There is a myriad type of structured and unstructured patient centered data that are featured with large volume and variety. The limitation of patients’ data is not being open source and strictly secured by federal (HIPAA) and internal regulations (Institution Review Board). HIPAA that passed in 1996 provides protection and confidential handling of protected health information, (1). Electronic healthcare record (EHR) system started by selling point of care dispensable prepackaged medications for physicians, then unveiled e-prescribing software, followed by addition of practice and care management systems, which adopted by most prestigious U.S. healthcare institutions (2). According to the Institute of Medicine at the National Academies of Science, physician diagnostic errors account for 10 percent of patients’ deaths, and 6 to 17 percent of hospital complications. In recent decades, malpractice claims payouts in the U.S. showed prevalence of 68.8 and 31.2 percent in outpatient and inpatient, respectively (3). Although, deep learning and artificial intelligence (AI) has solid contribution to diagnostic medicine (4), harnessing the power of AI has not been fully implemented in therapeutic medicine. Precision, personalized medicine is an unexplored field and requires widespread innovative AI approaches that require the replacement of current HIPAA regulation. Development of prescriptive analytic systems to provide optimum/best recommended therapy models for physicians while making their decision is inevitable. These innovative solutions would change the physician roles in healthcare industry can cause vanishing of some medical fields. Future progress may even cause vanishing of interventional radiologists and surgeons with AI trained robotics and applications.
Making a prognosis of particular disease was/is well correlated with physician experience (the senior the better physician). The shortness of descriptive data collection model (mostly name, gender, age, weight, height, patient complaint) limited the physician treatment options to be partly palliative. Later in the time with the development of genetic and computerized imaging science, volume and variety of data took attention of physicians who started considering holistic treatment approach and changed data storage options to computer systems. Electronic healthcare record (EHR) system started by selling point of care dispensable prepackaged medications for physicians, then disclosed e-prescribing software, and followed by addition of practice and care management systems (2) . Top 10 dominating electronic healthcare record (EHR) by overall market share, as reported by SK&A are Epic, eClinicalWorks, Allscripts, Practice Fusion, Next Gen Healthcare, GE Healthcare, Cerner (5). EHR enables physician to practice, tele-video chat with patients, communication with pharmacy, prescribe, review overall survival, complication, retrospective diagnose/exam, i.e. Prescription written through this portal enables dispensing of medicine by pharmacy. This system not only complies with HIPAA but also yield important insights in diagnostic field of medicine. EHR systems possess great potential for development of prescriptive analytics. However, the prevalence of diagnostic errors and malpractice issues enhanced the demand for AI solutions for precise, personalized medicine.
Descriptive analytics in health care system were/are of manually filled spreadsheets, computerized machines and databases. Currently, the de-identified data in EHR system is not limited with patient’s age, weight, height, race, gender, surgical operations, laboratory findings, prescribed medicine, alcohol/tobacco/drug usage, previous physical examination, acute/chronic disease, family history, patient complaint, blood sugar, cholesterol and enzyme level, tumor size, calcification in heart valves, genetic map, X-ray, ECG, EKG, CTs, MRI, and ultrasound images, i.e. Some of abovementioned current descriptive tools integrated with machine learning models to foretell/predict the disease from its signs and symptoms (diagnostics), which can be considered as predictive analytics and are not limited to mobile medical diagnostics, loT devices, and Chatbots. AliveCor’s Kardia App (featured with a sensor underneath the Apple watch/smartphone) feed visual EKG information to its machine learning model. Cognoa app predicts the risk level of a child to have autism or developmental delays. Babylon Health’s Chatbot provides recommendation to the patients based on the predicted symptoms found in the software. Healthy.io claims to offer FDA-approved home testing colorimetric urine analysis kit that identifies illness possibilities via scanned images. Zebra medical vision’s AI systems provide all-in-one solution for normal and abnormal findings that are automatically detected and analyzed for radiology (6). Arterys medical offers AI solutions for wide variety modality (CT, MRI, mamography, retinal, ultrasound) (7). Predictive analytics with early detection features provided solid solutions to diagnostic medicine and will likely to continue to grow and save the lives (8). Prescriptive analytics for treatment medicine still has not been fully explored. Prestigious U.S. hospitals have big data in their adopted EHR systems (like Epic); however HIPAA rules limit the access and postpone development of AI prescriptive solutions in medicine.
AI would change the treatment plan say from “pathway oncology” to “precise, personalized medical management”. It will reduce the errors made by doctors while executing their treatment plans and benefit them for not having malpractice cases, since the responsibility will be shared with AI solutions. Unleashing the data access to data scientists and further sub classification of current data stored in EHR would offer widespread treatment optimization tools for each individual patient. For instance, to develop personalized medicine, it would be very useful data know why a European patient with cardio valve insufficiency had bleeding while taking 80 mg Aspirin, and Boston Caucasian who have the same diagnose but has not showed any bleeding with 300 mg Aspirin.
Experience-based medicine will be vanished and replaced with an evidence-based, patient centric approach with the digitization of health care. Although, there are FDA approved useful library/open sources (DragoNN, Keras, TesnorFlow, Medical-ai-course materials, PyTorch, i.e.), Cancer Imaging Archieve, Genomic Data Commons Data Portal, EHR systems should deposit their large-scale databases to maximize the power of deep learning and AI (4). This progress will not only dramatically change clinical workflow (vanishing of family and primary care physician, diagnostic radiology) but also their role (reduces work time per patient from 30 minute to 5 minute, make physician to be health consultant rather than physical examiner). Another clear benefit of machine learning and AI solutions (that harness descriptive, predictive and prescriptive analytics) will be throwing diagnostic/treatment errors and malpractice issues to history, which overall increase the survival rate and create very profitable business model. This report recommends upgrading their database for recommending optimum treatment plans to each individual patient by using AI solutions. If HIPAA rules could have been loosened the rules to access to EHR systems, development of AI based precision, personalized medicine would be very beneficial for the patients.