Effect of demographic and injury etiologic factors on intensive care unit mortality after severe head injury in a low middle income country

Introduction

Low- and middle-income countries (LMIC) bear a disproportionate impact and burden of the global epidemic of head injury,[^1,2] with recent projections indicating a more dismal prospect in the not too distant future.[³] Road traffic accidents (RTA) are the most common cause of head injury related deaths and the World Health Organization estimates that between the years 2002 and 2020 the total number of traffic deaths will increase by 65% globally and by 80% in LMIC.[³]

Most head injuries are mild with only transient effects. However, when mortality or functional disability occur, they usually are the results of severe head injury.[⁴] After head injury, the risk of mortality among people living in LMIC is reported to be 100% higher than the risk of mortality among those living in high-income countries (HIC).[⁵] Furthermore, severe head injury has been identified as the most common cause of mortality after admission to the emergency room in Nigeria.[^6,7,8]

Multiple factors have been recognized as predictors of outcomes after moderate to severe head injury, including demographic,[^9,10] clinical,[¹¹] radiologic[¹²] and laboratory parameters.[^13,14] Their effects have been quantified and the information codified into risk scores, for example the corticosteroid randomization after significant head injury predictive model, which are appropriate to inform patient management.[^15,16,17] Since most of the studies were based on head injured patients resident in HIC, very little is known about the prognostic relevance of these factors in the setting of LMIC. It is likely that the nature of the relationship between predictors and outcomes may differ between both societies given the different patterns of epidemiologic presentation and management. For instance, severe head injury patients in LMIC are generally younger than their counterparts in HIC[²] and because of a longer life expectancy in HIC, head injury in HIC tends to show a bimodal distribution in age due to the effect of falls from a height in the elderly,[¹⁸] a pattern not seen in LMIC.

The objective of this study was to determine the effects of demographic and injury causative factors on mortality in severely head injured patients admitted into intensive care in a tertiary hospital of a LMIC.

Materials and Methods

The study was conducted at the University of Port Harcourt Teaching Hospital, a regional referral tertiary hospital in the Niger Delta region of Southern Nigeria. The hospital caters to a population of about 5 million. The records of all patients who were admitted into intensive care for severe head injury from 1^st January, 1997 to 31^st December, 2006 were retrieved and reviewed. Severity of injury was assessed using the aggregate score of head injured patients on the Glasgow Coma Scale (GCS) as at the time of presentation to the emergency department. Severe head injury was defined as GCS score ≤8.[¹⁹] For this study, we collected demographic data related to patient's age, gender, etiology of injury and mortality outcome at discharge from intensive care.

Descriptive analysis of the retrieved variables was performed using χ² test of proportion for analysis of categorical variables and the two-sample t-test for analysis of continuous variables. One-way ANOVA was used to test the hypothesis that age did not differ across etiologic factors. Logistic regression models were fitted to evaluate the association between age, gender, etiologic factors (predictor variables) and intensive care unit (ICU) mortality (outcome variable). The effect of age was estimated as the odds ratio (OR) over the difference between the 75^th and 25^th percentiles of cohort's age. We estimated the effect of age as the change in OR over the interquartile range, so as to allow for a more direct comparison of prognostic effect of age to those of gender and injury etiologic factors.[⁹] To test the likelihood that the effect of age on mortality is continuous and non-linear, we performed a likelihood ratio test to compare the main effects model (a model that assumes that the effects of predictors are linear) to a model, in which age was fitted using restricted cubic splines with four knots. Restricted cubic splines are smooth and flexible statistical functions that can be used to adequately approximate non-linear relationships and they have been recommended for prognostic analysis in traumatic brain injury to evaluate the effect of continuous variables on outcomes.[²⁰] To visually, examine the shape of the effect of age on ICU mortality, we obtained smoothed plots of estimates of the probability of ICU mortality by age based on the model with age fitted with spline function. Test of the overall interaction effect of age and gender on mortality was also performed. Estimates of effect sizes were expressed as OR with 95% confidence intervals (CIs). For all analyses, the level of statistical significance was set at the probability level of 0.05. The R software version 2.15.0 was used in performing all analysis, employing Harrell's rms library (R foundation for statistical computing, Vienna, Austria).

Results

Descriptive statistics

During the period under review, 231 patients were managed for severe head injury in the ICU, representing 38.7% of total ICU admissions. There were 169 males (73.2%) and 62 females (26.8%). The sex ratio was therefore 2.8:1. The mean age of the patients was 31.2 ± 15.5 years (range: 2-79 years). The proportion of patients who fall in the age bracket of 15-44 years was 64.9%. No statistically significant difference was seen on comparing the average age of males (32.20 ± 13.61 years) to that of females (28.32 ± 19.70 years), P = 0.092.

Etiology of injury

The causes of severe head injury were RTA in 194 (84.0%) patients, assault in 16 patients (6.9%), gunshot injury in 12 patients (5.2%) and fall from a height in nine patients (3.9%). Table 1 shows age and gender distribution of patients by etiology of injury. The mean age of patients were similar across etiologic factors (ANOVA P = 0.217). Males were at increased risk of severe head injury from all etiologic factors.

Outcome analysis

A total of 122 patients died in the course of ICU admission, representing a mortality rate of 52.8%. Post-ICU hospital stay for surviving patients was uneventful. Table 2 shows descriptive statistics of predictor variables by mortality outcome. Patients who died during ICU admission were likely to be older than those who survived (ANOVA P = 0.01). The results of multivariable analysis are as shown in Table 3. There was a statistically significant association between age and mortality (P = 0.01) in adjusted analysis. The risk of mortality increased by 56% between the ages of 21 years and 40 years, adjusting for patient gender and etiology of injury. Gender was not a statistically significant modifier of the effect of age in the model (P = 0.218). There was a significant departure from linearity in the effect of age (P = 0.033). Smoothed estimates plot of probability of ICU mortality against age indicated a U-shaped relationship between age and mortality [Figure 1]. The change point in the risk of mortality was around the age group of 20-30 years. There was no significant difference in the odds of mortality when males were compared with females (P = 0.817), adjusting for age and etiology of injury. The odds of mortality did not significantly differ across etiology of injury, adjusting for age and gender (P = 0.62). Patients who sustained gunshot injuries were at a higher risk of mortality compared with those who had a RTA, but this relation was statistically insignificant (OR: 2.21, 95% CI: 0.63-7.79).

Discussion

To our awareness, this study represents only a very few to describe the relationship between potential predictors and outcomes of severe head injury in a LMIC. Furthermore, it attempted to quantify the effect of demographic and etiologic factors on risk of ICU mortality. The pattern of head injury among study cohorts is typical of the picture seen in many LMIC, where head injury often occurred among young adults, commonly males who were involved in a RTA.[^{21,22,23,24,25}] Most patients in this study did not survive their head injury. The proportion of patients who died during ICU admission is similar to 56% reported by a study in Kenya.[²⁶] Other studies of severe head injury patients presenting to tertiary hospitals in other African countries[²⁷] and in other centers in Nigeria[²⁸] have reported higher mortality rates in the range of 70-90%. When contrasted with much lower mortality rates in more affluent societies, which are in the range of 16-38%,[^18,29,30] these findings underscore the impact of severe head injury on the most economically active populations of LMIC and the importance of putting in place effective accident prevention and injury control measures and organized trauma care systems, which are currently lacking in many of these countries, Nigeria inclusive.

Of note also are differences in management provided to affected patients in LMIC compared with HIC. In our setting, patients with severe head injury admitted into the ICU (which is a 10 bed unit with dedicated nursing and medical staff) received further management consisting of clinical monitoring for raised intracranial pressure, maintenance of adequate perfusion and oxygenation. Plain skull radiographs and cranial computed tomographic scan were often requested, but could not be performed in most patients because in our fee-for-service health-care system most patients are unable to afford these basic imaging investigations or services are just inaccessible for logistic reasons. Mannitol was administered for suspected raised intracranial pressure. Prior to 2005, patients with suspected cerebral edema received dexamethasone. Exploratory burr hole was performed if clinically indicated. More advanced, timely neurointensive care is available to patients in HIC.[^4,18]

We found a strong effect of age on the risk of ICU mortality. For long, age has been recognized as a critical factor in predicting outcomes after head injury.[^20,23] Among our patients, the risk of mortality increased by 56% over the 25^th and 75^th percentile of age (21-40 years). Analysis of a large cohort of moderate and severe traumatic brain injury enrolled into multinational randomized clinical trials conducted primarily in HIC found a 114% increase in unadjusted odds of 6 months functional disability over the same age percentile, in this case 21-45 years and 49% increase after adjustment for other predictors of poor outcome.[⁹] This previous study also found a continuous and non-linear effect of age as was the case in our present study. The u-shaped relationship between age and mortality risk among our patients is consistent with prior studies that documented higher mortality rates at extreme ages of life.[^31,32] The change point among our patients, which occurred in the age group of 20-30 years suggest that although adults in this age group are often the most vulnerable to a head injury in LMIC,[²⁶] they are most probably the least likely to experience ICU mortality. A similar analysis in HIC found a change point for functional disability at 6 months at the older age group of 30-40 years.[⁹]

In the present study, gender was not a significant predictor of ICU mortality. A number of previous studies have examined the prognostic role of gender under the premise that hormonal differences related to a neuroprotective effect of estrogen may modulate head injury outcomes.[^33,34,35] However, no consistent pattern has yet been identified. While a relatively old meta-analysis concluded that females experienced better outcomes after head injury, only one of the included studies used mortality as the end point.[³⁶] Some studies have reported a significant reduction in risk of mortality in menopausal patients compared with pre-menopausal patients, suggesting that female hormonal status may be relevant to understanding the prognostic effect of gender.[^34,35] More recent systematic review[³⁷] and meta-analysis[⁹] have concluded that women do not fare better than men functionally after moderate to severe traumatic brain injury.

The risk of mortality in study cohorts did not differ according to etiology of injury, which agrees with prior studies.[¹⁰] It is also likely this finding is artefactual as the small number of cases that is due to causes other than RTA in the present study indicates a lack of power to detect a statistically significant association. The study also has other noteworthy potential limitations. The study focused on severely head injured patients who survived to ICU admission at our center; hence the conclusions cannot be generalized to patients who died prior to admission or for long-term mortality outcomes. In addition, more robust findings could have been obtained in a multi-institutional study. Furthermore, the retrospective nature of our data may have potentially confounded the results, especially with regards to evaluating the effects of other predictors that could possibly be associated with ICU mortality. However, the focus in this study was on the effect of baseline demographic and etiologic factors.

Conclusion

The present study indicated a poorer outcome with advancing age after severe head injury among patients admitted into ICU in a LMIC. The effect of age was found to be continuous, but non-linear with a change point in the risk of mortality at an age group younger than that of more affluent societies. The effect of age was not modified by gender. The risk of ICU mortality was not related to patient gender or etiology of injury. Further research is needed to examine the temporal significance of these results in our setting and others in Nigeria and indeed LMIC and to explore the relationship of other predictors and outcome measures.

Source of Support:

Nil

Conflict of Interest:

None declared.