AUTHOR=Dabek Filip , Hoover Peter , Jorgensen-Wagers Kendra , Wu Tim , Caban Jesus J. 

TITLE=Evaluation of Machine Learning Techniques to Predict the Likelihood of Mental Health Conditions Following a First mTBI

JOURNAL=Frontiers in Neurology

VOLUME=Volume 12 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2021.769819

DOI=10.3389/fneur.2021.769819

ISSN=1664-2295

ABSTRACT=OBJECTIVE: Limited research has evaluated the utility of machine learning models and longitudinal data from electronic health records (EHR) to forecast outcomes of mild traumatic brain injury (mTBI). The objective of this study is to assess and compare various data science and machine learning techniques in anticipating mental health (MH) conditions among active duty Service Members (SMs) following a first diagnosis of mTBI. 

METHODS: Demographics and encounter metadata of 35,451  SMs who have sustained an initial mTBI were obtained. Patient demographics, ICD-9/10 codes, enhanced diagnostic related groups, and other risk factors were obtained from the encounter records from the year prior to mTBI and were utilized to develop feature vectors representative of each patient. To embed temporal information within each feature vector, various window configurations were devised.  These feature vectors, in conjunction with various machine learning (ML) models, were used to predicted the presence or absence of any MH condition post-mTBI.

RESULTS: When evaluating the ML models, neural network techniques showed the best overall performance in identifying patients with new or persistent MH conditions post-mTBI.  Various window configurations were tested and results show that dividing the observation window into three distinct windows [-365:-30,-30:0,0:14] provided the best performance. Overall, the models described in this paper identified the likelihood of developing MH conditions at [14:90] days post-mTBI with an accuracy of 88.2% (AUC: 0.82; AUC-PR: 0.66). 

DISCUSSION: Through the development and evaluation of different ML models we have validated the feasibility of designing algorithms to forecast the likelihood of MH conditions after the first mTBI.  Patient attributes including demographics, symptomatology, and risk factors proved to be effective features to employ when training ML models for mTBI patients.  When patient attributes and features are estimated at different time window, the overall performance increase illustrating the importance of embedding temporal information into the models. 

CONCLUSION: Predictive analytics can be a valuable tool for understanding the effects of mTBI, particularly when identifying those individuals at risk of negative outcomes. The translation of these models from retrospective study into real-world validation models is imperative in the mitigation of negative outcomes with appropriate and timely interventions.