How artificial intelligence will impact the future of healthcare

By David Fletcher
Vice President – Innovations, Streamline Health, Inc.

Recently, my colleague looked at how healthcare analytics has changed over the past 10 years. Now, let’s look at the future of artificial intelligence in healthcare.

Analytics and artificial intelligence go hand in hand. There’s isn’t an aspect of healthcare that AI solutions won’t touch over the course of the next decade. We’ve amassed an enormous amount of clinical, billing and operational data and now we are getting machines to learn historical patterns in order to make predictions about future events. From improving revenue to developing new cures for ailments, AI is here to stay – and that’s a good thing.

From a patient care perspective, the possibilities are endless.

Learn how automated pre-bill coding analysis can protect and improve revenue integrity.

Artificial intelligence can expedite diagnoses

Diagnosing ailments requires a highly skilled medical professional, but even in this specialized field, AI technology is making strides. Speaking with Becker’s Hospital Review, GE President and CEO Charles Koontz reported that deep learning systems can analyze x-ray images to quickly identify pulmonary disease.

The AI uses a special algorithm to scan a library of images at the University of San Francisco, which it then uses to cross-reference a new patient’s x-ray results. Koontz noted that this solution could save lives by prioritizing high risk patients in the hospital queue.

Another promising AI technology is IBM’s Watson. Speaking with CBS, Dr. Ned Sharpless of the University of North Carolina said that Watson was able to read roughly 25 million scientific papers in a week – an otherwise impossible task.

Without Watson, Sharpless’s team of cancer researchers was unable to produce a comprehensive list of clinical trials that could potentially help patients with highly complicated cases. Watson can not only read and process millions of papers, but also scan the internet for ongoing clinical trials that could benefit at-risk cancer patients.

“IBM’s Watson read 25 million scientific papers in a week.”

Heart attack prediction is another fascinating way that AI can help patients. Science Magazine reported that researchers at the University of Nottingham, in the U.K., have developed an AI that can successfully predict cardiac events significantly better than expert guidelines developed by human specialists.

To test the AI, researchers fed it patient case data from 2005. The AI looked at 22 factors that may predict heart disease including patient ethnicity, kidney disease history, arthritis rate and more. The researchers compared the AI’s predictions to actual patient data from 2015 and found that the AI was better at predicting heart disease than the clinical guidelines of the American College of Cardiology/American Heart Association.

Two graduates of Stanford Thrun Lab created a similar algorithm that can differentiate between malignant carcinoma and benign seborrheic keratosis. The first is a potentially deadly form of skin cancer, and the latter is a harmless wart-like growth.

Interestingly, this algorithm was built with a Google-developed algorithm designed to differentiate between images of cats and dogs. Using this as a base, the researchers were able to train the program to find skin cancer. The result: a smartphone app that can help people diagnose skin cancer without paying for an expensive doctor visit.

In the near future, AI systems may be able to identify all kinds of ailments. They may also play an integral role in developing new treatments.

Artificial intelligence and drug development

AI systems can hold infinitely more usable data than a human brain, which means that an AI could develop new drugs faster than a team of capable humans. Quartz Magazine reported on an AI called AtomNet that promises to develop new drug treatments for dangerous diseases like Ebola and multiple sclerosis.

In essence, AtomNet uses its vast data sets of molecular structures to predict several potential drugs that would interact with molecular structures of pathogens. The drugs must then be tested in rigorous clinical trials before they are deemed safe for use. AtomNet’s creators, Atomwise, hope that with each iteration their AI will get better at developing safe treatments for ailments that have long stumped medical professionals.

These are a few examples of ways that AI is changing clinical practice but AI is also being leveraged in healthcare administration as well. Early attempts at administrative AI such as computer-assisted coding (CAC) use natural language processing (NLP) but that is no longer enough in today’s AI enabled world. For AI to be truly revolutionary, the technology solutions need to learn from historical data.

Some organizations dodged the CAC bullet

Computer-assisted coding isn’t exactly a form of artificial intelligence, but some in the industry promised that it would improve coder productivity and efficiency using a spell checker-like system called NLP. Many organizations that invested in CAC technology were burned by low return on investment.

In many aspects, CAC simply failed to deliver. Based on feedback we receive from clients, many coders actually saw a decrease in productivity, and the accuracy of codes changed only slightly. Indeed, some experienced medical coders were made to feel in awe of CAC and therefore were hesitant to contradict the computer’s suggestions – even when human coders knew something was amiss.

This confusion and failure to make good on promises is a big problem for organizations invested in CAC. Even small coding mistakes can lead to massive losses in revenue. Organizations that lacked the resources to implement CAC actually dodged a bullet.

Experienced coders need a more elegant solution.

Our cloud-based platform offers 100% automated analysis of inpatient and outpatient coding and charge data accuracy prior to billing. 

Why artificial intelligence needs to leapfrog CAC

Artificial intelligence with machine learning is the clearest way forward for advancing medical coding. Administrators can measure improvement in terms of productivity, efficiency, revenue discovery and compliance risk. An advanced solution shouldn’t just assist coders – it needs to be able to handle most of the low complexity work.

But this isn’t to say that AI should take over medical coding. Even the most complex AI in the world could not completely replace a skilled human coder. Current AI systems are simply not equipped to make judgment calls.

In an ideal world, AI would accomplish most of the work, leaving only highly complicated cases for human coders. This is exactly what Streamline Health® eValuator™ does. By auditing all charts prior to billing, eValuator™ acts as a highly skilled AI auditor, flagging any charts that are likely to have mistakes. This frees human coders to focus their skills on complex cases while simultaneously lowering the overall risk to the provider.

Our expert auditors at Streamline Health are generating data every day that we, in turn, are using to discover and tune algorithms that enable a pre-bill audit for every encounter. The AI future is now.

Developing new cures, predicting diseases and building administrative efficiencies are just a few of the myriad ways AI will change the healthcare industry. In another ten years, AI solutions may be the gold standard by which all others are judged.