Article Text

Download PDFPDF

Death after head injury: the 13 year outcome of a case control study
  1. T M McMillan1,
  2. G M Teasdale2,
  3. C J Weir3,
  4. E Stewart1
  1. 1Psychological Medicine, Faculty of Medicine, University of Glasgow, Glasgow, UK
  2. 2NHS Quality Improvement Scotland, Glasgow, UK
  3. 3MRC Hub for Trials Methodology Research, Centre for Population Health Sciences, University of Edinburgh, Edinburgh, UK
  1. Correspondence to Professor T M McMillan, Psychological Medicine, Faculty of Medicine, University of Glasgow, Gartnavel Royal Hospital, 1055 Great Western Road, Glasgow G12 0XH, UK; t.m.mcmillan{at}


Background Head injury is common, and the risk of subsequent disability and death is high. Increased risk of death years after injury might be explained by factors associated with, but not a consequence of, the head injury. This unique prospective study investigates mortality over 13 years after injury.

Methods A cohort of n=767 with head injury was compared with two case control groups, matched for age, gender and deprivation, and in one control group, matched for duration of hospital admission following (non-head) injury.

Results Two-fifths of the head injury cohort had died. The death rate (30.99 per 1000 per year) was much higher than in community controls (13.72 per 1000 per year). More than 1 year after injury, the death rate in younger (15–54 years) adults was much higher than in community controls (17.36 vs 2.36 per 1000 per year) whereas in older adults the difference was more marginal (61.47 vs 42.36). Death rate was elevated after mild and after more severe head injury, including in younger adults after mild head injury (14.82 per 1000 per year mild head injury vs 2.21 community). Female gender and greater deprivation were not associated with increased death rates after head injury. Late after injury, deaths occurred from the same main causes as for the general population.

Conclusion Head injury is associated with increased vulnerability to death from a variety of causes for at least 13 years after hospital admission. There is a need to understand how head injury influences mortality, particularly in younger adults and after mild head injury.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Head injury accounts for the majority of trauma deaths in young adults.1 Previous work has largely focused on early mortality during hospital admission or in the first year after head injury, where a high death rate in those with severe head injury is well recognised.2–4 Some suggest that the risk of death continues to be raised thereafter but this is not a consistent finding.4–8 A definitive view is lacking because of important methodological shortcomings in existing studies.1 8 These include limitations associated with retrospective study design, potential biases from recruitment, including the use of unrepresentative samples from clinic or rehabilitation populations, military casualties or specific databases, or where the sample was restricted by age or severity of brain injury. These limitations may prevent agreement about factors that might be associated with survival such as gender and age at injury and with potential consequences of head injury such as self-harm.4–25

Surprisingly few studies have compared the occurrence of deaths after head injury with the death rate expected in the demographic population from which the head injured come. Where this has been attempted, an elevated risk of death more than 6 months after head injury has not been found consistently.4 7 8 26 However, these studies do not include comparators that are matched to the demographics of the head injured, who are often young, male and from socially deprived backgrounds.27 We report the results of a unique study in which late mortality in a prospectively identified cohort of people admitted to hospital after head injury was compared with matched hospitalised and community control groups.



A comparison of cumulative death rates between a cohort admitted to hospital after a head injury and two case control groups was conducted. The first control group was hospitalised for an injury other than head injury. The second was a community control group. Control groups were matched to the head injury cohort for age, gender and deprivation; the other injury group was, in addition, matched for duration of hospital admission and admission date. Information collected in head injury survivors at follow-up 1 year after injury was considered in relation to survival outcome.


The head injury group was recruited from people admitted with a head injury to the five general hospitals in Greater Glasgow. Survival outcome and cause of death were ascertained 13 years after admission from the General Register Office for Scotland (GROS). Previous reports describe outcome in this cohort at 1 and 5–7 years after injury.7 27 28


The Hospitalised Head Injury Study identified 2995 people admitted to a general hospital with a head injury between February 1995 and February 1996. Head injuries were identified by frequent visits by research workers to hospitals and by collaboration with hospital staff; in this way bias inherent in diagnostic coding inaccuracies was minimised.27

The original cohort that was selected for follow-up included all severe and moderate cases and a random sample of those with head injuries of mild and unclassified severity.27 Exclusions were not made on the basis of age. In this way the cohort selected for follow-up was able to represent the full range of severity of injury of hospital admissions with head injury over a 12 month period.

The Information Services Division (ISD) is Scotland's national organisation for health information, statistics and IT services. Tracing of the head injury cohort and identification of the two control groups at 13 years was undertaken by ISD, who provided the GROS information on the deceased. Information necessary for participant identification comprising name, surname, date of birth and postcode was obtained from the existing research database, and sent to ISD, except in two cases where date of birth was not recorded. ISD then identified survival outcome and cause of death from GROS data for up to 13 years after hospital admission. ISD identified two control groups, each of n=767, based on 1:1 matching with the head injury cohort. The other injury control group was identified from the Scottish Morbidity Records (SMR01). It comprised people admitted to Glasgow hospitals because of ‘other’ injury (not admitted due to head injury) in the same year as the head injury cohort and matched to the head injury group by gender, age, social deprivation (Scottish Index of Multiple Deprivation 200629 (SIMD) quintiles), date and duration of admission. Duration of admission was used as an approximation for severity of admitting condition. The community control group was identified via community health index numbers and focused on social deprivation by matching to the postcode area of the head injured group, by age and by gender.

Potential controls with a history of head injury were excluded. For the purpose of creating the control groups, a head injury was defined by ICD diagnoses in the following categories: ICD 9: 800–804 and 850–854; ICD 10: S02.0-S02.9 and S06.0-S06.9.

The other injury control group was defined by ICD-9 codes 805–848 and 860–959. Cases with the following codes were excluded: 925 crushing injury of face scalp and neck and 929 crushing injury of multiple unspecified sites, as intracranial injury would have been possible in such cases. Also excluded were injuries where there might have been associated anoxic or metabolic causes of brain injury: poisoning by drugs 960–979 (includes analgesics and psychotropic agents); toxic effects of non-medicinal substances 980–989 (includes carbon monoxide, solvents, alcohol and metals); ‘other and unspecified effects’ 990–995 (including heat exhaustion, drowning); and ‘complications of surgical care’ 996–999. In terms of ICD-10 codes the following were included: S00, S01, S03-S05, S10–99; T00-T19 (excluding T04; crushing injuries); and T33–35. Other codes not included refer to exclusions (such as poisoning) as for ICD-9.

Quantitative variables

Early and late effects of injury are distinguished by reporting outcome at 1 and 2–13 years. Age was an important determinant of outcome 7 years after injury.7 We subdivided age as ≤54 versus >54. These cut-offs were adopted on the basis of cut-offs used by the GROS in data published by them and used in our earlier work. Comparisons are made with GROS statistics for the Greater Glasgow or Scottish populations for 2002 (the midpoint of the 13 year follow-up) because the community control group was matched with the head injury group and therefore differs from the general population. Data on variables pertinent only to the head injury cohort were obtained from the existing research database, which in turn was obtained by research workers from hospital records and interview during admission to hospital at the study onset and at the 1 year follow-up.

Statistical methods

Cumulative death rates during follow-up are summarised by Kaplan–Meier estimates for each group. Within each study group multiple logistic regressions assessed the associations of the covariates age, gender and SIMD deprivation score with survival. The OR and its 95% CI were calculated for each covariate after adjusting for the others in the model. Interactions between each covariate and study group were tested to assess if associations varied significantly across groups. Overall and cause specific death rates over the entire 13 years of follow-up were summarised using the rate per 1000 per annum and its 95% CI.



Of the 767 cases in the original head injury cohort, 99% were successfully linked to survival outcome by ISD. The 10 cases that were not linked and their corresponding matched controls were omitted. Hence analyses are based on three groups each of 757. The control groups were comparable for age, gender and for deprivation on SIMD 2006 (see table 1).

Table 1

Demographic characteristics and matching of groups

Main findings

In the head injury cohort, 305/757 (40.3%) people had died within 13 years of injury. The proportion of deaths was higher in the head injury than in other injury (215; 28.2%) or community (135; 19.0%) groups (figure 1). This remained the case beyond the early period in year 1 where mortality was particularly high (deaths after year 1: 229, 33.6%; 168, 23.7%; 116, 15.7%, respectively; log rank test, χ2=65.35, df=2, p<0.0001). For other injury controls the death rate was higher than for community controls.

Figure 1

Kaplan–Meier curves with CIs for cumulative deaths over 13 years in head injury and control groups (n=757 per group at time 0).

The rate of death in the community control group over the 13 year period (13.72 per 1000 per year) was similar to that in the Glasgow population aged over 14 in 2002 (14.88 per 1000 per year). Death rate in the head injury group (30.99 per 1000 per year) was more than twice that for community controls or for the Glasgow population.

Factors associated with main findings

For all groups, older age at study onset was associated with greater risk of death at 13 years (table 2). A higher proportion of females than males died in each group (table 2). After adjustment for potential confounding effects from age and deprivation, an association between female gender and improved survival prospects was statistically significant for community controls. The absence of a gender advantage for females in the head injury and other injury groups did not reflect a greater risk of death in these groups relative to community controls, as shown by the non-significant interaction for gender between head injury (p=0.39) or other injury (p=0.12) and community controls. Relationships between survival and SIMD score in patients admitted to hospital for head injury or other injury were non-significant (table 2). A weak association was found in the community control group with a reduced survival prospect in people living in more deprived areas that persisted after adjusting for the potential confounding effects of age and gender (multiple logistic regression modelling).

Table 2

Associations between age at study onset, gender and Scottish Index of Multiple Deprivation with survival outcome 13 years after study entry

Higher rates of death during year 1 were evident in adults admitted to hospital after a head or other injury, than in community controls or Glasgow or Scottish populations (table 3). Death rates were higher relative to community controls and in younger than in older injury patients, particularly in the head injury group. More than 1 year after head or other injury, death rates were lower, and for older adults were closer to the rates for community controls and Greater Glasgow and Scottish populations. For younger adults more than 1 year after injury, death rates were higher for head injury (OR 9.40; 95% CI 5.35 to 16.50) and other injury (OR 4.32; 95% CI 2.40 to 7.80) groups compared with community controls. The death rate after head injury was higher than other injury (OR 2.17; 95% CI 1.50 to 3.14) (multiple logistic regression adjusting for gender and SIMD scores).

Table 3

Rates of death per year, per 1000 people in younger and older adults

Cause of death (see table 4)

The six main causes of death in the Greater Glasgow population account for 92–94% of deaths in each group (see table 4). Deaths from each cause were compared by group using 95% CIs for Kaplan–Meier survival curve estimates at 13 years. The frequency of deaths in the head injury group was significantly higher than for community controls for circulatory, respiratory, digestive, mental/behavioural and external causes and higher in the other injury group than in community controls for external causes.

Table 4

Primary causes of death per 1000 people per year for Greater Glasgow in 2002 and for each group 13 years after study onset

For 13 years after injury, the frequency of death from self-harm (ICD9 E950–958; ICD10 X60–84, Y87.0) did not differ between groups. Some cases of undetermined intent (ICD9 E980–989; ICD10 Y10–34, Y87.2) could have resulted from an intention to self-harm. Four cases of undetermined intent died during hospital admission as a result of their injuries and all of these were in the head injury group. Of the remaining 11 cases, eight were head injury and three were other injury controls. Hence the maximum number of deaths from self-harm after injury (self-harm plus undetermined intent) was 14, 11 and two for the head injury, other injury and community groups (rates of death per 1000 per year 1.42, 1.12 and 0.20, respectively). Five people died with dementia as a primary, secondary or tertiary cause. Of these, four were in the head injury group and one was in the other injury group. A further three (two in the head injured group and one in the other injury group) died from Korsakoff's disease. Seven people died from seizure disorder as a primary cause (six head injury and one other injury); of these, three head and one other injury died more than a year after injury and two (head injury) died less than 6 months after injury.

Additional findings in the head injured group

Further information was gathered prospectively for the head injury group. In year 1 the frequencies of death were higher after severe (28.3%) and moderate (12.1%) than after mild head injury (MHI: 5.6%) (χ2 test p<0.0001). In years 2–13, this effect was no longer significant, given a higher frequency of deaths in moderate (37.9%) and mild (32.4%) cases (severe, 32.4%) (χ2 test p=0.52). The death rate for MHI (27.85 per 1000 per year; 95% CI 24.60 to 31.22) was higher than for community controls (13.72; 11.67 to 15.96) and this difference was emphasised in younger adults (MHI 14.82; 11.75 to 18.29; community 2.21; 1.24 to 3.60) and was attenuated in older adults (MHI 59.13; 53.24 to 64.17; community 39.45; 34.37 to 44.50).

A previous history of head injury was not a risk factor for death (OR 1.05; 95% CI 0.74 to 1.50). A pre-injury admission for brain illness (ie, mental problems, stroke or other requiring medical attention; OR 2.22; 95% CI 1.43 to 3.42) or pre-injury mental health treatment (OR 1.94; 95% CI 1.26 to 2.99) was associated with death. Factors relating to lifestyle, namely history of physical disability (OR 6.41; 95% CI 4.44 to 9.26), habitual excess use of alcohol (OR 2.46; 95% CI 1.77 to 3.42) or living alone (OR 2.64; 95% CI 1.89 to 3.68) were associated with death outcome at 13 years.


Main findings

More than 40% of young people and adults admitted to hospital in Glasgow after a head injury were dead 13 years later. This stark finding is not explained by age, gender or deprivation characteristics. As might be expected following an injury, the highest rate of death occurred in the first year after head injury. However, risk of death remained high for at least a further 12 years when, for example, death was 2.8 times more likely after head injury than for community controls. Admission for non-head injury was also associated with a greater late risk of mortality but after head injury the risk of death was even higher. However, this finding does imply that factors associated with risk of injury in general or the consequences of an injury are related to longevity and emphasises the importance of comparison groups in studies of this kind. A second major finding is that in comparison with controls, younger adults are at greater risk of death after head injury, whereas for older adults, the risk relative to controls is marginal more than a year after injury. Thirdly, the rate of death is elevated even after a mild head injury and especially in younger adults.

Comparison with other studies

The finding that mortality after head injury is high during hospital admission or for up to a year after injury is not new.2 3 21 24 Studies with a later follow-up in survivors at 6 or 12 months are often cross sectional or retrospective and often include participants over a wide range of time since injury (eg, days to decades). Many use unrepresentative samples of the head injury population, such as those from neurosurgery or rehabilitation units, only include those with persisting disability or exclude cases by factors relevant to mortality, such as age or severity of injury.6 10–12 15 16 21 24 26 Few studies compare survival outcome after head injury with a comparator,4 6 12 26 and with the exception of our own, none used case matched controls. A recent 10 year outcome study reported no overall increase in mortality in survivors at 6 months after head injury in comparison with population data, similar in age and gender.4 When survival outcome in adults aged 16–65 years was considered separately, a slight but significant increase in mortality was found after head injury, and this effect might have been increased had there been a control for social deprivation.4 Another study, with a general population comparator, reported 1.95 times greater mortality in those surviving 1 year after injury. Here older people with head injury were at greater risk although age subgroups were not considered separately.26 The death rate in our cohort 1–13 years after head injury was higher than in these studies and, for example, was five times higher than in the study by Flaada and colleagues.4 Differences between studies in age distribution and rates of death in comparators may explain this, but only in part. Studies with26 and with no comparison group reported an association between older age at head injury and mortality.11–13 18–21 23 Our study also showed a high rate of death in older than in younger adults but relative to controls there was a high risk in younger adults and a marginal effect in older adults. We found no adverse risk associated with gender. In the literature there is no consensus9–17 and studies have not used comparators.

The main causes of death after head injury were similar to those found in the general population. Others also report high rates of death from respiratory, digestive25 and circulatory causes.5 6 12 The particularly high death rate from external causes is partly associated with deaths early after injury, and partly from drug and alcohol associated deaths. Relationships between head injury and suicide are difficult to adduce from the present study. This is partly because of the low frequency of actual or possible suicide, partly because of difficulty in ascribing intent and partly because of the overlap between sociodemographic risk factors for head injury and for suicide. These issues are reflected in the lack of consensus in the literature.5 6 12 22–25 The low frequency of deaths from dementia or seizures similarly makes it difficult to interpret the findings of a higher frequency after head injury.

Factors influencing late mortality

The severity of head injury was not associated with survival outcome beyond the first year after injury. Even when head injury was classified as mild by the Glasgow Coma Scale, death rate was more than twice as high as in community controls. Admission to hospital for injury (or factors associated with admission) contributed to a poorer survival outcome. Poorer survival outcome after head injury was related to factors associated with lifestyle prior to the injury (physical limitations, habitual alcohol excess and living alone) and to a history of mental health problems, as also reported elsewhere.20 23

The particularly high rate of death in younger adults is important. Thirteen years after injury the death rate in those aged 15–54 years was more than six times higher than in community controls and this was not an artefact of gender or deprivation. Factors associated with hospital admission are relevant, but when controlling for this in addition the death rate remained more than twice as high in the head injury than in the other injury group. Beyond the early period of up to 1 year after injury, the death rate in those aged over 54 years did not exceed that found in community controls. The reason for greater vulnerability in younger adults is unclear but requires further consideration, especially given the particularly high risk of head injury in younger adults27 together with economic concerns over work force planning and demographic trends towards an ageing population.30

Strengths and limitations

The prospective nature of the head injury cohort, the long follow-up, the completeness of the follow-up data and the comparison with matched controls make this study unique in the world literature. The study was limited by reliance on GROS cause of death information and by the absence of qualitative information in the control groups whose data were obtained retrospectively.

Implications of findings and future work

Survival beyond the early period after head injury has received relatively little attention. Although mortality is increased by factors associated with admission to hospital after (any) injury, after a head injury the risk is greatly elevated. Demographic factors do not explain the risk of death late after head injury, and there is a need to further consider factors that might lead to health vulnerability after head injury and in this way explain the range of causes of death. The elevated risk of mortality after mild head injury and in younger adults makes further study in this area a priority.



  • Funding The 13 year follow-up was funded by the Chief Scientist Office (CZG/2/397). The original prospective study was also funded by the Chief Scientist Office (K/OPR/2/2/D229).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the NHS Greater Glasgow and Clyde Research Ethics Committee (08/S0710/82).

  • Provenance and peer review Not commissioned; externally peer reviewed.