Mortality From Rural Traumatic Brain Injury Are Underestimated Without Inclusion of Prehospital Deaths Authors: Gareth Rutter, Maria Joseph, Richard Ferch, Mitchell A. Hansen, Zsolt J. Balogh Published: 2026 ABSTRACT ABSTRACTBackgroundRural residents are more likely to suffer moderate and severe traumatic brain injury (TBI) and experience delay to definitive care than urban patients. Prior studies comparing these populations were based on hospitalised patients only. We examined TBI outcomes incorporating pre‐hospital mortality to determine the association between location of injury and mortality risk.MethodsTwelve‐month (2019) retrospective cohort study was performed on all adult patients with TBI within one Trauma System. Demographics, geographical location of injury, mechanism, scene vital parameters and interventions, injury severity, polytrauma status, TBI morphology and outcomes were collected for all hospital and prehospital deaths. Uni‐ and multivariate analysis was conducted including binary logistic regression to examine associations between mortality and potential confounders of mortality.Results266 patients (145 hospitalised and 121 prehospital deaths) with median age 53.6 years (18–98), 72.6% males, scene Glasgow Coma Scale (GCS) < 14 (48.3%), were included. Ninety‐four (36%) patients were urban and, 167 (64%) were regional, rural or remote. Rural compared to urban patients had higher risk of death (adjusted OR 1.934, 95% CI 1.07–3.49) without differences in GCS, pupillary status, injury pattern, radiological TBI severity or initial management. The strongest risk factor for mortality was polytrauma (adjusted OR 7.804, 95% CI 4.155–14.62, p < 0.01).ConclusionsComprehensive assessment of TBI mortality indicates that rurality is associated with mortality and polytrauma status is a profound risk factor. These findings highlight that prehospital mortality is essential to monitor and optimisation of polytrauma care is a potential modifiable factor in TBI outcomes. Introduction Approximately 6000 patients are admitted to New South Wales (NSW) hospitals annually with traumatic brain injuries (TBI) [1]. The disease burden is disproportionately borne by rural patients, with the incidence of TBI in rural patients approximately three‐fold that of urban patients [1]. Moreover, rural patients are more likely to sustain severe TBI, with significant differences in mechanism [2]. Furthermore, time to definitive neurosurgical care is longer [3]. Despite this, it remains controversial whether delay in treatment affects outcome [4]. With respect to outcome, a retrospective cohort analysis found that mortality was 30% greater in rural patients with TBI in NSW in the 1980s [5]. However, more contemporary data, including a 2024 meta‐analysis, found no significant difference in mortality between rural and urban patients with TBI globally [2]. Notably, only two of the included studies examined pre‐hospital mortality, perhaps introducing an unrecognised selection and survival bias, by examining only patients with TBI who made it to hospital treatment. Hence, it is possible that poorer TBI outcomes amongst rural patients remain under‐reported, given the lack of pre‐hospital mortality data. The National Coronial Information System (NCIS) is an Australian national database of all reported deaths to the Coroner, including all sudden and unexpected deaths, such as those resulting from motor vehicle collisions (MVC) or falls. Therefore data from the NCIS would capture pre‐hospital mortality resulting from TBI and better represent the complete picture of TBI associated mortality. The aim of this study was to describe the comprehensive epidemiology of moderate‐to‐severe TBI in rural and urban patients within an inclusive trauma system served by one Level‐1 trauma centre with the only neurosurgical service in the region. We hypothesised that injuries in rural and remote locations with moderate‐to‐severe TBI lead to higher mortality than their urban counterparts. Methods All patients admitted to JHH and its referring hospitals (hereafter referred to collectively as the ‘JHH Trauma Service’) during the calendar year of 2019 were included. This period was selected due to the higher proportion of closed (i.e., completely accessible) cases within the NCIS dataset. Hospital cases were identified utilising ICD‐10 codes relating to TBI. Each identified case record was then reviewed by two investigators for inclusion and exclusion criteria, and de‐identified clinical, radiological, and outcome data was extracted from clinical documentation and the Trauma Registry. The NCIS database was queried for pre‐hospital deaths within the JHH Trauma System for the calendar year of 2019. All deaths were identified utilising ICD‐10 codes for TBI, a free text search for ‘head injury’, ‘traumatic brain injury’, ‘subdural haematoma’, ‘extradural haematoma’ ‘subarachnoid haemorrhage’, ‘diffuse axonal injury’, and mechanism of injury (MOI) codes including MVC and falls. Each autopsy report was then individually reviewed for inclusion and exclusion criteria. Included cases were cross‐referenced against hospital data to avoid duplication and de‐identified clinical data was extracted. Inclusion and Exclusion Criteria Inclusion criteria were patients above 18 years, with acute moderate‐to‐severe TBI, defined either clinically with a Glasgow Coma Scale (GCS) ≤ 13 or radiologically abnormal with brain CT. Patients who died before hospital with documented TBI were also included. Exclusion criteria were patients younger than 18 years, patients with GCS > 13 with a normal CT scan, or patients with non‐acute TBI and transfers from outside of the Trauma System. Data Variables For patients who were admitted to hospital, variables collected included demographics (age, gender), geographical location at time of injury, intoxication, mechanism of injury (MOI), pattern of injuries, injury severity as defined by the Injury Severity Score (ISS) mode of retrieval, direct admission or transfer, time to definitive hospital care, vital signs including GCS and pupillary status at scene and on arrival to hospital and pre‐hospital interventions (intubation, cardiopulmonary resuscitation, hyperosmolar therapy). Imaging on arrival was reviewed by two authors and injury patterns individually recorded, including presence, thickness in millimetres (mm) and compartments of haemorrhage, presence and location of skull fractures and presence and millimetres of midline shift and tonsillar herniation. Rotterdam scores were calculated [6]. Initial management was classified as either surgical, conservative or palliative. Subsequent clinical progression and mortality were also recorded. For patients who died pre‐hospital, variables were collected for the following: demographics (age, gender), geographical location at time of injury, intoxication and mechanism and pattern of injury. Injury severity (as defined by ISS) was not gathered in patients who died prior to hospital as insufficient data was available and these patients would in many cases have an indistinctive ISS of 75 due to their unsurvivable injury patterns. Injuries were considered polytrauma if the Abbreviated Injury Scale (AIS) was greater than 2 in at least 2 body regions [7]. Geographical location of injury was categorised on a five‐point scale from 1 (Major City) to 5 (Very Remote) using the third edition of the Australian Statistical Geography Standard—Remoteness Area (ASGC‐RA) [8]. Distance from the initial geographical location of injury to definitive neurosurgical care at John Hunter Hospital was calculated using Google Maps. Cause of death for patients who died after admission to hospital was determined using the cause listed on the death certificate. Cause of death for patients who died prior to hospital was determined as TBI when this was the only listed injury. However, in patients with polytrauma, the proximate cause of death could not be determined. Ethics The study was approved by the Hunter New England Human Research Ethics Committee (AU202409‐05). NCIS database access was approved by the Justice and Community Safety Human Research Ethics Committee for the Victorian Department of Justice and Community Safety (CF/24/22157). Statistical Analysis Statistical analysis was conducted using STATA/BE (Version 19, Statacorp, College Station, Texas). Patients were dichotomised as either urban (ASGC‐RA Category 1) or regional, rural and remote (ASGC‐RA categories 2–5). For univariate analyses across groups, independent samples t‐tests were used to compare continuous variables (e.g., age, SBP, DBP, HR, Shock index), Chi‐squared tests for associations between categorical variables (e.g., sex, mechanism of injury, intoxication status, polytrauma status) and Mann–Whitney U test for non‐parametric comparisons of ordinal variables (e.g., GCS, ISS, AIS Head). A significance level of p < 0.05 was used. Binary logistic regression was used to examine associations between covariates and mortality, with results reported as adjusted odds ratios. Two multivariate models were run, for all deaths, including patients who died prior to hospitalisation and then only for patients who died after hospitalisation. For all deaths including pre‐hospital mortalities, all collected parameters were included in the multivariate regression (age, rurality, polytrauma, intoxication status). For in‐hospital deaths, a multivariate logistic model was run with variables selected a priori based on clinical relevance and entered simultaneously into the regression model. Covariates included were age, rurality, distance from tertiary trauma centre, time to definitive neurosurgical care, polytrauma, intoxication status, shock parameters (SBP, DBP, SI and HR at scene and at arrival to tertiary trauma centre) and parameters of clinical severity (ISS, Head AIS ad GCS at scene). Stepwise regression analysis with backward elimination of least significant variables was conducted, ultimately leaving age, distance, GCS at scene, SBP at scene, ISS, Head AIS, Polytrauma and intoxication status. Results Five hundred and eighty‐nine patients were admitted with traumatic brain injury at John Hunter Hospital, 145 cases met inclusion criteria (Figure 1). From the NCIS database, 1357 coronial records were screened; and 121 cases met inclusion criteria (Figure 2). Flowchart of Inclusion/Exclusion Criteria JHH. Flowchart of Inclusion/Exclusion Criteria for Pre‐hospital deaths. Demographics Mean age of patients within the entire dataset was 53.6 years (range 18–98 years, standard deviation 21.5 years), 193 (72.6%) patients were male and 167 (64%) were injured at regional, rural or remote location (Table 1). TABLE 1Demographic and clinical characteristics of all patients.Age in years—mean (range, standard deviation)53.6(18–98, 21.5)Sex—no. of patients (%)Male193(72.6)Female73(27.4)Intoxication—no. of patients (%)None201(75.6)Single substance42(15.8)Polysubstance23(8.7)Polytrauma—no. of patients (%)Isolated injury138(51.9)Polytrauma128(48.1)Mechanism of Injury—no. of patients (%)Fall103(38.9)Pedestrian15(5.7)Pushbike6(2.3)MVC100(37.7)Quad Bike2(0.8)Jump from height5(1.9)Assault/other15(5.7)Gunshot19(7.2)Location of InjuryAustralian Statistical Geographical Classification Remoteness Area (ASGC‐RA Score)—no. of patients (%)RA 1 (Major City)94(36.0)RA 2 (Inner regional)118(45.2)RA 3 (Outer regional)44(16.9)RA 4 (Remote)1(0.4)RA 5 (Very remote)4(1.5)Distance (km) from Tertiary Trauma Centre—mean (range)223(0–1200) Demographic and clinical characteristics of all patients. Subgroup analyses demonstrated a statistically significant difference in age between rural and urban patients (50.8 vs. 58.8 years respectively, p < 0.01) (Table 2). TABLE 2Subgroup analysis between urban and regional rural and remote patients admitted to hospital.UrbanRuralc p Age in years—mean58.850.8 0.00 a Sex—no. of patients (%)Male66 (70.2)127 (73.9)0.53b Mechanism of Injury—no. of patients (%) 0.00 b Fall54 (52.4)49 (47.6)—Pedestrian versus other7 (46.7)8 (53.3)—Pushbike accident3 (50.0)3 (50.0)—Motor vehicle/Motorbike accident17 (17.0)83 (83.0)—Quad Bike/ATV Accident0 (0.0)2 (100.0)—Jump from height4 (80.0)1 (20.0)—Assault/other6 (40.0)9 (60.0)—Gunshot wound2 (10.5)17 (89.5)—Intoxicated—no. of patients (%)21 (22.3)44 (25.6)0.56b Polytrauma—no. of patients (%)34 (36.1)94 (54.7) 0.00 b Haemodynamic Parameters—mean (SD)SBP at Scene (mmHg)144 (32.6)122 (31.4) 0.00 a DBP at Scene (mmHg)87 (16.8)69 (19.6) 0.00 a HR at Scene87 (21.8)88 (31.2)0.90a Shock Index at Scene0.64 (0.29)0.77 (0.37)0.06a SBP at Hospital arrival (mmHg)145 (30.4)128 (20.9) 0.00 a DBP at Hospital arrival (mmHg)80 (16.9)78 (13.3)0.46a HR at Hospital arrival81 (23.5)85 (20.6)0.32a Shock Index at Hospital arrival0.60 (0.30)0.70 (0.29)0.08a ISS—median (IQR)21 (9)25 (12)0.26c AIS Head Severity Score—median (IQR)4 (1)4 (1)0.85c GCS at Scene—median (IQR)14 (4)13.5 (6)0.36c Pupils at Scene—no. of patients (%)0.27b Bilaterally reactive62 (93.9)65 (85.5)—Bilaterally small non‐reactive2 (3.0)1 (1.3)—Unilaterally fixed and dilated1 (1.5)4 (5.3)—Bilaterally fixed and dilated1 (1.5)5 (6.6)—Traumatic mydriasis/not assessable0 (0.0)1 (1.3)—Rotterdam—no. of patients (%)0.52b 110 (14.9)17 (21.8)—226 (38.8)25 (32.1)—312 (17.9)18 (23.1)—413 (19.4)9 (11.5)—56 (9.0)8 (10.3)—60 (0.0)1 (1.3)—SDHPresent—no. of patients (%)52 (77.6)53 (68.0)0.19b Size (mm)8.7 (1–36)6.5 (1–24)0.06b EDHPresent—no. of patients (%)10 (14.9)10 (12.8)0.71b Size (mm)10.3 (4–24)14.3 (4–37)0.34b MLS (mm)2.94 (0–25)2.19 (0–19)0.35a Initial Management—no. of patients (%)0.38b Palliated5 (7.5)6 (7.5)—Non‐operative50 (74.6)52 (65.0)—Operative12 (17.9)22 (27.5)— Note: Haemodynamic parameters, GCS, pupils, ISS and AIS scores and initial management will only encapsulate TTC patients (no data for NCIS patients).Abbreviations: IQR = Interquartile range; SD = Standard deviation. a Two‐sample t‐test. b Pearson chi2. c Two‐sample Wilcoxon rank‐sum (Mann–Whitney) z statistic with two‐tail p‐value. Subgroup analysis between urban and regional rural and remote patients admitted to hospital. Note: Haemodynamic parameters, GCS, pupils, ISS and AIS scores and initial management will only encapsulate TTC patients (no data for NCIS patients). Abbreviations: IQR = Interquartile range; SD = Standard deviation. Two‐sample t‐test. Pearson chi2. Two‐sample Wilcoxon rank‐sum (Mann–Whitney) z statistic with two‐tail p‐value. Mechanisms of Injury Mechanism of injury was fall in 103 patients (38.9%), followed by MVC in 100 patients (37.7%). A minority of TBI's were caused by gunshot in 19 patients (7.2%), assault in 15 patients (5.7%) and pedestrian‐related traffic collisions in 15 patients (5.7%) (Table 1). Subgroup analysis of injury mechanism by geographical location demonstrated statistically significant differences based on AGSA‐RA category. Rural patients were more likely to sustain a TBI from MVC, gunshot wound, or assault whereas urban patients were more likely to fall or jump from a height. Rural patients were also more likely to be polytrauma (54.7% vs. 36.1%, p < 0.01) (Table 2). Location of Injury and Retrieval The mean distance from the injury location to John Hunter Hospital was 223 km. Sixty four percent of the injuries occurred in regional rural or remote areas (ASGC‐RA category 2–5) (Table 1). Fifty‐four percent of patients were transported by road ambulance, 20.7% by helicopter and 20% by fixed wing aircraft. 63.6% of patients were directly admitted to JHH; 32.9% were transferred from a regional hospital. Multivariate linear regression demonstrated that of patients who were transferred to hospital, rural location of injury was correlated with longer time to definitive care (β = 0.71, 95% CI 0.47–0.94, p < 0.001); whereas there was no statistically significant correlation between time to definitive care and injury severity (measured by ISS, AIS‐head, GCS at scene), shock index or mortality (all p > 0.05). Injury Characteristics, At‐Scene Clinical Status and Pre‐Hospital Management Of patients admitted to hospital, the median GCS at the scene was 14 (interquartile range: 5). Deterioration in GCS between the scene and hospital was identified in 18 (16.8%) patients. Pupillary status at the scene was bilaterally reactive in 89.4%, unilaterally fixed and dilated in 3.5%, bilaterally fixed and dilated in 4.2%, bilaterally fixed but not dilated in 2.1% and 1 (0.7%) traumatic mydriasis. A deterioration in pupillary status between the scene and hospital was identified in 10 (7%) patients. There was no difference in initial GCS, pupillary status and deterioration of those between urban and rural patients. There was no difference between rural and urban patients in ISS and AIS‐head (Table 2). Rural patients had a statistically significant lower systolic blood pressure both at the scene and at the John Hunter Hospital. There was no statistically significant difference in the Shock Index between rural and urban patients at either the scene or at arrival to hospital (Table 2). Thirty‐one patients (21%) were intubated prior to admission to JHH. Hyperosmolar therapy was administered in 11 patients (7.7%), with mannitol more commonly used than hypertonic saline. Cardiopulmonary resuscitation was performed in 3 patients (2.1%) who were transferred to hospital. More rural location of injury was associated with more frequent pre‐hospital intubation ( z = 3.093, p < 0.05) and use of hyperosmolar therapy (z = 3.203, p < 0.05). Acute subdural haematoma was the most common injury, observed in 105 patients (72.4%) with mean depth of 7.6 mm (range 1–36 mm). Other compartments of haemorrhage are provided for reference in Table 3. Seventy‐six patients (76%) sustained a concomitant skull fracture. Mean midline shift was 2.5 mm (range 0–25 mm). Tonsillar herniation was present in five patients (3.5%). There was no significant difference in radiological patterns of TBI, nor Rotterdam score, between urban and rural patients (Table 3). TABLE 3Injury characteristics.Compartment of injurySubdural Haemorrhage (SDH)—no. of patients (%)105(72.4)SDH Depth mm—mean (range)7.6(1.0–36.0)Extradural Haemorrhage (EDH)—no. of patients (%)20(13.8)EDH Depth mm—mean (range)12.3(4.0–37.0)Subarachnoid Haemorrhage—no. of patients (%)100(69.0)Convexal32(22.1)Basal10(6.9)Mixed58(40.0)Contusion—no. of patients (%)90(62.1)DAI—no. of patients (%)7(4.8)Skull fracture(s)—no. of patients (%)76(52.4)Convexal14(9.7)Basal26(17.9)Mixed36(24.8)Midline shift (mm)—mean (range)2.5(0–25)Tonsillar Herniation—no. of patients (%)5(3.5)Rotterdam score—no. of patients (%)127(18.6)251(35.2)330(20.7)422(15.2)514(9.7)61(0.7) Injury characteristics. Management and Outcome Hundred and two (70.3%) hospitalised patients were managed initially non‐operatively, 6 of these later had surgery. Thirty‐two patients (22.1%) were managed surgically on admission; 15 (10.3%) had intracranial pressure monitor or external ventricular drain insertion, 12 (8.3%) had craniotomy and evacuation of haematoma, and 5 (3.5%) underwent craniectomy. Eighteen patients (7.6%) were palliated, 11 on arrival to the John Hunter Hospital and 7 during the course of their admission. Comparison of Pre‐Hospital and In‐Hospital Mortality Patients who died prior to hospitalisation were younger (38.6 vs. 69.2 years, p < 0.01), more frequently males (81% vs. 37.1%, p = 0.02), intoxicated (37.2% vs. 8.6%, p < 0.01), had sustained a polytrauma (73.6% vs. 34.3%, p < 0.01) and their injury location was more frequently rural (77.7% vs. 22.3%, p < 0.01) than those who made it to hospital admission (Table 4). TABLE 4Subgroup Analysis of Pre‐hospital and In‐hospital Mortalities.In hospitalPre‐hospital p Age in years—mean69.238.60.00 a Sex—no. of patients (%)Male13 (37.1)98 (81.0)0.02 b Mechanism of injury—no. of patients (%)0.00 b Fall25 (71.4)7 (5.8)—Pedestrian vs other1 (2.9)11 (9.1)—Pushbike accident0 (0.0)3 (2.5)—Motor vehicle/motorbike accident8 (22.9)70 (57.9)—Quad bike/ATV accident0 (0.0)0 (0.0)—Jump from height0 (0.0)5 (4.1)—Assault/other1 (2.9)6 (5.0)—Gunshot wound0 (0.0)19 (15.7)—Intoxicated—no. of patients (%)3 (8.6)45 (37.2)0.00 b Polytrauma—no. of patients (%)12 (34.3)89 (73.6)0.00 b Location of injuryDichotomised ASGC‐RA (Major City vs. Regional/Remote) 0.00 b RA 1 (Major City)19 (54.3)27 (22.3)—RA 2–5 (Regional or Remote)16 (45.7)93 (77.7)—Distance (km) from Tertiary Trauma Centre—mean (range)122 (0‐550)341.8 (0‐1200) 0.00 a a Two‐sample t test. b Pearson chi2. Subgroup Analysis of Pre‐hospital and In‐hospital Mortalities. Two‐sample t test. Pearson chi2. Cause of death for patients who died after admission to hospital was TBI in most patients (n = 30, 86%). In four cases, the cause of death was listed as TBI in association with other injuries and their subsequent complications. In one case, the cause of death was listed as myocardial infarction during admission. Relative Risks of Mortality Multivariable logistic regression for deaths in hospitalised patients (n = 80), demonstrated that mortality risk was independently associated with older age (adjusted odds ratio [aOR]:1.11, 95% confidence interval [CI]:1.04–1.18; p = 0.001). Higher GCS on scene was associated with lower risk of death (aOR:0.58, 95% CI:0.43–0.80; p = 0.001). Rurality, SS, AIS, scene SBP, polytrauma status, and substance use were not independently associated with mortality (Table 5). TABLE 5Multivariate binary logistic regression for in‐hospital mortalities only.aOR95% CI p Age1.1101.0426–1.1825 0.001 GCS at Scene0.5840.4276–0.7980 0.001 Location of Injury (Dichotomised)1.9730.3113–12.50280.471ISS1.0110.8343–1.22600.909AIS1.3870.2332–8.25900.719SBP at Scene0.9990.9738–1.02380.908Polytrauma1.2420.0739–20.8760.880Intoxication0.6400.1135–3.6080.613 Note: Observations n = 80, pseudo‐R 2 = 0.466. Multivariate binary logistic regression for in‐hospital mortalities only. Note: Observations n = 80, pseudo‐R 2 = 0.466. On multivariate logistic regression for all deaths including pre‐hospital deaths (n = 263), mortality was independently associated with older age (aOR:1.03, 95% CI:1.01–1.04; p = 0.001), rurality (aOR: 1.93, 95% CI: 1.07–3.49; p = 0.029), polytrauma (aOR: 7.81, 95% CI: 4.15–14.66; p < 0.001) and intoxication (aOR:2.69, 95% CI: 1.51–4.81; p = 0.001) (Table 6). TABLE 6Multivariate binary logistic regression for all mortalities, including pre‐hospital mortality.aOR95% CI p Age1.026(1.010–1.041)0.001Location of Injury (Dichotomised)1.934(1.071–3.492)0.029Polytrauma7.805(4.155–14.662)0.000Intoxication2.694(1.508–4.812)0.001 Note: Observations n = 263, pseudo‐R 2 = 0.186. Multivariate binary logistic regression for all mortalities, including pre‐hospital mortality. Note: Observations n = 263, pseudo‐R 2 = 0.186. Discussion It is well established that the incidence and severity of TBI is higher in rural and remote versus urban patients [1]. It is less clear whether this translates to outcomes as a recent meta‐analysis found no difference [2]. The present study set out to examine this complex question in a comprehensive way, including the often overlooked cohort of prehospital mortality. On univariate analysis of mortality in all patients, including pre‐hospital mortalities, those from rural or remote locations had higher risk of death compared to urban patients (RR 1.37, p = 0.02). Rurality is a strong independent relative risk factor for mortality, alongside with age, intoxication status and polytrauma. On multivariate analysis of mortality in all patients, including pre‐hospital mortalities, the statistically significant higher mortality risk in rural patients is again demonstrated (adjusted OR 1.934, p = 0.029). However, on multivariate analysis of in‐hospital mortalities only, the association between rurality and mortality association was lost (Table 5). This could explain why previous studies have not found difference between rurality and mortality [2]. These findings corroborate the results of an epidemiological study published in the 1980s, which also included pre‐hospital deaths [5]. Thus, we argue that pre‐hospital mortalities represent a ‘hidden burden of disease’ for traumatic brain injury overlooked by hospital centric studies, suggesting that rurality contributes disproportionately to pre‐hospital mortality. Other predictors for mortality on multivariate analysis of all deaths including pre‐hospital deaths, were older age, rurality, intoxication and polytrauma. Polytrauma itself was the strongest risk factor for death (adjusted OR 7.804, p < 0.001). Similar to rurality, this was not significant on multivariate analysis of in‐hospital deaths only, suggesting polytrauma, like rurality, seems to be a disproportionate contributor of death in patients dying prior to hospital. This may also explain why injury severity (as captured by ISS score), was not associated with mortality in our study. This variable was only included and analysed for patients who made it to hospital, as insufficient data was available to reliably calculate the score in patients who died prior to hospital. If they had been included, this would most likely have skewed to a higher ISS overall. These findings concord with the trauma literature, where polytrauma is in itself is recognised as a disease process where the outcome is disproportionately worse than what would be expected from the sum of each individual injury in part [9], and extracranial injuries can exacerbate intracranial injuries [10]. In a subgroup analysis comparing all rural and urban patients in this study, rural patients were younger (50.8 vs. 58.8), and most likely to be injured as a result of an MVC, while in urban locations from falls. In a further subgroup analysis of patients who died prior to hospitalisation versus within hospital, these patients were more likely to be rural, male, younger (38.6 vs. 69.2 years) and polytraumas. Their MOI was mostly MVC. This data suggests that there are two typical phenotypes of TBI mortality: young males with high energy polytrauma, predominately from rural locations and older patients with low energy injuries, predominately from urban locations. These findings could inform policy decisions regarding trauma bypass protocols, optimising transport of patients with TBI and better targeting public health interventions to relevant demographics. There are several limitations of this project. Firstly, it is a single trauma system retrospective analysis over the course of 1 year, which limits the generalisability of our data. Secondly, whilst time to definitive treatment was longer in rural patients, this did not translate to a significant association between time to treatment and mortality. It is likely that our study was underpowered to detect this difference given the relatively small sample size. This could be better captured in future adequately powered prospective studies. Importantly however, time to definitive care is likely multifactorial, not dependent solely on location of injury, but on other factors including allocation of resources and triaging of injury severity. The complexities of these were impossible to tease out with our current study design. These challenges were recently reviewed by Merakis et al. [3]. Moreover, the prehospital care of all patients was governed by standardised NSW Ambulance protocols. Whilst rural patients in our study were more likely to be intubated and to receive hyperosmolar therapy during transfer, these observed differences likely reflect the severity of the traumatic insult (with rural patients more likely to be polytraumas), and logistical considerations for transfer given their distance to definitive care. However, our study design could not capture these nuances. Finally, comprehensive conclusions are limited by inherent issues with pre‐hospital mortality data. The National Coronal Information System does not capture relevant contextual clinical factors such as resuscitation measures, physiological derangements and injury severity, or logistical factors such as time of medical arrival to scene of injury. Furthermore, there were a minority of cases that were excluded due to incomplete data or censored data (as they are active, open coronial cases). Moreover, within the pre‐hospital mortality dataset, it is not possible to determine whether patients with polytrauma died as a result of the intracranial injuries, their extracranial injuries or both. Thus, it would be more accurate to say they died ‘with’ a TBI than ‘of’ a TBI. This is an inherent limitation of the nature of autopsy reports, which comment on the presence but not severity of TBI. Conclusion This study suggests that rural patients are at higher risk of mortality from TBI compared to urban patients. Interestingly, in rural and urban patients who make it to hospital, there appears to be no statistically significant difference in time to definitive neurosurgical care, severity of injury, or pre‐hospital management to account for this disproportionate mortality. This relationship seems to only become clearly elucidated when pre‐hospital mortality is considered. Pre‐hospital mortality represents a hidden burden of disease in TBI, which requires more comprehensive characterisation. This pre‐hospital mortality burden is strongly associated with polytrauma status, so focused treatment of this complex disease in itself likely to improve TBI outcomes [9]. Author Contributions Zsolt J. Balogh: conceptualization, methodology, project administration, writing – review and editing, supervision. Maria Joseph: data curation, formal analysis, validation, writing – review and editing. Richard Ferch: conceptualization, supervision, writing – review and editing. Mitchell A. Hansen: conceptualization, methodology, supervision, writing – review and editing. Gareth Rutter: data curation, funding acquisition, investigation, methodology, project administration, writing – review and editing, writing – original draft.