Grade of Injury, not Initial Management, Is Associated with Unplanned Interventions in Liver Injury

Abstract

Introduction:

Risk factors for complications after liver injury do not distinguish between patients undergoing selective non-operative management (sNOM) vs operative management (OM) as the initial treatment strategy. Our objective was to identify risk factors for complications requiring an unplanned intervention following sNOM or OM. We hypothesized that patient undergoing sNOM will have fewer unplanned interventions.

Methods:

Adults presenting to a level I trauma center with grade III or higher liver injury over a period of 6 years were reviewed. Patient and injury factors, initial management strategy, subsequent complications and interventions were obtained. Bivariate analysis was performed between patients undergoing sNOM vs OM to determine factors associated with unplanned interventions, defined as intervention >48 hours after injury. Logistic regression was performed to identify independent risk factors for unplanned interventions.

Results:

One-hundred ninety one patients were identified: 105 (55%) grade III, 64 (34%) grade IV, and 22 (12%) grade V injury; 136 (71%) underwent sNOM and 55 (29%) underwent OM. Twenty-nine (15%) patients required an unplanned intervention: 26 percutaneous drainage, 10 ERCP, and 3 angiography; 12 had multiple procedures. Male gender, younger age, higher ISS, higher grade of injury, firearm mechanism, and initial OM (all p<0.05) were associated with unplanned interventions. Firearm mechanism and injury grade IV and V, but not initial OM, were independent risk factors for an unplanned intervention.

Conclusions:

Grade of liver injury, not the initial mode of treatment, was significantly associated with requiring an unplanned intervention for liver-related complications. Surveillance at 7–10 days, or prior to discharge, in the high-risk group may be able to capture those requiring unplanned intervention and readmission.

Introduction

Liver is the most frequently injured organ in the abdomen after a traumatic injury,¹ and 5% of all patients admitted after a traumatic injury have some degree of liver involvement.² Over the past 40 years, mortality rates associated with liver injury have steadily decreased for both penetrating and blunt injuries, which is attributable in large part to advances in selective non-operative management (sNOM) and improvements in critical care.³ Currently, sNOM is the preferred treatment for hemodynamically stable patients with liver injury, regardless of injury grade.⁴ As a result, studies looking at complications after liver injury have largely focused on outcomes for patients who underwent sNOM. Complications after sNOM increase with the grade of injury, with grade III injuries having a complication rate of 1%, grade IV with 21%, and grade V with 63%.⁴ These complications include bile leak, biloma, abscess, bleeding, hepatic necrosis, or liver failure that often requires unplanned interventions such as percutaneous drainage, ERCP, angioembolization, and/or surgery.⁵

Despite the advances in sNOM a subset of patients who have severe liver injuries present with unstable vital signs or have injuries not amenable to sNOM. These patients undergo operative management (OM) as their initial treatment strategy. However, modern risk factors for complications after liver injury includes “all comers” and does not distinguish between patients receiving sNOM vs OM. Specifically, it remains unknown whether an initial presentation with hemodynamic instability that necessitates an immediate laparotomy is indicative of a higher risk for complications, compared to a patient with a similar grade injury who undergoes sNOM. In addition, it remains unknown whether there is a population in which surveillance for a complication such as an arteriovenous malformation or a biloma would be beneficial. Current recommendations suggest that imaging should be performed when “clinically indicated” but do not address particular high-risk groups or timing of imaging.⁴

The primary objective of our study was to identify risk factors for complications that required an unplanned intervention following initial sNOM or OM. We hypothesized that patients who underwent an initial sNOM, would have fewer unplanned interventions, compared to patients who underwent an initial OM. Our secondary objective was to characterize the timing of unplanned interventions for severe complications and identify an appropriate temporal window for surveillance imaging.

Methods

Patient Population

We performed a retrospective review of prospectively collected data at an urban, academic Level I trauma center over a 6-year period from January of 2013 to December of 2018. We included consecutive patients age ≥ 18 years, who sustained grade III or higher liver injury. Patients who died within 24 hours of admission were excluded from analysis related to complications and delayed interventions, as these patients would not survive long enough to develop complications requiring a delayed intervention.

Data Collection

We utilized the trauma registry to obtain demographic factors such as age, gender, race, mechanism of injury, Glasgow Coma Scale (GCS) score upon arrival, and Injury Severity Score (ISS). Initial management for liver injury and subsequent complications and interventions were obtained from retrospective review of electronic medical records. Records were reviewed up to 90 days after admission to ensure that liver-related complications that occurred after discharge were captured.

Liver injury grade was obtained from the trauma registry and verified by chart review. Grade of liver injury was defined using the organ injury scale proposed by the American Association for Surgery of Trauma (AAST), and was based on intra-operative and/or abdominal CT findings.^6,7 For patients who had initial imaging performed prior to transfer to our facility, grade of injury was determined using the abdominal CT scan performed at the transferring facility.

Patients were categorized into sNOM or OM based on the initial management strategy. Selective NOM was defined as observation (with serial abdominal exams and laboratory evaluations), or angiogram with or without embolization. Patients who had an immediate laparotomy with or without angiogram were categorized in the OM group.

Outcomes

Primary outcome was unplanned interventions, defined as liver-related interventions 48 hours or more after the liver injury. For patients who underwent damage control operations and were left with an open abdomen, subsequent planned explorations, such as for the purposes of packing removal or planned washouts, were not considered to be unplanned. However, once a patient’s abdomen was definitively closed, any subsequent operation or procedure performed for a liver-related complication was considered unplanned. Operative notes for all subsequent laparotomies were carefully examined for the indication and all interventions performed at the operation. If an operation was performed for reason that was distinctly remote from the liver injury, it was not counted as a “liver-related” event. Any laparotomy for which the indication was unclear, such as a general washout for an abscess of unclear origin, was categorized as “liver-related.” Unplanned interventions that were collected included percutaneous drain placement, ERCP, angioembolization and/or liver-related laparotomies.

Secondary outcomes were hospital length of stay (LOS), intensive care unit (ICU) LOS, ventilator days, discharge disposition, inpatient death, and readmissions related to liver injury within 30- and 90-days after discharge.

Data Analysis

Data are presented as median (interquartile range) for continuous variables and n (percent) for categorical variables. Descriptive analysis was performed for the entire cohort of patients by liver injury grade. Bivariate analysis was performed between patients with and without unplanned interventions to determine factors associated with these interventions. Categorical variables were compared using the chi-square test and continuous variables were compared using the Kruskal-Wallis one-way analysis of variance test.

Logistic regression analysis was performed to determine independent risk factors associated with unplanned interventions. To create our logistic regression model, we utilized variables that were significant (p<0.05) from the bivariate analysis. For collinear variables, the variable that was had a stronger association with the outcome and was deemed to be more clinically relevant to our research questions were preferentially used in the model. For example, firearm injury was used instead of penetrating mechanism of injury since it was more strongly associated with an unplanned intervention in bivariate analysis. Also, liver injury grade was utilized in the model. Firearm injury and male gender were also highly collinear, so male gender was not included in the model. All data analysis was performed using STATA SE version 16.0 (College Station, TX). This study was performed under a protocol approved by the local Institutional Review Board.

Results

A total of 252 patients with grade III or higher liver injury were identified between 2013 and 2018. Sixty-one patients (24.2%) died within 24 hours of admission. Among the patients who died, eight (13.1%) had grade III injuries, 30 (49.2%) had grade IV, 20 (32.8%) had grade V, and 3 (4.9%) had grade VI liver injury. Thirty (49.2%) of the patients who died within 24 hours arrived to the ED in cardiac arrest and were unable to be resuscitated. Twenty-five (41.0%) other patients were taken to the OR directly from the ED and died in the operative theater. All patients who died within the first 24 hours of admission were excluded from subsequent analysis as they would not have been able to manifest a complication. Of the remaining 191 patients, 120 (62.8%) were male and median age was 30 (24–41). Seventy-three (38.2%) patients had a penetrating mechanism of injury, of which 57 (78.1%) were due to firearms. Initial treatment strategy was sNOM in 136 (71.2%) patients, with observation only for 109 (80.1%) patients, and angiography with or without embolization for 27 (19.9%) patients. OM was performed in 55 (28.7%) patients, which included liver cauterization, packing, hepatorrhaphy, or resection. Within the OM cohort, 11 (20.0%) patients also received angiography with or without embolization as an adjunct to the operation.

One hundred five (55.0%) patients had grade III, 64 (33.5%) had grade IV, and 22 (11.5%) had grade V liver injury. Penetrating mechanism of injury was more common in higher severity of liver injury, although this was not significant (33.3% of grade III vs 39.1% of grade IV vs 59.1% of grade, p=0.077). Patients with higher severity of liver injury were more likely to receive OM (23.8% of grade III vs 26.5% of grade IV vs 59.0% of grade V, p<0.001). (Table 1)

Table 1 –

Demographics for Patient Population by Injury Grade

	Grade III n=105 (55.0%)	Grade IV n=64 (33.5%)	Grade V n=22 (11.5%)	p
Male Gender	64 (61.0%)	41 (64.1%)	15 (68.2%)	0.791
Median age (IQR), years	31 (25–40)	30 (23–45)	25 (20–43)	0.428
Penetrating Mechanism	35 (33.3%)	25 (39.1%)	13 (59.1%)	0.077
Mode of Injury				0.186
Firearm	24 (22.9%)	23 (35.9%)	10 (45.5%)
MVA	59 (56.2%)	35 (54.7%)	8 (36.4%)
Fall	7 (6.7%)	1 (1.6%)	0 (0.0%)
Stabbing	10 (9.5%)	2 (3.1%)	3 (13.6%)
Physical Assault	3 (2.9%)	2 (3.1%)	0 (0.0%)
Other	2 (1.9%)	1 (1.6%)	1 (4.5%)
Median ISS (IQR)	19 (14–22)	29 (24–34)	42 (35–43)	<0.001
Median GCS (IQR)	15 (14–15)	15 (13–15)	15 (6–15)	0.238
Initial Management				<0.001
sNOM	80 (76.2%)	47 (73.4%)	9 (40.9%)
OM	25 (23.8%)	17 (26.6%)	13 (59.1%)

IQR – Interquartile Range; MVA – Motor Vehicle Accident; ISS – Injury Severity Scale; GCS – Glasgow Coma Scale; sNOM – Selective Non-Operative Management; OM – Operative Management

Twenty-nine (15.2%) patients received unplanned interventions after initial treatment for liver injury. Three (10.3%) patients received angiography, 26 (89.7%) patients received percutaneous drain placement, and 10 (34.5%) patients received ERCP; 12 (41.4%) patients received multiple interventions. The median number of days to unplanned interventions were as follows: angiography 10 (2–25) days, percutaneous drain placement 15 (9–24) days, ERCP 16 (11–22) days, and operative exploration 13 (8–25) days. Overall, the median number of days to any unplanned intervention was 11 (9–17) days. Bivariate analysis showed that male gender, age, ISS, mechanism of injury and particularly firearm injury, grade of liver injury, and initial management strategy were all associated with an unplanned intervention following initial management. (Table 2)

Table 2 –

Bivariate Analysis of Patients Requiring a Delayed Procedure

	No Delayed Intervention n=162 (84.8%)	Delayed Intervention n=29 (15.2%)	p
Male Gender	96 (59.3%)	24 (82.8%)	0.016
Median age (IQR), years	32 (24–43)	25 (23–34)	0.036
Penetrating Mechanism	54 (33.3%)	19 (65.5%)	0.001
Firearm Injury	10 (6.2%)	17 (58.6%)	<0.001
Median ISS (IQR)	22 (17–32)	29 (22–35)	0.031
Liver Injury Grade			0.001
Grade III	97 (59.9%)	8 (27.6%)
Grade IV	50 (30.9%)	14 (48.3%)
Grade V	14 (8.6%)	7 (24.1%)
Initial Management			0.001
sNOM	97 (59.9%)	8 (27.6%)
OM	65 (40.1%)	21 (72.4%)

IQR – Interquartile Range; MVA – Motor Vehicle Accident; ISS – Injury Severity Scale; sNOM – Selective Non-Operative Management; OM – Operative Management

Nine (4.7%) patients in the cohort died during initial hospitalization; 7 patients had grade IV or V liver injury, and 5 patients underwent initial OM. Eleven (37.9%) patients were readmitted within 30 days for a liver-related complication, of whom 7 patients required a procedure during the readmission. No additional readmissions occurred in the subsequent 60 days (Table 3). Seven patients were initially discharged to home and were readmitted with a subsequent unplanned intervention; all of these patients had grade III or IV injuries, and all had a penetrating mechanism of injury, with 6 of 7 from firearms.

Table 3 –

Outcomes of Hospitalization

	No Delayed Intervention n=162 (84.4%)	Delayed Intervention n=29 (15.6%)	p
Discharge Disposition			0.424
Home	113 (69.8%)	21 (72.4%)
SNF	15 (9.3%)	5 (17.2%)
PM&R	22 (13.6%)	3 (10.3%)
LTAC	3 (1.9%)	0 (0.0%)
Expired	9 (5.6%)	0 (0.0%)
Median LOS (IQR), days	7 (4–12)	17 (13–32)	<0.001
Median ICU LOS (IQR), days	3 (2–6)	8 (3–13)	0.001
Median Ventilator Days (IQR), days	0 (0–3)	4 (0–7)	0.004
30-day Readmission for Liver-Related Complication	2 (1.2%)	11 (37.9%)	<0.001

SNF – Skilled Nursing Facility; PM&R – Physical Medicine and Rehabilitation; LTAC – Long-Term Acute Care; LOS – Hospital Length of Stay; IQR – Interquartile Range; ICU – Intensive Care Unit

The logistic regression model is shown in Table 4. In this model, firearm injury and liver injury grades IV and V were independently associated with need for unplanned intervention. Initial management was not independently associated with an unplanned intervention. Age was also not associated with unplanned intervention.

Table 4 –

Regression Analysis

	Odds Ratio	95% Confidence Interval	p
Age	1.0	0.9–1.0	0.167
Firearm Injury	3.1	1.2–8.2	0.019
Livery Injury Grade
Grade III	1.0	Reference
Grade IV	3.1	1.2–8.1	0.023
Grade V	4.3	1.3–14.6	0.019
Initial Operative Management	1.0	0.3–2.8	0.960

Discussion

Liver injury in the modern era continues to be a proxy for severe life-threatening injury; in our initial patient cohort of grade III or higher injuries, one quarter of patients died within 24 hours. Patients who needed unplanned interventions differed from those who did not in several ways, which should allow clinicians to identify a high-risk population and maintain a high index of suspicion for development of complications. Those who needed unplanned interventions were younger, had more penetrating injuries, and had more severe injury. These patients also tended to be treated initially with OM. After adjustment for the above factors, the initial management strategy, (OM or sNOM) was not associated with delayed complications requiring procedures. Instead, the grade of liver injury was significantly associated with requiring an unplanned intervention for liver-related complications. Additionally, we determined that the median day for an unplanned intervention was 11 days. Surveillance at 7–10 days or prior to discharge for the high-risk group may be able to obviate some readmissions or allow for earlier discharge.

Much of the modern literature on liver injury exclusively examines patients who underwent sNOM. This assumes that patients treated without an operation are inherently different from those operatively managed, perhaps because of their initial injury severity and hypotension. For example, the current Eastern Association for the Surgery of Trauma practice management guideline for liver trauma only addresses patients who are nonoperatively managed, leaving no guidance for patients who presented unstable or who required abdominal operations for other concomitant injuries.⁴ Kozar et al. examined the complications of sNOM after high-grade blunt liver injury, and concluded that complication rates in grade IV and V liver injury is expected to be 21% and 63%, respectively.⁸ Our study showed similar complication rates of 25% and 50% in survivors with grade IV and V liver injury who underwent initial sNOM. Interestingly, this was also similar to complication rates of 21% and 23% seen in survivors who were treated initially with OM. No treatment modality is immune to the complications such as bile leaks, bilomas, arteriovenous malformations, bleeding, or infection. In our adjusted analysis, grade of injury was significantly associated with complications requiring an unplanned intervention, while the initial treatment strategy was not.

Currently there is a lack of consensus on routine use of CT scan for surveillance in patients with liver injury, especially those treated with sNOM. In general, recommendations mention “selective” surveillance but the selection criteria continue to be nebulous⁴. Cuff et al. concluded that routine follow-up CT scan is not necessary after liver injury in patients treated with sNOM, but selection should be based on the severity of injury, presence of associated intra-abdominal pathology, and clinical parameters.⁹ In a systematic review performed by Mebert et al., out of the 920 CT scans performed to detect complications, only one patient underwent intervention as a result of the CT scan.¹⁰ Conversely, Barbier et al. do recommend follow-up CT scan. In their study, they were able to identify liver and pulmonary complications in 15% and 28% of patient respectively, however they did not recommend a time frame in which the follow-up CT scan would be beneficial.¹¹ Bala et al. focused on the bile leaks after high grade liver injury and recommended that optimal time for follow-up CT scan is 7–10 days after the initial injury in order to reduce complications associated with bile leaks.¹² This aligns with the timeline estimate that we found in our study, based on the timing of complications and unplanned interventions seen in our study population. Prospective studies would be required to evaluate the efficacy of such a screening protocol.

Over time, the mortality rate following liver injury has decreased, especially after blunt liver injury^5,13,14. This has largely been attributed to the increased use of sNOM and a more selective OM. Our institution was an early adopter of sNOM, including for penetrating trauma, and also commonly use angiography as an adjunct to laparotomy¹⁵. We agree that the shift towards adoption of sNOM has contributed to the overall decrease in early deaths from liver trauma. However, if sNOM was the only reason for improvement of outcomes, then the expectation in our analysis would be that the OM group would have more delayed complications as well. In our data, late deaths were multifactorial and occurred on at a median of hospital day 8 (IQR 5–9). Several patients died early with open abdomen and one died of fulminant hepatic failure; these patients, based on our definition of unplanned interventions, would not have been eligible for unplanned interventions since they had open abdomen and continued to require “planned” washouts. In our adjusted analysis, the initial OM was not associated with more complications, but the grade of liver injury was associated. This suggests that the decrease in mortality over time has been affected not only by sNOM but also other modern techniques, such as having the option to add angiography to operations in the OM group. It is also likely that improvements in critical care and management of sepsis are factors in the improvement of mortality over time.

There are several limitations of our analysis. In this retrospective, single center study, we performed a thorough chart review to identify all delayed complications that were attributable to the original liver injury. However, it is possible that complications could be miscategorized. Also, many patients had multiple injuries which likely put them at additional risk for complications. Because injury severity score and liver injury grade were highly associated and collinear, we elected to include one of these variables in our regression model and cannot make a determination as to the magnitude of any contribution of other injuries on the development of complications that require treatment. However, our study was able to identify simple, clinically relevant, and statistically significant factors for risk stratification for use in the future.

In modern trauma care, sNOM is the current standard of care for the appropriately selected patient.^4,16 In many cases, observation alone is sufficient, but for those patients with concern for arterial bleeding, angiography with or without embolization can be used as monotherapy or in conjunction with an operative intervention.¹ However, no treatment modality is immune to complications such as bile leaks, bilomas, arteriovenous malformations, bleeding, or infection. Patients with higher risk for complications that require an unplanned intervention include patients with higher grade injury and penetrating trauma, specifically from firearms. In adjusted logistic regression, factors such as age and initial OM were not associated with the development of complications. It is possible that in the future we could prevent readmissions and identify complications earlier with a surveillance CT scan at 7–10 days in high-risk patients. Application of these findings may help clinicians identify patients at high risk of complications and identify those complications earlier.