I need Google Search for my closet.
In some areas, I am well-organized, but then there’s my closet. I can never seem to find that one piece of clothing I want. You know, the comfortable long-sleeved shirt, the old pants I use for working outside, or the one dress shirt that looks great with my new sport coat…. I need help.
Speaking of Google, you’re probably aware of the big headlines surfacing last week in the Wall Street Journal and elsewhere about “Project Nightingale” — where Google uploaded a large set of patient data from Ascension Healthcare to work on their “Patient Search” capability.
Initially, this made news because of the perceived risk to health privacy of the patients involved, as well as the lack of disclosure. While “Project Nightingale” has many real issues to explore, the practice of moving data around in this way is not terribly unusual. There is no particular reason that there is more risk in this project than in others by traditional EMR companies that move the same data to the cloud, or to data centers anywhere. Perhaps more illuminating are the challenges Google and other companies face in becoming trusted guardians of our healthcare data. From a technology perspective, there is no doubt that Google has the capability to keep data highly secure – there’s really only the question of whether we, as consumers, trust them to do so.
In this blog post, I want to focus specifically on the use case described in the project: Google’s use of amazing search technology on medical records.
I hope…I really hope…my medical records are better organized than my closet! Yes, my spouse’s encyclopedic wardrobe memory and object permanence are the primary reasons I look respectable in public. But a Google Search of my closet would be a lot more convenient and make life easier for both of us for two key reasons:
One, I’m not terribly consistent in what I call things. “The purple tie with the squiggles on it” is not exactly a standardized lexicon. Standardized taxonomy in naming everything would enable better organization and lend better structure to my closet. Which raises Reason Two – my closet fits the true definition of unstructured data.
Medical professionals spend a notable portion of their lives learning standard terminology in medicine – specifically to prevent ambiguity and confusion when communicating medical issues. As a country, we spend billions of dollars implementing structured Electronical Medical Record (EMR) systems, then upgrading them to communicate using SNOMED, ICD-10, and other industry standard code sets – all to help in the organization and taxonomy of patient medical records.
So, in a patient care use case, what would this search tool do?
What it will NOT do is mine across all the different doctors I’ve seen across different cities and health systems. (Oh, this is theoretically possible from a technology perspective…but it would involve bridging everyone’s medical records with social and demographic data to correctly map my life and piece my history together. That, or me doing it manually…not terribly likely.)
I am reasonably confident that the EMR in my current doctor’s office is well-structured, organized, and coded so my doctor can make informed, relevant care decisions based on my current health data. That’s exactly what I want from my doctor.
It’s not that there isn’t important information in my medical past – it’s just that a Google Search capability is not the key to finding it. Either, I, as an informed healthcare consumer, will be able to point to where the data is located (e.g., I had surgery at that hospital in Minnesota in 2001), or I really don’t want a search appliance able to find it, out of concern over loss of privacy. If I have not already “claimed” an old medical record as my own, do I want the contents of that record being exposed to a search engine for a purpose that I have not authorized?
To be very clear, my skepticism here is about the utility of a Patient Search engine in the specific context of individual patient care. I have been around healthcare long enough to envision many uses of such a tool in the administration or the economics of healthcare.
There is a stunning amount of raw healthcare data available digitally in one form or another. The U.S. healthcare system has created an enormous industry around finding conditions, diagnoses, and procedures that were not properly coded into the medical record, but not in the context of individual patient care. Instead, the financial profile across the population has become a critical driver in many parts of healthcare. A missed diagnosis code in a hospital procedure can change reimbursement by thousands of dollars in the case of diagnosis-related reimbursement. Failure to code a routine health check performed as part of primary care visit could result in the loss of quality bonuses. Leaving off a behavioral diagnosis from a patient chart in a given year can result in a substantial loss of premium for payers in many government programs, due to the impact on risk adjustment. Companies such as Innovalon and Change Healthcare mobilize armies of people to go around reading and codifying medical charts every year to help doctors and plans achieve quality scores and favorable risk adjustment treatment. There has already been a surge of use cases in Natural Language Processing and machine learning to aid in this process, and I can easily see how Google might be able to bring disruptive change to this insiders’ corner of the healthcare industry. Consistent coding is absolutely critical to enabling effective value-based healthcare, quality measurement, and risk adjustment of premiums. Google could be an innovator and accelerator to this necessary process and could provide substantial help in reducing these administrative costs that are currently built into the system. However, let’s not pretend that Google is the only game in town. Systems and payers have already invested heavily in processes and technology to close these gaps on a go-forward basis.
So, what will Project Nightingale do for Ascension? Many systems are looking for ways to monetize the healthcare data they have gathered and stored. In addition to the administrative value I talk about here, there may be value in such a project for research initiatives – an entire other area worthy of discussion and rife with controversy. However, when it comes to changing patient care, I will continue to expect that my doctor, and their supporting systems, are looking at the data necessary to provide me with the best care.
In the meantime, I’d take that closet search tool…
Healthcare is an industry ripe for innovation – especially around new technologies that activate the massive digital data available today. Knowing which technologies most effectively address a specific healthcare challenge is a huge barrier to entry in this market. Within the VedaPointe platform, we are striving for absolute transparency at all layers of use. This ensures that health systems and payers alike have comprehension of data utilization and can count on our expertise in choosing the right digital solution to solve their most meaningful challenges.