Identification of risk factors for mortality and delayed oral dietary intake in patients with open drainage due to deep neck infections: Nationwide study using a Japanese inpatient database

Data on risk factors for deep neck infection including descending necrotizing mediastinitis (DNM) have been limited. Using a nationwide database, the aim was identifying the factors related to patient death and delay in recovering oral intake.


| INTRODUCTION
A deep neck infection (DNI) is a serious disorder that often spreads to other organs and sometimes proves fatal. The most crucial initial treatment strategy in the management of DNIs is prompt assessment and selection of patients requiring surgical intervention for drainage. Descending necrotizing mediastinitis (DNM) originating from a DNI is a relatively rare, but rapidly progressive, destructive disease that is often fatal. [1][2][3][4] Despite improvements in diagnostic methods, surgical technique, and intensive care protocols, DNM has been reported to be a life-threatening condition with mortality rates up to 40%. [5][6][7] Although the etiology and clinical outcomes of treating DNI and/or DNM have been extensively reported, no study, other than a few sporadic case reports published in the Japanese language, has addressed dysphagia after surgical treatment. [8][9][10] We thus recently reported a logistic regression analysis addressing risk factors related to swallowing recovery after surgical intervention for DNI in a retrospective cohort of 123 patients. 11 However, these studies have several limitations attributed to their relatively small sample size from a single medical center. Moreover, the infrequency of DNM has been one reason for the difficulties in prospective investigations of these patients; no precise therapeutic schema has been proposed, and the optimal treatment for DNM is still debated.
The present study used a nationwide inpatient database in Japan to investigate the factors associated with mortality and delayed recovery of oral dietary intake in patients who underwent surgical drainage for DNI, including DNM.

| Data source
Data were extracted from the diagnostic procedure combination (DPC) database, a national inpatient database in Japan. The details of the DPC database have been described previously. 12,13 Briefly, the DPC database includes administrative claims data and detailed medical data, collected for all inpatients discharged from participating hospitals. The number of hospitals that participated in the DPC projects included more than 80% of acute inpatient care in Japan. [12][13][14][15][16] To optimize the accuracy of recorded diagnoses, responsible physicians are obligated to record the diagnosis with reference to the medical charts. The dates of hospital admission, surgery, discharge, bedside procedures, and drugs administered are recorded using a uniform data submission format. 13 All patient identifiers were removed from this database. Because of the anonymous nature of the data, the need for informed consent was waived. Study approval was obtained from the Institutional Review Board of the Tokyo Medical and Dental University (protocol number: M2000-788-15).

| Patient selection and characteristics
From the 38.2 million inpatients in the DPC database over a total of 5 years (between April 2012 and March 2017), patients satisfying the following inclusion criteria according to surgical interventions encoded with original Japanese codes 17,18 were selected: (1) underwent incisional drainage for DNI (Japanese code: K384-2) or retropharyngeal abscess (Japanese code: K387, corresponding to intra-pharyngeal drainage for DNI) and (2) underwent mediastinotomy (Japanese code: K502).
According to the World Health Organization and previous reports using the DPC database, 17,18 BMI was classified as underweight (<18.5 kg/m 2 ), normal (≤18.5 to <30 kg/m 2 ), and obese (≥30 kg/m 2 ). In addition to these groups, the 774 cases with missing values for calculation of BMI were categorized into the "missing" group. There were also missing values for smoking status (337 cases), and patients were thus subdivided into the following three groups: smoking, nonsmoking, and missing.

| Outcome measurement
The primary outcome was survival at discharge. Regarding the subgroup analysis focusing on patients who were alive at discharge, the secondary outcome was the interval between admission and presumed full recovery of oral intake without nutrition from tube feeding or intravenous hydration, according to our previous retrospective, cohort study. 11

| Statistical analysis
Differences in continuous variables and frequencies between the groups were evaluated using the Kruskal-Wallis or Mann-Whitney U test and the chi-squared test, respectively. Multiple logistic regression analysis was performed to identify risk factors associated with survival at discharge (primary outcome) and delayed recovery of postoperative swallowing (secondary outcome). In terms of the secondary outcome, patients were divided into two groups according to median duration (9 days) of recovering oral intake without support of tube feeding or intravenous hydration support. Inclusion of independent variables in the model was based on earlier research and existing knowledge 19 in terms of the following risk factors for aggravation of DNI 11,20-23 : age, sex, obesity, smoking habits, DNM, comorbid DM and sepsis, repeated surgical interventions, undergoing tracheotomy and/or mechanical ventilation, and duration of empirical antibiotic therapy. A p value <0.05 was considered significant. All statistical analyses were conducted using EZR software. 24 3 | RESULTS

| Demographic data
A total of 6405 cases were identified, as shown in Figure 1. Of the 6040 patients who underwent drainage for a DNI and/or retropharyngeal abscess, 13 and 1232 cases were excluded because their surgeries were attributed to an injury or foreign body and related to resection of pharyngeal tumor or hemostasis after laryngopharyngeal surgeries, respectively. Of the 365 patients who underwent mediastinotomy, 154 patients were selected after excluding 108 patients who underwent surgery for resection of a mediastinal tumor, 6 patients with mediastinitis attributed to injury or a foreign body, and 97 patients whose mediastinitis was attributed to a thoracic lesion including the esophagus, because ICD-10 codes addressing acute infections in the head and neck region were lacking. Of the 154 patients who underwent mediastinotomy related to DNIs, 44 underwent both drainage for a deep neck abscess and mediastinotomy. Finally, 4949 patients from 635 hospitals were selected.
Of all of the 4949 patients, the 4791 patients who were alive at discharge were divided into two groups according to the median interval (9 days) between admission and full recovery of oral intake without nutrition from tube feeding or intravenous hydration. Namely, Group A and Group B included the cases with intervals less than <10 days and ≥10 days, respectively. Another 158 patients who died in hospital were assigned to Group C. Table 1 summarizes the baseline characteristics of each group of patients. The patients were categorized into three groups according to the clinical outcome, as follows: (A) alive at discharge undergoing nutritional support for 9 days after admission; (B) alive at discharge undergoing nutritional support for more than 10 days after the first operation; and (C) died in hospital.
F I G U R E 1 Flow diagram of patient selection. DPC, diagnostic procedure combination T A B L E 1 Baseline characteristics of patients categorized by survival at discharge and prolonged nutritional support by tube feeding or intravenous hydration The median ages of Groups A, B, and C were 56, 66, and 73 years, respectively. Group C showed a significantly higher interquartile range for age  years) than the other groups (p < 0.001, Kruskal-Wallis test). Conversely, both the male to female ratios and obesity evaluated by the BMI of the three groups were not significantly different.
Smoking status was also not significantly different.
Concerning disease comorbidity, the prevalence of DM was higher in Group C (69.6%) than in Group A (20.7%) and Group B (40.0%) (p < 0.001). In addition, comorbid sepsis was significantly higher in Group C (51.3%) than in Group A (5.2%) and Group B (19.7%) (p < 0.001). In terms of the presence of DNM, the prevalence was higher in Group C (29.1%) than in Group A (3.6%) and Group B (18.1%) (p < 0.001).
Regarding surgical interventions, the prevalence of repeated surgical interventions (more than one surgical drainage) was higher in both Group B (13.5%) and Group C (11.4%) than in Group A (1.8%) (p < 0.001). The prevalence of subsequent tracheotomy was greater than 40% in both Group B and Group C, significantly higher than in Group A (18.9%) (p < 0.001). Moreover, there was a significant difference in patients undergoing mechanical ventilation and/or intensive care unit (ICU) care among the groups (p < 0.001).
As for postoperative care, the prevalence of receiving rehabilitation therapy for oral dietary intake was lower in Group A than in Group B (p < 0.001). Regarding the duration of antibiotic therapy, the mean durations of Groups A, B, and C were 9, 18, and 20 days, respectively (p < 0.001, Kruskal-Wallis test). Table 2 presents the characteristics of the patients with and without DNM. The non-DNM group had a younger interquartile range (41-73 years) than the DNM group (p < 0.001, Mann-Whitney U test). The prevalence of DM was significantly higher in the DNM group (54.55%) than in the non-DNM group (28.5%). Comorbid sepsis was also significantly higher in the DNM group (39.6%) than in the non-DNM group (10.1%). Moreover, there was a significantly higher prevalence in the DNM group of mechanical ventilation (65.45%) and/or ICU care (49.3%). Regarding in-hospital mortality, 8.4% of DNM group patients died in hospital, significantly more than in the non-DNM group (2.55%).

| Risk factors contributing to inhospital mortality
Logistic regression analysis showed that the following factors were significant (with crude odds ratios [

| Interval between admission and achievement of oral intake
The present study evaluated information obtained from a nationwide database, and it was difficult to assess Notes: Patients were categorized into three groups according to outcomes related to survival at discharge and prolonged interval for achieving oral intake, as follows: (A) alive at discharge with nutritional support for 9 days after admission; (B) alive at discharge with nutritional support for more than 10 days after the first operation; and (C) died in hospital. Abbreviations: BMI, body mass index; No., number of patients.
a Chi-squared test, conducted for other than unknown cases for each characteristic. **Kruskal-Wallis test at p < 0.001.
whether patients had been on nasogastric tube feeding for many years before admission and whether patients were discharged after gastric tube removal. Of the 4791 patients who were alive at discharge (Groups A and B in Table 1), 36 underwent tube feeding until the day of discharge. Moreover, these 36 patients included all five patients who underwent nasogastric tube feeding throughout their admission. Regarding the other 4755 patients, they did not undergo tube feeding or had stopped before the day of discharge, and they were thus included in the analyses. Table 4 shows the distribution of intervals between admission and achievement of oral intake in the 4755 patients who were alive at discharge (Groups A and B in Table 1) categorized by DNM, which were subgroup analyses without Group C. More than half of the 4261 patients without DNM (56.5%) achieved oral dietary intake <10 days after admission. Conversely, the rate was only 18.2% in the 494 patients with DNM. In addition, 56.5% of the patients with DNM could not achieve oral dietary intake in 20 days.
In addition, logistic regression analysis showed that sex, obesity, and smoking did not differ according to the delay in oral dietary intake (

| DISCUSSION
In this study, a total of 4949 patients throughout Japan who underwent drainage surgery for DNIs were investigated using a Japanese nationwide inpatient database. To the best of our knowledge, this is the first study to investigate the factors affecting mortality and delay in oral dietary intake in patients with DNIs, including DNM, in a nationwide clinical setting. The present study was also unique in that it compared these two outcomes by logistic regression analyses using the same clinical risk factors as independent variables. Interestingly, a few factors were associated with both in-hospital mortality and delay in oral dietary intake. Comorbid DM and sepsis were significantly associated only with in-hospital mortality. Conversely, several factors (DNM, repeated surgical interventions, ICU admission, and rehabilitation for oral dietary intake) were found to be significantly related only to delayed oral dietary intake. Advanced age has been reported to be a risk factor for developing DNIs. 25,26 The present study showed that age ≥75 years was significantly associated with both mortality (adjusted OR 5.57, 95%CI 2.8-11.1) and longer recovery of oral dietary intake (adjusted OR 1.88, 95%CI 1.51-2.33). The presence of systemic disease was the most important predisposing factor for the severity of DNI. Of them, comorbid DM is a well-known risk factor. The present study showed that DM was correlated with mortality, with an adjusted OR of 2.47 (95%CI 1.69-3.62). These results are consistent with a recent systematic review and meta-analysis, showing that DM was associated with a higher prevalence of multi-space spread of infection, complications, and failure to identify pathogens, with risk ratios of 1.96, 2.42, and 1.29, respectively. 22   is that this factor was adjusted by other variables, because the crude OR of this factor was 2.59 (95%CI 2.27-2.94). Similar to sex, obesity evaluated by BMI was relatively uniform among the three groups. Specifically, the prevalence of patients with obesity (BMI ≥30 kg/m 2 ) was 4-5% in all groups, and it was not a risk factor for either mortality or delay in oral dietary intake. Although the relationship between the severity of DNIs and obesity has yet to be determined, a previous report showed that obesity (BMI ≥30 kg/m 2 ) was correlated with longer hospitalization. 27 The DPC database has missing data for evaluating BMI in 7% (337/4949) of cases. In the DPC data, the codes corresponding to each surgery, clinical procedure, and medication are almost complete because they are compulsory items for reimbursement of health care costs defined in a central system in Japan to check the adequacy of the data (All-Japan Federation of the National Health Insurance Organization). However, other information submitted with claims that is not directly related to charges may be missing, such as height/weight and smoking index. 28 These less accurate data may also be responsible for the failure to identify obesity and smoking as risk factors for both mortality and delay in oral dietary intake in the present study. The smoking index was missing in 15% (754/4791) of cases in the DPC data, similar to the previous study. 28 The mortality rate for all patients with DNIs was 3% (158/4949), and that for patients with DNM was 8% (46/550). The latter rate is consistent with recent metaanalyses 6,7 demonstrating mortality of 9-10% for the combined cervical and transthoracic approach compared to mortality of 47-50% for cervical drainage only in cases of infection to the inferior mediastinum. Interestingly, comorbidity with DNM was not found to be a significant factor for mortality, with an adjusted OR of 1.14 (95%CI 0.76-1.73). These results could be attributed to the dissemination of strategies in Japan for transthoracic drainage in patients with DNM with extension below the tracheal bifurcation. 11,29 In the present cohort, 28% (154/550) of the patients with DNM underwent transthoracic drainage. The other 72% (396/550) of patients underwent incisional or intra-pharyngeal drainage. In contrast to the results for mortality, the present study showed that DNM was a significant factor for the delay in oral dietary intake, with an adjusted OR of 1.41 (95% CI 1.04-1.92). In addition to severe infection, systematic debridement and broad opening of the involved fascial planes are considered to contribute to limited laryngeal elevation and/or severe scar contraction specifically around the cricopharyngeal muscle. 11 Therefore, patients with DNM should receive postoperative care with careful attention to the risk of delayed recovery of oral intake.
Of the systemic diseases contributing to mortality, several previous reports addressed death due to complications with sepsis. 25,[30][31][32][33] In contrast to the outcome related to DNM, the present study showed that sepsis was a significant factor only for mortality, with an adjusted OR of 3.32 (95%CI 2.29-4.82), not for a delay in oral dietary intake, with an adjusted OR of 1.15 (95%CI 0.88-1.51). Therefore, sepsis, rather than DNM, should be considered a risk factor for mortality in treating DNIs and DNM.
In the present study, tracheotomy contributed only to a delay in oral dietary intake, with an adjusted OR of 1.69 (95%CI 1.44-2.00), which was similar to the result for DNM. Previous reports advocated tracheotomy for DNI cases with severe airway obstruction and/or presumed difficult re-intubation specifically managed by otolaryngologists. 6,[34][35][36][37] These results are consistent with our previous study and imply that a disturbance in swallowing function may be encountered after tracheotomy, presumably resulting from desensitization of the larynx after diversion of the air passage or fixation of the larynx. 11,38 It is plausible that undergoing repeated surgery is related to a delay in oral dietary intake, with an adjusted OR of 1.71 (95%CI 1.17-2.49). Interestingly, this factor had the opposite effect on the risk for mortality, with an adjusted OR of 0.60 (95%CI 0.35-1.05). Singhal 39 reported that CT imaging of the neck and chest should be performed with any clinical deterioration of the patient or empirically 48-72 h after an operative drainage procedure to identify any progression of the infection. Because initial drainage is often inadequate in patients specifically with DNM, the present results suggest surgical interventions with repeated drainage. However, these patients should receive postoperative care with careful attention to the risk of delay for recovering oral intake.
Postoperative care has been reported to play a critical role in patients with DNIs and DNM. 5,31,40 Appropriate and careful selection of patients for ICU admission is sometimes fundamental for these patients to manage severe sepsis and/or septic shock and every possible complication, both at the beginning and after surgery. 31,40 In the present study, ICU admission was not found to be a crucial factor related to mortality. One hypothesis for these results is that the above-mentioned critical care might contribute to the survival of patients with severe morbidities.
In contrast to the above-mentioned risk factor, mechanical ventilation was a significant risk factor for both mortality and delay in oral dietary intake, with adjusted ORs of 3.96 (95%CI 2.51-6.23) and 1.91 (95%CI 1.52-2.40), respectively. One background hypothesis is that the deteriorated conditions of the patients led to endotracheal intubation in the first place. According to the previous reports, most cases with DNM showed general deterioration and received support from an artificial respirator for an extended period. 2,40 Similar to the results for mechanical ventilation, the duration of empirical antibiotic therapy (days) was also a significant risk factor for both mortality and delay in oral dietary intake, with adjusted ORs of 1.00 (95%CI 1.00-1.02) and 1.18 (95%CI 1.17-1.20), respectively. Along with surgical drainage, high-dose intravenous antibiotics are the mainstay for the management of DNIs. These results correspond to the longer duration of intravenous antibiotic therapy in more critically ill patients with infection.
Finally, if patients can recover from their severely ill condition, rehabilitation for dysphagia might be indicated. Although only a few reports have addressed rehabilitation for DNIs or DNM because of the absence of detailed swallowing evaluation, 11,41 the present nationwide study showed that approximately 10% of patients with DNIs received rehabilitation therapy. This was a significant factor related to a delay in oral dietary intake, with an adjusted OR of 2.04 (95%CI 1.43-2.91). These results correspond to the situation where the right person (suffering from dysphagia) is in the right place (undergoing rehabilitation).
Several limitations of this study need to be acknowledged. First, this study was based on a retrospective cohort study using a national Japanese database, and generalization of the results outside Japan may not be appropriate. Second, comorbidities are less accurately recorded in administrative claims databases than in planned prospective studies. Third, the absence of records on vital signs, blood tests, and blood and bacteriological cultures in the DPC database precluded a more rigorous definition of septic shock, as discussed in the previous report. 13 Moreover, DPC data come from an inpatient database, and it is difficult to evaluate the delay from onset to intervention.
Within these limitations, the current study with the largest retrospective cohort using a nationwide database has several advantages. It was found that clinical risk factors differed between mortality and delay in oral dietary intake other than age ≥75 years, sepsis, mechanical ventilation, and duration of empirical antibiotic therapy. Other than these factors, DM contributed only to mortality. Conversely, DNM, repeated surgery, tracheostomy, and intensive care unit admission contributed to a delay in oral dietary intake, but not mortality. Further research including multicenter prospective studies of DNM and DNI is necessary to establish precise therapeutic approaches for managing DNIs and the optimal treatment for DNM.