We can thank Alan Turing (British mathematician and logician) in the 1950’s for kicking-off Artificial Intelligence (AI) and questioning whether machines had the ability to think or not. AI wasn’t fully conceptualized and named until the Dartmouth Summer Research Project of 1956, which was a 6–8-week workshop comprised of mathematicians and scientists discussing the simulation of intelligence by machines.
It’s only been since the early 70’s that the healthcare industry started adopting and applying the concepts to biomedical research and thus kicked off the proliferation of AI research as it applies to medicine.
Case managers working in Utilization Management (UM) are typically working 8 to 12 hour shifts, performing 20+ chart reviews per day for medical necessity with intermittent updates of information throughout the day. Very few hospitals have adequate staffing for 24/7/365 coverage of UM. Therefore the average medical record only has a few “touches” by UM during the hospital stay, leaving long periods of time where significant events may be unrecognized, and not acted upon. For example, the physician places an observation order shortly after UM confirms the original inpatient order is appropriate. How long will it be before the status change is noticed? Will a delay in noticing the change, lead to negative impacts to the hospital and it’s revenue cycle?
The current system leads to missed opportunities within the UM department. Humans simply cannot monitor the medical record 24/7, or have immediate recall of the extensive variations in rules and regulations. Healthcare is not static, and throughout an episode of care, the data is continually changing. Unfortunately, there is a lack of system awareness, and slow response to key information found in new progress notes, vital signs, orders, or other data sets. Additionally, UM rules and regulations may vary by the payer, the state, the clinical condition, and other factors. Criteria are frequently updated and it is difficult for the individual to maintain a full understanding of the rules and regulations. Compounding the risk for missed opportunities, is the fragmentation of UM care. Even with the best documentation, knowledge is lost during transfers from one unit to another, or with handoffs from one case manager to another. Combined, these issues frequently lead to delayed or missed opportunities to have a positive impact on length of stay, status determinations, risk for denial of service, and other key performance metrics.
AI can be structured to support and strengthen the status determinations made by the UM team. This may be through relative risk calculations, references to care standards, or the inclusion of relevant regulations. Alternatively, AI may be employed to continuously monitor the episode of care, alerting the UM team when there are key actionable events, such as a transfer to the intensive care unit, or placement of a discharge order.
AI often excels at performing mundane, repetitive tasks, consistently and tirelessly, thereby freeing up time for the UM team to focus on higher priority issues. There are many things to assess when integrating AI, we've created a checklist of considerations to help guide a deeper design.
For more information and to dive deeper into more AI biases, check out our recent blog post titled “Utilization Management (UM) and the Biases of Artificial Intelligence (AI)”.
We are optimistic that AI will be a positive and revolutionary force in hospital utilization management. However, this potential may be lost if AI programs are poorly designed to meet the complex needs of the real world. Our hopes are that institutions will design thoughtfully and proactively in partnership with their case managers and physician advisors. When considering the addition of an AI to the UM process, make sure it is designed for the result you want, that it can produce this result, is user-friendly, and fits within the UM workflow. Whether the AI is a snapshot in time or a process of continuous monitoring, the system should adapt to our needs, rather than requiring us to adapt to its.