Article Text

Download PDFPDF

Research paper
Mortality and morbidity 15 years after hospital admission with mild head injury: a prospective case-controlled population study
  1. T M McMillan1,
  2. C J Weir2,
  3. J Wainman-Lefley1
  1. 1Institute of Health and Wellbeing, MVLS, University of Glasgow, Glasgow, UK
  2. 2Centre for Population Health Sciences, University of Edinburgh and Edinburgh Health Services Research Unit, Edinburgh, UK
  1. Correspondence to Professor T M McMillan, Institute of Health and Wellbeing, University of Glasgow, Gartnavel Royal Hospital, 1055 Great Western Road, Glasgow G12 0XH, UK; thomas.mcmillan{at}


Objective To investigate mortality rate in a population of adults admitted to hospital with mild head injury (MHI) 15 years later.

Design A prospective case control, record linkage study.

Participants 2428 adults with MHI and an equal number of community controls (CC) were case-matched for age, gender and social deprivation. A further control group admitted with a non-head injury was in addition matched for duration of hospital admission. Controls with a history of head injury prior to study entry were excluded.

Main outcome measures Death or survival 15 years poststudy entry.

Results Mortality per 1000 per year after MHI (24.49; 95% CI 23.21 to 25.79) was higher than in CC (13.34; 95% CI 12.29 to 14.44; p<0.0001) or ‘other injury’ controls (OIC) (19.63; 95% CI 18.43 to 20.87; p<0.0001). Age at injury was important: younger adults (15–54 years) with MHI had a 4.2-fold greater risk of death than CC; in adults aged over 54, the risk was 1.4 times higher. Gender and social deprivation showed a similar association with death in the MHI and control groups. Repeated head injury was a risk factor for death in the MHI group. The frequency of hospital admission with systemic disease preinjury and postinjury was higher in both injury groups than in CC and higher in MHI than OIC. Prospective data in the MHI group suggest an association between preinjury lifestyle and mortality. Causes of death after MHI were similar to those of the control groups.

Conclusions Adults hospitalised with MHI had greater risk of death in the following 15 years than matched controls. The extent to which lifestyle and potential chronic changes in neuropathology explain these findings is unclear. Lifestyle factors do contribute to risk of death after MHI and this finding has implications for lifestyle management interventions.


Statistics from


Head injury is very common with incidence rates in Europe estimated to average 235/100 000.1 A high mortality rate in the first year after admission to hospital with a severe head injury is well recognised.2 ,3 Less is known about later mortality, and even less about late survival after mild head injury (MHI), which accounts for 95% of all head injuries.4 Clinical guidelines tend to focus on early recovery and in general terms conclude that good recovery from any disability is expected within days or a few weeks after injury and recommend a brief follow-up comprising advice, education and reassurance.5 By the very fact that an injury is labelled in hospital as mild, there is reason for healthcare professionals and for the public to expect no reduction in life expectancy. Given the high incidence of MHI this deserves to be examined, especially in view of the elevated risk of death late after more severe head injury.4 ,6 ,7

The head injury population differs demographically from the general population, with over-representation of males, younger adults and those with greater social and economic deprivation.2 ,3 These features can result in poor rates of recruitment and high loss to follow-up; other limitations in studies of long term outcome after a head injury have included restriction to samples that are not representative of the head injury population such as those attending for rehabilitation, with head injury arising from sports injuries or military service.8 We present here the 15-year survival outcome of an unselected population of hospital admissions with MHI that takes account of these methodological shortcomings and in addition we consider history of hospital admission before and after injury.


  1. To determine the death rate in a population of people with an MHI 15 years after admission to hospital.

  2. To determine whether death rate is elevated after MHI in comparison with matched control groups.

  3. To determine factors associated with death and survival in a prospective population of people with MHI.

  4. To compare cause of death in people with MHI to matched controls.


A prospective case control, record linkage study.


Patients with an MHI were identified from a cohort that prospectively identified all young people and adults admitted to hospital in Glasgow between February 1995 and February 1996.2 The Glasgow Coma Scale (GCS) score at the time of admission was used to identify those with an MHI, defined by a total score of 13–15.2 The data in the present study were obtained by electronic tracing of these patients by the Information Services Division (ISD) of the National Health Service for Scotland.

Information was obtained about participant health as reflected in admissions to hospital in the 15 years before and 15 years after the admission for MHI. For each admission, data were extracted from Scottish Morbidity Records-01 (SMR-01) concerning the date, duration of admission and diagnoses as classified by WHO International Classification of Diseases system in use at that time (ICD-9 or ICD-10). Information on the occurrence of death and the date and cause of death between the time of injury and our census point 15 years after injury was obtained from the records of the General Register Office for Scotland. ISD identified two control groups matched for the factors described below.

Mild head injury

The population in the original study comprised 2995 case records of all severities of head injury. Of these, 32 who were not resident in the Glasgow area and 235 with moderate-severe head injury were excluded as were 59 where severity of injury was unclassified.2 To prevent double counting of the same individuals, 132 records of readmissions with a further MHI during the recruitment period in 1995–96 were excluded. This left a total population of 2537, defined as Glasgow residents admitted to Glasgow Hospitals with MHI (classified as GCS of 13–15) and prospectively identified in the Hospitalised Head Injury Study.2

Of these 2537, we traced all but 21 (99.2%). Hence, we passed information on 2516 individuals to ISD for record linkage to SMR-01 on the basis of: first name; last name; gender; date of birth; postcode; hospital admitted to. Anyone who was not linked by ISD or where the record linkage was uncertain, it was checked via our paper records and, if necessary, by using the unique Community Health Index9 identifier by NHS Greater Glasgow, and then rechecked by ISD. By this process, all but six MHI cases were record linked, yielding 2510 for case control matching (figure 1).

Figure 1

Flow chart of mild head injury patients included in the study.

Figure 2

Kaplan-Meier Curves: Cumulative deaths with 95% confidence intervals.

Case control groups

The ISD created two control groups matched to the MHI group. The community control (CC) group was identified using unique Community Health Index numbers. It was matched on a case by case basis to the MHI group by gender, age and social deprivation (matching to postcode). The ‘other injury’ controls (OIC) group comprised people admitted to Greater Glasgow hospitals because of an ‘other’ injury not involving their head, in the same year as the MHI cohort. They were also matched to the MHI group by gender, age, social deprivation (Scottish Index of Multiple Deprivation 2006 (SIMD 2006)10 quintiles derived from postcodes) and duration of admission. Duration of admission was used as a proxy to match severity of admitting condition. OI was defined by diagnostic codes from the ICD-9 and ICD-10 as detailed previously.4 Deaths (all groups) were considered using the main diagnostic categories listed in ICD-9 and ICD-10. In the MHI and OI groups, only cases who survived the index injury until after hospital discharge were included. This avoided inclusion of cases who died from other causes during their admission.

Matches to the 2510 people in the MHI group were found for all but 82 cases. Only MHI cases with both OIC and CC matches were included in analyses. In this way, there were three groups, each with n=2428 for analyses. This represents case control matching to 96.5% of the 2516 MHI cases who were traced and to 95.7% of the 2537 people in the MHI population overall (see figure 1).

Statistical analysis

Survival was summarised in the three groups using Kaplan–Meier survival curves and compared using a log-rank test, stratified for matching indicator. The proportional hazards assumption of the Cox model was not satisfied for group comparisons or covariate associations and so logistic regression models for death at 15 years were fitted. Where data from all three groups were analysed, the logistic regression model was stratified by matching indicator. Associations are reported as ORs (95% CIs). Within each study group, associations among the age, sex and social deprivation matching variables and death at 15 years were assessed in a single logistic regression model. Then, in a separate logistic regression model for each matching variable, data from all groups were modelled, assessing the interaction between matching variable and group to compare formally across groups, the strength of its association with death. Finally, given the prospective nature of the head injury study, some preinjury information obtained at the time of injury was only available for the MHI group (previous admission with head injury, brain illness or mental health treatment; physical limitation; habitual alcohol excess; homelessness). These factors and the age, sex and social deprivation matching variables were entered into a forward stepwise multiple logistic regression model for death at 15 years in the MHI group. Death rates per 1000 per year with exact 95% CIs were calculated to allow interpretation of results on the absolute as well as the relative scale.

Counts of hospital admissions and the number of days of inpatient stay prestudy and poststudy entry were analysed by Poisson regression. A mixed model was applied to accommodate the matching of participants across groups. Associations from the Poisson regressions were reported using rate ratios and 95% CIs. Differences across groups in the change in admission rates between the prestudy and poststudy entry periods were assessed by testing for a group by study period interaction.


Demographic and clinical factors

The groups were matched for age, gender and SIMD 2006 quintile. The median age in 1995 was 39 years (range 14–98) for hospitalised groups and 41 (range 15–98) for CC; in each group, 1868 were male (76.9%). In each group, the majority were from the more deprived SIMD quintiles 4 and 5 (76.8%) with 14% in quintiles 1 and 2. The MHI was due to a fall in 1052 (43.3%), assault in 881 (36.3%), road traffic accident in 216 (8.9%) and other cause in 276 (11.4%). The cause was unknown in three individuals. The OI was due to a fracture (760; 31.3%), open wound (740; 30.5%), abrasion/contusion/laceration or tear (470; 19.4%), dislocation or sprain (150; 6.2%) or other cause of injury (308; 12.7%).

In the MHI group, the death rate per 1000 per year was 24.49 (95% CI 23.21 to 25.79). Over the 15-year follow-up, 36.7% (891/2428; 95% CI 34.9 to 38.7) of the MHI group had died. The death rate was higher in the MHI group than in either control group over years 1–15, and almost twice as high as in the CC group (13.34; 95% CI 12.29 to 14.44; p<0.0001). The death rate in the OIC group (19.63; 95% CI 18.43 to 20.87; p<0.0001) per 1000 per year was higher than in the CC group (figure 2).

Increasing age (p<0.0001), male gender (p<0.01) and greater social deprivation (p<0.001) were each associated with greater risk of death in each of the three groups. In a separate logistic regression model for each matching variable, the interaction between variable and group was not significant for gender (p=0.93) or deprivation (p=0.43); it was only significant for age category (p=0.0057). This significant effect is due to a weaker association between age and death in MHI than in CC.

The absolute death rates are higher in both younger and older MHI age categories than in controls (table 1). However, compared with the CC group, the relative risk of death after MHI is 4.2 times higher in younger and 1.4 times higher in older age categories. This difference in age-related risk compared with CC is present, but less marked in OIC (2.4 times greater in younger and 1.3 greater in older). As can be seen from table 1, in those aged <55 at study entry, the death rate is higher in the MHI than in the OIC group.

Table 1

Deaths in 15 years per 1000 per year by age category (95% CIs)

Characteristics of the MHI group

In addition to age and social deprivation, the presence of habitual alcohol excess at the time of injury (death rate per 1000 per year: present 35.34, 95% CI 33.01 to 37.66; absent 18.81, 95% CI 17.07 to 20.63), the number of previous admissions with head injury (one or more 28.36, 95% CI 25.58 to 31.20; none 23.36, 95% CI 21.92 to 24.83) and preinjury physical disability (present 41.32, 95% CI 38.78 to 43.78; absent 15.85, 95% CI 14.26 to 17.52) independently predict death at 15 years. The regression model comprising these five factors accounts for 49.6% of the variance (table 2).

Table 2

Association between patient characteristics and death at 15 years for the MHI group only (multiple logistic regression: forward stepwise variable selection)

Systemic disease, repeated head injury and mortality

In addition to the demographic matching of control groups, preinjury health history was considered by comparing hospital admissions with systemic disease across groups (table 3). Systemic disease was defined as any admission not diagnosed as external injury. The MHI group had more admissions to hospital preinjury with systemic disease, but not for more days in total than the OIC group; both hospitalised groups were admitted more often and for more days than CC (table 4).

Table 3

Mean (SD), median and range (in square brackets) for the total number of hospital admissions with systemic disease in 15-year time periods before and after study entry

Table 4

Group differences in incidence and duration of hospital admission with systemic disease for 15 years before and 15 years after study entry (Poisson regression)

Poststudy entry the MHI group was admitted to hospital with systemic disease more often but for shorter periods of time than the OIC group. Both of the hospitalised groups were admitted more often and for longer with systemic disease than the CC group poststudy entry (table 4).

A change in admission patterns between prestudy entry and poststudy entry might be expected if the hospital admission at the time of study entry has a pivotal effect on health. In comparison with CC over the same time period, there was a greater increase in the number and duration of admissions due to systemic disease in the MHI and OIC groups postinjury versus preinjury (table 5).

Table 5

Group differences in incidence and duration of hospital admission with systemic disease from 1980 to 2011, comparing prestudy versus poststudy entry to community controls (Poisson regression)

There might be a cumulative effect of repeated head injury on mortality outcome. Data on hospital admissions with head injury were available poststudy entry for all groups and prestudy entry for the MHI group only (all controls were selected on the basis of no head injury prestudy entry). The MHI group was admitted more often with a further head injury between study entry and 2011 than the CC (ratio 18.72, 95% CI 12.03 to 29.12; p<0.0001) or OI (ratio 3.77; 95% CI 2.99 to 4.76; p<0.0001) groups. The total length of stay for head injury as a result of further head injuries was longer in the MHI than in the control groups and although less markedly so, longer in OIC than in CC groups (table 6).

Table 6

Total duration of admissions with further head injury after study entry; frequency (%) per group

Cause of death

The six categories of cause of death in table 7 comprise 93% of deaths in the MHI group. Less than 1% of deaths in each group were accounted for by any other single ICD category. Death rates were higher in the MHI and OIC groups than in the CC group for all causes except neoplasm. The MHI group had more frequent deaths the OIC group for digestive and mental/behavioural causes. The OIC group had more frequent deaths than the CC group for all causes except circulatory and neoplasm (table 7).

Table 7

Cause of death by group over 15 years (rate per 1000 per year and 95% CI)

The frequency of death from self-harm (ICD-9 E950–958; ICD-10 X60–84, Y87.0) after hospital discharge was low in MHI (0.60 per 1000 per year), OIC (0.52) and CC (0.19) groups. Death from ‘undetermined intent’ (ICD-9 E980-989; ICD-10 Y10-34, Y87.2) after discharge could for some have resulted from an intention to self-harm. After combining the numbers with undetermined intent and those with self-harm, the rates remain low (MHI 1.07; OIC 0.74; CC 0.22). In all, 148 people in total died with dementia as a primary, secondary or tertiary cause (MHI 1.32 per 1000 per year; OIC 1.35; CC 0.91). If considering only those aged under 65 at injury (and hence less likely to have dementia at the time of injury), the dementia rates are MHI 0.37/1000/year, OI 0.17/1000/year and CC 0.14/1000/year. Eleven people died from seizure disorder as a primary cause (seven MHI, three OIC, one CC).


Main findings

People admitted to hospital after an MHI were at an increased risk of death for at least the next 15 years. Their death rate of 2.49% per year resulted in more than a third dying within the 15-year postinjury follow-up period. The cumulative death rate after MHI was higher than in either control group throughout the 15-year follow-up. In a population with this demographic profile, admission to hospital with an injury of any kind was associated with greater mortality after discharge, but admission with MHI adds to this risk. The accident groups are distinguished by the enhanced risk of death in the MHI subgroup aged under 55. Overall, the MHI group has a history of head injury prestudy and poststudy entry, and differs from both control groups by having more repeat head injuries (at least poststudy entry) and more frequent hospital admissions in general. In this way, the MHI group was less healthy preinjury and postinjury than controls, at higher risk of cumulative effects of head injury and at greater risk of death. Some information, available only for the MHI group, supports a view that preinjury this population have less healthy lifestyles, given their greater likelihood of disability and history of head injury prestudy entry. Following the index MHI, the trajectory of ill health increased suggesting a pivotal effect of this hospital admission, although this effect was not specific to MHI as it was also found in the OIC group. Death in the MHI group resulted from a wide range of causes, again consistent with a less healthy lifestyle. What the present study has not been able to identify is whether differences in death rates between MHI and OIC groups reflect general health and lifestyle factors only, the development of long term neuropathology associated with head injury, or both.

Comparison with other studies

There is very little evidence on late mortality, years after an MHI. A multicentre US sample reported a standardised mortality ratio of 1.73 for adults with MHI,6 which is not dissimilar to the ratio of rates per 1000 in the present study when comparing MHI with CC (1.84). A few studies have reported long term survival after severe head injury, and suggest increased mortality rates.7 ,11 Two further studies reported an increased risk of death in adults up to 13 years years after mild to severe head injury.4 ,6 In two papers, we reported an elevated risk of death after mild and severe head injury at 7 and at 13 years after injury; these studies overlap with the present study to the extent that MHI included were a representative sample of the population with MHI reported here.4 ,12

Associations between factors present at the time of injury and subsequent mortality have been investigated in several studies. Older age at injury has been associated with higher rates of disability and death after head injury,13 ,14 but studies here fail to control for the strong association in the general population between age and risk of death. In the present study, we used age-matched controls and found that both younger and older adults with MHI were at greater risk of death than controls. Although the absolute rate of death was higher in older adults, if taking into account the age-associated rate of death in CC, the relative association between MHI and death was much greater in younger adults. This is consistent with mortality findings recently reported in a cohort with mild-severe head injury, where a twofold greater risk of death was found in younger adults with MHI compared with controls.6

The lack of evidence for an effect of the extent of social deprivation or of gender on mortality risk compared with the matched CC is consistent with our earlier findings in a cohort comprising the broad range of head injury severities.4 It has been argued both that female15 and male7 gender is associated with poorer late outcome after head injury. Our results suggest that after adjusting for age and social deprivation, the risk of death for females with MHI over the 15-year period was reduced by a third compared with males. However, this effect was not specific to head injury, given the similar association in hospitalised OIC and CC groups.

Strengths and limitations

The study was highly successful in tracing and record linking more than 95% of the population admitted to Glasgow Hospitals with MHI in a 12-month period in 1995–1996. The features of the study population are typical of those reported for head injury, with a majority male and with greater social deprivation.1 ,3 The definition of MHI used here (GCS score of 13–15) is very widely employed. Duration of post-traumatic amnesia is often only accessible retrospectively and is often used as an index of severity. It was not recorded in this study and by this index, some of the population may have been judged to have had an injury that was worse than ‘mild’.16 The population was however typical of those treated as though they were mild and represent the mild end of the head injury severity spectrum attending hospital, given their GCS and short hospital stay. The extent to which an excessive alcohol habit or physical disability explains the greater risk of death in the MHI group cannot be stated with certainty because this information was not available for the demographically-matched controls. Finally, our findings represent a population from Glasgow in 1995 and there is a need to consider the extent to which they may generalise geographically. Health in the West of Scotland is recognised to be poorer than found in much of Europe.17 The reasons for this are not precisely clear, but are thought to reflect de-industrialisation; including the institution of economic policies in the 1980s that affected employment, housing and culture differently in UK regions and in this way affected health behaviours.17 ,18 This model is thought to have potential relevance to other economies in the Western developed world that have faltered in recent years.17 ,18

Why is mortality risk increased after MHI?

Although preinjury lifestyle factors do not entirely explain the selective changes in health pre-MHI and post-MHI, they are likely to contribute to mortality risk. Preinjury history of habitual alcohol abuse or physical disability was associated with increased risk of death and both are known as risk factors for mortality after severe head injury.6 ,7 ,19 ,20 On the basis of admissions with systemic disease, both hospitalised groups were less healthy than CC preinjury, and health as defined in this way became selectively worse in the MHI group postinjury, supporting this interpretation. Is it the case that the MHI provokes lifestyle changes that further increase the risk of death or does the head injury directly cause chronic pathology? The pivotal effect of admission to hospital with MHI on the frequency of later hospital admissions supports the former interpretation. There is also growing evidence for the latter including from neuroimaging21 ,22 and neuropathology studies on MHI. Bigler23 reported a single MHI case who presented with cognitive impairment and died of cardiovascular disease 7 m postinjury; autopsy revealed microscopic evidence in the absence of gross neuropathology for neuroinflammatory changes consistent with trauma. Recent reviews point to chronic neuropathology after repeated concussions in sports including in boxers, where autopsy studies report amyloid β, white matter degeneration and cell loss.24 These findings have resonance with the present study where the risk of repeat head injury was 18.72 times higher postinjury in the MHI group than in CC. There is also an important parallel between clinical outcome studies and neuropathology findings, with approximately a third of patients demonstrating a dynamic deterioration in outcome in the years following a head injury8 and a similar proportion showing chronic and marked neuroinflammation many years after injury in autopsy studies.25 Although controversial, several studies point to an increased risk of neurodegeneration leading to dementia including after MHI.26 ,27 The overall incidence of dementia in the present study was too low to give sufficient statistical power to test this hypothesis. The position is the same for self-harm/harm of undetermined intent and seizures.

It may be that changes in lifestyle and/or chronic neuropathology increase the risk of systemic disease and death after MHI. It is known that survival from head injury can lead to chronic illness28 and there is evidence for the occurrence of systemic disease (especially cardiovascular and respiratory) in more than half of cases after severe head injury.29 The association between systemic disease and less severe head injury is less well evidenced, although Hilz et al30 recently reported reduced autonomic cardiovascular modulation in adults 20 months after MHI in the absence of cardiovascular presentations. In the present study, participants in each group had a similar distribution of the major categories of cause of death reported in Scotland, with higher rates in the MHI group in all except for neoplasm. This is broadly consistent with recent long term mortality studies on severe head injury.4 ,6 ,7 ,19

Future research needs to further tease apart these potential causes of elevated mortality risk. This will require a greater understanding of long term neuropathological changes after head injury and after repeated head injury in prospective studies that include clinical, cognitive and biological data and information on accumulated life stress. This may enable us to predict more precisely who is at risk of poorer outcome and premature death.

Implications for practice

The MHI population is heterogeneous and the risk of mortality is likely to be predicated by pre, post and injury factors. For this reason, current advice to those attending hospital with MHI needs to be altered from a focus purely on monitoring and actions related to acute symptom recovery to include advice and offer of intervention regarding lifestyle management such as alcohol use, exercise and diet. Lifestyle interventions are recognised as key in secondary prevention and mortality reduction including for coronary heart disease, stroke and mental health,31–33 but have been neglected after head injury perhaps with dire consequences.


The authors thank Laura Marchbank and Michael Fleming of ISD, NHS National Services Scotland, for identifying the matched controls and performing the record linkage. Thanks are also due to Professor Sir Graham Teasdale for comments on an earlier draft.



  • Funding The study was funded by the Chief Scientist Office for Scotland; grant number CZH/4/674. CJW was supported in this work by NHS Lothian via the Edinburgh Health Services Research Unit.

  • Contributors TMMM was the primary investigator for the follow-up study and took the lead role in the study and in writing the manuscript. CJW provided statistical advice, analysed the data and contributed to the manuscript. JW-L was the research worker on the project and commented on the manuscript.

  • Competing interests None.

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

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.