patients underwent immediate splenectomy, splenorrhaphy, or embolization, 1381 patients (62%) had successful nonoperative management, and 252 patients (11%) failed nonoperative management. This is consistent with other smaller reports, primarily from high-level trauma centers, with overall success rates of nonoperative management ranging from 50% to 70%.20- 22 Another way these data can be considered is that of the 1633 patients in whom nonoperative management was initially planned, 1381 patients (85%) succeeded and 252 patients (15%) failed. Of the 252 patients who failed nonoperative management, 62 (25%) failed longer than 48 hours after ED presentation, with 31 (50%) of 62 of these patients failing later than 5 days from injury, and 18 failing after index hospital discharge for an overall postdischarge late failure rate of 1.1%. Fortunately, none of these very late failures died, an important point considering that the risk of postdischarge death from delayed rupture is often used to condemn “early” discharge or nonoperative management in general, and it has been the source of legal challenges to care. Of the 62 patients failing nonoperative management greater than 48 hours after admission (Figure), 50% failed between days 3 and 5, and 72.5% failed within the first week. This is consistent with the recently concluded prospective multi-institutional trial of the American Association for the Surgery of Trauma that followed up 300 patients with splenic injury via a structured practice guideline that included strict bed rest for 5 to 7 days.23 In this study, 6% of the patients failed nonoperative management, but in only 3.3% was the cause ongoing bleeding from the spleen, which prompted laparotomy and splenectomy between days 4 and 5 after injury.
A remarkable 96.4% of these patients in our study were cared for at a designated trauma center. This is an example of an inclusive trauma system at work. Washington State has a total of 95 community hospitals, and 14 other hospitals including private specialty hospitals, state or military hospitals, and Veterans Affairs hospitals.24 Seventy-seven of these are designated trauma centers, of which 28 are level I, II, or III centers. Washington State is unique in its inclusion of level IV (n = 34) and level V (n = 15) centers in its trauma network. Ninety percent of the splenic injuries were managed in level I and III trauma centers, and all of the deaths occurred in these hospitals, and the vast majority of the deaths (by implication the most seriously injured) occurred in the level I and II trauma centers.
These data also suggest that the level III and IV trauma centers are less likely to effectively manage nonoperatively a patient with a splenic injury. After adjusting for age and ISS, the 2 factors we found that significantly correlated with failure of nonoperative management, the level III and IV trauma centers were still twice as likely to perform splenectomy overall and twice as likely to fail attempts at nonoperative management. This most likely represents a relative uncomfortableness with this management strategy, or perhaps a lack of effective resources for close observation of these patients, although we have not explored other reasons for this discrepancy such as age or training of surgeons, splenic injury pattern, or financial implications.
The study has some significant limitations. As a population-based study, it has limited access to individual medical records. The institutional review board restrictions have thus far precluded us from investigating in detail the records of individual patients outside of our own institution. This has limited our ability to define the exact injury pattern of the spleen, the exact findings at laparotomy, and the cause of death. The decision to define nonoperative management as any patient who did not get a splenic intervention (angioembolectomy or laparotomy) within 4 hours of presentation to the index hospital was arbitrary. It could be argued that those patients who had a laparotomy between 4 and 8 hours of ED admission were not intended to be nonoperatively managed, but they were simply slow to get to an operation. However, as given in Table 1, this cohort of 100 patients is more similar to those patients successfully managed nonoperatively than it is to those patients managed with urgent splenectomy. We elected to include pediatric patients in this study in an effort to most completely represent a population study. However, children may be more aggressively nonoperatively managed, and the vital signs of children were not included in the analysis of factors influencing the success or failure of nonoperative management. And finally, we do not present computed tomographic scan or operative assessment of injury severity grading of splenic injury, as is commonly reported. Although this information would most likely mimic numerous other reports showing an association between the grade of splenic injury and performance of splenectomy, other factors, most notably patient physiology, influence the decision to attempt nonoperative management regardless of splenic ISS. Nevertheless, computed tomographic grading of splenic injury severity may be predictive of late failure, but we do not have that data.
This study demonstrates that nonoperative management of splenic injury was successful in 62% of all splenic injury patients in Washington State over a recent 7-year time frame, with 27% of patients having an immediate splenectomy and 11% a delayed splenic intervention representing a failure of nonoperative management. These numbers represent a true population-based reference for comparison. Age older than 55 years and a high ISS were predictive of failure of nonoperative management and arguably warrant 5 to 7 days of in-hospital observation for older, more severely injured patients in whom nonoperative management is attempted. Of those patients in whom nonoperative management failed, 75% fail within 2 days of injury, 88% within 5 days, and 93% within 1 week of injury. While 1.1% of patients failed after hospital discharge, none died, suggesting that signs and symptoms of ongoing or renewed bleeding can be recognized and prompt successful delayed interventions. Clear, understandable, and direct patient discharge information regarding this possibility is essential. Nonoperative management is more effectively practiced in higher-level trauma centers. This study does not provide an explanation for this observation, and it should not imply that all patients with spleen injuries must be managed at level I or II trauma centers. The very low mortality among spleen injury patients in level III through V trauma centers suggests that statewide resources are effectively being used. This article will help define target benchmark goals for nonoperative management and guidelines for in-hospital observation.
Correspondence: Lisa K. McIntyre, MD, Department of Surgery, Harborview Medical Center, 325 Ninth Ave, Box 359796, Seattle, WA 98104 (firstname.lastname@example.org).
Accepted for Publication: January 31, 2005.
Previous Presentation: This paper was presented at the 112th Scientific Session of the Western Surgical Association; November 8, 2004; Las Vegas, Nev; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.
John Weigelt, MD, Milwaukee, Wis: McIntyre et al try to determine if we can predict which patients will fail nonoperative management of a splenic injury. They used large clinical and administrative databases to identify patients with splenic injuries in the state of Washington. A definition of failure for nonoperative management is chosen, and the data analyzed. A failure rate of 15% is found and associated with an age older than 55 years, severe injury by ISS, and hospital resources dedicated to injury care. Their analysis appears correct, but is it of value for the practicing surgeon? I have many questions; I will limit them to a few.
Why did you choose the 4-hour interval to define the immediate splenectomy group? You admit in the article that the 4- to 8-hour group might simply be mobilization of resources for any given patient. If these patients are excluded as failures, then your failure rate is 10%. Which number would you like us to take home?
Your database only has 186 patients older than 55 years. Do you believe this number is adequate to state your conclusion as a clinical tenet?
What about patients with only a splenic injury? Would this population be a better denominator to tease out risk factors for failed nonoperative management?
Finally, can we look at your data another way? Using blood pressure as a marker for hemodynamic normality, only 5% of your entire patient population had a systolic blood pressure less than 90 mm Hg. What do you really think the indication for splenectomy was in this population, and would an appropriate clinical protocol help all of the hospitals regardless of resources dedicated to injury?
One last question: If—I hope it does not happen—one of our distinguished senior members of the Western Surgical Society has a splenic injury while traveling through Seattle, will you use this data to remove their spleen immediately?
I think this is a provocative paper. I thank the authors for sharing it with us and look forward to their answers.
Dr McIntyre: Thank you for your comments, Dr Weigelt. Regarding your questions, the first one being our arbitrary cut-off being 4 hours as the period that we use to define immediate splenectomy. This was an arbitrary cut-off; it seemed like a reasonable period for which patients who do have abnormal physiology as the main reason for immediate surgery to undergo splenectomy. When we looked at the various patient characteristics in these patients who underwent surgery between 4 and 8 hours from their ED presentation, we did find that their presenting hemodynamics and other characteristics were much more similar to those patients who were successfully managed nonoperatively than those who underwent immediate splenectomy. Our database actually has a total of about 260 patients that were older than 55 years. True, 186 patients were those who were managed with a nonoperative plan, but this just demonstrates that a higher percentage of patients who are in this age category are more likely to undergo immediate splenectomy. In terms of trying to analyze only patients with an isolated splenic injury, it may be an interesting cohort of patients to examine, but we felt that by including all patients with splenic injuries whether or not they had other injuries, this was a more representative population of patients that we see. In terms of the 27% of patients who underwent splenectomy as their initial treatment, in our article it is noted that the systolic blood pressure of 90 mm Hg is actually only their presenting blood pressure. So, we do not know how many of these patients also had systolic pressure less than 90 mm Hg recorded at anytime during either their emergency room course or after admission to the hospital. Also, we did not have access to other variables such as what the fluid requirements were in these patients to keep their systolic blood pressure normal, and we did not have any information regarding hematocrits or blood transfusion totals, all of which, of course, influence management. Finally, I would say that if any of our senior members did have the misfortune of having a splenic injury in our state, I would say that again, the most important consideration in deciding whether a patient should undergo immediate splenectomy is really based on the patient’s physiology at the time, not just a single point in time, but their course after presentation. It just needs to be known that those patients older than 55 years and those with a high ISS will have a greater chance of failure of nonoperative management, if this is chosen, and these patients need to be watched more carefully.
Juan A. Asensio, MD, Los Angeles, Calif: I would like to congratulate you on a very nicely presented paper that seeks to shed further light as to how we can refine criteria for nonoperative management of adult patients with splenic injuries. Now, I presume that all of these patients were actually adult patients, although from your presentation this was not clear. I have some questions for you. How many of your patients were adult and how many were pediatric patients? Of your 252 patients who failed nonoperative management, do you know how many of them were subjected to angiography and angioembolization? I think what is most important—and I did not see this in your presentation—is that there were no AAST [American Association for the Surgery of Trauma] organ injury scale injury grades for your splenic injuries. I think that for patients who sustain grade IV and grade V injuries this is also a well-known risk factor for failure of nonoperative management. So, I would like to know if you had any data that correlated AAST organ injury scale grades for splenic injuries with operative management. And finally, data has been shown that dates way back to the 1990s about associated injuries in adult patients; these range between 11% and about 15%. So, I would like to know, of those who failed nonoperative management, how many of them actually had other associated nonsplenic injuries that would have mandated surgical intervention had they been discovered.
Dr McIntyre: I would like to clarify that our study did include pediatric patients. One of our comparison age groups was between 0 and 14 years. Of the 252 patients who failed, we had few, I believe fewer than 5 patients who underwent angioembolization; most patients underwent splenectomy when they failed. We did not have the injury severity of the spleen available to us from our trauma registry, and we are hoping to pursue this information with a second institutional review board to do a more thorough medical record review to try to gather that sort of information. Likewise, we do not have information as to the reason for laparotomy in the cohort that failed, whether other associated injuries were found at laparotomy or whether the splenic injury itself was impetus for the operation.
Thomas H. Cogbill, MD, LaCrosse, Wis: I congratulate the authors on bringing a large statewide experience to us, but that large statewide experience does also present some methodology problems, and I wonder if you would comment on this. Basing your enrollment criteria on discharge diagnoses, you stated that patients who did not have an immediate splenectomy must have had “presumed” nonoperative management. But this does not consider when the diagnosis of splenic injury was made during the course of the hospitalization. How many patients had a delayed diagnosis of their splenic injury and, therefore, were never initially managed by interventional nonoperative techniques, but instead required more precipitous operative techniques at the time the splenic injuries were actually discovered?
Dr McIntyre: In terms of the patients who were readmitted and underwent splenectomy, it is true that it is unclear whether at the previous admission a splenic injury was even diagnosed. This certainly could actually represent patients who had a delayed diagnosis rather than a failure of planned nonoperative management. We did not have that data available to us in the trauma registry.
Thomas S. Helling, MD, Kansas City, Mo: That was a very nice study. You mentioned that the risk of splenectomy in level III and level IV trauma centers was twice as high as level II and level I trauma centers. Do you suspect that some of those patients had been transferred out to higher levels of care and, therefore, the denominator was lower in your calculation?
Dr McIntyre: We made it a point to ensure that the patients who were readmitted to other hospitals did not represent transfers from other institutions. None of the patients who were in the level III or level IV trauma centers were subsequently transferred to higher levels.