Is the buzz around healthcare AI solutions largely hype, or can they deliver measurable benefits? Lest you think it’s too soon to tell, check out the following.
According to a new report from market analyst firm Frost & Sullivan, AI and cognitive computing will generate $150 billion in savings for the healthcare business by 2025. Frost researchers expect the total AI market to grow to $6.16 billion between 2018 and 2022.
The analyst firm estimates that at present, only 15% to 20% of payers, providers and pharmaceutical companies have been using AI actively to change healthcare delivery. However, its researchers seem to think that this will change rapidly over the next few years.
One of the most interesting applications for healthcare AI that Frost cites is the use of AI in precision medicine, an area which clearly has a tremendous upside potential for both patients and institutions.
In this scenario, the AI integrates a patient’s genomic, clinical, financial and behavioral data, then cross-references the data with the latest academic research evidence and regulatory guidelines. Ultimately, the AI would create personalized treatment pathways for high-risk, high-cost patient populations, according to Koustav Chatterjee, an industry analyst focused on transformational health.
In addition, researchers could use AI to expedite the process of clinical trial eligibility assessment and generate prophylaxis plans that suggest evidence-based drugs, Chatterjee suggests.
The report also lists several other AI-enabled solutions that might be worth implementing, including automated disease prediction, intuitive claims management and real-time supply chain management.
Frost predicts that the following will be particularly hot AI markets:
- Using AI in imaging to drive differential diagnosis
- Combining patient-generated data with academic research to generate personalized treatment possibilities
- Performing clinical documentation improvement to reduce clinician and coder stress and reduce claims denials
- Using AI-powered revenue cycle management platforms that auto-adjust claims content based on payer’s coding and reimbursement criteria
Now, it’s worth noting that it may be a while before any of these potential applications become practical.
As we’ve noted elsewhere, getting rolling with an AI solution is likely to be tougher than it sounds for a number of reasons.
For example, integrating AI-based functions with providers’ clinical processes could be tricky, and what’s more, clinicians certainly won’t be happy if such integration disrupts the EHR workflow already in existence.
Another problem is that you can’t deploy an AI-based solution without ”training” it on a cache of existing data. While this shouldn’t be an issue, in theory, the reality is that much of the data providers generate is still difficult to filter and mine.
Not only that, while AI might generate interesting and effective solutions to clinical problems, it may not be clear how it arrived at the solution. Physicians are unlikely to trust clinical ideas that come from a black box, e.g. an opaque system that doesn’t explain itself.
Don’t get me wrong, I’m a huge fan of healthcare AI and excited by its power. One can argue over which solutions are the most practical, and whether AI is the best possible tool to solve a given problem, but most health IT pros seem to believe that there’s a lot of potential here.
However, it’s still far from clear how healthcare AI applications will evolve. Let’s see where they turn up next and how that works out.