Prediction of Surgical Outcome for Acute Traumatic Brain Injury in Older Adults
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.349603
Source of Support: None, Conflict of Interest: None
Keywords: Craniotomy, geriatric, outcome, subdural hematoma, traumatic brain injury, traumatic parenchymal lesion
Increasing patient age is strongly associated with a rising incidence of injury and a higher mortality and morbidity rates, mainly due to traumatic brain injury (TBI). The aging population trends currently observed in high-income countries, if extended into the low-middle income countries in the future, the global public health burden of TBI in older adults will increase significantly. In our institute's registry of TBI in 2005, there were 5.2% patients older than 60 years, and in 2015, it rose to 9.6%. With the current definition of the elderly defined as older adults aged ≥65 years, the proportion of older adults constitute 5.6% of cases of TBI at our hospital.
In developed nations, decisions for surgery are largely based on the established clinical and imaging indications like midline shift, hematoma thickness, and time between injury and treatment. Craniotomies are done for elderly patients for a head injury with optimism in developed countries. However, in developing countries with limited resources for perioperative treatment, decisions on surgical treatment in these patients are generally difficult and depend on operating surgeons' experience and the institutional and social background of patients. In our institute, there are differing opinions regarding indications for surgery for an elderly patient with TBI. Except for an unsurvivable patient, i.e., Glasgow coma scale (GCS) 3, nonreacting pupils, an absent brain stem reflex surgery is offered to the vast majority of elderly patients on compassionate grounds. In our uncontrolled observation, we have observed remarkably high mortality and severe disability rate in elderly patients with severe TBI who underwent craniotomy. This study aimed to identify the predictors of mortality after craniotomy for acute TBI in elderly patients.
This study was done at the National Institute of Mental Health and Neurosciences, a tertiary-level referral center for treating neurological, neurosurgical, and psychiatric patients. It has a separate facility for trauma cases that serve patients from Bangalore and neighboring districts. Patients with TBI, who have a different registration process, are evaluated by neurosurgery residents. Their data are entered in a structured “head injury pro forma,” which consists of comprehensive clinical and computed tomographic (CT) scan findings. A qualified neurosurgeon on duty verifies these data. Data of all patients aged ≥65 years who underwent craniotomy for acute TBI, over a period from January 1, 2015 to October 2019, were retrospectively reviewed. All these patients underwent head CT scanning. The CT scan findings were confirmed independently from images archived in the picture archival communication system by one of the senior residents of neurosurgery. The preoperative CT scan findings were used for analysis. Only those patients who underwent index surgery for acute TBI were included. The patients who underwent surgery for sequelae of head injury were excluded, e.g., chronic subdural hematoma, repair of traumatic cerebrospinal fluid leaks, and cranioplasty following a decompressive craniectomy. The standard clinical and imaging variables for TBI were recorded. The medical comorbidities, indication for surgery, and intraoperative complications were also recorded. The outcome of interest was survival at 6 months after surgery.
To determine the prognostic factors of survival at 6 months, in geriatric trauma patients who had undergone surgery, a total of 35 variables were considered. The continuous variables were tested for normality, in the survived and nonsurvived group. The ordinal variables with five or more categories can often be used as continuous without any harm to the analysis. The variable motor score was used as an ordinal approximation of a continuous variable.,, Since the variables did not satisfy the normality assumption, the analysis was performed with the nonparametric method. The association of categorical variables with the survival of patients was tested using the Chi-square test. Fisher's exact test was performed when the percentage of expected frequency was less than 5 in more than 20% of the cells.
The variables that showed significance in univariate analysis were considered for the prediction of survival in multivariable binary logistic regression. The effects were evaluated using logistic regression, after checking for the assumptions of dependent variables for linearity, multicollinearity, and heteroscedasticity. A backward likelihood ratio method with stepwise selection based on Akaike Information Criteria and P value was performed. All statistical analysis was done using SPSS ver 22 and R ver 3.6.1.
During the study period, a total of 7544 patients underwent surgery for acute trauma, of which 265 (3.5%) were older adults (age ≥65 years). The CT scan images were not available for two patients for review, and 6 months' outcome was not available for 57 patients leaving 206 (77.7%) patients for analysis. The age of patients ranged from 65 to 80 years. Only 32 (15.5%) of the patients were ≥75 years age [Table 1]. The commonest surgical procedure performed was craniotomy and evacuation of supratentorial subdural hematoma (SDH) with or without evacuation of the traumatic parenchymal lesion (contusion or intracerebral hemorrhage) in 129 (65.6%) of cases followed by craniotomy and evacuation of supratentorial extradural hematoma in 26 (12.6%), and craniotomy and evacuation of supratentorial traumatic parenchymal hemorrhage in 40 (19.4%) cases. The other surgical procedures performed were craniotomy for evacuation of posterior fossa hematoma in three cases, evacuation of supratentorial EDH and SDH in the same patient in three cases, and surgery for depressed skull fracture in five cases. Twenty-two (10.7%) patients also required a decompressive craniectomy for persisting brain bulge after the evacuation of the intradural hematoma. Twenty-five (12.1%) patients required reoperation mainly for the removal of new hematoma or infection [Table 2]. One hundred and four (50.5%) patients had one or more medical comorbidities; the commonest was diabetes mellitus in 54 (26.2%) cases [Table 3]. Only seven (3.4%) patients were taking antithrombotic. The in-hospital mortality was 46 out of 206 (22.3%), and six months mortality was 116 out of 206 (56.3%) [Figure 1]. Among the 90 survivors at 6 months, GOS was available for 81 (90%) patients. As the GOS was assessed telephonically, the category of GOS could not be concluded based on available information for nine patients. Good recovery was seen in 57 (70.5%), moderate disability in 16 (19.8%), and severe disability in 7 (8.6%) patients. Only one (1.2%) patient survived in a vegetative state at 6 months.
Mann–Whitney U test showed that, among clinical and CT scan features, GCS at admission and motor score were significantly different between the groups, whereas age was not [Table 1]. None of the patients with GCS 3 or 4 survived. Other variables significant were a pupillary reaction, basal cisterns, and traumatic SAH. Only one patient had bilaterally dilated nonreactive pupils; he did not survive. Hypotension before surgery, major extracranial injuries, and midline shift did not show any difference in this section. Among the surgical parameters, the main procedure for surgery and reoperation showed a statistically significant difference. Patients who underwent surgery for supratentorial SDH + traumatic parenchymal lesion had higher mortality. Six patients underwent reoperation for the evacuation of ipsilateral hematoma, and all six died. However, the location of surgery, decompressive craniectomy, and episodes of intraoperative hypotension were not associated with survival [Table 2]. For medications and medical comorbidities, the survival showed a significant association with the presence of medical comorbidity with diabetes mellitus showing a significant difference between the survival and nonsurvival groups, whereas medication did not show any association. The GCS score at discharge, motor score at discharge, and length of stay were the other variables that were significant.
The effect of GCS on admission, motor score, pupillary reaction, basal cisterns, traumatic SAH, medical comorbidities, GCS at discharge, and motors core at discharge taken together, on the survival of the patients, were evaluated using logistic regression. The only patient with bilaterally dilated pupil was clubbed with unilateral dilated pupil category for logistic regression. The final model showed that pupillary reaction and motor score contributed significantly toward the survival of the patients. In describing the multivariable logistic regression analyses, the raw regression coefficients (b) and their SEs are provided [Table 4]. The odds ratio (OR) and a 95% confidence interval (CI) for each significant effect are given in [Table 5]. From [Table 5], it can be observed that the odds of death are nearly three times more for patients with dilated pupil compared to a nondilated pupil, and the odds of dying are less by 72% for a unit increase in motor score.
The ORs calculated provide the odds of death with a P value <0.05 considered as significant. The significant predictors in logistic regression were pupillary reaction (b = 1.05 ± 0.48; OR = 2.86, 95% CI = 1.12–7.29) and motor score (b = -1.26 ± 0.28; OR = 0.28, 95% CI = 0.16–0.49), using which the Log Odds of death can be represented as
Log Odds of Death = 6.15 + 1.05×pupillary reaction -1.26×motor score
The probability of death for each combination of pupillary reaction and motor at discharge was calculated using the following formula:
The results are given in [Table 6]. It was observed that the probability of death was higher for patients with dilated and nonreactive pupil (code = 2) as compared to the nondilated and reactive pupil (code = 1). Similarly, as the motor at discharge increased, the probability of death decreased, keeping the pupillary reaction constant. The likelihood of death was least (0.411) in patients with nondilated pupil and a motor score of 6. All other combinations of the factors had a probability higher than 0.5, the highest being dilated pupil with motor score 1. The area under the curve (AUC) for our model was 0.801 with 95% CI 0.733–0.869. Based on AUC, we can say that that the model has 80% predictive power to discriminate survivors and nonsurvivor patients [Figure 2].
Elderly patients with TBI experience higher morbidity and mortality than do younger patients. There is often an assumption of the futility of surgery in these patients. However, a subset of older adults with TBI may recover well, suggesting that chronological age alone is inadequate prognostic markers. A Swiss group defined specific criteria in place that would limit craniotomy to elderly patients with acute SDH and Karnofsky performance scale score of 80 or more and living independently, no known dementia, no comorbidities that had a survival time of less than 12 months, and desire to proceed with surgery and consent to do so. In our institute, we do not have the criteria for surgery for elderly patients with TBI. We found that 6 months mortality for surgery for acute TBI in unselected elderly patients was 56.3%, which was extremely high. The 6 months mortality for surgically treated elderly patients has been reported from 16.7 to 62%.,,,,, The variability in mortality among centers is due to the selection of patients. The centers reporting lower mortality chose those patients for surgery who were more likely to survive. Another reason for this variability was the inclusion of only acute SDH, which is the most frequent traumatic mass lesion seen in elderly patients and has been associated with higher mortality compared to EDH.
The most frequent indication of surgery in our present study was acute SDH with or without other parenchymal lesions. Acute SDHs are the commonest traumatic mass lesions in elderly patients with TBI.,,, The mortality of acute SDH with parenchymal lesion was 71% as compared to 20–58% for other traumatic lesions in our study. The higher mortality in acute SDHs is reported in other studies as well.,,,
Many studies have reported prognostic factors for outcome after surgery for acute TBI in older adults.,,,,,,,,,,,,,, These factors are age, GCS, pupillary reaction, midline shift, basal cisterns compression, traumatic SAH, medical comorbidities, use of anticoagulants or antiplatelets, frailty, etc. In the univariate analysis, we also found the influence of following factors in the prediction of mortality: GCS, motor score, pupillary reaction, and presence of medical comorbidities, traumatic SAH, surgery for supratentorial SDH and traumatic parenchymal, and need for reoperation. Though some studies have found worse outcomes in very elderly patients, i.e., age >75 or 80, we did not find increasing age as a predictor of mortality.,,, We did not find the effect of the use of antithrombotic in determining mortality, as reported by others.
On multivariable logistic regression, we found that only pupillary reaction and motor score were significant predictors of mortality at 6 months. The accuracy of our model was 80%. The popular prognostic model International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) for outcome prediction may not be valid in the elderly population. Wan et al. found that actual outcomes in a cohort of elderly patients with severe TBI were slightly better than the expected results as predicted by the IMPACT model. Bobeff et al. developed a predictive model using logistic regression, proposed as Elderly Traumatic Brain Injury Score to predict 30-day mortality or vegetative state. The variables included in the score were motor score, platelet count, red blood cell distribution width coefficient of variation, and comorbidity. The model accurately predicted 88.8% of outcomes. However, their patients' population included nonoperated cases and chronic subdural hematomas that underwent only burr hole craniostomy. Alford et al. developed Subdural Hematoma in the Elderly score for prediction of 30-day mortality. The variables included age, GCS, and SDH volume. However, they covered all ranges of subdural hematomas, i.e., acute, subacute, and chronic.
Our model has strength in terms of a limited number of variables and simplicity. The accuracy of our model is good. This model may be applicable, particularly in developing countries, as the availability of perioperative care and neurorehabilitation is limited. Our model can triage patients for surgery and identify the patients who are unlikely to survive. Predictions from this model can be used for counseling the family or caretakers for making decisions regarding the need for surgery or for the continuation of care. In our study, we found that none of the elderly patients with GCS 3 or GCS 4 survived. Only one patient had bilaterally dilated pupils, and he did not survive. There may be case for aggressive management of young patients with these poor prognostic factors, but in elderly patients one may have to counsel the relatives before offering surgery.
The limitations of our study were retrospective analysis and non-inclusion of non-operative cases. We did not have information about medical complications after surgery. Though the elderly population is increasing even in developing countries, the older adults who undergo surgery are fewer, and an adequate number of patients for the prognostic model will require multicenter study over many years. The patients who do not undergo surgery are observed and referred to other hospital for medical management. Only operated patients are admitted; hence, we did not have information about nonoperative cases. We did not analyze age as categorical variable by dividing in different age groups. The age of our patient ranged from 65 to 80 years. In developed countries, there are many octagenerians. The studies on TBI in older adults from developed countries bifurcate age as ≤80 or >80, where they have found difference in outcome., We did not use functional outcomes for developing the model as the functional outcome was not available for all the patients. The GOS was not available for 10% survivors. The preinjury functional status was not documented in files of patients; hence, it was difficult to assess GOS at follow-up. Moreover, there is a lack of clarity on the use of standard outcome assessment tool GOS for elderly patients. Many older adults have functional limitations, and to attribute postinjury disability to injury-related factors is not accurate. A model that can predict meaningful survival would be more useful in practice. Unfortunately, in TBI, the prediction of unfavorable outcome is more accurate than the prediction of favorable outcome.
Declarations–Author BID is a member of National Institute for Health Research (NIHR) Global Health Research Group on Neurotrauma, UK. This research was commissioned by the National Institute for Health Research (NIHR) Global Health Research Group on Neurotrauma (project 16/137/105) using UK aid from the UK Government. The views expressed in this publication are those of the author (s) and not necessarily those of the NIHR or the Department of Health and Social Care.
It was a retrospective review of data and does not require approval from our institute's ethics committee.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6]