A Statistical Investigation Into the Management Strategy of Head Injured Patients

Anderson, Elaine Margaret (1988) A Statistical Investigation Into the Management Strategy of Head Injured Patients. MSc(R) thesis, University of Glasgow.

Abstract

Radiologists and neurosurgeons debate the need to refer all head injured patients for radiography. Whilst radiologists have constructed a management strategy for referring recent head injured patients for computerised tomography scanning and X-ray, neurosurgeons have devised guidelines for the management of patients with a head injury for which the presence/absence of a skull fracture is an important feature for admission to hospital. To examine whether individuals with a high risk of a skull fracture could be Identified, a study based on 3424 patients from the Accident and Emergency Department at the Monklands District General Hospital was carried out. Twelve variables for each attender were considered. By employing different discrimination procedures, it was hoped that a classification rule with low error rates could be identified to determine the presence of a skull fracture. The second section of this thesis deals with estimating the risk of a head injured patient developing an intracranial haematoma in an attempt to reduce the number of unnecessary admissions. Data were available on 8504 head injured patients from Accident and Emergency Departments in Great Britain plus 988 head injured patients from the Haematoma Study at the Southern General Hospital. The medical background to the two questions posed in this study are described in more detail in Chapter 1. In Chapter 2, a comprehensive examination of the twelve variables recorded at Monklands District General Hospital is carried out. Combining categories of some variables and the construction of the new variable Glasgow Coma Sum from the three variables Eyes Open, Motor and Verbal are discussed. The application of the linear logistic regression model to the two class discrimination problem (absence/presence of a skull fracture) is considered In Chapter 3. The performance of this method is assessed using error rates - sensitivity and specificity, the Youden Index and the area under the receiver operating characteristic curve. The linear logistic regression model employing the three variables - Glasgow Coma Sum, Headache/Vomiting and Facial Injury - seem to perform well In terms of the three aforementioned methods of assessment. At the end of this chapter, the linear discriminant is contrasted with the linear logistic regression model utilising the subsets {COMASUM, VOM, FAC} and {COMASUM, VOM, FAC, SCALP}. Both models appeared to perform equally well. In Chapter 4, an alternative procedure to the linear logistic regression model for the discrimination problem based on classification trees is described. A comparison of the two methods is made using the Brier Score. Although only small differences existed, the classification tree approach is preferable on the basis of being a simpler method in practice for allocating future patients and being able to handle missing data. To answer the second question, the absolute and relative risks of developing an intracranial haematoma are considered in Chapter 5. The calculation of confidence intervals for both types of risks are described. Two approaches for calculating the confidence Intervals for relative risks are considered. In Chapter 6, four features of the data set - Cause of Injury, Glasgow Coma Sum, Sex and Skull Fracture - are employed to estimate the risk of a head injured patient developing an Intracranial haematoma. Risks are also extended to Include children. Only Skull Fracture and Glasgow Coma Sum were useful for considering future management of a head injured patient. An adult head injured patient with no skull fracture and Glasgow Coma Sum of 15 has a low risk of a haematoma. Patients who have no skull fracture and Glasgow Coma Sum of 9-14 or 3-8 or have a skull fracture present with Glasgow Coma Sum of 15 have intermediate levels of risk, while patients with a skull fracture and Glasgow Coma Sum of 9-14 or 3-8 have a high risk of developing an intracranial haematoma.Children suffering from a head Injury with no skull fracture and Glasgow Coma Sum of 15 have a low risk of developing an intracranial haematoma, while children with a skull fracture and Glasgow Coma Sum of 3-8 have a high risk of developing an intracranial haematoma.Ammendments to the existing guidelines for admission or transferral to a Neurosurgical Unit plus appropriate management strategy for referring head injury patients for radiography are discussed In Chapter 7.

• Item Type: Thesis (MSc(R))
• Qualification Level: Masters
• Additional Information: Adviser: Gordon D Murray
• Keywords: Biostatistics, Medicine, Health care management
• Date of Award: 1988
• Depositing User: Enlighten Team
• Unique ID: glathesis:1988-76491
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Nov 2019 14:16
• Last Modified: 19 Nov 2019 14:16
• URI: https://theses.gla.ac.uk/id/eprint/76491