The regulation of AI in healthcare is becoming a hot topic for anybody building or using software in this industry. Considering our CEO, Andrei Ismail, holds a PHD in AI, we wanted to hop on the bandwagon and discuss the actual benefits, as well as the legal and technological implications, of AI in healthcare vs. the magical all-powerful notion it’s been in media for some time now. In this blog post, we’re discussing the legalities and ways to operate within the law to apply this powerful technology.
While some may argue it’s too advanced for current healthcare settings, artificial intelligence is a major upcoming trend — so healthcare tech providers should get on board sooner rather than later.
But we must establish clear rules and good practices early to avoid making mistakes and giving the entire promising field a bad rep. Finding a balance between new ideas and established rules of compliance lets us integrate AI and improve our outcomes.
Let’s see what the latest regulations for artificial intelligence in healthcare say, and what that means for our work.
AI for Healthcare Software: A Promising Prospect
(If You Approach it Right)
AI is improving at performing human tasks faster and cheaper than actual people. It could aid with diagnoses and treatments, offering customized care, and performing administrative work. These capabilities come from its ability to quickly analyze heaps of data, letting workers focus on the actual patients.
We also have robotic assistants, patient engagement platforms, and natural language processing (NLP). To learn more about everything machine learning and AI for healthcare can do, read this paper from the Future Health Journal
The potential is massive, but according to a study on the perception of AI by healthcare workers in marginalized communities, we must consider the context where we apply it. Few healthcare professionals are trained and comfortable with AI, even in privileged contexts. We must also remember these systems aren’t infallible.
As a technology, artificial intelligence in healthcare brings these risks:
- Security and compliance concerns. With AI analyzing patient data, there’s a risk of privacy breaches or data leaks. Since PHI protection is paramount, this could make organizations lose their HIPAA compliance status.
- Algorithm biases. AI algorithms may incorporate biases from the data used to train them. It could result in unequal treatment or misdiagnosis for some patients, like racial minorities or individuals from disadvantaged backgrounds.
- Dependence on AI algorithms. Relying too heavily on AI for diagnostics and treatment decisions may lead to errors or misdiagnosis. While it can process information, it lacks the human ability to interpret context and nuances.
Naturally, regulations of AI in healthcare seek to abolish some of the risks while keeping the benefits.
What AI Laws for Healthcare Software Are in Place?
In 2021, the FDA took to regulating medical devices that use machine learning and AI for healthcare. Here’s what they said:
- AI proposals should outline what the system can learn and how it implements this learning safely and effectively.
- Good machine learning practices should prevent biases and involve people in the AI learning process. That way, we won’t depend on black box algorithms we don’t understand.
- The right to explanation suggests we should be able to ask AI systems why they made specific treatment recommendations. This principle increases transparency.
- Manufacturers of SaMD tools should collect real-world performance information. That way, we can keep improving systems until they reach peak efficacy.
Although these regulations of AI for healthcare deal with devices, similar principles may apply to SaaS. Notably, California’s AB 311 law adapts them to require AI to explain reasons for decisions, non-automated systems to check the decisions, develop safeguards, and draft policies that describe the tools, their risks, and steps taken to mitigate them.
The recent Executive Order issued by President Joe Biden tackles AI for healthcare. Here’s what it says:
- The order requires sharing safety test results with the government to ensure AI systems for healthcare are safe.
- Emphasis lies on developing privacy-preserving techniques to safeguard patient health data.
- Artificial intelligence in healthcare must address discrimination and bias to ensure fair access and outcomes for all patients.
- The best practices of using AI for healthcare must be considered, along with addressing job displacement.
- Government guidance influences the regulation of AI in healthcare, ensuring responsible use.
In a nutshell, existing legislation prioritizes safety, security, and ethical use. AI must be as transparent as possible and ensure positive, equitable outcomes for patients. However, executive orders aren’t set-in-stone laws. We expect to see new safety and usage standards, and we have some idea what they may say.
What New AI Safety Standards Can We Expect?
The European Commission, the European Medicines Agency (EMA), and the International Coalition of Medicines Regulatory Authorities (ICMRA) collaborate on regulations of AI in healthcare. As the FDA is a member organization of the ICMRA, we expect these standards to apply to the US:
- Oversight. The businesses that use AI for healthcare must have an oversight committee that manages risks. Regulators will designate Qualified Persons to ensure compliance.
- Explainability. Whenever possible, AI laws ask developers to avoid black box models. This article from ScienceDirect discusses explainability in more detail.
- Data integrity. The rules of data usage and storage will be tighter. The information must be strictly relevant and collected only when necessary. The least sensitive data must be used in each context.
- Human-centricity. Humans must be involved in the development and use of AI tools. Beyond modeling itself, patient-reported outcomes should be part of evaluations.
- Internationalization. We should develop and standardize machine learning practices for the healthcare industry. The principles are quality, transparency, and reliability.
These guidelines come together in the European Commission Artificial Intelligence Act, the first-ever AI law in place. It aims to create a unified definition of AI systems and adopt a risk-based approach to its usage in healthcare. Systems with unacceptable risk levels will be banned, while lower risk levels will be subject to regulations.
The US and UK AI Safety Institutes formalized cooperation on April 1st, 2024. In the US, the National Institute of Standards and Technology (NIST) will establish security standards for public technology that uses AI for healthcare.
Some rules remain murky, but we see the general trends of AI use in healthcare today and into the future. Let’s see how it affects our jobs in particular — the creation of such software.
How to Safely Use AI for Healthcare Software Development
Artificial intelligence in healthcare software offers immense benefits in terms of efficiency, accuracy, and patient care. However, we must obey AI laws and use the technology ethically to maximize its potential.
Ensure your healthcare software has robust data security measures to protect PHI and stay HIPAA-compliant. Encrypt data in transit and at rest, install access controls, and perform security audits. Companies like Medcrypt have this in the bag, offering complete API solutions for better cybersecurity.
Prioritize validation of AI algorithms to guarantee their accuracy and reliability in clinical settings. Think about rigorous testing and validation against large datasets and using peer-reviewed research. Google DeepMind excels in this regard, investigating the impact of AI and partnering with research labs for better coverage.
The final aspect deals with the end-users themselves. Train healthcare professionals on ways to use your AI-powered software and interpret its results. Consider workshops, online courses, or hands-on training sessions.
Pro tip: Artificial intelligence is developing, and the same goes for regulations of AI in healthcare. Don’t become complacent post-release, but stay up-to-date with laws and your users for the best results. Visit our healthcare security and compliance page to learn more.
How Vitamin Harnesses AI for Healthcare Software Development
We conduct a Vitamin Sanity Check before adopting any new technology, AI included. Our experts evaluate if this tech is suited to solve a particular business problem or if traditional approaches are more appropriate.
For instance, AI isn’t a substitute for human expertise. While algorithms can analyze data and flag potential issues, they still need human doctors to interpret the results and make the final call. Similarly, AI-driven chatbots can provide basic information and assistance but can’t replace human interaction in patient care.
On the other hand, NLP has revolutionized many aspects of software development, including coding. AI-powered code completion tools, like GitHub Copilot or Microsoft IntelliSense, use NLP to suggest code snippets as developers type. These tools improve coding productivity, especially for large codebases or complex libraries.
AI-driven testing tools can automate the software QA process, including test case generation, test execution, and result analysis. For example, AI can inspect user interactions with an app to generate test cases that cover these scenarios.
And how do we abide by the regulations of AI in healthcare?
- Security. Integrating AI into healthcare brings challenges concerning sensitive PHI. Our cybersecurity team is well-equipped, offering the safest way forward.
- Continuous learning. We keep abreast of the latest trends in artificial intelligence for healthcare, letting us devise innovative solutions.
- Collaboration with experts. Working with healthcare experts ensures our AI-powered tech is clinically relevant, user-friendly, and aligned with best practices in care delivery.
Our Final Thoughts
The potential benefits of AI in medicine are vast, and we should use it to improve healthcare delivery and patient outcomes. Since AI laws make a lot of sense, following them doesn’t just keep you on the right side of the law, but ensures you’re using it in the best possible way.
Our point is — don’t be afraid to explore artificial intelligence in healthcare software but know your AI laws first. It is the only scenario that brings positive outcomes.
Are you considering whether AI would fit into your latest healthcare project? Get in touch and let our experts gauge whether using it would be safe, viable, and worthwhile.
