Predicting lethal outcome after severe head injury -- a computer-assisted analysis of neurological symptoms and laboratory values

Classification and Regression Tree for Prediction of Outcome after Severe Head Injury Using Simple Clinical and Laboratory Variables

Many previous studies have constructed several predictive models for outcome after severe head injury, but these have often used expensive, time consuming, or highly specialized measurements. The goal of this study was to develop a simple, easy to use a model involving only variables that are rapidly and easily achievable in daily routine practice. To this end, a classification and regression tree (CART) technique was employed in the analysis of data from 345 patients with isolated severe brain injury who were admitted to Asclepeion General Hospital of Athens from January, 1993, to December, 2000. A total of 16 prognostic indicators were examined to predict neurological outcome at 6 months after head injury. Our results indicated that Glasgow Coma Scale was the best predictor of outcome. With regard to the other data, not only the most widely examined variables such as age, pupillary reactivity, or computed tomographic findings proved again to be strong predictors, but less commonly applied parameters, indirectly associated with brain damage, such as hyperglycemia and leukocytosis, were found to correlate significantly with prognosis too. The overall cross-validated predictive accuracy of CART model for these data was 86.84%, with a cross-validated relative error of 0.308. All variables included in this tree have been shown previously to be related to outcome. Methodologically, however, CART is quite different from the more commonly used statistical methods, with the primary benefit of illustrating the important prognostic variables as related to outcome. This technique may prove useful in developing new therapeutic strategies and approaches for patients with severe brain injury.

Brain and Spinal Cord Injury

Up to 50% of all trauma deaths in the United States involve significant injury to the brain or spinal cord. This chapter highlights the public health significance of traumatic brain and spinal cord injury and examines methodological issues in studies of the epidemiology of these injuries. It addresses methodological challenges in epidemiologic and clinical studies of brain and spinal cord injury, including difficulties in case ascertainment, differing approaches to brain injury classification, and measurement issues in brain injury severity and outcome scales. The chapter summarizes scientific literature addressing demographic and lifestyle risk factors for brain injury including age, sex, and alcohol consumption. External causes of traumatic brain and spinal cord injury are also discussed, including transportation-related injuries and increasingly recognized sports-related brain injuries.

Side effects and responsibility of medical informatics

Medical informatics systems have the ultimate goal to improve the quality of health care. However, these systems have also the potential to compromise the quality of health care if they are misused, intrinsically faulty or entail unexpected side effects. The purpose of the paper is to discuss examples from the experience of the author, where well-intended medical informatic applications proved to have potentially harmful effects or side effects. It is argued that medical informaticians (MI) have an extended professional responsibility, which covers not only the state of the art technical planning, an implementation of information processing systems in medicine, but also the final result for the patient. In order to discuss professional duties of medical informaticians, a modification of the Software Engineering Code of Ethics and Professional Practice developed by ACM/IEEE is proposed as a guideline. For several examples, these guidelines are used to analyze possible actions and professional responsibility.

Incidence of fever in the rehabilitation phase following brain injury

There appears to be a high incidence of fever after brain injury. The most common cause for fever is infection. The incidence of fever occurring as a result of hypothalamic thermoregulatory dysfunction after brain injury is less clear. This study retrospectively reviewed the charts of 286 subjects with brain injuries. Subject subpopulations were divided into traumatic brain injuries, anoxic brain injuries, and brain injuries resulting from aneurysmal subarachnoid hemorrhage. Fever events were described as any core temperature greater than 99.9 degrees F. Most subjects suffered a severe brain injury and had an average acute hospital length of stay ranging from 35.4 to 60 days. The average rehabilitation length of stay ranged from 38.4 to 45.1 days. Twenty-four percent of subjects experienced fevers, with each of the populations having similar occurrence rates. Unexplained fever events were found in the traumatic brain injury (7%) and aneurysmal subarachnoid hemorrhage (8%) subpopulations only. No unexplained fever event was associated with a temperature greater than 100.8 degrees F.

Plasma glucose levels and outcome after aneurysmal subarachnoid hemorrhage

Plasma glucose levels were studied in 616 patients admitted within 72 hours after subarachnoid hemorrhage (SAH). Glucose levels measured at admission showed a statistically significant association with Glasgow Coma Scale scores, Botterell grade, deposition of blood on computerized tomography (CT) scans, and level of consciousness at admission. Elevated glucose levels at admission predicted poor outcome. A good recovery, as assessed by the Glasgow Outcome Scale at 3 months, occurred in 70.2% of patients with normal glucose levels (< or = 120 mg/dl) and in 53.7% of patients with hyperglycemia (> 120 mg/dl) (p = 0.002). The death rates for these two groups were 6.7% and 19.9%, respectively (p = 0.001). The association was still maintained after adjusting for age (> or < or = 50 years) and thickness of clot on CT scans (thin or thick) in the subset of patients who were alert/drowsy at admission. Increased mean glucose levels between Days 3 and 7 also predicted a worse outcome; good recovery was observed in 132 (73.7%) of 179 patients who had normal mean glucose levels (< or = 120 mg/dl) and 160 (49.7%) of 322 who had elevated mean glucose levels (> 120 mg/dl) (p < 0.0001). Death occurred in 6.7% and 20.8% of the two groups, respectively (p < 0.0001). It is concluded that admission plasma glucose levels can serve as an objective prognostic indicator after SAH. Elevated glucose levels during the 1st week after SAH also predict a poor outcome. However, a causal link between hyperglycemia and outcome after delayed cerebral ischemia, although suggested by experimental data, cannot be established on the basis of this study.

The effect of glucose administration on carbohydrate metabolism after head injury

The role of intravenous infusion of glucose in limiting ketogenesis and the effect of glucose on cerebral metabolism following severe head injury were studied in 21 comatose patients. The patients were randomly assigned to alimentation with or without glucose. Systemic protein wasting, arterial concentrations of energy substrates, and cerebral metabolism of these energy substrates were monitored for 5 days postinjury. Both groups were in negative nitrogen balance, and had wasting of systemic proteins despite substantial protein intake. Blood and cerebrospinal fluid (CSF) glucose concentrations were highest on Day 1, but remained higher than normal fasting levels on all days of study, even in the patients who received no exogenous glucose. Although there were no differences in blood or CSF glucose concentrations in the two groups of patients, the glucose group had higher plasma insulin levels, with a mean +/- standard deviation of 14.8 +/- 7.3 microU/ml compared to 10.3 +/- 4.2 microU/ml in the saline group. The blood concentrations of beta-hydroxybutyrate, acetoacetate, pyruvate, glycerol, and the free fatty acids were higher in the saline group than in the glucose group. Cerebral oxygen consumption was similar in the two groups, while the cerebral metabolism of glucose and of the ketone bodies was dependent on whether glucose was administered. In the glucose group, glucose was the only energy substrate utilized by the brain. In the saline group, the ketone bodies beta-hydroxybutyrate and acetoacetate replaced glucose to the extent of 16% of the brain's total energy production. Cerebral lactate production and CSF lactate concentration were lower in the saline group. These studies suggest that administration of glucose during the early recovery period of severe head injury is a major cause of suppressed ketogenesis, and may increase production of lactic acid by the traumatized brain by limiting the availability of nonglycolytic energy substrates.

Outcome in 134 patients with prolonged posttraumatic unawareness. Part 1: Parameters determining late recovery of consciousness

A retrospective study of 134 patients in a condition of prolonged unawareness state (that is, in coma for over 1 month) following brain trauma was conducted in order to identify prognostic factors. Eight easily evaluated parameters were found to be significant for predicting nonrecovery of consciousness. The following six features were present during the early posttraumatic phase (that is, during the 1st week after trauma): fever of central origin; diffuse body sweating; disturbances in antidiuretic hormone secretion; abnormal motor reactivity; respiratory disturbances; and diffuse nonneurological injuries. The first three features were manifestations of hypothalamic damage. Two factors, evident at a late phase following injury (after the 1st week posttrauma), namely late epilepsy and communicating hydrocephalus, were also significant in predicting nonrecovery.

Improved outcome prediction based on CSF extrapolated creatine kinase BB isoenzyme activity and other risk factors in severe head injury

The present study of 43 patients with severe head injury shows that outcome prediction can be markedly improved by combining an appropriate marker of the degree of initial brain damage and other risk factors. The patients were classified into three groups according to their actual outcome after 6 months: death (22 patients); persistent vegetative state or severe disability (eight patients); and moderate disability or good recovery (13 patients). By applying stepwise logistic discriminant analysis to the patients' data, five significant risk factors were selected: degree of neurological damage assessed by cerebrospinal fluid (CSF) extrapolated creatine kinase BB isoenzyme activity, Glasgow-Liège Coma Scale score, age, incidence of thoracic injury, and intracranial pressure (ICP). Extrapolated creatine kinase BB activity had the highest prognostic ability (67%). Uncontrollable elevated ICP proved to be systematically associated with death, whereas its absence was not necessarily indicative of a favorable outcome. The combination of the five variables yielded a total prognostic efficiency of 91%. The percentages of correctly predicted patients for the three outcome groups were, respectively, 100%, 50%, and 100%. Thus, half of the persistently vegetative and severely disabled patients were identified by the selected factors.

The Glasgow-Liège Scale. Prognostic value and evolution of motor response and brain stem reflexes after severe head injury

In 1982, we developed a new coma scale, the Glasgow-Liège Scale (GLS), which combines the Glasgow coma scale (GCS) with the quantified analysis of five brain stem reflexes. After severe head injury, the two most important parameters for determining the degree of encephalic disturbances are motor responses (M) and brain stem reflexes (R). The object of this study was to further evaluate the prognosis ability of M and R on admission and during the first month after injury. The study is based on 141 patients. The highest score, during the first day, was less than or equal to 7 on the GCS and 12 on the GLS. Using a multiple group logistic discriminant analysis, we confirmed that, in the first 24 hours, the study of brain stem reflexes appears to be the one factor with the best prognostic ability. We also showed that the prognostic value of certain signs is optimal for a limited period. If, within the two variables M and R, recovery follows an identical pattern whatever the final outcome may be, the restructing speed differs for each outcome. M follows an exponential curve spread over a long period explaining its importance in the course of time. On the other hand, R follows a linear model with straight lines more or less parallel for each outcome. R evolves over a short period of time. These clinical findings give us the opportunity to discuss the physio-pathology of head injury.