Penetrating abdominal trauma (PAT) in South Africa is amongst the most prevalent worldwide. In 2013, interpersonal violence was ranked 5th in all-cause mortality in Cape Town [1]. Groote Schuur Hospital (GSH) is a government-funded, tertiary teaching hospital situated in Cape Town, South Africa. It is the chief academic hospital of the University of Cape Town and one of the busiest trauma referral hospitals in the world, with an estimated 10,000 patients being seen in the trauma unit annually, 70% having sustained intentional injuries of which 57% is of a penetrating mechanism. The current global trend in PAT has seen management shift towards selective conservatism [2,3,4,5,6,7,8,9]. The purpose of this study is to compare the presentation, management and outcomes of patients managed non-operatively versus operatively (both immediate and delayed) in our unit to provide evidence-based guidance for the safe implementation of selective conservatism in PAT.

Methods

Setting and data sources

GSH is the central referral hospital in the Cape Metro West health district, which serves an estimated catchment area of 2,292,000 uninsured patients. It has 975 beds of which 50 beds are dedicated to trauma. The validated, electronic Trauma Health Record (eTHR) was implemented at GSH in January 2014, whereby electronically generated records replaced all previous handwritten record keeping [10]. Data variables required by eTHR are collected prospectively by clinicians in real time including data required by the Abbreviated Injury Scale (AIS) classification system—a consensus-derived, anatomically based, 7-digit injury scoring system used to generate the ISS [11, 12].

Patients presenting to the centre with PAT during the study period of 2 years between 1 May 2015 and 30 April 2017, were identified from eTHR and prospectively entered into a REDCap database designed to specifically audit the outcomes of patients with PAT [13]. Penetrating abdominal trauma was defined as any penetrating wound between the 5th intercostal space and the pubis anteriorly, and the angle of the scapula down to the creases of the buttock posteriorly. The abdomen was further subdivided into the following zones (with borders) for descriptive purposes: thoracoabdominal (from the 5th intercostal space to the costal margin); anterior abdomen (from the xiphoid to pubis, between the anterior axillary lines); pelvis (iliac crests superiorly down to the perineum inferiorly); and back/flank (posterior to the anterior axillary lines). The management pathway, i.e. immediate laparotomy or non-operative management (NOM), as well as the need for additional imaging, was determined by the trauma surgical trainee following adequate assessment in the trauma resuscitation bay. Any subsequent laparotomy before the discharge of the NOM cases, were then classified as a delayed laparotomy (DOM) and recorded appropriately. There were no exclusion criteria. Figure 1 illustrates these definitions with the institutional algorithm for PAT. All patients with a ‘negative’ abdominal examination and/or a ‘negative’ CT investigation, were admitted for abdominal observation. Patients with left thoracoabdominal (TA) PAT wounds were subjected to abdominal observations (as described in the discussion below) before undergoing a diagnostic laparoscopy to exclude an occult diaphragm injury.

Fig. 1
figure 1

Institutional algorithm for penetrating abdominal trauma

Full size image

Data analysis

Descriptive analysis of the PAT database included: basic demographics; admission of illicit drug use; presenting vital signs; blood investigations; number of penetrating traumatic insults; penetrating wound positions; presence of peritonism and/or evisceration; indication for laparotomy; radiological investigations and interventions; operative or nonoperative management; laparotomy findings- therapeutic, non-therapeutic, or negative; abdominal visceral injuries and associated injuries. When recording the indication for laparotomy, the most clinically urgent reason was used. However, when the indication was unclear, no entry was made, and should two indications have been seen as equally urgent, both were recorded. The presence of peritonism was recorded separately. Injury severity was described by the Injury Severity Score (ISS); the Revised Trauma Score (RTS); Abbreviated Injury Scale (AIS); Penetrating Abdominal Trauma Index (PATI); and Kampala Trauma Score (KTS) [11, 12, 14,15,16]. Outcome variables included Intensive Care Unit (ICU) admission; hospital length of stay (LOS); hospital readmission; relaparotomy; in-hospital complications; and mortality. Complications were defined as any deviation from the normal postoperative course requiring the need for: pharmacological; endoscopic; interventional radiological; or surgical treatment, and were categorised according to the Clavien-Dindo classification system [17]. Multi-trauma injury patients were classified as having an AIS of greater than or equal to three in at least two organ systems.

Comparisons were then made between subgroups of patients managed operatively, both immediately and delayed (DOM), as well as non-operatively (NOM). Univariate predictors of DOM were also explored. The positive predictive value (PPV) of operative indications for therapeutic laparotomy was calculated and outcomes between the subgroups were compared.

For the descriptive analysis, continuous data were summarised using means and 95% confidence interval (95% CI) if normally distributed, whereas medians and interquartile range (IQR) was used for non-normally distributed data. For inferential statistics, parametric tests were performed such as Chi-square tests or where appropriate the non-parametric equivalent. Furthermore, univariate logistic regression was performed to determine associations for delayed non-operative management and reported with odd ratios and 95% CIs. A p value of < 0.05 was considered statistically significant. STATA® 14 software was utilized for statistical analysis.

Results

Presenting characteristics and injury profile

Descriptive assessment

In total, 805 consecutive patients with penetrating abdominal trauma were managed over the 2 year period. There were 502 (62.4%) and 303 (37.6%) patients with gunshot (GSW) and stab wounds (SW), respectively. The overwhelming majority were young men (94.7%), with a mean age of 28.3 years (95% CI 27.7–28.9) and median ISS of 13 (IQR 9–22). There was a preponderance of poly-assault, and more than one penetrating injury was present in 68.2% of SW; and 64.0% of GSW. Cumulative abdominal wall penetrative wounds were abdominal (anterior and back/flank) in 696 (86.5%); thoracoabdominal in 332 (41.2%); and pelvic in 192 (23.9%) injuries. Three hundred and sixty-eight (45.7%) presented with peritonism, more commonly in GSW (59.4%) compared to SW (23.1%) (p < 0.001). The three most commonly injured organs overall were: colon—177 (22.0%); liver—189 (23.5%) and small bowel—241 (29.9%). Multi-trauma injuries were found in 40.9% of cases overall. This included 106 (35%) SW, and 223 (44.4%) GSW patients (p = 0.005). Table 1 summarises the presenting characteristics and injury profile of the cohort.

Table 1 Presenting features and injury profile of 805 patients with PAT
Full size table

Comparative assessment

Successful non-operative management was achieved in 304 (37.8%) patients and 501 (62.2%) were managed operatively. Of these, 477 (59.3%) underwent immediate laparotomy, and 24 (3.0%) delayed operative management. Table 2 compares the presentation; management; and outcome of the cohort compared by the three possible operative outcomes i.e. immediate laparotomy (IOM); delayed operative management (DOM); and non-operative management (NOM). The young, male demographic was maintained across the three groups with no significant deviations. Patients managed by immediate laparotomy were more likely to have sustained a GSW; anterior abdominal wound; and higher median ISS. The DOM & NOM groups were more likely to have undergone a CT scan, but there was no significant difference in the use of interventional radiology between the three groups. CT scans were obtained preoperatively in some patients for investigation of the urinary tract system in the presence of haematuria. This allows for Grade 3 and lesser renal injuries to be managed conservatively intraoperatively by not entering Gerota’s fascia.

Table 2 Presentation, management and outcomes by operative strategy
Full size table

On univariate analysis, number (OR 1.05; 95% CI 0.92–1.19); location (OR 1.1; 95% CI 0.79–1.25); and mechanism of injuries (OR 0.49; 95% CI 0.22–1.11) were not associated with DOM (i.e. failed NOM). Rates of therapeutic laparotomy were achieved in 90.3% of the immediate; and 80.3% in the DOM cohorts. On arrival, immediate laparotomy was performed for haemodynamic instability in 44 (5.5%) patients (PPV 93.2%); peritonism in 298 (37.2%) patients (PPV 93.3%); unreliable clinical assessment in 21(2.6%) patients (PPV 66.7%); radiological findings in 54 (6.7%) patients (PPV 87.0%); and for organ evisceration in 29 (3.6%) patients (PPV 89.7%). In the DOM cohort, 2 (8.3%) became haemodynamically unstable (PPV 100%): 14 (58.3%) developed peritonism (PPV 85.7%): 10 (41.7%) had positive radiological findings (PPV 87.5%); and 8 (33.3%) developed signs of sepsis (PPV 87.5%). Table 3 describes PPV of operative indications for therapeutic laparotomy in the immediate laparotomy, and DOM cohorts.

Table 3 PPV of operative indications for therapeutic laparotomy
Full size table

Nineteen diagnostic laparoscopies were performed to exclude an occult diaphragm injury. Six (31.5%) diaphragm injuries were detected; three were repaired laparoscopically and three were repaired via laparotomy.

The mortality rate was 11.3% in immediate laparotomy; 1.3% in the NOM, and 0% in the DOM subgroups (Table 2). This trend differed slightly with the complication rate: immediate laparotomy—41.5%; DOM—33.3%; followed by NOM group—10.0%. Although the rates of mortality and morbidity of the NOM cohort are far lower, the rate of complications, ICU admission, reoperation and missed injuries were no different in the immediate laparotomy; and DOM cohorts (p > 0.05). All complications were classified by Clavien-Dindo grade and are presented by operative strategy in Table 4. The grade of complication did not differ by operative strategy (p > 0.05).

Table 4 Clavien-Dindo complication grading by operative strategy
Full size table

Discussion

The concept to selective conservatism for PAT has gained traction ever since Shaftan’s landmark paper in 1960 when he challenged the dogma of mandatory explorative laparotomy [2]. Various subsequent studies stretching from opposite corners of the world have further supported this selective approach with excellent results [3,4,5,6,7]. These aforementioned authors all concluded that patients presenting with PAT in the absence of haemodynamic instability; peritonism; organ evisceration; positive radiological findings; or an unreliable clinical examination can be managed expectantly, without increased morbidity or mortality. However, due to the rapid evolution of this approach, finer details and smaller subgroup analysis is needed to convince sceptics.

Here we present one of the largest; and most detailed prospective series to date, comparing various operative strategies side by side, in a composite PAT cohort (SW and GSW). We used these three operative strategies, namely: immediate laparotomy; NOM; and DOM (i.e. “failed” NOM) to dissect out the outcomes of every sub-group and compare alongside one another.

Various authors have noted that mandatory laparotomy policies for penetrating abdominal trauma result in unnecessary laparotomy rates ranging from 5.3 to 27% for GSW, and 23 to 53% for SW [18, 19]. This implies that nearly a quarter of GSW and almost half of all abdominal SW do not require a laparotomy, however, our findings were closer to a third and two thirds, respectively.

Given these striking findings, the important learning objective for any trauma unit managing PAT must be how to implement selective NOM to decrease the number of non-therapeutic laparotomies in PAT (both SW and GSW) without an increased missed injury rate; and without causing increased mortality or morbidity in the failed NOM patients. This manuscript demonstrates this with our protocolised approach involving an initial, methodical assessment and resuscitation utilizing protocolized guidelines, (such as that proposed in Advanced Trauma Life Support™) followed by the decision whether the patient must proceed for immediate surgery; further investigation; or observation alone [20].

Indications for immediate laparotomy include haemodynamic instability (indicating ongoing haemorrhage); peritonism (suggesting abdominal contamination); organ evisceration or impalement; unreliable clinical examination; or radiologically confirmed bladder or ureteric injuries [19, 21, 22].

Once the aforementioned indications for mandatory laparotomy have been excluded, the patient can be considered for NOM, either immediately or following further investigation. In 2015 Navsaria et al. proposed a selective rather than mandatory use of CT scans in abdominal GSW management, where absolute indications for imaging included right upper quadrant / right thoracoabdominal injury (to exclude liver injury); and haematuria (to exclude urogenital injury) [8]. In this study presented here, all except one of the indications for immediate laparotomy had a PPV of greater than 87%, namely, unreliable clinical examination, which had a PPV of just 66.7%. This is, however, not an unexpected finding for this indication, if referencing the literature [21]. The reason for this is that even with the advances in technology, CT scans only display a sensitivity of 91–97%, specificity of 96–98%, and accuracy of 96–98% for detecting intra-abdominal injuries that actually require laparotomy in patients with abdominal GSWs [23,24,25]. A negative CT scan thus does not exclude an intrabdominal injury. Therefore, in patients with PAT and concomitant head injuries and/or high-spinal cord injuries (where the serial abdominal examination is unreliable), despite the high rate of negative explorations, a laparotomy is justified.

Non-operative management should rarely (if ever) involve discharging a patient directly from the emergency department. This practice is not safe until the sensitivity, specificity and accuracy of CT for penetrating abdominal trauma is 100%. Thus, we and other authors feel it should be routine practice to admit all patients with PAT for further observation [21].

During this admission the patient is kept nil per mouth, maintaining hydration with an isotonic intravenous crystalloid solution. Anti-biotics are withheld, and analgesia administered as necessary. Routine four-hourly vital sign observations and regular serial clinical assessment (preferably by the same clinician) must be carried out. After 24hrs of observation, should the patient’s abdominal examination or haemodynamic status not deteriorate, the patient can be fed. Should they not develop foregut symptoms (nausea of vomiting) they can be considered for discharge home [26, 27]. Further investigation during this admission is at the discretion the treating clinician. However, fever and a rise in WCC need to be interpreted with caution. Failure of NOM prompts operative intervention and is usually signalled by peritonism; ongoing blood loss; or concern of sepsis. Of those managed by DOM in our cohort, 2 (8.3%) became haemodynamically unstable (PPV 100%); 14 (58.3%) developed peritonism (PPV 85.7%); 10 (41.7%) had positive radiological findings (PPV 87.5%); and 8 (33.3%) developed signs of sepsis (PPV 87.5%). The aforementioned significance of radiology in decision making in this DOM group is contrary to a recent systematic review and metanalysis of GSW in PAT [28].

Historically PAT to varying abdominal regions were managed differently, with clinicians being more hesitant to treat back and pelvic injuries non-operatively, fearing occult hollow-viscous injuries [22, 29,30,31]. However, our experience as well as prior publications refute this, indicating that GSW and SW to the anterior abdomen; back; flank and pelvis can all be approached with the common principle described above [31,32,33,34,35]. We found that although anterior abdominal wound is associated with immediate laparotomy, unlike a recently published systematic review and meta-analysis, abdominal wound position is not associated with DOM (i.e. “failed NOM”) [28]. The abdominal region injured may however, guide the choice of diagnostic adjuncts in the work-up [21, 31,32,33,34,35,36,37,38,39].

Unsurprisingly, higher injury severity scores (ISS) and GSW were independently associated with immediate laparotomy, however, neither were significant in differentiating NOM from DOM cases in this study. In other words, both GSW and the more severely injured patients have a higher chance of needing an immediate laparotomy, but should that not be clinically indicated, their chances of NOM success are no different to the rest.

Contrary to Zafar et al. reporting that failed NOM is associated with increased mortality, two subsequent systemic reviews by Lamb et al. in 2014, and Al Rawahi et al. in 2018, showed no difference in outcome between those undergoing early and late laparotomies, which is in keeping with our findings [28, 40, 41]. In this series, successful non-operative management was achieved in 304 (37.7%) patients, with a significantly lower morbidity and mortality rate compared to the operatively managed patients. This is in-spite of the fact that the four NOM mortalities recorded were unrelated to PAT (one fatal CVA; and three transhemispheric GSW, for which the patients could better be described as palliated, rather than managed non-operatively). However, in addressing the abovementioned concern, it is more important to note that the DOM subgroup presented here, did not have worse outcomes (mortality; complications; ICU admissions; reoperations; or missed injuries) than those managed by immediate laparotomy (p > 0.05). Thus, implying that selective conservatism in PAT in a high volume tertiary referral trauma centre for civilian SW and GSW is not only effective but safe too. Numerous other publications support this finding, by having shown that patients managed with selective non-operative management compare favourably to operative management as they have shorter admission periods and equivalent, or reduced mortality [8, 9, 26, 42].

Our findings must be interpreted within the study’s limitations. This is a single centre study at a high volume trauma unit, which experiences a high incidence of PAT. Whether these findings are generalizable and similar results can be expected in lower volume trauma units or areas with a low incidence of penetrating trauma is unknown, however, selective NOM has been reported to be safe in lower volume centres too [43]. The prospective nature of this study was enabled by the real-time electronic trauma registry. This is a clinician entered the database and although no data validation of the PAT patients occurred in this study, validation studies of this database have been performed previously [10].

Conclusion

Patients presenting with PAT in the absence of haemodynamic instability; peritonism; organ evisceration; unreliable clinical examination; and positive radiological findings can be managed expectantly without increased morbidity or mortality. Figure 2 summarises the current study findings.

Fig. 2
figure 2

Institutional algorithm for penetrating abdominal trauma

Full size image