Transfer Is Associated with a Higher Mortality Rate in Necrotizing Soft Tissue Infections

Patients and Methods

This study was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center.

A retrospective chart review was performed of all patients treated at two large urban, tertiary referral hospitals. Patients were included on the basis of their admission International Classification of Diseases (ICD)-9 code (728.86) or ICD-10 code (M72.6) for NSTI from 2012–2015. Patients were considered transfers if they came from a different emergency department or had inpatient status at a different hospital. Patients were excluded if they did not require any further debridements at the tertiary center or were less than 18 years of age. All operative reports were reviewed to ensure the post-operative diagnosis of a NSTI matched the ICD-9/10 code. Patients who did not have a post-operative diagnosis of an NSTI by the operating surgeon were excluded from the study. Patients who were placed on comfort care without any prior surgical debridement were included just on the basis of the ICD-9/10 code. Patients who were treated primarily at our institutions were referred to as P-NSTI, whereas those who were transferred were referred to as T-NSTI.

The primary endpoint was in-hospital death. Secondary outcomes were timing of transfer, debridement status prior to transfer, route of transfer, quick sequential organ failure assessment (qSOFA) score, and intensive care unit (ICU) level of care (mechanical ventilation, hemodialysis, and vasopressor use). Patients also were compared on their microbiologic data and use of appropriate empiric antibiotics on presentation. The qSOFA scoring system [10] was used as an indicator for severity of disease on presentation to the tertiary center [11]. The qSOFA scoring system consists of systolic blood pressure ≤100 mm Hg, respiratory rate ≥22 breaths/min, and Glasgow Coma Scale (GCS) <15. Patients are given one point for each criterion. In cases where the GCS was not explicitly recorded at time of presentation, patients were considered to have a GCS of 15 if they were described as alert and oriented. If they were described as lethargic or somnolent, they were considered to be GCS <15. The ICU level of care was used as a marker for disease severity during their hospital admission [12]. Early hospital transfer was defined as ≤1 hospital day at the transferring hospital, whereas late hospital transfer was defined as >1 hospital day. Transfer patients also were categorized by whether they underwent debridement prior to transfer and into survivors and non-survivors. Survivors were defined as those who survived to hospital discharge. Empiric antibiotic regimens at the time of presentation were considered appropriate if they covered gram-positive, gram-negative, and anaerobic organisms [4]. Antibiotic regimens that did not cover methicillin-resistant Staphylococcus aureus were not considered appropriate.

The characteristics of T-NSTI and P-NSTI patients were compared using the Pearson χ² and Fisher exact tests where appropriate for categorical variables. The Wilcoxon rank sum test was used for continuous variables. To determine the independent association between transfer status and in-hospital death, we conducted a multivariable regression model. The patient-related factors of survivors and non-survivors (listed in Table 2 below), which demonstrated a univariable association with a particular outcome p < 0.3, were included as potential predictors in the regression model.

Transfer patients were then divided into survivors and non-survivors and compared to see if there were any differences with regard to timing of transfer, debridement status prior to transfer, and route of transfer in addition to the variables listed in Table 1. Multivariable regression was not conducted on the transfer patients because of the small number of patients.

All statistical analysis was performed with SPSS v 25 (IBM, Aramonk, NY). A p value <0.05 was considered statistically significant.

Results

A total of 138 patients met our study criteria. There were 39 (28.0%) transfer patients and 99 (72.0%) patients who were treated primarily at our institutions. Demographic and peri-operative characteristics in the two groups are listed in Table 1. The T-NSTI patients were more likely to be smokers, have a higher qSOFA score, and require hemodialysis. The P-NSTI patients were more likely to have diabetes mellitus. A total of 70 P-NSTI patients (72.2%) were treated at a Level 1 trauma center and 33 (80.5%) of the T-NSTI patients were transferred to a Level 1 trauma center for their care (p = 0.39) (Table 1).

A total of 129 patients (93.5%) had intra-operative cultures obtained, and microbiologic data were available. Of these patients, 9 (7.0%) had no growth on cultures, 11 (8.5%) had a monomicrobial infection, and 109 (84.5%) had a polymicrobial infection. Transfer patients were more likely not to have an intra-operative culture obtained and to have no growth in their cultures (Table 1). The P-NSTI patients were more likely to receive appropriate empiric antibiotics compared with T-NSTI patients on the day of presentation (85.9% versus 69.2%; p = 0.02) (Table 1).

There was a significant difference the in-hospital mortality rate between T-NSTI and P-NSTI patients (35.9% versus 14.1%; p < 0.01). Non-survivors were significantly more likely to have baseline chronic kidney disease (CKD) (Table 2) and to have a higher qSOFA score (2 [1–2] versus 1 [0–1]; p < 0.01) on presentation to the tertiary center and require various levels of ICU care (Table 2). There were no differences in the microbiologic data and appropriate antibiotic administration on presentation between survivors and non-survivors for the whole cohort (Table 2).

It was difficult to determine the adequacy of the first debridement for patients debrided prior to transfer. Only 1 (2.6%) of the transfer patients had a note from the tertiary center indicating that the first debridement was inadequate at the index hospital. A total 38 (97.4%) of the transfer patients had a debridement within 24 hours of presenting to a tertiary center. The one transfer patient who did not have a debridement at a tertiary center was placed on comfort care on arrival. The total number of debridements at the index hospital could not be obtained accurately in all cases. However, the median number of tertiary center debridements in the P-NSTI group was 3 (2.0–4.5) versus 2 (1.0–5.0) in the T-NSTI group (p = 0.06).

A multivariable logistic regression model was performed to determine independent predictors of in-hospital death (Table 3). Cirrhosis (odds ratio [OR] 58.54; 95% confidence interval [CI] 3.29–104.84; p = 0.01), need for hemodialysis (OR 6.93; 95% CI 1.38–34.83; p = 0.02), and transfer status (OR 5.33; 95% CI 1.02–28.30; p = 0.04) were independent predictors of in-hospital death (Table 3). Smoking (OR 0.07; 95% CI 0.01–0.69; p = 0.02) and treatment at a Level 1 trauma center (OR 0.18; 95% CI 0.03–0.97; p = 0.04) were independent protectors against in-hospital death (Table 3). The logistic regression model was statistically significant p < 0.01 and explains 66.7% (Nagelkerke R²) of the variance in in-hospital deaths and correctly classified 89.8% of cases. The in-hospital mortality rate was significantly lower at the Level 1 trauma center than at the non-trauma tertiary center (15.5% versus 34.3%; p = 0.02).

Transfer patients were grouped into survivors and non-survivors. Demographic and peri-operative characteristics are depicted in Table 4. The two groups had notable differences in their characteristics. Furthermore, non-survivors were more likely to present with a higher qSOFA score (1.5 [1–2] versus 1 [0–1]; p = 0.02) on presentation to the tertiary center and require some level of ICU care (Table 4). Survivors were more likely to be treated at a Level 1 trauma center than were non-survivors (96.2% versus 61.5%; p = 0.01). There were no differences in timing of transfer and debridement status prior to transfer in the two groups (Table 4). However, a subset analysis found that patients who were transferred early without any debridement had a trend toward a lower in-hospital mortality rate than all other transfers (21.4% versus 40.0%; p = 0.21). In addition, patients transferred directly from the emergency department (ED) had a trend toward a better mortality rate than patients who had inpatient status at the index hospital (25.0% versus 43.5%; p = 0.23).

A total of 32 (82.1%) of the 39 reasons for transfer were either “need higher level of care” or “care not available at transferring hospital,” 1 (2.6%) was attributable to test/procedure/treatments not being available at the transferring hospital, and 6 (15.4%) patients were transferred because necessary specialist physicians were not available.

Discussion

We found that T-NSTI patients had a significantly higher mortality rate than P-NSTI patients. Although there were notable differences in demographics and peri-operative characteristics in the two groups (see Table 1), our multivariable logistic regression model determined that transfer status was an independent predictor of in-hospital death (see Table 3). We found that the timing of transfer and debridement status prior to transfer were not different among survivors and non-survivors (see Table 4), but there was a trend toward a better in-hospital mortality rate in patients who were transferred early and without prior debridement. Survivors were more likely than non-survivors to be treated at a Level 1 trauma center for both the overall cohort and transfer patients.

The findings of current literature evaluating transfer status using large national databases have been mixed. Two studies utilized the Nationwide Inpatient Sample (NSI) to evaluate the correlation of transfer status and outcomes in patients with NSTI [8,13]. Both of these studies found that transfer patients had significantly higher in-hospital mortality rates than patients who were not transferred. Mills et al. performed a risk factor analysis using the National Surgical Quality Improvement Program (NSQIP) database for patients who underwent operations for NSTI [2]. They found that direct admits (e.g., non-transfers) had a significantly lower 30-day mortality rate than patients who were transferred from an outside facility.

These results contradict those of Ingraham et al., who found that interhospital transfer was not an independent risk factor for death within 30 days [9]. Although those investigators also used the NSQIP database, their study differed in that they confined their analysis to patients whose index operations at the receiving hospital were classified as emergencies in an attempt to exclude patients who were transferred for wound care. Furthermore, it was thought this selected for transfer patients who had their index operation at the receiving center. If the assumption holds true, their study suggests that transfer patients who have their index debridement at the tertiary center have a mortality rate similar to that of patients treated primarily at a tertiary center.

The findings set forth by Ingraham et al. may seem counterintuitive, as NSTIs have long been thought to require immediate surgical debridement [5,14]. Therefore, it can be argued that the expected outcome would have been a higher mortality rate for transfer patients who did not have a debridement done at the index hospital. Although we did not observe a difference in the in-hospital mortality rate according to debridement status prior to transfer, we found a striking trend toward an improved in-hospital mortality rate in patients who were transferred early without any prior debridement compared with all other transfers. We also found a trend toward a better mortality rate in patients who were transferred directly from the ED compared with those transferred with inpatient status. These trends appear to be consistent with the findings set forth by Ingraham et al. and require further evaluation with larger, multi-institutional studies, as they may have implications for the overall management of transfer patients from referring hospitals.

We also found that P-NSTI patients were more likely to receive appropriate antibiotic regimens than the T-NSTI patients at the time of presentation (see Table 1). This may imply that the physicians at tertiary centers are more familiar with the disease, diagnosis, and treatment of NSTIs. Counterintuitively, an appropriate antibiotic regimen was not associated with survival (see Table 2). However, we suspect that antibiotics are playing a limited role in the initial treatment of this disease compared with surgical debridement. Although we were unable to assess the adequacy of the debridements at the index hospital accurately, 97.4% of the transfer patients required a debridement within 24 hours of presenting to the tertiary center. One patient had a note from a tertiary center that explicitly stated that there had been inadequate debridement at the index hospital. Therefore, it is possible that the mortality difference between the P-NSTI and T-NSTI groups is related to the adequacy of the first debridement.

Interestingly, we found that treatment at a Level 1 trauma center was associated with a better in-hospital mortality rate than treatment at a non-trauma tertiary center. Currently, the reasons for the outcome discrepancy between these two institutions is unclear. The two institutions are treated with similar surgical personnel (attendings and residents). Therefore, the discrepancy may be related to the resources available at one hospital versus the other. This suggests a role for regionalization of transfer of NSTI patients to a Level 1 trauma center or equivalent. Whereas regionalization of trauma care has led to significant reductions of trauma-related deaths [15,16], the role of regionalization along with its implementation in emergency general surgery (EGS) patients remains unclear. Future studies are needed to investigate the reasons for the outcome discrepancies and whether regionalization to Level 1 trauma centers for EGS patients would be beneficial.

In this study, we also found that cirrhosis and the need for hemodialysis were significantly associated with in-hospital death. Cirrhosis has been shown to be a risk factor for the development of NSTIs [17] along with being a risk factor for death [18]. The need for hemodialysis also was significantly associated with in-hospital death for both the overall cohort and the transfer patients. The need for hemodialysis in sepsis is associated with death [19] and likely represents the overall physiologic deterioration in patients with NSTI. Therefore, it is imperative that patients receive proper surgical and intensivist care prior to requiring hemodialysis, whether it be at a tertiary center or elsewhere.

Surprisingly, we found that smoking status was an independent protector against in-hospital death in the regression model. It is unclear why such an association exists in our cohort. However, previous literature has suggested that current smoking was independently associated with a lower mortality rate among patients with pneumonias [20,21]. In these studies, smokers were more likely to be infected by low case-fatality rate (CFR) pneumococcal serotypes, indicating that smoking may influence susceptibility to more harmful species. The microbiology data of smokers were not collected in this study, but this is an area of interest for future studies. In addition, the components of cigarette smoke are immunosuppressive [22]. Therefore, it may have decreased the patients' ability to generate a vast inflammatory response, decreasing their risk of overwhelming sepsis.

Our study has several important limitations. Its inherent nature has a retrospective chart review allows us to evaluate only data that have already been collected. Furthermore, we found that T-NSTI patients had higher qSOFA scores than P-NSTI patients (see Table 1). This suggests that T-NSTI patients had poorer physiologic status on presentation to the tertiary hospital. Although qSOFA score was not an independent predictor of death in the regression model (see Table 3), there likely exists a selection bias, as T-NSTI patients were more ill at the time of presentation. More studies are needed to understand better why transfer patients are presenting in poorer physiologic condition to a tertiary center.

Furthermore, physicians at tertiary centers are likely to encounter NSTIs more often than those at smaller community hospitals. Therefore, they may have a higher index of suspicion to make the diagnosis earlier. Unfortunately, we had access only to the dates on which patients presented, not the exact times. Therefore, we were not able to assess accurately the timing of diagnosis in the transfer group. Lastly, there are various reasons a patient will be transferred from one hospital to another. This could be attributable to unfamiliarity with the disease, surgeon unavailability, or inadequate resources [23,24]. Therefore, there likely are many factors involved in the decision to transfer a patient. We attempted to discover the reason for transfer in these patients because it may have implications for overall outcomes. However, the majority of reasons we found were vague, not specific enough to make a reasonable assessment. Further studies are needed with alternative strategies to understand if the reasons for transfer are associated with morbidity and death in these patients.

Conclusion

Transfer status is an independent predictor of in-hospital death in patients with NSTI. Timing of transfer and debridement status were not associated with in-hospital death. However, there was a trend toward a better mortality rate for patients who were transferred early without any prior debridement. The need for hemodialysis was an independent predictor of in-hospital death, whereas treatment at a Level 1 trauma center and smoking status were independent protectors. Larger multi-institutional studies are needed to clarify the factors contributing to an increased mortality rate in these patients.