So, I have this question, are we drowning in data and still starving for insight? See, I think we have an I-Care data crisis. Now, I'm an optimist about the future of I-Care AI. That's why I do this job.
But I have to be brutally honest about our true present limitations. Well, this is one of those factor fiction. Let me give you the facts. We as an industry are standing in our own way.
I mean, we generate petabytes. No, by the way, that's a lot of information of up-dynamic data, both OCT scans, funds photography, biometry measurements. But the problem is, we have it all locked away in these isolated digital silos that no one can access. Now, the grand vision of large language models and predictive AI transforming our clinical decisions is simply being choked out by the single self-inflicted wound, a catastrophic lack of data standardization, and most importantly, connectedness.
We as an industry are our single biggest problem. Now, the channels of acquiring this efficiently large and high-quality data set to train a robust LLM is not a problem of volume. It's a problem of connectedness. Right now, we're embarrassingly suffering from a true data disaster in my opinion.
Our data resides in tens of thousands of non-connected silos that we call electronic health records that really have evolved little in the last 30 years since their development. There are two inherent problems with this aging and somewhat archaic technology. First, is a version of non-standardized digital chaos. Within these digital filing systems, information we collect is not standardized.
For example, one practice may record mild dry eye as a free text quote that may be downloaded from an AI scribe, another system may use a structured ICD-10 code, and a third might use some proprietary scale that works in that office and that location for that provider. Now this non-uniformity renders aggregated data, virtually useless for AI training, unless we spend countless man hours cleaning, normalizing, and harmonizing that data. A model can't learn if it cannot consistently define the disease. For still, the systems themselves, or at least those in charge of these systems, are really hostile to connection.
Now we have dozens of proprietary EHR vendors often making data transfer difficult, incomplete, and very expensive. This is not just a failure of interoperability between systems, but a failure of connectedness within systems. For example, a patient record might be fractured across two different clinics that use the same EHR simply because the practices operate as separate entities with no mandated data sharing infrastructure. This insulated or insular approach is the very antithesis of modern large scale data science.
This fragmentation is not just annoying, it's a direct inhibitor to the evolution of care. We can't aggregate the necessary evidence to identify game changing trends because the data is scattered and really non-comparable. Finally, in my opinion, this is simply unacceptable in a world of globally integrated data. The real tragedy here is the inability to create a holistic, multimodal data set across the industry that's required to advance evidence-based medicine and eye care.
True clinical advancement requires fusing all aspects of the patient's journey. Now there are three main areas of fusion that need to occur, so let's break them down. First, we need to overcome our diagnostic disconnect. We cannot easily link the precise quantitative data from a diagnostic technology, let's think like an OCT vascular density map, to a provider's clinical observation, hey, the macular democat better, and the patient's subjective history, oh, my vision's getting better.
These remain three separate, distinct files often in separate systems or separate parts of the system with separate non-linked identifiers making it nearly impossible to correlate various pieces of the holistic picture. Point number two, there's a void in genotype phenotype correlation. The true frontier of evidence-based medicine lies in connecting the patient's genotype, i.e. the genetic makeup, to their phenotype, their observable disease, presentation, and progression.
Now, this is impossible when clinical data is siloed and genetic data is managed if we want to call it that, by a third party that has no standardized consent-driven pipeline to bring the two pieces together for large-scale analysis. And the third point about fusion is perhaps the most agrarious gap is our collective failure to rigorously and uniformly track treatment success or failure over time. For example, we initiate therapy, let's say it's a specific anti-vegethe agent. A glaucoma surgery protocol, or maybe it's my opium management.
But the outcome is buried in this unstructured follow-up notes. Now, LMS thrive on identifying patterns of response and non-response, but without clean longitudinal data that definitively links intervention to outcome, we simply can't develop the robust predictive evidence-based medicine treatment protocols needed to revolutionize patient care. That, in my opinion, is a crime. We're forcing clinicians to practice based on small-scale child data, or we're shat, our own data set with an end of one, i.e.
my opinion, rather than the collective real-world experience of the entire global eye care community. Now, undoubtedly, we need solutions. Well, never, my parents always taught me, don't complain about a problem unless you have a solution. So let's throw some of those out.
Those solutions demand a unified effort involving technology, regulation, a complete change of mindset. That's the toughest one. Now, we can overcome this fragmentation, unlock the potential of LLMS and AI if we meet a few goals. So let's talk about our first one.
First, we need mandatory healthcare interoperability adoption. I mean, we have to move beyond information blocking regulations, i.e. HIPAA, and the proprietary mindsets of data gatekeeper companies, EHRs, to mandate the adoption of HIR. Human or, I should say, healthcare interoperability systems by all EHR vendors and device vendors.
Crucially, the industry must collaboratively establish a specific and rigorous HIR profile to ensure the standardized capture of key data elements, i.e. visual acuity, intracular pressure, OCT, layer thickness, so that myel diabetic retinopathy observations are coded and identified equally everywhere. This will require not just acceptance standardization, but an also an entirely new and more expansive code set than the ICD-10, much less that generally archaic and simplistic 5-digit CPT code can ever hope to capture. So in review, first step, we need a different coding system and we need a way that everybody is saying the same thing.
No fear, those solutions are coming. A key component to interoperability is number two, which is the development and universal acceptance of an open architecture knowledge network. Instead of physically moving sensitive patient data, we have to adopt a centralized data knowledge center. This center would allow for centralized models to access and be trained across thousands of decentralized data silos.
The data never leaves the practices system, but the knowledge and the updated model weights are shared providing scale needed for training OEMs while preserving privacy. Now this is going to be a problem. Don't get me wrong is that lots of data silos don't want to share that information, so this might be a right for disruption where some other company can then say, well, this is the way we're going to do it and EHRs have to reconsider what they do. Now this knowledge system is going to require some AI harmonization tools to fight AI's data problem.
For example, new LLM tools can be developed specifically to ingest non-standardized free text, whether it be free text or whether it be in the form of ambient scribes that are converting the talk to text, and structure data from disparate sources that automatically harmonize and normalizes this data into a single, clean format. This acts as a necessary bridge to use legacy EHR data that currently exists. However, it may be, as an alternative, just as efficient to just develop an entirely new knowledge system that does not include the incomplete EHR data sets of today. Maybe starting new is less expensive than trying to convert the old.
Now to entice all parties, we've got number three, which is we have to have regulatory bodies like CMS and other payers that consider establishing financially and quality-based incentives, high-destructural longitudinal tracking of data treatment endpoints. For example, when an anti-VEDGF injection is given, the system would require specific and standardized data inputs regarding anatomical and functional outcomes that every highlight, bold print, follow-up, create every follow-up visit, creating a definitive success failure dataset needed for EBM protocol development. Now this course would be much easier if interoperated and standardized data over a single universal knowledge system were really accepted as norm. However, the most fundamental solution, last but not least, will be transparent, patient directed, data portability.
Now let me repeat that. Transparent, so everyone can see what's going on. Patient directed, put the patient back in control. Data portability, so data can move wherever it needs to move in a transparent way.
So I believe that empowering the patient is the key to the future. Establishing an infrastructure that allows a patient to easily and securely aggregate and share their own complete interconnected record includes all clinical imaging and genetic data via a whole bunch of APIs will bypass many of the current practice level data sharing bottlenecks that we have to endure. This places the ultimate control and responsibility for data flow with the individual. Creating knowledge while upholding privacy.
In other words, imagining a situation where the patient controls their own data can send where they want to and can see who touches it and how it's being used and give their okay if they want it used or not. Now the biggest point I want to make in conclusion is that solving the IKR data crisis is not a technical challenge. Everything I said very simple to do. It's a political and cultural one.
IE, it's a human problem. It's time to tear down the data silos and build an interoperable data infrastructure that are patients and the providers and the future of IKR deserve. There are personal health information systems in development that will do just that. As the health care industry, we must be ready to get out of our own way and work together to connect everything.
Obviously the charge for a better knowledge system to evolve health care.