Data source

In this study, the inpatient medical claims data between January 1997 and December 2008 were retrieved from the Taiwan's National Health Insurance Research Database (NHIRD). NHIRD was established by the Taiwan's National Health Research Institute to enhance medical research. The NHIRD have has been used extensively in various published studies[19]. The database comprises detailed information about clinical visits for each insured subject, including diagnostic codes according to the clinical modification of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9CM) code. The confidentiality of individuals was protected using encrypted personal identification to avoid the possibility of the ethical violations related to the data. Exemption was obtained from the institutional review board of Chi Mei Medical Center.

Patient selection and definition

The inpatient medical claims data during 2000–2007 was analyzed in this study. TBI patients were defined as those claiming medical expenditure applications for TBI surgery. The patients with ESRD were identified using the catastrophic illness certification records, as patients who underwent hemodialysis for more than three months could apply for a catastrophic illness card. A total of 74226 TBI patients received surgery, and of those, 1104 patients with ESRD were selected as the study sample. We then randomly selected 3312 TBI patients without ESRD (three for every patient with TBI) who were matched with the study cohort in terms of age and gender. The disease records for these patients were consistent with the format of ICD-9CM code. The co-morbidities of hypertension (HTN), diabetes mellitus (DM), myocardial infarction (MI), stroke, and heart failure (HF) were based on the records from one year before the date of TBI diagnosis. HTN was coded by hypertensive disease (ICD-9CM: 401-405), hypertensive encephalopathy (ICD-9CM: 437.2), and hypertensive retinopathy (ICD-9CM: 362.11). The DM was coded as follows: diabetes mellitus (ICD-9CM: 250), polyneuropathy in diabetes (ICD-9CM: 357.2), diabetic retinopathy (ICD-9CM: 362.0), and diabetic cataract (ICD-9CM: 366.41). MI was coded by acute myocardial infarction (ICD-9CM: 410) and old myocardial infarction (ICD-9CM: 412). Stroke was coded by cerebrovascular disease (ICD-9CM: 430-438). HF was coded by heart failure (ICD-9CM: 428), rheumatic heart failure (ICD-9CM: 398.91), acute myocarditis (ICD-9CM: 422), cardiomyopathy (ICD-9CM: 425), hypertensive heart disease with heart failure (ICD-9CM: 402.01, 402.11, and 402.91), and hypertensive heart and chronic kidney disease with heart failure (ICD-9CM: 404.01, 404.03, 404.11, 404.13, 404.91, and 404.93).

Measurements

The main event in this study was one-year mortality as TBI patients with ESRD had short survival after surgery. Survival time was calculated from the date of TBI surgery to the date of death or to one year after admission. Demographic and clinical characteristics, including age, gender, level of urbanization[20], DM, HTN, MI, stroke, heart failure, length of stay (LOS), length of ICU stay, length of ventilation (LOV), and tracheostomy, were also used to estimate the risk of one-year mortality. For stratified analysis, patients were further classified into three groups according to their length of ICU stay: shorter group (first quartile), in-between group (first quartile to third quartile), and prolonged group (third quartile). This approach has been used on some previous studies for the distribution of ICU length of stay [21], [22].

Statistical analysis

Comparisons between TBI patients with/without ESRD were performed using Pearson's chi-square test or Fisher's exact test for categorical variables and the Student's t test or Wilcoxon rank-sum test for continuous variables. The Kaplan-Meier plot was constructed to describe the proportion of patients who were still alive after one year, and the log-rank test was used to compare the mortality risk difference between two subgroups. The Cox proportional regression model was applied to estimate the hazards of mortality adjusted by potential confounding factors. To test the hypothesis that for TBI patients, the effect of pre-existing ESRD on mortality varied by age and the length of ICU stay, the model was also stratified by age and the length of ICU stay group to evaluate the potential risk factors. A p-value <0.05 was considered significant. All analyses were performed using Statistical Analysis System (SAS) statistical software (version 9.3; SAS Institute, Inc., Cary, NC). The Kaplan-Meier curves were plotted using STATA (version 12; Stata Corp., College Station, TX).

High mortality rate was observed in TBI patients with ESRD who underwent surgical intervention compared with patients without ESRD

Several studies have reported that acute kidney injury may develop due to renal ischemia following hypoperfusion as a complication of traumatic brain injury [23], [24]. However, pre-existing ESRD effects on TBI patients were not evaluated. Our finding indicated that TBI patients with ESRD had a higher percentage of hospitalization and one-year mortality, ICU days, length of ventilation, and comorbidities than patients without ESRD. The hazard ratio of one-year mortality, which was adjusted by age, gender, and clinical variables for TBI patients between those with ESRD and those without ESRD, is 2.85 (95% C.I.: 2.56–3.17). The results provide evidence supporting a new concept that pre-existing ESRD is a risk factor of mortality after TBI surgery. In addition, ESRD is associated with increased mortality among patients who are admitted to the ICU, and this effect is mostly a result of comorbidity [10], [25]. Our results are also consistent with previous reports that ESRD patients have a greater incidence of comorbidities such as diabetes, stroke, hypertension, heart disease than patients without ESRD. This may support that the conclusion that ESRD patients use a significantly greater amount of resources than do non-ESRD patients [26], [27]. Therefore, the investigation of the impact of neurosurgical interventions on healthcare costs and economic burden in dialysis patients, who are a growing population with high co-morbidity, is a very important issue for the future.

In addition, our results are consistent with those of Chapman et al. [28], who reported that the majority of deaths occurred within the first month of ICU admission, depending on the medical etiology. In the current study, we found that the majority of deaths occurred at 0.86 months (IQR: 0.26–8.91), with a one-month post-ICU admission mortality rate of 52.36% (49.44%–55.33%). In our study, the one-year mortality rate of 76.63% (74.10%–79.08%) was higher than the previously reported 40% to 65% for critically ill ESRD patients [7], [29], [30]. Our results are also consistent with those of Rhodes et al. and Dhanani et al. who indicated that patients who had pre-existing ESRD after major surgery had a higher mortality than patients without ESRD [8], [31]. Furthermore, in the current study, the one-year mortality trend test from 2000 to 2007 between TBI patients with ESRD and TBI patients without ESRD revealed significant differences (76.63% vs. 29.77%). These results imply that despite aggressive surgical intervention, the one-year mortality remained high and stable (76.63%). Thus, patients with pre-existing ERSD who suffer TBI and require neurosurgical intervention should be considered to have a serious and critical condition. This important information can be useful to surgeons, families, intensivists and nephrologists when considering surgical intervention and defining goals for treatment.

Studies have shown that hypertension, DM, MI, stroke, and heart failure are associated with higher mortality in TBI patients [32], [33]. However, how these comorbidities affect the prognosis in TBI patients with ESRD is still unknown. This is the first study to examine the effect of comorbidities on mortality among TBI with ESRD after controlling for potential risk factors. With the insult of ESRD, hypertension, DM, MI, and heart failure are not significantly associated with higher mortality for TBI patients.

Pre-existing stroke is a risk factor of mortality for ESRD patients with TBI who underwent surgical intervention

The overall mortality of spontaneous intracerebral hemorrhage in ESRD patients was 61% to 71.4%, and the mortality rate after stroke was 46.6% at one month and 64.3% at one year [17], [18]. Thus, stroke is an important risk factor of mortality for ESRD patients. In addition, the study by Thompson and colleagues demonstrated that the cerebrovascular disease had high association with mortality [32]. Our findings also indicate that pre-existing stroke is a significant risk factor for ESRD patients with TBI who underwent surgical intervention. Among these patients, those with preexisting stroke displayed a 29% increase in mortality risk compared with those without preexisting stroke (HR: 1.29, 95% C.I.: 1.08–1.55).

However, we believe that the effects of different types of intracranial hemorrhage, such as epidural hemorrhage, subdural hemorrhage, and intracerebral hemorrhage, on TBI mortality rate deserve to be investigated in the future. Therefore, we recommend the link between pre-existing stroke and TBI, effects of severity, subtypes, and external causes of TBI should be examined in future studies.

Pre-existing DM and MI effects on the mortality in ESRD patients who underwent TBI surgical intervention in the younger (age<65) and older (age≥65) population, respectively

Age has long been recognized as an independent adverse outcome predictor following TBI in various studies. Older age is associated with poorer outcome in patients with TBI [34]–[36]. Consistent with previous studies, we found the long-term mortality rate starting from one month (60.45% vs. 65.66%) to one year (73.45% vs 82.32%) after TBI in ESRD patients was significantly different between patients <65 y/o and patients ≥65 y/o. We also found significantly elevated one-year mortality rates in the older dialysis patients compared with the younger patients after TBI surgery (HR: 1.21, 95% C.I.: 1.05–1.39). These findings are consistent with the US Renal Data Service annual report that indicates that among dialysis patients age 65 and older, the mortality is twice as high as that in the general population [37]. Given that societal aging may increase the number of surgical cases of ESRD with TBI and that the older population presents a high mortality risk, neurosurgeons should be more cautious when performing surgical intervention in older ESRD patients if TBI has occurred.

Currently, diabetic nephropathy is the leading cause of ESRD in Germany and Taiwan [27], [38]. Many studies have reported an association between pre-existing DM at the initiation of dialysis and a poor outcome in ESRD patients undergoing dialysis [38], [39]. We further elucidated that for TBI patients with ESRD, pre-existing DM is a risk factor of one-year mortality (HR: 1.35, 95% C.I.: 1.12–1.63) for patients <65 years of age. One possible hypothesis is that younger ESRD patients have a longer follow-up related to DM, so a poorer health condition leads to a higher mortality risk for those who undergo surgical intervention. We believe that the management of TBI patients with ESRD derived from DM will represent a serious problem in the future.

In addition, pre-existing MI was associated with a significantly increased risk of mortality (HR: 1.61, 95% CI: 1.07–2.42) in the older population (age≥65). The reason for this result might be that pre-existing MI in older patients is a key factor to determine mortality compared with other comorbidities. Because we evaluated neither the time interval since MI nor the severity of MI, which would influence cardiac function and mortality, we believe this information encourages us to pay more attention to these new high-risk groups during the decision-making process for surgical intervention.

Length of ICU stay could be a mortality predicator and a surrogate of disease severity for ESRD patients who underwent TBI surgery

Previous studies found that the TBI patients with prolonged ICU stays have greater disease severity or poor functional outcome according to the different scores [40]–[44]. Since the TBI patients were unstable after TBI surgery at the beginning of ICU stay, and longer ICU lengths of stay might be associated with greater disease severity, examining the risk factors for mortality by different lengths of ICU stay among TBI patients is meaningful. Results of stratified analysis indicated that patients with different lengths of ICU stays have different risk factors for mortality: for patients with short ICU stays, age was an independent risk factor for mortality, and it was similar as former studies that age is the most frequently documented risk factor [45], [46]; for those with prolonged ICU stays, HTN was a protective factor, and this “ reverse epidemiology” had also been reported on the some studies for ESRD patients [47], [48]; for in-between groups, stroke and LOV were associated with increased risk of death, but tracheostomy was associated with lower risk.

One reason for tracheostomy need may be related to general clinical practices and the general consensus of neurosurgeons that tracheostomy is not required for patients who have a critical condition and less opportunity to survive. The previous studies indicated that the benefits of tracheostomy have been demonstrated for critically ill patients, especially for patients with limited neurological recovery, without spontaneous ventilation, excessive secretion, and long-term functional outcome [49]–[51]. However, the association between the tracheostomy and mortality is still unclear. A meta-analysis review indicated that the tracheostomy could reduce the risk of mortality, but it is no significant [21]. Only one study demonstrated that tracheostomy could significantly reduce mortality [52]. In our study, the effect of tracheostomy on mortality presented the protective information in the different ICU stays, only the in-between ICU stays group of TBI patients with ESRD who received tracheostomy displayed a 32% reduction in the one-year mortality rate compared with those who did not receive tracheostomy (HR: 0.68, 95% C.I.: 0.49–0.93). Since trauma patients are unstable at short length of ICU stay, tracheostomy may not be applied according the guideline[49]–[51]. Thus, it could not be significant for protecting the mortality. Considering of the severity of TBI, TBI patients with ESRD within prolonged ICU days may have the poor functional outcomes. The non-significant protective effect on mortality is consistency as previous studies [21], [22], [41]. Therefore, the tracheostomy may still have some benefit for some restricted patients. This information can be provided to the families of patients if they have doubts about the tracheotomy procedure in ESRD patients who underwent TBI surgery.