Two chronic illnesses—heart disease and diabetes—cost the United States billions of dollars annually, yet the advancement of new technology and analytics have the potential to cut ‘costly and unnecessary’ hospitalizations, while simultaneously improving patient care and outcomes.
Researchers at Boston University’s Center for Information and Systems Engineering are part of an ongoing effort to harness the capabilities of machine-learning algorithms to problems in healthcare. Collected data and statistics indicate that hospitalizations due to heart disease and diabetes could be predicted a year in advance, with an accuracy rate of almost 82%, allowing providers the opportunity for early intervention.
Hospitals provide patients’ electronic health records (EHRs), comprehensive documents that contain all critical patient information, which would be utilized in conjunction with algorithms to predict who might face hospitalization. This in turn leads to hospital intervention, with the chance to treat the disease in a more aggressive manner in outpatient settings, in addition to avoiding burdensome and costly hospitalizations.
The team at Boston University found that using the entirety of a patient’s EHR, which can contain as many as 200 factors, leaders to superior prediction results than any previous studies. Moreover, the approach using an algorithm can be easily duplicated and scaled, and applied to a very large number of patients.
Both health and cost benefits of applying machine-learning analytics in medicine are vast; a study based on a year’s worth of hospital admissions, conducted by the U.S. Agency for Healthcare Research and Quality, estimated that 4.4 million of U.S. hospital admissions—a total of $30.8 billion in costs—could have been prevented: approximately half of all unnecessary hospitalizations.
Yet the efforts at Boston University are only ‘the tip of the iceberg;’ other companies like Google with extensive experience in collecting and analyzing data have demonstrated interest in the domain of machine learning. A host of technologies, including implantable medical devices, fit trackers, and smart watches already capture health data, and lifestyle and wellness choices.
The potential to predict future hospitalizations with over an 80% rate of accuracy through the sole use of medical records points to the enormous capabilities of machine learning, analytics, and data-driven personalized medicine.
One comment
30 million DR patients should be screened in the US every year. A very high percentage do not have access to screening. Smartphone based retinal screening for diabetic retinopathy connected to a machine learned data base is a reality. Would love to speak with researcher who are focused on retinal screening for observation of DR, Glaucoma and Hypertension.