•  
  •  
 

SMU Data Science Review

Abstract

Abstract. Current detection methods for Dementia and Alzheimer’s disease include cerebral spinal fluid (CSF) markers and/or the use of positron emission tomography (PET) imaging, both being high-cost, highly invasive testing methods. The need for low-cost, minimally invasive methods to prescreen individuals for cognitive impairment has been a challenge for many years. Today’s costs associated with an annual screen for all adults 65 and above using current methods (CSF, PET) reach well beyond trillions of dollars per year. Motivated by the limited accessibly and high costs, an alternative tool presented within this paper demonstrates an effective rule out screening for Dementia and Alzheimer’s disease. Leveraging Electronic Health Records (EHR) data, low-cost computing, modern statistical modeling, and useable Machine Learning algorithms were able to derive a screening tool that effectively detects 98 percent of individuals without Dementia and Alzheimer’s disease. Approximately 43,000 EHR patient records’ totaling 5,000 patients from the University of North Texas Health Science Center were evaluated using this new rule out screening method, which consists of traditional Machine Learning models: Random Forest, AdaBoost, SVM combined with the application of Natural Language Processing of physician’s notes. The findings from this study help define a new paradigm in medical practice where an effective rule out screening method for Dementia and Alzheimer’s disease can be used as an initial screening tool. This study effectively cuts the cost of current detection methods by 75 percent, gives access to all adults age 65 and above while still leaving expensive methods as secondary lines of detection.

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Share

COinS