AI models must be built on more complete and global data sets

While artificial intelligence begins to transform health care, a large question is often overlooked: hospitals, health systems and computer suppliers form AI on the right data?
Many AI models are strongly based on data from American and European sources. As such, this can create biases that limit treatment options. Precious ideas of other parts of the world are left out. Actually, Research has shown that biased data sets can contribute to disparities in health care and ignore effective treatments available outside the United States
John Orosco has a lot of experience in AI and data sets thanks to his work as CEO of Red Rover Health. The company specializes in the simplification of the integration of the DSE via a platform which uses secure restful APIs to connect third-party software with DSE systems. RESTFUL APIs are a type of web protocol interface that allows a consumer to execute to recover data or publish updates to a source system.
All this is designed to allow health care organizations to improve existing DSEs with better breed systems, improve access to patient data in real time and rationalize clinical workflows.
We spoke with Orosco on the main challenge with AI and data, AI reaching its full potential in health care by being trained on more diverse and global data, AI Connection with genomics and precision medicine and why AI models should consider therapies without flow flows to provide patients with the best possible treatment.
Q. What is the main challenge today with artificial intelligence and data?
A. The main problem with AI in health care today is not the technology itself – is that we are still in the early stages of its evolution. Large language models continue to ripen at a quick pace, and although they already have incredible promises, it is clear that we have only scraped the surface.
The first indicators suggest that AI is there to stay and that it will fundamentally reshape how we approach automation, decision -making and productivity in industries, especially in health care.
But as powerful as these models become, their effectiveness depends on access to data. AI cannot be as good as the information with which it must work. And in health care, where data is often fragmented on different systems, buried in notes not structured or locked behind obsolete interfaces, it is a real challenge.
Integration is not only useful – it is essential. Without access to complete and well connected data sources, the full potential of the AI is actually neutralized. It becomes a brilliant tool that can only see part of the image.
Thus, the real objective at the moment should not only be on what AI could do in the future, but on what we can do today to prepare for this future. This means decomposing data silos, building an intelligent infrastructure and ensuring that LLMs have access to such relevant and high quality data as possible.
While the models continue to improve – and they will do it – this foundation will be what will determine the real value that we can unlock. In short, the models ripen quickly – now it’s up to us to make sure that the data is ready for them.
Q. You think that AI in health care can only reach its full potential if it is trained on more diverse and global data than it is today. Please develop.
A. It can only reach its full potential if it is formed on various sets of global data. Currently, a large part of the data used to train LLM come from specific regions, mainly in the United States, although this may seem a good starting point given the amount of data on health care in the United States, it is in fact limiting.
AI training only on regional or national data Connectors in cultural, systemic and clinical biases in this region. It gives us a narrow lens through which AI understands medicine and health, and this fundamentally restricts its usefulness.
Take the United States, for example. Our health system tends to promote certain processing approaches, such as prescribing medication or recommend surgery. On the other hand, other countries could rely more on natural remedies, alternative therapies or the different care pathways.
If AI is only trained on data based on the United States, it will naturally reflect and strengthen these processing models, even when other approaches could also be or more effective in different contexts. This is one of the reasons why many American patients who can afford to wonder abroad – because they believe that there are effective treatments available outside the limits of FDA approvals or US clinical standards.
If we really want AI to support better results for health on a global scale, we must think beyond borders. This means the formation of models on a wide range of data from different countries, cultures and care models. It is not only a question of volume, it is a question of variety. Various data makes the AI smarter, more adaptable and ultimately more equitable.
Without this, we risk building technically advanced but functionally narrow tools. If we want the AI to reflect the complete spectrum of human health and treatment possibilities, we must give it a more complete image of the world.
Q. You say that genomics and precision medicine can offer more personalized care. What is AI and data connection?
A. There is a powerful link between AI, data and the future of personalized care thanks to genomics and precision medicine. Consider the human body as an operating system. Each of us works on our own single source code, which is our DNA.
The cartography of the genome is essentially decorating this system. He tells us how we are wired to respond to certain drugs, how we metabolize the drugs and even the conditions that we could be predisposed. Despite this information, a large part of modern medicine always adopts a test and error approach.
We prescribe treatments, then say things like: “Let’s see what you feel in a week.” It is intrinsically imprecise and often ineffective or even risky.
This is where AI can play a transformative role. When AI is trained on genomic data in combination with other clinical data such as processing protocols, laboratory results and real world evidence, it becomes much more precise in its predictions and recommendations.
By including genomics in the data mixture, AI can help identify the most effective processing for each individual before trials and errors even start. He can also help avoid serious side effects by reporting drugs that a person is likely to be poorly metabolizing or not answering at all.
The future of precision medicine depends on this type of integration. The genomic data in itself is precious, but its full potential is only achieved when combined with wider and large -scale data sets by AI. When this happens, we get closer to the care that is not only personalized but proactive, predictive and safer. AI becomes the engine that transforms data into insights into perspicacity, and genomics become a fundamental layer in truly individualized care.
Q. You also suggest that AI models should consider therapies without flow flows to provide patients with the best possible treatment. What do you mean here?
A. What I mean is that AI models should extend their point of view beyond local and traditional treatment protocols, especially when these protocols are defined by regional regional organs. Too often, AI systems are trained on data sets which only reflect what has been approved or reimbursed in a country, generally based on regulatory or insurance parameters.
Although this may have a sense from the point of view of compliance, this limits the potential of AI to offer patients a really complete view of the available treatment options. It is not because therapy is not approved by the FDA or is not covered by the assurance that it lacks merit. In fact, it could be widely accepted and effective in another country.
Therapies not in accordance with the runners or alternatives used in the world should not be ignored by AI simply because they fall outside the local medical game. Patients deserve to know what is available – not only in their postal code or insurance network, but around the world.
Of course, access and reimbursement are real obstacles, and there are political and regulatory complexities, especially here in the United States, but the role of AI should be to inform and widen the conversation, not to reduce it.
If a patient sees a processing recommendation generated by AI which includes promising therapy used internationally, he can discuss it with his doctor and make an informed decision together.
At the end of the day, the AI should serve as a impartial guide, not forced by local policies or insurance limitations. The empowerment of patients with a broader vision of what is possible can lead to more personalized and thoughtful care. It will not be easy to implement this inclusive state of mind on a global scale, but the fact of not doing it means that AI will still not cost its real potential to support better health results.
Follow Bill’s successful coverage on LinkedIn: Bill Siwicki
Send him an email: bsiwicki@himss.org
Healthcare It News is a publication of the Himss media.
Now look: AI chief officers need an in -depth understanding of technologies and clinical operations