Surgical intervention and patient factors associated with poor outcomes in patients with traumatic brain injury at a tertiary care hospital in Uganda

Charis A. Spears BA1,3, Syed M. Adil BS1,3, Brad J. Kolls MD, PhD, MMCi1,2, Michael E. Muhumza MBChB, MMed4, Michael M. Haglund MD, PhD, MACM1,3,5,6, Anthony T. Fuller MD, MScGH1,3,5,6, and Timothy W. Dunn PhD1,5,7,8
View More View Less
  • 1 Duke University Division of Global Neurosurgery and Neurology, Durham;
  • | 2 Department of Neurology,
  • | 3 Duke University School of Medicine, Durham, North Carolina;
  • | 4 Department of Neurosurgery, Mulago Hospital, Kampala, Uganda;
  • | 5 Department of Neurosurgery, Duke University Medical Center, Durham;
  • | 6 Duke University Global Health Institute, Durham;
  • | 7 Duke Forge, Duke University School of Medicine, Durham; and
  • | 8 Department of Statistical Science, Duke University, Durham, North Carolina
Free access

OBJECTIVE

The purpose of this study was to investigate whether neurosurgical intervention for traumatic brain injury (TBI) is associated with reduced risks of death and clinical deterioration in a low-income country with a relatively high neurosurgical capacity. The authors further aimed to assess whether the association between surgical intervention and acute poor outcomes differs according to TBI severity and various patient factors.

METHODS

Using TBI registry data collected from a national referral hospital in Uganda between July 2016 and April 2020, the authors performed Cox regression analyses of poor outcomes in admitted patients who did and did not undergo surgery for TBI, with surgery as a time-varying treatment variable. Patients were further stratified by TBI severity using the admission Glasgow Coma Scale (GCS) score: mild TBI (mTBI; GCS scores 13–15), moderate TBI (moTBI; GCS scores 9–12), and severe TBI (sTBI; GCS scores 3–8). Poor outcomes constituted Glasgow Outcome Scale scores 2–3, deterioration in TBI severity between admission and discharge (e.g., mTBI to sTBI), and death. Several clinical and demographic variables were included as covariates. Patients were observed for outcomes from admission through hospital day 10.

RESULTS

Of 1544 patients included in the cohort, 369 (24%) had undergone surgery. Rates of poor outcomes were 4% (n = 13) for surgical patients and 12% (n = 144) among nonsurgical patients (n = 1175). Surgery was associated with a 59% reduction in the hazard for a poor outcome (HR 0.41, 95% CI 0.23–0.72). Age, pupillary nonreactivity, fall injury, and TBI severity at admission were significant covariates. In models stratifying by TBI severity at admission, patients with mTBI had an 80% reduction in the hazard for a poor outcome with surgery (HR 0.20, 95% CI 0.04–0.90), whereas those with sTBI had a 65% reduction (HR 0.35, 95% CI 0.14–0.89). Patients with moTBI had a statistically nonsignificant 56% reduction in hazard (HR 0.44, 95% CI 0.17–1.17).

CONCLUSIONS

In this setting, the association between surgery and rates of poor outcomes varied with TBI severity and was influenced by several factors. Patients presenting with mTBI had the greatest reduction in the hazard for a poor outcome, followed by those presenting with sTBI. However, patients with moTBI had a nonsignificant reduction in the hazard, indicating greater variability in outcomes and underscoring the need for closer monitoring of this population. These results highlight the importance of accurate, timely clinical evaluation throughout a patient’s admission and can inform decisions about whether and when to perform surgery for TBI when resources are limited.

ABBREVIATIONS

GCS = Glasgow Coma Scale; GOS = Glasgow Outcome Scale; HD = hospital day; HIC = high-income country; HR = hazard ratio; ICP = intracranial pressure; ICU = intensive care unit; LMICs = low- and/or middle-income countries; MNRH = Mulago National Referral Hospital; MOI = mechanism of injury; moTBI = moderate TBI; mTBI = mild TBI; RTI = road traffic incident; SSA = Sub-Saharan Africa; sTBI = severe TBI; TBI = traumatic brain injury.

OBJECTIVE

The purpose of this study was to investigate whether neurosurgical intervention for traumatic brain injury (TBI) is associated with reduced risks of death and clinical deterioration in a low-income country with a relatively high neurosurgical capacity. The authors further aimed to assess whether the association between surgical intervention and acute poor outcomes differs according to TBI severity and various patient factors.

METHODS

Using TBI registry data collected from a national referral hospital in Uganda between July 2016 and April 2020, the authors performed Cox regression analyses of poor outcomes in admitted patients who did and did not undergo surgery for TBI, with surgery as a time-varying treatment variable. Patients were further stratified by TBI severity using the admission Glasgow Coma Scale (GCS) score: mild TBI (mTBI; GCS scores 13–15), moderate TBI (moTBI; GCS scores 9–12), and severe TBI (sTBI; GCS scores 3–8). Poor outcomes constituted Glasgow Outcome Scale scores 2–3, deterioration in TBI severity between admission and discharge (e.g., mTBI to sTBI), and death. Several clinical and demographic variables were included as covariates. Patients were observed for outcomes from admission through hospital day 10.

RESULTS

Of 1544 patients included in the cohort, 369 (24%) had undergone surgery. Rates of poor outcomes were 4% (n = 13) for surgical patients and 12% (n = 144) among nonsurgical patients (n = 1175). Surgery was associated with a 59% reduction in the hazard for a poor outcome (HR 0.41, 95% CI 0.23–0.72). Age, pupillary nonreactivity, fall injury, and TBI severity at admission were significant covariates. In models stratifying by TBI severity at admission, patients with mTBI had an 80% reduction in the hazard for a poor outcome with surgery (HR 0.20, 95% CI 0.04–0.90), whereas those with sTBI had a 65% reduction (HR 0.35, 95% CI 0.14–0.89). Patients with moTBI had a statistically nonsignificant 56% reduction in hazard (HR 0.44, 95% CI 0.17–1.17).

CONCLUSIONS

In this setting, the association between surgery and rates of poor outcomes varied with TBI severity and was influenced by several factors. Patients presenting with mTBI had the greatest reduction in the hazard for a poor outcome, followed by those presenting with sTBI. However, patients with moTBI had a nonsignificant reduction in the hazard, indicating greater variability in outcomes and underscoring the need for closer monitoring of this population. These results highlight the importance of accurate, timely clinical evaluation throughout a patient’s admission and can inform decisions about whether and when to perform surgery for TBI when resources are limited.

In Brief

Neurosurgical capacity in Uganda is increasing rapidly, but the relationship between neurosurgical intervention and traumatic brain injury (TBI) outcomes in the country has not been quantified. Here, the authors reveal significant associations between neurosurgical intervention and reduced risk of acute poor outcomes for mild and severe TBI at a national referral hospital in Kampala, Uganda. Moderate TBI patient outcomes were more variable, underscoring a need for closer monitoring of this population.

Traumatic brain injury (TBI) affects an estimated 69 million people annually.1 Low- and/or middle-income countries (LMICs) suffer a disproportionate three times greater number of cases annually than high-income countries (HICs).1 This disparity is exacerbated by especially high mortality in LMICs, where the odds of death in patients with severe TBI (sTBI) are more than twice the odds in comparable patients in HICs.2 A major contributor to death and other poor outcomes in patients with TBI is secondary injury to cerebral tissue as intracranial pressure (ICP) increases.3 While patients in HICs may initially receive medical management for elevated ICP and undergo neurosurgical intervention in refractory cases, in LMICs, delayed presentation to hospitals with the capacity for neurosurgical care4 and specialized medical management often necessitates early decompression to prevent secondary injury.5 However, in Sub-Saharan Africa (SSA), where there is an estimated 1 neurosurgeon per 2.62 million people,6 the necessity for neurosurgical intervention greatly exceeds the available resources.

In Uganda, where collaborative efforts with providers from HICs have substantially increased neurosurgical capacity,7–9 there are still only 12 neurosurgeons serving over 43 million people.10 However, compared to the in-hospital mortality rate of up to 77% for patients with sTBI in other countries in SSA,11–14 one study of patients admitted in 2008–2009 noted a mortality rate of 26% at the Mulago National Referral Hospital (MNRH), which employed all but one of Uganda’s formally trained neurosurgeons.15 This rate was comparable to the 6-month mortality rate of 30% observed in HICs,2 possibly because of the relatively high neurosurgical capacity at this institution. Notwithstanding, of the 22% of patients with sTBI who had undergone surgery, only half survived,15 compared to a study in Malawi that documented a mortality rate of 7% among patients who had undergone exploratory burr hole procedures for moderate to severe TBI.12 This highlights the necessity for further investigation into the impact of surgery on in-hospital mortality and other outcomes in this setting.

We previously performed a survival analysis comparing clinical outcomes in patients who did versus patients who did not undergo surgical intervention for TBI, the first to our knowledge.16 That study, which used data from a tertiary care referral center in Moshi, Tanzania, showed that patients undergoing surgery had a reduced hazard for a poor outcome compared to that in nonsurgical patients. This association was observed to varying degrees depending on TBI severity. There were no in-house neurosurgery-trained physicians during the study period, and emergency burr hole procedures were most frequently performed. However, the MNRH has a higher TBI caseload and greater surgical complexity and overall neurosurgical capacity. Therefore, our objective in the present study was to investigate whether neurosurgical intervention is associated with a decreased hazard for TBI-related mortality and other acute outcomes in the relatively higher-resource—but still resource-constrained—setting of the MNRH. Further, we assessed whether this association differs according to TBI severity and explored the contributions of additional demographic and clinical factors. These data can aid providers in decision-making about triage, ongoing clinical evaluation, and allocation of scarce resources.

Methods

Study Design

In this retrospective, cross-sectional observational study, we performed Cox regression analyses to determine the association between surgery for TBI and the acute occurrence of poor outcomes among patients admitted to the neurosurgery ward at the MNRH. The hazard ratio (HR) provided by the Cox model measures the probability at any given time within a defined period that an individual will have a poor outcome after an intervention compared to an individual who does not undergo the intervention. This allowed comparison of the associations between surgery and the acute risk for poor outcomes in groups of patients with differing TBI severity. Published recommendations17–19 in performing survival analyses and STROBE guidelines20 were followed in reporting results.

Study Setting

We used data from the MNRH in Kampala, Uganda. The largest public hospital in Uganda, the MNRH treats approximately 1500 TBI cases annually, including patients from adjacent countries in East Africa. Neurosurgical procedures are performed by 5 neurosurgery consultants (the equivalent of attendings), 6 neurosurgery residents, and 2 casualty (emergency) department providers trained in neurosurgery.

The registry, which contains data collected since June 2016, includes over 60 variables regarding demographics, clinical assessment, diagnostics, and management collected throughout the hospital course.21 The database includes patients admitted to the neurosurgery ward and does not include those transferred directly to the intensive care unit (ICU) or managed nonoperatively or operatively in the casualty department.

Patient Population

Patients with TBI admitted to the MNRH neurosurgery ward between July 2016 and April 2020 were considered. Because our primary interest was acute TBI, patients with chronic subdural hematoma were not included. Patients were excluded if data were missing for admission date, discharge date, or whether or not they had undergone surgery for TBI, as these variables were used to calculate time to event, outcome, and treatment group, respectively (Fig. 1). Because each patient’s last recorded Glasgow Coma Scale (GCS) score was used to determine outcome in lieu of a Glasgow Outcome Scale (GOS) score, patients without a GOS score were excluded if they also did not have a GCS score recorded within 2 days of discharge (see Outcome below). Patients were also excluded if data were missing for key covariates, if surgery type was unspecified or unknown, or if there were clear errors in data entry for dates of admission, discharge, or death.

FIG. 1.
FIG. 1.

Flowchart for selection of study population. *Does not include chronic subdural hematoma.

Exposure and Explanatory Variables

Patients were divided into treatment groups depending on whether or not they had undergone surgery for TBI (i.e., the exposure variable). Surgery type was primarily decompressive (including craniotomy or craniectomy with or without hematoma evacuation) or nondecompressive (including debridement, defect repair, depressed skull fracture elevation) in nature. Surgery type was not included as a covariate because of complete separation in the patients with mild TBI (mTBI); that is, no poor outcomes were observed for any mTBI patients undergoing nondecompressive surgery.

Potential explanatory variables included age, sex, self-reported alcohol abuse (defined as > 2 alcoholic beverages per day) in the 2 weeks prior to the injury, mechanism of injury (MOI), TBI pathology per preoperative CT findings, pupillary reactivity, and TBI severity at admission. The categories for MOI and TBI pathology were consistent with those in our previous study16 and are widely used in the literature. “Other” TBI pathologies included intraparenchymal, intraventricular, and subarachnoid hemorrhage (all comprising fewer than 10 surgical patients); extracranial injuries; and unspecified pathologies (particularly when preoperative CT was unavailable). Because of the small sample size, the “unknown” category was combined with the “other” category in the regression analysis. As few patients had two nonreactive pupils, this category was combined with the “one pupil nonreactive” category in the regression analysis. Patients were stratified into TBI severity groups on the basis of their admission GCS score using the widely accepted classifications of mTBI (GCS scores 13–15), moderate TBI (moTBI; GCS scores 9–12), and sTBI (GCS scores 3–8).22

One potential confounder was that delays to surgery in this setting varied with TBI severity, which could differentially impact outcomes.4 However, a properly adjusted Cox model with surgery as a time-varying variable should control for the delay to surgery as a confounder because survival is assessed relative to all patients at risk at a given time. Nevertheless, we performed two sensitivity analyses for surgery delays: we fit a logistic regression model to measure associations between time to surgery and poor outcomes, and we included time to surgery as a time-varying covariate in the unadjusted Cox model for the entire cohort. Because delay to surgery was not significantly associated with poor outcomes in this study and did not yield a significant HR, we did not include it in the adjusted models.

Another potential confounder was the possibility for a change in TBI severity between admission and surgery to affect the likelihood of both surgery and a poor outcome. To address this, a sensitivity analysis was performed to assess changes in TBI severity from admission to the time of surgery, as estimated by the last recorded GCS score within 1 day of surgery (Supplemental Table 1). As most patients (81%, n = 298) had no change, change in TBI severity was not included as a covariate.

Outcome

A composite outcome measure was used. The GOS23 was dichotomized into good outcomes (GOS scores 4–5) and poor outcomes (GOS scores 1–3), consistent with our previous survival analysis16 and precedent in the literature.24 However, as the GOS score was available for only 45% (n = 692) of patients, the following were also considered poor outcomes: death (when not recorded as a GOS score), sTBI at both admission and within 2 days of discharge, or deterioration in TBI severity from admission to within 2 days of discharge (e.g., from moTBI at admission to sTBI before discharge). Conversely, improvement in TBI severity and maintaining mTBI or moTBI from admission to discharge were considered good outcomes. The 2-day window was chosen to prevent inaccurate representation of rates of outcomes due to an unrecorded change in TBI severity before discharge.

Statistical Analysis

Cox models with time-varying surgery variables were created using standards detailed in the literature,25 with hospital admission as the study start time. As surgery was introduced with a time delay relative to admission, it was included as a time-varying variable that only classified patients as surgical at the time of surgery (i.e., the surgery variable for a patient was changed from “0” to “1” only after surgery had occurred). A model comparing surgery to no surgery and a separate model adjusted for covariates were created, along with individual models for mTBI, moTBI, and sTBI. The surgery curves in the Kaplan-Meier plots were generated by grouping all surgical patients together at the start of the study.

Patients were observed for poor outcome events from admission until discharge, in-hospital death, or hospital day (HD) 10, whichever came first. As it is recommended that time-to-event analyses continue until 10%–20% of patients remain without an outcome,17 HD 10, the first day at which < 10% (actual: 9.6%) of patients remained, was chosen as our time cutoff. Patients without an outcome at HD 10 were not considered to have a poor outcome event (i.e., data were right censored).26 Data were also censored whenever a patient was discharged with a good outcome.

Demographic and clinical data were summarized with descriptive statistics. Chi-square tests were used to compare distribution of demographic variables between groups. The lifelines, pandas, NumPy, SciPy, and Matplotlib packages in the Python programming language version 3.6 (Python Software Foundation) and the ggplot2 package in R 3.5.3 (R Foundation for Statistical Computing) were used to perform data analysis and graphically depict data. All tests were two-sided, and statistical significance was set at p < 0.05.

Institutional Review Board

Ethics approval was obtained from the Duke University Health System in Durham, North Carolina, and the MNRH Research and Ethics Committee in Kampala, Uganda. As this was a secondary analysis of a de-identified data set, informed consent was not required from study participants.

Results

Study Participants

The selection of study participants is outlined in Fig. 1. From an initial cohort of 2305 patients, a final cohort of 1544 was included after the application of all exclusion and inclusion criteria.

Demographics and Clinical Characteristics

Table 1 summarizes patient characteristics. Twenty-four percent (n = 369) of patients had undergone surgery. Most patients were younger than 40 years of age (82%, n = 1261), were male (84%, n = 1304), and had been injured in a road traffic incident (RTI; 61%, n = 937). Only 49% (n = 181) of surgical patients had been involved in an RTI, compared to 64% (n = 756) of nonsurgical patients. The proportion of falls was greatest in patients with an age ≥ 50 years (26%, n = 36) and in patients with moTBI (12%, n = 52; Supplemental Table 2).

TABLE 1.

Demographic and baseline characteristics of 1544 patients with TBI

VariableNo. (%)p Value
TBI SurgeryNo TBI SurgeryAll Patients
No. of patients369 (24)1175 (76)1544 (100)
Age in yrs
 <1884 (23)223 (19)307 (20)0.15
 18–29156 (42)465 (40)621 (40)
 30–3972 (20)261 (22)333 (22)
 40–4925 (7)117 (10)142 (9)
 ≥5032 (9)109 (9)141 (9)
Sex
 Female47 (13)193 (16)240 (16)0.10
 Male322 (87)982 (84)1304 (84)
History of alcohol abuse
 No251 (68)814 (69)1065 (69)0.89
 Yes94 (25)285 (24)379 (25)
 Unknown24 (7)76 (6)100 (6)
MOI
 RTI181 (49)756 (64)937 (61)0.21
 Assault147 (40)288 (25)435 (28)
 Fall41 (11)131 (11)172 (11)
TBI pathology
 Subdural hematoma53 (14)89 (8)142 (9)<0.001
 Epidural hematoma165 (45)124 (11)289 (19)
 Other/unknown151 (41)962 (82)1113 (72)
Pupillary reactivity
 Both reactive341 (92)1077 (92)1418 (92)0.47
 One nonreactive20 (5)80 (7)100 (6)
 Both nonreactive8 (2)18 (2)26 (2)
TBI severity at admission
 Mild (GCS 13–15)227 (62)690 (59)917 (59)0.15
 Moderate (GCS 9–12)109 (30)336 (29)445 (29)
 Severe (GCS 3–8)33 (9)149 (13)182 (12)
Surgery type
 Decompressive298 (81)
 Nondecompressive71 (19)
Outcome
 Poor13 (4)144 (12)157 (10)<0.001
 Good356 (96)1031 (88)1387 (90)

Boldface type indicates statistical significance.

Fifty-nine percent (n = 917) of patients had presented with mTBI, 29% (n = 445) with moTBI, and 12% (n = 182) with sTBI. A similar distribution of TBI severity was observed among nonsurgical patients; however, 62% (n = 227) of surgical patients had presented with mTBI, 30% (n = 109) with moTBI, and 9% (n = 33) with sTBI. Overall, 81% (n = 298) of surgical patients had undergone decompressive surgery; this proportion constituted 79% (n = 179) of those with mTBI, 83% (n = 90) of those with moTBI, and 88% (n = 29) of those with sTBI.

Outcomes

The distribution of each outcome among surgical and nonsurgical patients by TBI severity is presented in Fig. 2. Ten percent (n = 157) of patients had a poor outcome, comprising 4% (n = 13) of surgical patients and 12% (n = 144) of nonsurgical patients. The mortality rate, which accounted for the majority of poor outcomes, was 9% (n = 139) in the overall cohort, 3% (n = 12) in surgical patients, and 11% (n = 127) in nonsurgical patients. Mortality was highest in patients with sTBI, reaching 39% (n = 58) in nonsurgical patients and 15% (n = 5) in surgical patients. Notably, no patients had GOS scores 2–3, 0.7% (n = 11) deteriorated in TBI severity, and 0.5% (n = 7) remained in the sTBI category. Among patients with a good outcome, many (40%, n = 612) had GOS scores 4–5, and several maintained mTBI or moTBI (36%, n = 552). The proportion of patients with GOS scores 4–5 was higher for surgical patients for all TBI severity groups.

FIG. 2.
FIG. 2.

Distribution of outcomes among the overall cohort and the cohort stratified by TBI severity. No patients had an outcome of GOS scores 2–3. Numbers of patients with poor outcomes were 13 for surgical patients, 144 for nonsurgical patients, and 157 for the overall cohort. *Indicates poor outcome.

Factors Affecting the Hazard for a Poor Outcome

The Kaplan-Meier curves for the overall population and each TBI severity group are presented in Fig. 3. The unadjusted and fully adjusted Cox models are summarized in Table 2. In the unadjusted model, surgery was associated with a significantly reduced hazard for a poor outcome (HR 0.41, 95% CI 0.23–0.72). After adjusting for other covariates, the HR was similar (HR 0.41, 95% CI 0.22–0.74).

FIG. 3.
FIG. 3.

Kaplan-Meier plots for overall cohort and each TBI severity group. Kaplan-Meier plots were generated for models that included surgery as a non–time-dependent treatment variable. All surgery patients were grouped together at the start of the study. Figure is available in color online only.

TABLE 2.

Unadjusted and fully adjusted Cox models for overall study population

VariableHR (95% CI)p Value
Unadjusted model (n = 1544)
 No surgeryRefRef
 Surgery0.41 (0.23–0.72)0.002
Fully adjusted model (n = 1544)
 Surgery
  No surgeryRefRef
  Surgery0.41 (0.22–0.74)0.003
 Age: 1-yr increase1.02 (1.02–1.03)<0.001
 Sex
  FemaleRefRef
  Male0.84 (0.53–1.32)0.44
 History of alcohol abuse
  NoRefRef
  Yes1.38 (0.98–1.95)0.07
  Unknown1.26 (0.73–2.17)0.41
 MOI
  RTIRefRef
  Assault0.91 (0.63–1.33)0.64
  Fall1.88 (1.22–2.90)0.004
 TBI pathology
  Subdural hematomaRefRef
  Epidural hematoma0.85 (0.47–1.54)0.60
  Other/unknown0.99 (0.63–1.55)0.96
 Pupillary reactivity
  Both reactiveRefRef
  One or both nonreactive4.73 (3.24–6.89)<0.001
 TBI severity at admission
  Mild (GCS 13–15)RefRef
  Moderate (GCS 9–12)1.70 (1.11–2.61)0.02
  Severe (GCS 3–8)3.20 (1.99–5.14)<0.001

Ref = reference.

Boldface type indicates statistical significance.

In the fully adjusted model, a 1-year increase in age was associated with a 2% increase in the hazard for a poor outcome (HR 1.02, 95% CI 1.02–1.03). Falls were associated with an increased hazard (HR 1.88, 95% CI 1.22–2.90). Patients with pupillary nonreactivity were nearly five times as likely to have a poor outcome as patients with both pupils reactive (HR 4.73, 95% CI 3.24–6.89). Patient sex, self-reported history of recent alcohol abuse, and TBI pathology were not significant covariates in the overall model.

TBI Severity and Poor Outcomes

Compared to patients with mTBI, the HR was 1.70 (95% CI 1.11–2.61) in patients with moTBI and 3.20 (95% CI 1.99–5.14) in patients with sTBI after adjusting for other covariates (Table 2). Table 3 summarizes the results of the Cox models for mTBI, moTBI, and sTBI. Patients with mTBI had an 80% reduction in the hazard for a poor outcome with surgery (HR 0.20, 95% CI 0.04–0.90), whereas patients with sTBI had a 65% reduction (HR 0.35, 95% CI 0.14–0.89). Patients with moTBI had a statistically nonsignificant 56% reduction in the hazard for a poor outcome with surgery (HR 0.44, 95% CI 0.17–1.17).

TABLE 3.

Cox models stratified by TBI severity at admission

VariablemTBI, n = 917moTBI, n = 445sTBI, n = 182  
HR (95% CI)p ValueHR (95% CI)p ValueHR (95% CI)p Value  
Surgery  
 No surgeryRefRefRefRefRefRef  
 Surgery0.20 (0.04–0.90)0.040.44 (0.17–1.17)0.100.35 (0.14–0.89)0.03  
Age: 1-yr increase1.04 (1.02–1.06)<0.0011.03 (1.02–1.05)<0.0011.02 (1.00–1.03)0.02  
Sex  
 FemaleRefRefRefRefRefRef  
 Male1.09 (0.39–3.07)0.871.44 (0.64–3.24)0.380.57 (0.28–1.16)0.12  
History of alcohol abuse  
 NoRefRefRefRefRefRef  
 Yes1.25 (0.61–2.57)0.541.51 (0.84–2.73)0.171.16 (0.66–2.03)0.60  
 Unknown0.48 (0.06–3.70)0.480.77 (0.22–2.69)0.681.54 (0.77–3.06)0.22  
MOI  
 RTIRefRefRefRefRefRef  
 Assault0.66 (0.30–1.47)0.311.04 (0.55–1.99)0.900.98 (0.53–1.80)0.95  
 Fall0.80 (0.29–2.18)0.662.39 (1.20–4.77)0.012.00 (0.96–4.16)0.07  
TBI pathology  
 Subdural hematomaRefRefRefRefRefRef  
 Epidural hematoma0.27 (0.07–1.12)0.070.60 (0.20–1.78)0.361.52 (0.65–3.57)0.34  
 Other/unknown.82 (0.31–2.17)0.690.64 (0.28–1.46)0.291.10 (0.56–2.17)0.79  
Pupillary reactivity  
 Both reactiveRefRefRefRefRefRef  
 One or both nonreactive12.81 (3.72–44.04)<0.0016.22 (3.33–11.62)<0.0013.94 (2.33–6.66)<0.001  

Boldface type indicates statistical significance.

Falls were associated with an increased hazard for a poor outcome only in patients with moTBI (HR 2.39, 95% CI 1.20–4.77). Age and pupillary nonreactivity were significantly associated with an increased hazard in all TBI groups. TBI pathology, sex, and history of alcohol abuse were not significant in any model.

Time to Surgery, Discharge, and Outcome

Supplemental Table 3 summarizes time to surgery and discharge for the overall cohort and stratified by TBI severity. The overall median time to surgery was 2.0 (IQR 1.0–4.2) days. Patients with sTBI had a median time to surgery of 0.9 (IQR 0.4–1.8) days compared to 2.1 (IQR 1.1–4.7) days for those with mTBI and 2.1 (IQR 1.0–4.2) days for those with moTBI. However, delay to surgery was not a significant covariate in an unadjusted Cox model for the entire cohort (HR 1.00, 95% CI 0.98–1.01).

The overall median time to discharge was 3.3 (IQR 1.6–5.7) days. Surgical patients had a median length of stay of 5.3 (IQR 3.6–7.7) days, compared to 2.6 (IQR 1.3–5.1) days for nonsurgical patients. The median time to outcome was 5.3 (IQR 3.6–7.7) days for surgical patients and 2.6 (IQR 1.3–5.1) days for nonsurgical patients.

Discussion

In this study, we used survival analyses to assess the associations between surgical intervention for TBI and acute morbidity and death in a low-resource setting. Building on previous work by evaluating outcomes in an LMIC with relatively high neurosurgical capacity, we determined the associations of various factors with the hazard for a poor outcome. These data may offer guidance regarding surgical triaging. This is key in LMICs, where CT imaging studies cannot be reliably obtained to inform decision-making and where resources to perform surgery are limited.

Overall mortality among our study cohort was low (9%), confirming previous results.4 At 15%, mortality among sTBI patients who had undergone surgery was much higher. However, this was an improvement over the previously reported 50% mortality rate among admitted sTBI patients who had undergone surgery in 2008–2009,15 possibly due in part to efforts to increase neurosurgical capacity and improve perioperative management at the MNRH.8,9

Differential Association of Surgery With Poor Outcomes

We confirmed previous results16 that surgery is associated with a reduced hazard for a poor outcome and that this association differs depending on TBI severity at admission. Notably, not all studies of TBI in LMICs have demonstrated a benefit for neurosurgical intervention; in logistic regression models built on data from 356 patients in Malawi, surgery was not a significant predictor of outcome.27 This discrepancy may reflect differences in surgical training, operative volume, resource limitations, and injury patterns across different regions of SSA.

In the present study, mTBI was associated with the greatest reduction in hazard with surgery, though the statistical significance of this finding is unclear given that the confidence intervals overlapped with those for the moTBI and sTBI cohorts (note that significant HRs calculated from a larger data set would likely be more accurate). This enhanced benefit for mTBI complements the results of our previous survival analysis, which yielded HRs of 0.17 for moTBI, 0.20 for mTBI, and 0.47 for sTBI in patients with an outcome within the first 3 days of admission, although the confidence intervals also overlapped for all three models.16 Conversely, sTBI patients had an HR similar to that of mTBI patients in this study. This result may be partly explained by differences in neurosurgical capacity, study populations, and statistical analysis methods between the two studies (e.g., availability of certain covariates within the databases and our inclusion of a time-varying surgery variable in the models in this study, precluding potential deflation of HRs).

Need for Close Monitoring of Patients With moTBI

The effect of surgery on the hazard for a poor outcome was smallest—but not clinically insignificant—for patients with moTBI. That this finding was not statistically significant underscores the variability in outcomes in this population.28 As an intermediate group, patients with moTBI are often grouped with those with mTBI or sTBI in the literature, and they have a variable clinical course. However, their status could rapidly change for the worse.29 Therefore, these patients should be closely monitored for early changes in the clinical picture that could indicate a need for surgery.

Ideally, moTBI patients would be imaged immediately upon presentation and in the case of clinical deterioration, as ICP monitoring is not readily available in this setting. However, while CT scans are eventually obtained for most patients, they are expensive and often inaccessible, leading to delays in imaging.4 A previous study performed at the MNRH showed a median time interval of almost a day between neurological evaluation and CT scanning for moTBI patients, significantly different from the median of approximately 12 hours for those with mTBI and a few hours for those with sTBI. Both mTBI and moTBI patients who died experienced significantly longer delays at this interval than patients who survived. However, the median time between CT and surgery, which was approximately 96 hours for mTBI, 36 hours for moTBI, and 12 hours for sTBI, was significantly different between the groups but was not associated with mortality.4 While delay to surgery was not associated with poor outcomes in this study, the similarity in overall time to surgery for mTBI and moTBI patients—taken together with the previous results—suggests that these groups are initially triaged similarly. However, moTBI patients have higher mortality rates, particularly when imaging is delayed,4 and have more variable outcomes after surgery. Therefore, a key target for improving outcomes in moTBI is earlier imaging, especially in patients with a lower GCS score. We are currently exploring inexpensive, noninvasive bedside imaging with a near-infrared device and noninvasive ICP monitoring as potential solutions to the issues of diagnosis and monitoring.

Other Risk Factors for Poor Outcomes

Looking at the entire cohort, we noted that other risk factors for poor outcomes—besides not undergoing surgery and having a more severe TBI—were increasing age, pupillary nonreactivity, and falls. The hazard of older age is consistent with findings in the literature30 and aligns with results from the aforementioned study in Tanzania.16 Falls were most common in the moTBI group, and when controlling for TBI severity, MOI was significant only in this group. Therefore, these patients may have suffered a more serious fall (e.g., from a ladder or stairs), or there may be an interaction with age or another covariate.

Pupillary nonreactivity is generally associated with poorer outcomes in TBI, especially when both pupils are nonreactive.31 Pupil reactivity is also included as a variable in two major TBI prognostic models, CRASH (Corticoid Randomisation After Significant Head Injury)32 and IMPACT (International Mission of Prognosis and Analysis of Clinical Trials).33 That this factor maintained its significance in all models confirms its importance as a prognostic factor in TBI. Importantly, neurosurgical intervention also maintained its significance across all groups after controlling for TBI severity, further highlighting the critical role of surgery as a modifiable positive prognostic factor in LMICs.

Study Limitations

While survival analyses are often employed to assess the efficacy of an intervention in randomized controlled trials, we performed a retrospective analysis in patients who were not randomized to surgery. Therefore, the causal effects of surgery can only be tentatively assessed, as certain differences between surgical and nonsurgical patients may have been unaccounted for by our models.26 Similarly, relatively small sample sizes precluded the inclusion of certain covariates in the Cox models, and other potential confounders (e.g., polytrauma, comorbidities, equipment availability, socioeconomic factors) and interactions were not investigated, which may have reduced the validity of our models.

A key limitation of retrospective database analysis is that relevant data may be unavailable for large numbers of patients. In our registry, the GOS score was included as an outcome measure because of its validity and widespread use34 but was not recorded for most patients, necessitating the use of different outcome measures to increase the sample size. The heterogeneity of outcomes may have impacted the validity of our results. However, the GCS score may be predictive of TBI outcomes within certain parameters, namely very high or very low GCS scores and a broad definition of outcomes (e.g., “good” and “poor"),24 and the clinical significance of the outcome measures may mitigate this limitation. For many surgical patients, the TBI pathology was unspecified. Further, time from injury to admission, which is often delayed and variable in this population,4 was unavailable in the database during the study period. Efforts to improve data collection are underway.

Additionally, the TBI registry at the MNRH currently does not capture sTBI patients who undergo immediate surgery in the casualty unit (emergency department) given the severity of their injuries or patients who are transferred directly to and from the ICU, so the most severely injured patients are underrepresented, and rates of poor outcomes for sTBI patients may be underestimated. However, the ICU at the MNRH comprises only 6–8 beds, which are typically reserved for patients considered to be inoperable. Therefore, few patients are excluded from the database because of transfer to the ICU. While several other critically ill patients are not included in the database, interpretations of our data must take into consideration that our registry comprises only those patients who are admitted to the neurosurgery ward. Despite the limitations of the database, our finding that nondecompressive or decompressive surgery is potentially beneficial for certain patients with acute TBI is an important step toward further understanding the impact of surgery on TBI outcomes in LMICs.

Conclusions

The results of this study affirm the potential benefits of surgical intervention with respect to acute outcomes in TBI in a low-resource setting. This study also highlights the need for earlier imaging and closer monitoring of moTBI patients, who may be more prone to poor outcomes when they are triaged similarly to mTBI patients. In the context of a high TBI disease burden and associated morbidity and mortality compounded by shortages of neurosurgical providers and resources, these data may ultimately facilitate triage and management decisions. However, further research is necessary to better understand which individual patients may benefit the most from specific surgeries, particularly in the long term, and how to balance need and benefit to fairly and optimally allocate scarce resources. Current efforts include building and validating a prognostic model using these data and investigating how providers in this setting use prognostication in treatment and resource-allocation decisions.

Acknowledgments

We thank the members of the Duke University Division of Global Neurosurgery and Neurology and the MNRH for facilitating the conduction of this investigation.

Disclosures

Dr. Haglund has received from NuVasive and LifeNet clinical or research support for the study described.

Author Contributions

Conception and design: Dunn, Spears, Haglund, Fuller. Acquisition of data: Muhumza. Analysis and interpretation of data: Dunn, Spears, Adil, Kolls, Haglund, Fuller. Drafting the article: Dunn, Spears, Adil. Critically revising the article: Dunn, Spears, Adil, Kolls, Haglund, Fuller. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Dunn. Statistical analysis: Dunn. Administrative/technical/material support: Dunn, Spears, Muhumza. Study supervision: Muhumza, Haglund, Fuller.

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

Previous Presentations

This work has been presented as an oral presentation at the virtual World Federation of Neurosurgical Societies Global Neurosurgery Symposium held on September 11, 2020.

References

  • 1

    Dewan MC, Rattani A, Gupta S, et al. Estimating the global incidence of traumatic brain injury. J Neurosurg. 2019;130(4):10801097.

  • 2

    De Silva MJ, Roberts I, Perel P, et al. Patient outcome after traumatic brain injury in high-, middle- and low-income countries: analysis of data on 8927 patients in 46 countries. Int J Epidemiol. 2009;38(2):452458.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 3

    Hawthorne C, Piper I. Monitoring of intracranial pressure in patients with traumatic brain injury. Front Neurol. 2014;5:121.

  • 4

    Vaca SD, Kuo BJ, Nickenig Vissoci JR, et al. Temporal delays along the neurosurgical care continuum for traumatic brain injury patients at a tertiary care hospital in Kampala, Uganda. Neurosurgery. 2019;84(1):95103.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Clavijo A, Khan AA, Mendoza J, et al. The role of decompressive craniectomy in limited resource environments. Front Neurol. 2019;10:112.

  • 6

    Karekezi C, El Khamlichi A, El Ouahabi A, et al. The impact of African-trained neurosurgeons on sub-Saharan Africa. Neurosurg Focus. 2020;48(3):E4.

  • 7

    Fuller A, Tran T, Muhumuza M, Haglund MM. Building neurosurgical capacity in low and middle income countries. eNeurologicalSci. 2015;3:16.

  • 8

    Haglund MM, Kiryabwire J, Parker S, et al. Surgical capacity building in Uganda through twinning, technology, and training camps. World J Surg. 2011;35(6):11751182.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Haglund MM, Warf B, Fuller A, et al. Past, present, and future of neurosurgery in Uganda. Neurosurgery. 2017;80(4):656661.

  • 10

    Central Intelligence Agency. The World Factbook 2020. Accessed November 13, 2020. https://www.cia.gov/library/publications/the-world-factbook/geos/ug.html

    • Search Google Scholar
    • Export Citation
  • 11

    Staton CA, Msilanga D, Kiwango G, et al. A prospective registry evaluating the epidemiology and clinical care of traumatic brain injury patients presenting to a regional referral hospital in Moshi, Tanzania: challenges and the way forward. Int J Inj Contr Saf Promot. 2017;24(1):6977.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Eaton J, Hanif AB, Mulima G, et al. Outcomes following exploratory burr holes for traumatic brain injury in a resource poor setting. World Neurosurg. 2017;105:257264.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 13

    Aenderl I, Gashaw T, Siebeck M, Mutschler W. Head injury—a neglected public health problem: a four-month prospective study at Jimma University Specialized Hospital, Ethiopia. Ethiop J Health Sci. 2014;24(1):2734.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14

    Djientcheu V, Nguifo Fongang EJ, Owono Etoundi P, et al. Mortality of head injuries in Sub-Saharan African countries: the case of the university teaching hospitals of Cameroon. J Neurol Sci. 2016;371:100104.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15

    Tran TM, Fuller AT, Kiryabwire J, et al. Distribution and characteristics of severe traumatic brain injury at Mulago National Referral Hospital in Uganda. World Neurosurg. 2015;83(3):269277.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Elahi C, Rocha TAH, da Silva NC, et al. An evaluation of outcomes in patients with traumatic brain injury at a referral hospital in Tanzania: evidence from a survival analysis. Neurosurg Focus. 2019;47(5):E6.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Pocock SJ, Clayton TC, Altman DG. Survival plots of time-to-event outcomes in clinical trials: good practice and pitfalls. Lancet. 2002;359(9318):16861689.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18

    Stel VS, Dekker FW, Tripepi G, et al. Survival analysis I: the Kaplan-Meier method. Nephron Clin Pract. 2011;119(1):c83c88.

  • 19

    Stel VS, Dekker FW, Tripepi G, et al. Survival analysis II: Cox regression. Nephron Clin Pract. 2011;119(3):c255c260.

  • 20

    Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. PLoS Med. 2007;4(10):e297.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Kuo BJ, Vaca SD, Vissoci JRN, et al. A prospective neurosurgical registry evaluating the clinical care of traumatic brain injury patients presenting to Mulago National Referral Hospital in Uganda. PLoS One. 2017;12(10):e0182285.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    Andriessen TMJC, Horn J, Franschman G, et al. Epidemiology, severity classification, and outcome of moderate and severe traumatic brain injury: a prospective multicenter study. J Neurotrauma. 2011;28(10):20192031.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet. 1975;1(7905):480484.

  • 24

    McNett M. A review of the predictive ability of Glasgow Coma Scale scores in head-injured patients. J Neurosci Nurs. 2007;39(2):6875.

  • 25

    Crowley J, Hu M. Covariance analysis of heart transplant survival data. J Am Stat Assoc. 1977;72(357):2736.

  • 26

    Bull K, Spiegelhalter DJ. Survival analysis in observational studies. Stat Med. 1997;16(9):10411074.

  • 27

    Purcell LN, Reiss R, Eaton J, et al. Survival and functional outcomes at discharge after traumatic brain injury in children versus adults in resource-poor setting. World Neurosurg. 2020;137:e597e602.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Lund SB, Gjeilo KH, Moen KG, et al. Moderate traumatic brain injury, acute phase course and deviations in physiological variables: an observational study. Scand J Trauma Resusc Emerg Med. 2016;24:77.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29

    Godoy DA, Rubiano A, Rabinstein AA, et al. Moderate traumatic brain injury: the grey zone of neurotrauma. Neurocrit Care. 2016;25(2):306319.

  • 30

    Thompson HJ, McCormick WC, Kagan SH. Traumatic brain injury in older adults: epidemiology, outcomes, and future implications. J Am Geriatr Soc. 2006;54(10):15901595.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31

    Krieger D, Adams HP, Schwarz S, et al. Prognostic and clinical relevance of pupillary responses, intracranial pressure monitoring, and brainstem auditory evoked potentials in comatose patients with acute supratentorial mass lesions. Crit Care Med. 1993;21(12):19441950.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32

    Perel P, Arango M, Clayton T, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336(7641):425429.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33

    Murray GD, Butcher I, McHugh GS, et al. Multivariable prognostic analysis in traumatic brain injury: results from the IMPACT study. J Neurotrauma. 2007;24(2):329337.

  • 34

    McMillan T, Wilson L, Ponsford J, et al. The Glasgow Outcome Scale—40 years of application and refinement. Nat Rev Neurol. 2016;12(8):477485.

Supplementary Materials

Illustration from Fan et al. (pp 1298–1309). Copyright Jun Fan. Published with permission.

  • View in gallery

    Flowchart for selection of study population. *Does not include chronic subdural hematoma.

  • View in gallery

    Distribution of outcomes among the overall cohort and the cohort stratified by TBI severity. No patients had an outcome of GOS scores 2–3. Numbers of patients with poor outcomes were 13 for surgical patients, 144 for nonsurgical patients, and 157 for the overall cohort. *Indicates poor outcome.

  • View in gallery

    Kaplan-Meier plots for overall cohort and each TBI severity group. Kaplan-Meier plots were generated for models that included surgery as a non–time-dependent treatment variable. All surgery patients were grouped together at the start of the study. Figure is available in color online only.

  • 1

    Dewan MC, Rattani A, Gupta S, et al. Estimating the global incidence of traumatic brain injury. J Neurosurg. 2019;130(4):10801097.

  • 2

    De Silva MJ, Roberts I, Perel P, et al. Patient outcome after traumatic brain injury in high-, middle- and low-income countries: analysis of data on 8927 patients in 46 countries. Int J Epidemiol. 2009;38(2):452458.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 3

    Hawthorne C, Piper I. Monitoring of intracranial pressure in patients with traumatic brain injury. Front Neurol. 2014;5:121.

  • 4

    Vaca SD, Kuo BJ, Nickenig Vissoci JR, et al. Temporal delays along the neurosurgical care continuum for traumatic brain injury patients at a tertiary care hospital in Kampala, Uganda. Neurosurgery. 2019;84(1):95103.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5

    Clavijo A, Khan AA, Mendoza J, et al. The role of decompressive craniectomy in limited resource environments. Front Neurol. 2019;10:112.

  • 6

    Karekezi C, El Khamlichi A, El Ouahabi A, et al. The impact of African-trained neurosurgeons on sub-Saharan Africa. Neurosurg Focus. 2020;48(3):E4.

  • 7

    Fuller A, Tran T, Muhumuza M, Haglund MM. Building neurosurgical capacity in low and middle income countries. eNeurologicalSci. 2015;3:16.

  • 8

    Haglund MM, Kiryabwire J, Parker S, et al. Surgical capacity building in Uganda through twinning, technology, and training camps. World J Surg. 2011;35(6):11751182.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Haglund MM, Warf B, Fuller A, et al. Past, present, and future of neurosurgery in Uganda. Neurosurgery. 2017;80(4):656661.

  • 10

    Central Intelligence Agency. The World Factbook 2020. Accessed November 13, 2020. https://www.cia.gov/library/publications/the-world-factbook/geos/ug.html

    • Search Google Scholar
    • Export Citation
  • 11

    Staton CA, Msilanga D, Kiwango G, et al. A prospective registry evaluating the epidemiology and clinical care of traumatic brain injury patients presenting to a regional referral hospital in Moshi, Tanzania: challenges and the way forward. Int J Inj Contr Saf Promot. 2017;24(1):6977.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Eaton J, Hanif AB, Mulima G, et al. Outcomes following exploratory burr holes for traumatic brain injury in a resource poor setting. World Neurosurg. 2017;105:257264.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 13

    Aenderl I, Gashaw T, Siebeck M, Mutschler W. Head injury—a neglected public health problem: a four-month prospective study at Jimma University Specialized Hospital, Ethiopia. Ethiop J Health Sci. 2014;24(1):2734.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14

    Djientcheu V, Nguifo Fongang EJ, Owono Etoundi P, et al. Mortality of head injuries in Sub-Saharan African countries: the case of the university teaching hospitals of Cameroon. J Neurol Sci. 2016;371:100104.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15

    Tran TM, Fuller AT, Kiryabwire J, et al. Distribution and characteristics of severe traumatic brain injury at Mulago National Referral Hospital in Uganda. World Neurosurg. 2015;83(3):269277.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Elahi C, Rocha TAH, da Silva NC, et al. An evaluation of outcomes in patients with traumatic brain injury at a referral hospital in Tanzania: evidence from a survival analysis. Neurosurg Focus. 2019;47(5):E6.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Pocock SJ, Clayton TC, Altman DG. Survival plots of time-to-event outcomes in clinical trials: good practice and pitfalls. Lancet. 2002;359(9318):16861689.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18

    Stel VS, Dekker FW, Tripepi G, et al. Survival analysis I: the Kaplan-Meier method. Nephron Clin Pract. 2011;119(1):c83c88.

  • 19

    Stel VS, Dekker FW, Tripepi G, et al. Survival analysis II: Cox regression. Nephron Clin Pract. 2011;119(3):c255c260.

  • 20

    Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. PLoS Med. 2007;4(10):e297.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Kuo BJ, Vaca SD, Vissoci JRN, et al. A prospective neurosurgical registry evaluating the clinical care of traumatic brain injury patients presenting to Mulago National Referral Hospital in Uganda. PLoS One. 2017;12(10):e0182285.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    Andriessen TMJC, Horn J, Franschman G, et al. Epidemiology, severity classification, and outcome of moderate and severe traumatic brain injury: a prospective multicenter study. J Neurotrauma. 2011;28(10):20192031.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet. 1975;1(7905):480484.

  • 24

    McNett M. A review of the predictive ability of Glasgow Coma Scale scores in head-injured patients. J Neurosci Nurs. 2007;39(2):6875.

  • 25

    Crowley J, Hu M. Covariance analysis of heart transplant survival data. J Am Stat Assoc. 1977;72(357):2736.

  • 26

    Bull K, Spiegelhalter DJ. Survival analysis in observational studies. Stat Med. 1997;16(9):10411074.

  • 27

    Purcell LN, Reiss R, Eaton J, et al. Survival and functional outcomes at discharge after traumatic brain injury in children versus adults in resource-poor setting. World Neurosurg. 2020;137:e597e602.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Lund SB, Gjeilo KH, Moen KG, et al. Moderate traumatic brain injury, acute phase course and deviations in physiological variables: an observational study. Scand J Trauma Resusc Emerg Med. 2016;24:77.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29

    Godoy DA, Rubiano A, Rabinstein AA, et al. Moderate traumatic brain injury: the grey zone of neurotrauma. Neurocrit Care. 2016;25(2):306319.

  • 30

    Thompson HJ, McCormick WC, Kagan SH. Traumatic brain injury in older adults: epidemiology, outcomes, and future implications. J Am Geriatr Soc. 2006;54(10):15901595.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31

    Krieger D, Adams HP, Schwarz S, et al. Prognostic and clinical relevance of pupillary responses, intracranial pressure monitoring, and brainstem auditory evoked potentials in comatose patients with acute supratentorial mass lesions. Crit Care Med. 1993;21(12):19441950.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32

    Perel P, Arango M, Clayton T, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336(7641):425429.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33

    Murray GD, Butcher I, McHugh GS, et al. Multivariable prognostic analysis in traumatic brain injury: results from the IMPACT study. J Neurotrauma. 2007;24(2):329337.

  • 34

    McMillan T, Wilson L, Ponsford J, et al. The Glasgow Outcome Scale—40 years of application and refinement. Nat Rev Neurol. 2016;12(8):477485.

Metrics

All Time Past Year Past 30 Days
Abstract Views 832 832 0
Full Text Views 680 679 174
PDF Downloads 178 178 31
EPUB Downloads 0 0 0