Electronic Health Reporter

By Lu Zhang, founder and managing partner of Fusion Fund.

In theory, artificial intelligence has the potential to transform the healthcare industry for the better in a number of important ways. AI and machine learning’s predictive capabilities, for example, can improve accuracy in diagnosis and treatment. AI can also be put to use for boosting efficiency in the areas of administration and operations.

However, it appears that many healthcare institutions are reluctant to put this promise to the test. Compared to other industries, AI adoption in healthcare is not keeping up — this isn’t altogether surprising. Healthcare has several barriers to entry for AI and ML that are unique to its industry and need to be considered by anyone innovating in the space.

Why AI Adoption in Healthcare Has Fallen Behind

When it comes to the collection and storing of data — which is essentially the lifeblood of an effective AI solution — there are both regulatory and privacy concerns that need to be taken into consideration. In the United States, the Health Insurance and Portability Accountability Act (or HIPAA) has specific rules around the ways in which patients’ private information can be managed and shared. The General Data Protection Regulation (or GDPR) plays a similar role in the European Union. Beyond these legal questions, providers are also keenly aware of how important it is to ensure that sensitive patient data is kept secure and private for the sake of maintaining good reputations and healthy relationships with patients.

What’s more, even if all regulatory and privacy concerns can be addressed, there are still other issues at play when it comes to data. Specifically, the time and effort it takes to collect the amount needed to ensure AI is delivering accurate and unbiased analysis can be immense. The most efficacious way to get enough data quickly would be to share it among organizations. This is easier said than done, thanks to a lack of standards around data storage and a reluctance to entrust patient information with third parties.

But data isn’t the only obstacle to AI adoption. AI’s typical lack of transparency is another. In an environment where there’s hardly any room for error, it’s crucial to understand not only the thinking behind a decision, but also the variables that were used in the decision-making process. With modern AI, gaining access to this information can be difficult or, in some cases, even impossible.

Perhaps the most difficult barrier to entry for AI, however, is also a more subtle one: mindset. Many healthcare professionals are reluctant to bring in new technology that might disrupt carefully calibrated workflows or even replace human healthcare workers with computer counterparts. AI’s role in healthcare is to complement existing professionals — not replace them — but it’s understandable to be concerned about the ways in which AI could change how the industry operates.

With all these challenges and concerns, it’s perhaps no surprise that AI adoption is happening at a slower rate. However, none of these impediments should be seen as deal breakers. The healthcare industry needs to open its eyes to the benefits of AI. This technology will allow employees to deliver better patient outcomes quicker, easier, and cheaper. In the future, the doctors and nurses who know how to use AI tools will replace those who don’t. If AI developers want to help make this future happen, they need to take some steps to tailor the adoption process to the needs of healthcare professionals.

3 Ways Providers Can Make AI Tech Easier to Implement in Healthcare Environments

1. Work on augmenting workflows — not disrupting them.

The healthcare system is a complex organism that relies on tried-and-true processes to keep it running smoothly. Inserting AI into the equation has the potential to disrupt this equilibrium, undercutting any benefits it might bring.

However, AI also has the potential to make these processes more efficient and less prone to errors. AI and ML have the power to make healthcare operations truly intelligent. Creators who offer specific solutions for hospitals looking to improve their existing administrative and operational workflows can provide great entry points for wider AI adoption.

2. Focus on compatibility.

For ML to provide accurate diagnoses and treatment options, it needs diverse training data to learn from. While some larger organizations might be able to provide this data on their own, many others will lack the amount of data necessary for algorithms to be effective. Ideally, these institutions could aggregate data with other providers to create robust AI models that are not only more accurate, but also more generalizable with less bias.

To do this, the method for storing and processing this data would need to be compatible across companies. To provide AI solutions that work within the healthcare industry, developers should steer clear of proprietary data formats. Instead, they should embrace standards where they exist within the industry while encouraging the adoption of those standards where they’re missing.

3. Offer natural language processing.

The beauty of NLP systems is that they can be put to use in a variety of ways without needing a large amount of healthcare-specific data to be effective. NLP can be used to organize clinical documents as well as analyze unstructured notes to offer digestible summaries and actionable insights.

NLP can also be used to augment the work of experts in other areas, such as to improve the interpretation of patient imaging by radiologists. This type of AI solution is a simple and effective way to move the needle on AI adoption.

Even if adoption has been relatively slow, AI is already proving itself a valuable part of healthcare where it has been implemented. From diagnosis to health risk assessment to new drug development, AI has proven itself a useful tool in the world of healthcare. But for it to reach its full potential, AI developers need to create solutions that address the legitimate concerns of healthcare professionals and patients in order to get everyone on board and help the industry move forward.