Leading insurers are increasingly relying on packaged AI solutions such as IBM Watson to automate responses to frequently asked questions and power chatbots in portals. In one case, a leading insurer reports being able to address all AEP additional call volumes without adding customer service representatives.
All insurers struggle to manage provider information. Blockchain provides an intriguing vehicle that will transform management of key provider information by decentralizing the maintenance and upkeep of provider data. A key challenge for insurers will be agreeing on a common facility and standard, rather than individually maintaining their own version of provider data truth.
Disruptors are well on their way to replicating large portions of what pathologists and ophthalmologists typically handle. A leading diagnostics firm is on the verge of releasing the equivalent of a digital AI pathologist to provide a preliminary cancer diagnosis of digital images of tissue, while an AI startup has created a similar facility for early blindness detection by reading digital eye scans. Both technologies report 94% (or better) accuracy in diagnoses when tested against large samples of past physical scans diagnosed by trained physicians. Obviously, regulatory impediments remain to using these solutions in everyday clinical settings, however it is clear AI pattern matching techniques already rival clinicians with extensive experience. It is only a matter of time before they stake their claim alongside human colleagues.
As smartphones grow more powerful, real-time monitoring for specific health events is more common. At a Fortune 50 health insurer and managed care company, blood sugar and insulin data is collected to create a customized health assistant for people with diabetes. Taking it a step further, some universities are using machine learning models to develop an artificial pancreas. If the patient takes regular readings, the data can be applied to train a recurrent neural net and identify behavior patterns that lead to emergency room visits.
Doctors have a limited set of data points related to each patient, but our mobile devices have the power to capture vast amounts of information. By combining social data with additional biometric data, we have the ability to detect risks earlier and feed better health recommendations – plus determine which behaviors lead to negative health issues.
Recent studies suggest over 18% of the adult population in the United States suffers from some mental illness. Over time, psychiatrists have identified common mental health ‘red flags’ their patients exhibited throughout the course of treatment. With the advancement of machine learning algorithms, a UN project is attempting to proactively analyze social media streams for specific words and phrases in context to help detect and diagnose mental health issues earlier than ever before. In a related effort, the use of neuro-linguistic programming (NLP) will help evaluate semantic coherence and syntactic complexity of texts produced by patients to identify indicators of psychosis-based disorders with around 75% accuracy.
Health insurance companies and hospitals hold vast amounts of personal health information (PHI), but what’s being done with this wealth of knowledge? Undoubtedly, patterns exist and could be identified with the use of AI to help flag high-risk patients and potentially save lives. Any such endeavor would require extraordinary caution to avoid legal concerns, but the societal benefits are worth exploring.
Real-life implementation of AI is not without obstacles – between regulatory concerns and data privacy issues, there are many hurdles to overcome. As researchers and developers continue to tackle this space, AI advancements will grow leaps and bounds over the course of the next decade. We will inevitably see the healthcare industry mature and patient outcomes improve.
Not sure on your next step? We'd love to hear about your business challenges. No pitch. No strings attached.
