The debates in Washington and on the presidential campaign trail seemed to never end, but as Den Howlett optimistically opined, a new administration with a rebellious streak means that we may be on the cusp of a healthcare revolution that is both "nightmare and bonanza" for technology firms.
Although the problems in healthcare, notably exploding costs, inadequate coverage and inefficient processes, are multifaceted and resistant to quick fixes, IT will play a significant role in the solution.
There are two areas especially ripe for disruptive innovation and where political, business and technology trends are coming together to catalyze meaningful change: medical records and clinical diagnostics.
Health data quagmire
Electronic health records (EHR) are the healthcare industry's version of a mirage: an information oasis that continuously seems to recede into the barren distance. For a variety of tragic reasons over the past 18 months, I have become more familiar with the state of EHR dysfunction than I would like, suffering through its failings up close.
Having been subjected to countless re-entry of the same basic personal information and medical history while simultaneously observing the primitive state of application UIs and witnessing the futility of medical professionals as they try and find routine patient information, I can say EHR systems resemble Access on Windows 3.1 more than Google on a smartphone.
The sad state of EHRs isn't news to the doctors that must abide them every day. A 2015 AMA survey found that:
...compared to five years ago, more physicians are reporting being dissatisfied or very dissatisfied with their EHR system.
More than half of respondents are dissatisfied with their EHR system, 43% said that EHR created "productivity challenges" while 54% found that they increased operating costs.
In an overly charitable understatement, the head of the American College of Physicians Division of Governmental Affairs and Medical Practice admitted:
While EHR systems have the promise of improving patient care and practice efficiency, we are not yet seeing those effects.
Frozen in a time computer users would rather forget, EHR UIs don't help, the systems mostly fail at their primary job of aggregating and sharing information. However, if there's one thing IT is good at, it's managing and acting upon lots of data. When millions of customers can quickly find long-lost email attachments, search and summarize years worth of Amazon orders and use IFTTT to automatically generate a notification every time a particular celebrity Tweets, EHRs present a golden opportunity for IT innovation.
IBM and the FDA have recently announced a partnership to study the use of Bitcoin-like distributed ledger blockchains for the exchange of "owner mediated data" from a variety of sources including EHR, clinical trials, genomic records, mobile devices, wearables and other connected health appliances.
By providing an open, trackable and unalterable record of data custody, blockchains address many significant security concerns that have impeded EHR data collection and collaboration. As Shahram Ebadollahi, VP at IBM Watson Health puts it,
The healthcare industry is undergoing significant changes due to the vast amounts of disparate data being generated. Blockchain technology provides a highly secure, decentralized framework for data sharing that will accelerate innovation throughout the industry.
(For more on healthcare blockchain, see this Q&A between Jon Reed and a data scientist for a cloud service targeting healthcare data interchange).
Clinical data analytics - Watson to the rescue?
Another area ripe for disruption and where IT innovation can significantly improve both provider efficiency and quality is the analysis of clinical data to assist in health diagnosis and treatment. IBM, through its Jeopardy-winning, question answering system, Watson, has also staked out a position in the burgeoning market for healthcare decision support.
Unlike EHR, clinical data is often unstructured - 80% by IBM's estimate - in the form of research reports, narratives of patient symptoms and medical imagery. Furthermore, health data is in the midst of its own data explosion, where large health networks manage tens of petabytes of EHRs, images and annotations and the aggregate nationwide reaching zettabyte scale.
The combination of size and disorganization create the ultimate needle-in-a-haystack problem for doctors and nurses trying to diagnose often-ambiguous symptoms, cull standard protocols and recent research to identify the cause and best treatment and not waste time and money while putting patients through uncomfortable and unnecessary tests. It's a problem tailor-made for data aggregation, analytics and machine learning.
One of the early applications of Watson technology is to cancer treatment where the software can examine DNA sequences from a patient's healthy and tumor cells to identify mutations. As this IBM white paper details,
The cognitive system then examines the body of scientific literature to identify potential treatment pathways and determine their likelihood of success. The results often include treatments normally administered for a different form of cancer that may have not been traditionally considered. For example, a patient with lung cancer might be presented with a treatment option for stomach cancer because the individual happens to have a mutation found in several forms of stomach cancer.
As with image analysis and tagging software like that used by Facebook and Google, the beauty of an ML system like Watson is the ability to build on prior learnings. As the corpus of medical data and research grows, the computational horsepower increases with the addition of ML optimized hardware (see my overview here) supporting more complex ML models, the accuracy and precision of the diagnostic system significantly improves.
Radiology is another promising area for ML in providing diagnostic assistance. Applying ML algorithms like the convolutional neural networks (CNNs) used to pick out the lion amongst a group of cat pictures, can assist radiologists in identifying ambiguous imaging anomalies. According to Dr. Nick Bryan, Professor Emeritus at University of Pennsylvania School of Medicine,
ML algorithms incorporated into computer assisted detection/diagnosis (CADD) products are now detecting pulmonary nodules, diagnosing colonic polyps and screening for breast cancer, with much more to come. In the not so far future, ML will play a central role in radiology, becoming part of routine workflow and providing daily real-time clinical diagnostic support. I predict that within 10 years no medical imaging study will be reviewed by a radiologist until it has been pre-analyzed by a machine.
The systemic problem of health cost inflation is a mix of good and bad: both tremendous leaps in medical treatment and diagnostic and pharmaceutical technology paired with burdensome bureaucracy and inefficient processes. However, the proximate catalyst for change is a new administration in Washington unafraid to gore political sacred cows. I agree with Den Howlett when he notes that some in the healthcare business will see this as an opportunity, not a threat.
My hunch is that progressive providers will use these types of ‘threat’ as a good reason to perform their own preventative medical treatment by way of evaluating cost cutting options, facilitated by automation.
While insurance reform addresses the cost transfer between patient and provider, it has proven to do little to bend the cost curve by improving the efficiency of treatment. Such is the opportunity for IT: to bring Moore's Law technological improvements not just to diagnostic equipment and drug discovery, but to both clinical and patient data collection, management and analysis.
Programs like Watson Health are encouraging first steps at solving the challenges of an explosion of diagnostic data, but EHRs need a reboot. Blockchain and distributed ledgers seem promising vehicles for backend medical data interchange, however there is much more work to do on simplifying EHR data access and analysis.
For both medical records and diagnostics, I expect cloud services to be a preferred deployment platform due to the ready availability of scalable capacity, specialized hardware, sophisticated hardware and secure infrastructure. Whether they can satisfy regulatory and security requirements will be a challenge, but a challenge that IT is well used to meeting.
But who will step up and mandate change? A Trump government is making healthcare a priority and while the incoming administration may be naive, it could be that very naivety helps it discover or force novel technology approaches upon this moribund and underperforming service industry.