Why Healthcare AI is Missing the Point

Artificial intelligence (AI) is being rapidly adopted throughout the healthcare industry as organizations make more substantial investments in their AI infrastructure, the number of tech-forward pilot programs continues to grow, and new AI tools continue to be launched at an increasing rate. It’s being rightfully utilized as a potential solution for longstanding industry difficulties because it absolutely has the ability to help broaden access to care, improve decision-making, and strengthen organizational efficiency in an environment of ever-increasing limitations. 

However, realizing this potential demands more than just simply implementing novel technologies. It requires a clear understanding of the intended overall impact those technologies are expected to have and then designing a system which supports the need. Presently, many healthcare organizations are adopting AI by asking themselves what process they can refine, instead of asking what result they are striving for. While this difference appears minor, it fundamentally alters how technology is chosen, implemented, and even evaluated.

When the emphasis is on a single task, AI simply becomes ancillary to existing workflows instead of integrating within them. For example, Ai can be implemented as a chatbot that can help manage the volume of calls, or as part of a model to forecast missed appointments, or as a tool to expedite documentation. Each of these applied usages may operate as intended, yet separately they may fail to create substantial change. Not because the AI tools themselves are deficient, but because the system around them remained the same. AI may speed up a process, but if that process doesn’t ultimately align with the organization’s broader objectives and operational infrastructure, the result is merely faster inefficiency or unmet outcomes.

This pattern is becoming apparent in real-world usage. A recent study revealed that doctors who used AI diagnostic tools with inherent biases were less accurate in their diagnoses, which illustrates how tools not thoughtfully integrated into an organizational change-management process can unintentionally affect care outcomes. Another example is the failure of AI models due to incorrect or incomplete data such as what was seen in the ineffective EHR-integrated AI sepsis detection tool or the faulty Medicare Advantage AI model which had a 90% error rate. Furthermore, predictive models that have been widely adopted have had difficulty maintaining their performance outside of controlled settings, limiting their usefulness when applied to the complexities of routine patient care. 

These instances don’t indicate inherent flaws in AI, but they do emphasize a more crucial point: AI operates precisely as it is programmed, yet not automatically in accordance with the outcomes an organization seeks. And this is a very important distinction. Throughout healthcare, the divide isn’t between the technology and its capabilities, it’s between narrow and broad operational focus. 

The effects of this approach are clear. The industry is producing more insight than ever, but still working to convert that insight into consistent, organization-wide results. 

For AI to be effective, organizations must alter their primary focus. The beginning point isn’t the technology, nor is it the process. It is the ultimate organizational objective and how the technology fits into and furthers that objective.  Once this objective is clearly defined, AI can be strategically employed to support it. Technology then becomes a component of a unified approach, rather than a series of unrelated tools. 

This change in focus can also shift how success is assessed. Rather than asking if a tool functions as it should, organizations should be asking if outcomes are improving. Is the workflow more efficient, is the patient experience improving, and is time to care decreasing? These are only some of the broader operational metrics that are important. 

Organizations that are experiencing significant improvements related to AI are not necessarily using more advanced tools. They are just using the tools they have with more intention. They align technology with objectives, incorporate it into workflows, engage and continually improve both the tool and the system based on performance in the real world. 

AI is not underperforming due to a lack of potential. It is underperforming because it is too often being asked to fix individual processes instead of improving systemic performance. Healthcare doesn’t necessarily require more refined processes, but it does require more aligned systems. When organizations shift from improving how work is done to redefining why it is done, AI will cease to just accelerate activity and begin to fully deliver impact.

Photo: ismagilov, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.