Outcome analysis of burns patients after admission to burns intensive care unit in a tertiary regional referral centre
Goh SY, Thong SY, Win MTM, Ng SY
Background: The clinical course of severely burned patients may be stormy and the prognosis tends to be poor in patients with multiple comorbidities and those with inhalational injury. The aim of this study is to develop an objective and reliable predictive model for mortality in patients with major burns. This will help us identify the important factors influencing outcomes and allows more evidence-based prognostication.
Methods: Adult patients admitted to the burns intensive care unit (BICU) in a major tertiary referral center from 2008-2011 are selected. Demographic factors, types, severity and complications of burn injury as well as outcomes are reviewed.
Results:In the 4-year period, 181 patients were admitted to BICU. Mean age (SD) was 41 (16) years old. Mean (SD) total body surface area burn was 37.2 (30.2%). Mortality was 39.5%. Mean (SD) length of stay in the BICU and hospital for patients who eventually survived were 8.4 (13.4) and 28.5 (37.9) days respectively.
Lower airway burns has a significant relationship with the development of renal failure after multivariate analysis (Odds ratio 5.1, Confidence interval 1.1- 24.0). Greater total body surface burns, development of acute respiratory distress syndrome and older patients with more extensive burns predispose to mortality as shown in table 2.
In our cohort of patients, the probability of death may be estimated by this equation:
Probability of death= (1+ey)-1
y= -7.008+0.04(TBSA) +1.791(ARDS)*+0.054(Age+TBSA)
*= ARDS (0=no, 1=yes)
We have developed a predictive model for mortality in major burn patients. This may be useful in prognosis during early stages of care.
Survival after burns injury has improved tremendously over the last few decadeswith the refinement of fluid resuscitation, better intensive care and early surgical excision1as some of the strategies that have significantly influenced patient outcomes. These advancements have contributed to lowering mortality rates in burns patients in Singapore to 4.5% between 2003-20052. Despite these advancements, however, overall mortality rates of patients with major burns remain high. Numerous factors such as age, percentage body surface area burns and inhalational injuries3 have been found to influence the prognosis and outcomes in this group of patients. The combination of these predictive factors into scoring systems that would yield an expected mortality rate for each given patient has been the subject of many studies4-6.However, most of these studies have not been performed or validated in our local population.
A robust predictive model would be useful for clinicians as a more evidence-based approach for counselling and prognostication at an early stage of treatment. We can even plan further treatment and intervention based on prognosis and other clinical factors. A standardized model will also provide an opportunity for audit and a basis against which new treatment modalities may be compared.
Therefore, the objective of this study was to identify the prognostic variables influencing outcome in patients admitted to our burns intensive care unit and to develop a predictive model for mortality in patients with major burns.
The burns centre at the Singapore General Hospital is a major tertiary referral centre for burns injury in Singapore as well as the Southeast Asia region. Burn patients presenting at the Accident and Emergency department are assessed by the plastic surgerical team, who decide if the patient requires admission to the specialised burns unit or the burns intensive care unit. The extent and depth of burns were assessed and documented.All patients requiring intensive care, such as those with major burns, are haemodynamically unstable, or have sustained inhalational injury requiring mechanical ventilation, are managed by a team consisting of at least a plastic surgical specialist and an anaesthetist.
These patients received fluid resuscitation according to the Parkland’s formula. Adequacy of fluid therapy was assessed by endpoints such as hourly urine output, arterial blood pressure and central venous pressure. Early enteral nutrition, mechanical ventilation and vasoactive support were initiated as required. Early surgery for escharotomy, burns excision and grafting were carried out as early as possible.
This study was approved by our centre’s institutional review board. The medical records of all burns patients admitted to the burns intensive care unit at the Singapore General Hospital over a 4-year period between January 2008 and December 2011 were reviewed retrospectively. Information including demographics, comorbidities, mechanism of injury, total body surface area (TBSA) burned, incidence of inhalation injury, complications such as organ failure, length of hospital stay and mortality were recorded and entered into a database.
Statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS version 17, SPSS Inc., Chicago, IL). Data are presented as mean and standard deviationfor continuous variables and proportions for categorical variables. Univariate comparisons of proportions and means were respectively done using Chi Square test, Fisher exact test and t test. Logistic regression and linear regression analysis was applied to study the relationship between the variables and mortality and length of stay, respectively. Variables with a plausible relationship with mortality or p< 0.05 in univariate analysis were included in the model. Linked variables were removed in a stepwise fashion if p >0.05.
During the 4-year study period, a total of 182 patients were admitted to the burns intensive care unit (BICU).These patient characteristics are presented in table 1. The mean age of these patients was 40.5 +/- 16 years old, with males comprising a larger percentage of the cohort (79.1%).
Mean total body surface area (TBSA) burned was 37.2 +/- 30.2%. Most of the patients were healthy prior to their burns, with only a small minority having any significant medical issues. A high proportion of the patients suffered inhalational burns (83%). However, only 26 out of the 182 patients (14.3%) fulfilled the criteria for ARDS. Other significant patient characteristics and details of their burns injuries are listed in Table 1. Flame burn was the commonest injury etiology in our patient population (88.4%). Table 2 shows the mechanism of burn injuries suffered by our group of patients.
Length of stay
The mean length of stay was 20.9 days. The survivors spent a mean of 8.4 days in ICU, as opposed to 6.9 days for the non-survivors. The majority of patients who survived to discharge were discharged home (table 3).
TBSA, albumin level and sepsis were found to be significant predictors of LOS, yielding a final predictive model of:
LOS = 41.608 – 0.234(TBSA) – 0.919 (albumin) + 16.14 (sepsis*)
Where *=presence of sepsis (yes=1, no=0)
Out of the 182 patients, a total of 65 patients or 35.9 % did not survive the hospital stay. The mean age of these patients was 42.1 years. Mean TBSA involved was 66.8% (compared with a mean of 20.6% in the survivor group). In our analysis, we found that the non-survivor group had significantly larger TBSA burns (p<0.001) and a higher incidence of ARDS (p<0.001) and renal failure (p<0.001). Renal failure was defined as acute kidney injury necessitating inpatient dialysis. (Table 4)
To develop the predictive model, we analysed the variables in univariate analysis. Variables with a plausible relationship with mortality or p<0.05 in univariate analysis were included in the logistic regression model. The following variables were entered in the model and were stepwise removed if p<0.05: diabetes, COPD, asthma, ischaemic heart disease, circumstances of injury, presence of lower airway burns, presence of ARDS, transfers from secondary hospitals, TBSA, age plus TBSA, highest creatinine, creatine kinase levels, albumin levels and presence of sepsis. In the final model, TBSA, ARDS and age+TBSA were found to be significant variables related to mortality.
A predictive model for mortality was thus developed as follows:
Probability of death = (1+ey)-1
Where y = -7.008 + 0.04 (TBSA) + 1.791 (ARDS)* + 0.054( Age+TBSA)
* = ARDS (no=0, yes =1)
In this study, we elucidated the risk factors that contribute to mortality in patients admitted to the BICU, and developed a predictive model for mortality incorporating these factors. Previous mortality studies have sought to establish prognostic variables associated with burn mortality. In 1961 Baux described in a French thesis a simple empiric formula, stating that mortality rate was the sum of age and percentage area burned7. Inhalational injury was then found to be an important predictive factor and this was included in a mortality probability equation reported by Clark et al in 19868. The abbreviated burn severity index (ABSI), which is in widespread use, assigns numerical values according to the severity of 5 prognostic variables (age, gender, inhalational injury, %TBSA and presence of full thickness burns). The sum of these variables yields a predicted mortality rate9. Similar predictive factors have been found to be relevant in several other studies3,10,11. However, these studies generated highly variable predictive models, highlighting the need for individualised models for different patient groups.
Our study is unique as it is one of only a handful of studies based on an Asian population2,12-14 , with even fewer attempting to develop predictive mortality models5. In addition, we only included burns intensive care unit patients in our study, as opposed to the entire burn population. Advancements in medical care and aggressive early excision and grafting of burns have led to a global reduction in burn mortality in recent times. In Singapore, the overall mortality for burns victims was found to be 4.5% between 2003 and 200512. This improvement has also been evident in severe burns victims, with mortality falling annually from 60% in 2000 to 30% in 20032. However, death rate remains high in this group of patients, and it is our aim to look at the outcomes and predictive factors determining mortality in this susceptible group, and to develop a clinically relevant predictive model targeted at them.
Our study found that TBSA, age+TBSA and ARDS were significant predictive factors affecting mortality in our ICU patients. Inhalational injury, generally accepted as a prognostic factor8,9, was not found to be significantly associated with death in our patient group. There could be several explanations for this finding. In our series, 151 or 83% of our patients were diagnosed with inhalational burns. This is a very high percentage compared to most of the other studies, though it is not entirely surprising. Our patient cohort consisted entirely of ICU patients. This unique group of patients have either sustained major burns or inhalational burns that have required ventilatory support. . The remaining 17% of patients that might have been in ICU due to reasons other than inhalational injuries may not have been sufficient in number to demonstrate a survival advantage, if any. Secondly the lack of universally accepted diagnostic criteria means that the diagnosis of inhalational injury can vary widely between different institutions and intensivists, using either clinical examination or fibreoptic bronchoscopy, or a combination of both. Hence it has been suggested that, the need for invasive ventilation, the determination of which is far less complicated than the diagnosis of inhalational injury, may be a better marker for mortality risk6. In our study, we found that the presence of acute respiratory distress syndrome (ARDS), a common complication associated with major burns or severe inhalational injury, was directly associated with death. This could be an indication that the severity of inhalational injury, rather than the presence of it, may be a more suitable prognostic indicator for death, particularly in the group of patients requiring intensive care management.
Since this is a retrospective study, we were not only able to collect data relating to the patient’s characteristics at presentation, but also susbequent clinical data as treatment progressed, such as development of ARDS, sepsis and renal failure. Clearly these factors are important as response to therapy is a vital determinant to patient outcomes. However, inclusion of these variables may not be possible at patient presentation, the point at which prediction of mortality is sometimes vital for determining the direction and aggressiveness of therapy. Perhaps future prospective trials can look at determining predictive factors that influence mortality at various stages of treatment, creating a superior prognostic tool with which we may advise patients and families, guide therapy and perform internal audit and research.
Clinicians utilising any kind of predictive model to estimate mortality probability in the clinical setting should always proceed with caution. These may be used as a tool to aid clinical decisions regarding treatment but should not replace sound clinical judgment. Neither should the outcomes be judged solely upon whether the patient lives or dies, without scant consideration for the quality of life after the hospital stay. These endpoints are a lot more difficult to assess, and in the absence of a standardized tool the incorporation of quality of life indicators into prognostic scoring systems may still be a long way away.
In our study, we have developed a predictive model for mortality in our cohort of burn patients admitted to the burns intensive care unit. The next step would be to validate the model in future prospective studies. A validated model can potentially help teams involved in tough clinical decisions to prognosticate and formulate treatment plans for severely burned patients. It also serves to show that further studies need to be done to validate and come up with a more robust model. We did not find a significant relationship between inhalational injury and mortality in our study, a finding consistent with several other studies.