Since the premiere of the ChatGPT in the autumn of 2022, there has been much talk about potential generative AI applications in healthcare. The first hospitals have already employed prompt engineers while others are experimenting with gen AI to help doctors search for relevant data, make decisions, or simplify making notes in the electronic medical records.
In September 2023, Oracle integrated gen AI into its Cerner electronic medical record system. Oracle Clinical Digital Assistant enables doctors to schedule an appointment and labs, check information in the EHR, search relevant clinical information, or write a prescription – just by using voice commands. Recently, Google Cloud announced new AI-powered search capabilities to help physicians access patient’s information. Instead of clicking through the EHR to find out what medications the patients have been taking over the last ten years, the doctor can now ask AI to summarize the data.
Epic and Microsoft aim to create an AI-driven co-pilot for doctors “to integrate conversational, ambient, and generative AI technologies across the Epic electronic health record (EHR) ecosystem.” Meanwhile, Mayo Clinic is deploying Microsoft 365 Copilot – a new generative AI platform based on large language models (LLMs).
Gen AI and LLMs combined with ambient technologies and voice recognition systems to improve physicians’ experience and reduce administrative workload are becoming new features of EHR systems. Is it the right time to implement them?
Promises vs. unknowns
So far, some of the latest AI-based solutions are being adopted by large clinics. However, the race for healthcare IT systems developers to embed generative AI solutions in EHR systems has started, – and there will be even more AI coming soon to healthcare.
When press releases of the tech companies promise “enhanced productivity,” “advanced patient outcomes,” or “reduction in manual processes,” healthcare managers feel lost. There are too many questions that remain unanswered.
Gen AI solutions to help create notes in EHRs are so new that evidence of their benefits and physician satisfaction is lacking: The risks are high, while the potential ROI is hard to quantify. At the same time, most healthcare facilities have a backlog of investments in baseline IT infrastructure development and data security. Where to get money for AI if any return on investment is uncertain?
Where to get money for AI if the return on investment is uncertain?
Not every healthcare facility can afford – due to lack of expertise and human resources – to implement projects like the Mayo Clinic, which was an early adopter of Google Cloud’s Enterprise Search on Generative AI App Builder (the tool allows, for example, to build custom chatbots and semantic search applications).
AI-driven systems are expensive and require further investments in IT infrastructure, such as cloud storage capacity or voice recognition systems. The more data, the more money must be put into cybersecurity. Even without AI, medical facilities are very cautious about data-driven innovations due to restrictive data processing legislation.
On top of that, there is one more dilemma: technologies that are cutting-edge today may be obsolete tomorrow – by analogy to smartphones, which can be thrown in the trash after five years because they are neither updated nor serviced anymore.
How to avoid overinvesting or falling behind
Before feverishly implementing AI’s hottest innovations, every healthcare facility should create a long-term gen AI strategy addressing issues like budget and human resources planning, skill building, goals, and critical problems AI is to solve. Instead of chaotic purchases of AI applications suggested by startups and IT providers, consider how the technology can serve your organization first.
According to a survey by Bain & Company, only 6% of the US healthcare systems have established generative AI strategies. One of the barriers is the unclear legal framework.
The following important assignment for healthcare leaders is creating a solid data infrastructure and high EHR maturity. AI-based applications augmenting EHR’s functionality will reimage how data in electronic health records are collected, entered, processed, or searched. A must is a cybersecurity strategy aligned with the AI strategy. Every hospital should be aware of the risks: biased outcomes resulting from biased data, hallucinations that can lead to inaccurate outputs, and mistakes in data processing that might be harmful.
McKinsey’s experts suggest that “organizations must learn how to use gen-AI platforms, evaluate recommendations, and intervene when the inevitable errors occur. AI should augment operations rather than replace them.” Healthcare organizations must consider training programs to enhance the skills of their employees and strive to streamline the integration of generation-AI-powered applications for frontline staff, ensuring user-friendliness without imposing additional work demands or detracting from patient care responsibilities.
Building generative AI tools by healthcare facilities requires a partnership with an experienced tech company. According to McKinsey, healthcare managers must always check if the solution complies with regulatory requirements and data security standards.
Generative AI solutions not only help to perform some tasks but also impact the workflows and roles of professionals. Before investing in AI, leaders should think about the long-term consequences of AI.