AI in Healthcare: 5 barriers and solutions for integration

Looking at my LinkedIn feed, most are mentioning AI, ChatGPT and its future effects on the healthcare system. Everything from diagnosing diseases, developing new treatments, streamlining clinicians’ roles and improving patient care seem to be within its wheelhouse. 

However, we have to be realistic in integration and the actual process of change in healthcare. There still remain barriers which will slow down that uptake. 

1. Data privacy and security

AI systems require large amounts of data to train and learn. In healthcare this includes sensitive patient information, such as medical records, test results, and images. We don’t even know and haven’t seen the extent of what will happen if much of this data is used by unauthorized individuals or organizations. This could lead to identity theft, financial fraud, or even physical harm to patients.

2. Bias in the data

Data bias has been a longstanding concern as AI systems are only as good as the data they are trained on. If we are already reliant on the data we have, it could lead to inaccurate diagnoses or treatments. This has implications especially for historically marginalized populations - and we know we don’t have the right research data in certain instances.

3. Lack of transparency

If the system is complex and difficult for healthcare providers to understand how the system arrived at its conclusions, it will lead to a lack of trust. Or even an inability to use the system to its fullest potential. I’ve noted how some telehealth programs make more work for clinicians, not less. This has the same potential. 

4. Patient trust

On the other side of the equation, patients may not understand the safety and reliability of AI. This can be especially worse if their clinicians’ don’t. Who would be willing to use it in these cases ? 

5. Regulation and governance

The use of AI in healthcare is still in its early stage and there are currently no clear regulations or guidelines governing its use. Once again, tech has been faster than regulation. This has led to lack of reimbursement, standards, clinical evidence and guidance. 

I’m always a realist, however, and know that AI is being used in healthcare. And should, once some of these barriers are cleared. Here are some ways I see we can address those challenges:

• Data privacy and security: Tighter security and understanding the risks of AI data. Yes, most of this is standard. However the recent breach of telehealth data is a prime example of unintended consequences leading to security issues. Find out everywhere your data might be going and ensure it is safe. This can mean strong encryption and access controls and working with AI developers to ensure that their systems are designed with privacy and security in mind.

• Data bias: This is a long standing and more complicated issue, however, selecting data carefully can go a long way to improving the future of AI. This data should be representative of the population that the system will be used to serve and AI companies should develop techniques for detecting and mitigating bias in AI systems.

• Transparency and Clinician input: Clinicians can be trained on new modalities - to make it easier for them to use to its fullest capability. Also to help ensure that clinicians want to use it. Having clinicians put their input on it will improve its clinical utility and engagement. 

• Public trust and understanding: There has to be clear goals to patients on what the AI is being used for. Using clinical data and tracking health patterns is useful - patients want to know this too so they can also better use it. Having health and digital product education will go a long way to improve patient engagement and willingness to use it. 

• Regulation and governance: These will come as discussions between healthcare, AI companies, government agencies and policymakers are already starting. At the #AMA’s Digital Medicine Payment Advisory Group, AI is already a topic to prepare for reimbursement. Clinical guidelines have to be created as well as laws that protect consumers. 

AI is an exciting new forefront in medicine but we need to address these challenges. None of them are new - they are the same ones we deal with during every new digital health change. Starting earlier on these issues could potentially have it integrate - and revolutionize medicine - more quickly.