Read this if you are in healthcare and interested in AI.
AI in healthcare is here. While it’s the early days and there is a lot still to figure out in terms of accuracy and risk, it’s undeniable that AI has huge potential in solving some of healthcare’s greatest challenges.
AI to help alleviate staffing shortages
A February 2023 report to the US Senate drafted by the American Hospital Association painted a startling, though not new, picture of the healthcare staffing shortage: “The result of mounting pressures on the healthcare workforce has created a historic workforce crisis complete with real-time, short-term staffing shortages and a daunting long-range picture of an unfulfilled talent pipeline.” While this report focused on providers such as physicians and nurses, hospitals nationwide are struggling with filling positions across the board. There are just not enough people to fill the need—now and for the foreseeable future.
AI has the potential to address this critical staffing shortage in a number of ways. The area where most people think of AI efficiencies is in administrative tasks. These, as anyone in healthcare knows, abound in the daily work. Reducing the administrative burden on clinicians means more time with patients, and more patients that can be seen. In this scenario, it’s also possible that this reduced administrative workload will reduce burnout, improve the clinician experience, potentially keep more clinicians in the field—and encourage more to join. This, in turn, can reduce recruiting and retention costs and costs associated with outsourced and travel staff.
It has been shown that nurses can spend up to 35% of their time on data entry, documentation, and charting. Generative AI can assist in these tasks, freeing up nurses for direct patient care.
Risks to consider: The biggest risk here is the protection of personal health information (PHI). Safeguards need to be in place to remain Health Insurance Portability and Accountability Act (HIPAA) compliant. Common AI platforms are often hosted by third-parties, or are open-source systems that require proper due diligence to help ensure that any PHI in AI systems is secured. Often, medical billing systems contain PHI that is overlooked as a risk for HIPAA compliance, but can easily result in large scale breaches. Using AI to help with such administrative tasks will save time, making sure that the correct IT controls are put in place and are frequently evaluated for operating effectiveness.
AI to improve patient experience and outcomes
An added benefit of using AI for administrative tasks and reducing the workload on providers is that patients will have better experiences when providers have better experiences. Patients will, theoretically, have more time with their providers and more focused attention. Another way AI can help with patient experiences and outcomes is through AI diagnostic and image-interpreting tools.
With AI’s ability to sift through and analyze huge amounts of data, it can assist doctors in diagnosing patients and even predict what illnesses a patient is likely to develop—in some cases, with life-saving results. A March 2023 article in the Wall Street Journal reported that AI was able to diagnose sepsis in hospitalized patients two hours earlier on average than humans, which reduced the sepsis mortality rate by 18%.
AI is also proving to be adept at analyzing images. Using Natural Language Processing (NLP), AI can provide high-quality analysis of X-rays and MRIs, leading to precise early diagnoses. GE Healthcare reported a 30% increase in speed and enhanced image quality using this technology.
Risks to consider: As with any technology, results depend on the quality and quantity of data that is being analyzed. Patients may also feel uncomfortable being diagnosed by AI. Any diagnosis should be verified by a human and any technology should be thoroughly vetted and tested prior to use. There is a particular risk when using AI in emergency room situations, where the treating physicians may not have the patient's full medical history for backend algorithms to consider. This could result in a misdiagnosis. However, in a clinical setting, the results of AI diagnosis tools are close to the accuracy rate of a doctor (about 90%). The biggest challenge to the medical profession will be getting patients to trust the use of AI. In a study by Harvard Business Review, surveys showed that patients felt their medical conditions were unique and overwhelmingly preferred meeting with a doctor versus being diagnosed by AI.
AI and revenue cycle optimization
At BerryDunn’s Hospital Summit in October, audience members were asked where they saw AI having the greatest impact in the next five to 10 years. The top answer, with 35% of votes, was in the revenue cycle area. According to a recent article published by the Healthcare Financial Management Association, there is the potential for $9.8 billion in savings by automating revenue cycle functions. Many organizations are already using Robotic Process Automation (RPA) to aid in their revenue cycle functions. Many are now adding predictive analytics and automation using AI.
AI tools can help revenue cycle operations in a number of ways. One example is using AI to prompt providers on when and how best to document. It can also be used in medical coding to suggest appropriate codes based on relevant clinical data. One hospital reports that they’ve saved over $1 million using AI in their revenue cycle.
Risks to consider: As with most AI tools, the biggest risks are related to data privacy, security concerns, and concern about the accuracy of data. Like any investment an organization makes in technology, new systems need to be fully vetted and tested, and organizations must follow proper change management practices so that the new technology is properly designed and implemented. Rushing to implement AI may result in incorrect outputs and increase the organization’s risk exposure.
The healthcare industry is already a front-runner in using AI and there’s an exciting future ahead. Risk management is key to launching any new technology and particularly AI. Organizations should prioritize the following: Inventorying all uses of AI at the organization, having clear policies and procedures in place, and doubling down on compliance.
If you have any questions about AI or your specific situation, please don’t hesitate to reach out to the BerryDunn team. We are here to help.