Data mining applied to the cognitive rehabilitation of patients with acquired brain injury

Acquired brain injury (ABI) is one of the leading causes of death and disability in the world and is associated with high health care costs as a result of the acute treatment and long term rehabilitation involved. Different algorithms and methods have been proposed to predict the effectiveness of rehabilitation programs. In general, research has focused on predicting the overall improvement of patients with ABI. The purpose of this study is the novel application of data mining (DM) techniques to predict the outcomes of cognitive rehabilitation in patients with ABI. We generate three predictive models that allow us to obtain new knowledge to evaluate and improve the effectiveness of the cognitive rehabilitation process. Decision tree (DT), multilayer perceptron (MLP) and general regression neural network (GRNN) have been used to construct the prediction models. 10-fold cross validation was carried out in order to test the algorithms, using the Institut Guttmann Neurorehabilitation Hospital (IG) patients database. Performance of the models was tested through specificity, sensitivity and accuracy analysis and confusion matrix analysis. The experimental results obtained by DT are clearly superior with a prediction average accuracy of 90.38%, while MLP and GRRN obtained a 78.7% and 75.96%, respectively. This study allows to increase the knowledge about the contributing factors of an ABI patient recovery and to estimate treatment efficacy in individual patients.

Injury scaling research. Final report.

National Highway Traffic Safety Administration, Washington, D.C.

Pedestrian head injury reduction concepts. Final report.

Mode of access: Internet.

Physical and psychological factors predict outcome following whiplash injury

Predictors of outcome following whiplash injury are limited to socio-demographic and symptomatic factors, which are not readily amenable to secondary and tertiary intervention. This prospective study investigated the predictive capacity of early measures of physical and psychological impairment on pain and disability 6 months following whiplash injury. Motor function (ROM; kinaesthetic sense; activity of the superficial neck flexors (EMG) during cranio-cervical flexion), quantitative sensory testing (pressure, thermal pain thresholds, brachial plexus provocation test), sympathetic vasoconstrictor responses and psychological distress (GHQ-28, TSK, IES) were measured in 76 acute whiplash participants. The outcome measure was Neck Disability Index scores at 6 months. Stepwise regression analysis was used to predict the final NDI score. Logistic regression analyses predicted membership to one of the three groups based on final NDI scores (< 8 recovered, 10-28 mild pain and disability, > 30 moderate/severe pain and disability). Higher initial NDI score (1.007-1.12), older age (1.03-1.23), cold hyperalgesia (1.05-1.58), and acute post-traumatic stress (1.03-1.2) predicted membership to the moderate/severe group. Additional variables associated with higher NDI scores at 6 months on stepwise regression analysis were: ROM loss and diminished sympathetic reactivity. Higher initial NDI score (1.03-1.28), greater psychological distress (GHQ-28) (1.04-1.28) and decreased ROM (1.03-1.25) predicted subjects with persistent milder symptoms from those who fully recovered. These results demonstrate that both physical and psychological factors play a role in recovery or non-recovery from whiplash injury. This may assist in the development of more relevant treatment methods for acute whiplash. (c) 2004 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

Predicting functional capacity outcome 12 months after hospitalized injury

Background: There is a recognized need to move from mortality to morbidity outcome predictions following traumatic injury. However, there are few morbidity outcome prediction scoring methods and these fail to incorporate important comorbidities or cofactors. This study aims to develop and evaluate a method that includes such variables. Methods: This was a consecutive case series registered in the Queensland Trauma Registry that consented to a prospective 12-month telephone conducted follow-up study. A multivariable statistical model was developed relating Trauma Registry data to trichotomized 12-month post-injury outcome (categories: no limitations, minor limitations and major limitations). Cross-validation techniques using successive single hold-out samples were then conducted to evaluate the model's predictive capabilities. Results: In total, 619 participated, with 337 (54%) experiencing no limitations, 101 (16%) experiencing minor limitations and 181 (29%) experiencing major limitations 12 months after injury. The final parsimonious multivariable statistical model included whether the injury was in the lower extremity body region, injury severity, age, length of hospital stay, pulse at admission and whether the participant was admitted to an intensive care unit. This model explained 21% of the variability in post-injury outcome. Predictively, 64% of those with no limitations, 18% of those with minor limitations and 37% of those with major limitations were correctly identified. Conclusion: Although carefully developed, this statistical model lacks the predictive power necessary for its use as a basis of a useful prognostic tool. Further research is required to identify variables other than those routinely used in the Trauma Registry to develop a model with the necessary predictive utility.

DEVELOPMENT AND VALIDATION OF THE ANTERIOR CRUCIATE LIGAMENT INJURY-RISK-ESTIMATION QUIZ (ACL-IQ)

Over 2 million Anterior Cruciate Ligament (ACL) injuries occur annually worldwide resulting in considerable economic and health burdens (e.g., suffering, surgery, loss of function, risk for re-injury, and osteoarthritis). Current screening methods are effective but they generally rely on expensive and time-consuming biomechanical movement analysis, and thus are impractical solutions. In this dissertation, I report on a series of studies that begins to investigate one potentially efficient alternative to biomechanical screening, namely skilled observational risk assessment (e.g., having experts estimate risk based on observations of athletes movements). Specifically, in Study 1 I discovered that ACL injury risk can be accurately and reliably estimated with nearly instantaneous visual inspection when observed by skilled and knowledgeable professionals. Modern psychometric optimization techniques were then used to develop a robust and efficient 5-item test of ACL injury risk prediction skill—i.e., the ACL Injury-Risk-Estimation Quiz or ACL-IQ. Study 2 cross-validated the results from Study 1 in a larger representative sample of both skilled (Exercise Science/Sports Medicine) and un-skilled (General Population) groups. In accord with research on human expertise, quantitative structural and process modeling of risk estimation indicated that superior performance was largely mediated by specific strategies and skills (e.g., ignoring irrelevant information), independent of domain general cognitive abilities (e.g., metal rotation, general decision skill). These cognitive models suggest that ACL-IQ is a trainable skill, providing a foundation for future research and applications in training, decision support, and ultimately clinical screening investigations. Overall, I present the first evidence that observational ACL injury risk prediction is possible including a robust technology for fast, accurate and reliable measurement—i.e., the ACL-IQ. Discussion focuses on applications and outreach including a web platform that was developed to house the test, provide a repository for further data collection, and increase public and professional awareness and outreach (www.ACL-IQ.org). Future directions and general applications of the skilled movement analysis approach are also discussed.