Trauma has become one of the most leading causes of death and disability in the world. Physiological features such as vital signs monitored during traumatic events can be informative about the dynamics of trauma which can help develop interventions. As a result, interest in flexible statistical approaches for monitoring vital signs has gained root of late in the field of statistics. This has generated a growing interest in flexible methods for modelling and making valid inference for such data sets.
An alternative non-parametric approach to flexible models attractive for vital signs data, which allows development of fast computational schemes is the Gaussian process regression. This project will develop Gaussian process regression methods with variational Bayes computation approaches, which is able incorporate individual’s environmental, physiological and trauma specific information for modelling traumatic systolic blood pressure.
Background of the Study
Traumatic injury is defined as the physical injuries of sudden inception and gravity which require quick medical attention.
Traumatic injuries represent a substantial public health problem that has the potential to result in long-term disability or death. Traumatic injuries have serious implications on the individuals and their families, and the Nation at large. According to WHO (2012), Traumatic injuries accounts for a significant proportion of the global burden of disease, causing 9% of all deaths worldwide.
Recent advances in wearable sensing technology and communication infrastructure, several prototype medical devices have been made available for tracking traumatic events both online and offline, leading to proliferation of time series data of traumatic incidents. With the increasing availability of such that, in many scientific applications, the field of statistics has assumed greater importance. Interest in flexible statistical approaches for monitoring vital signs has been engineered recently within the statistical community.
The outcome of trauma may cause systemic shock called “shock trauma”, and may need correcting physiological disorders and interventions to save life. Traumatic injuries occur through various mechanism that may result to an individual obtaining penetrating, blunt and burn form of trauma. These mechanisms include motor vehicle collision, fall, sports injuries, homicide, suicide, assault, bites, animal attacks, natural disasters and other physical injuries which can occur at home, on the street, or while at work.
For developing countries such as Ghana, injury contributes a larger proportion to cause of death and years of life lost (Krug et al. 2000). Comprehensive surveillance systems are expected to assist in effectively understanding the burden of both injury and acute medical presentations. In Ghana, no such surveillance systems are currently fully functional to allow researchers and policy makers to follow trends over time. For example, only a small percentage of pedestrian injuries are included in police statistics; not all injured persons receive formal medical care; and several other vital statistics are inadequately captured.
Physiological features such as vital signs monitored during traumatic events can be informative about the dynamics of trauma and its relationship with other factors; thus, providing a principled way for developing pragmatic interventions. Due to technological advancements, traumatic vital signs data varied complexity in terms of size, dimension, type and covariates informative about the vital sign dynamics, are available in diverse applications for data analysts to analyse. Nevertheless, standard statistical methods are highly challenged in terms of their direct application to such data sets. This has generated a growing interest in flexible methods for modelling and making valid inference for such data sets. This study will develop Gaussian process regression methods with variational Bayes computation approaches, which is able to incorporate individuals’ environmental, physiological and trauma specific information for modelling traumatic systolic blood pressure.
Statement of Problem
According to Greve and Zink (2009), assessing of vital signs such as systolic blood pressure (SBP), heart rate (HR) and respiratory rate (RR) help to understand the physiological response and their appropriateness to injury response in trauma patients’ injuries and depending on the readings can help to decide the management strategy for the trauma patient.
Injury-induced alteration in initial physiological responses such as hypertension and heart rate (HR) has a significant effect on mortality. Bhandarkar et al. (2017), investigate the injury-induced early blood pressure (BP) and HR changes and their association with mortality and found that mortality among female trauma patients were more than males. Patients with normal BP and HR had a lesser mortality rate as compared to patients with abnormal BP and HR findings. Also, mortality was higher among hypotension-bradycardia patients. They therefore concluded that specific precautions in prehospital care should be given to trauma patients with these findings (Bhandarkar et al., 2017).
Non-parametric approach to vital sign monitoring has gained much attention in literature. In the framework of similarity-based modelling, Pimentel et al. (2013) developed an empirical approach for modelling multivariate vital-sign of patients who underwent upper-gastrointestinal surgery for detecting deterioration in physiology. The scheme segments the vital-sign data into normal and abnormal groups based on their admission information. Since similarity-based methods rely on the knowledge of a normality model, the density of the normal group is established via kernel density estimation technique. Velardo et al. (2014) proposed kernel density (KDE)-based approach with automatic alert thresholds for modelling both univariate and multivariate vital-sign of chronic obstructive pulmonary patients in order to prioritize them for clinical review. The univariate approach sets threshold based on appropriate percentile of estimated cumulative distribution function of each vital-sign.
An attractive alternative non-parametric approach to flexible models for vital signs data, which allows development of fast computational schemes is the Gaussian process regression. A Gaussian process is natural generalization of multivariate Gaussian random vectors to infinite dimensions. They provide priors for unknown functional random variables for Bayesian non-parametric modelling. They are used for various modelling tasks involving physiological data (Rasmussen and Williams, 2006). Recently, Cheng et. al. (2018) developed a flexible statistical framework in the sparse Gaussian process regression framework with Bayesian scheme computation for medical time series data on hospital patients, and called their scheme MedGP. Khalid et al. (2013) developed an automated Bayesian Gaussian process approach for detecting deterioration in a univariate vital sign. To consider the possible correlations between multiple health indicators, (D?richen et al. 2015) proposed multi-task Gaussian process models for biomedical applications for monitoring multiple vital signs jointly. Though the above methods are appeal, no covariate information is incorporated into the model. However, the health of an individual is influenced by several factors including environmental, economic, social and biological.
Physiological features such as vital signs monitored during traumatic events can be informative about the dynamics of trauma and its relationship with other factors; thus, providing a principled way for developing pragmatic interventions. This project will develop Gaussian process regression methods with variational Bayes computation approaches, which is able to incorporate individual, environmental, physiological and trauma specific information for modelling traumatic systolic blood pressure.
Objectives of the Study
The specifics objectives of the study are to;
- Develop full Bayesian Gaussian process regression models for times series data of traumatic events
- Develop approximate Bayesian computation methods in the Variational Bayes framework
- Develop simulation-based solutions for the developed models
- Apply the developed methodologies on real traumatic data sets.
Significance of the Study
Health monitoring is useful in many ways. Firstly, deterioration in health can be detected on time for pragmatic intervention to be provided preventing many adverse events requiring emergency. This implies less pressure to intensive care units and reduction in cost of care. Secondly, the result will provide better knowledge of the characteristics involving trauma which will be useful for an adequate planning for care, costs reduction, and for the establishment of a preventive policy for the society.
Concepts and Definitions
- A traumatic event is an incident that causes physical, emotional, spiritual, or psychological harm that makes one feel threatened, anxious, or frightened.
- Trauma is an emotional response someone has to an extremely negative event (American Psychological Association: APA).
- Systolic blood pressure measures the pressure in one’s blood vessels when the heart beats (Centers for Disease Control and Prevention: CDC)
- A Gaussian process is a collection of random variables, any Gaussian process finite number of which have a joint Gaussian distribution (Rasmussen & Williams, 2006).
List of Abbreviations
- WHO World Health Organisation
- MoH Ministry of Health
- KATH Komfo Anokye Teaching Hospital
- AEC Accidents and Emergency Centre
- RTI Road Traffic Injuries
- IED Improvised Explosive Devices (IED)
- CDC Centers for Disease Control
- MVC Motor Vehicle Collision
- Organization of the Study
This thesis is divided into five chapters under the heading: Introduction, Literature Review, Methodology, Analysis and Summary, Conclusions and Recommendations.
Chapter one is the introduction of the thesis. It presents the background of the study, statement of the problem, objectives, significance of the study, definition of terms and the outline of the study.
The Chapter two reviews literature related to the study. It describes some studies already made in the Gaussian process and Bayesian paradigm.
Chapter three follows with the methodology. This entails the technique for data collection, data collection instrument or tool, population and sampling and data analysis.
Cite this essay
Traumatic Injury. (2019, Dec 05). Retrieved from https://studymoose.com/chapter-oneintroductiontrauma-has-become-one-of-the-most-example-essay