Vice President of Clinical Data Science, Dr. Ashwin Prakash, recently sat down with us to talk about the investment CHS is making in artificial intelligence and machine learning as a way to enhance safety, quality and patient outcomes.
We designed CHS’ clinical scorecard to shine a light on the measures behind our clinical outcomes and how well we are performing on them. The scorecard provides a monthly glance at key measures like nurse turnover rate, CAUTI, stroke door-to-needle time, etc., and helps identify improvement opportunities. Having said that, the monthly cadence only provides a retrospective look at how we did over a 12 month period.
What we needed next was something more tactical, something that let us know how we are doing on a daily basis and provides the ability to drill down – so we created clinical dashboards. As patient data flows into the dashboards, we’re able to see near real-time performance. The dashboards allow us to implement processes, make adjustments to those processes, and see what sort of impact those adjustments have on our metrics before they land on the scorecard. In other words, the monthly cadence of the scorecard is now broken out to a daily cadence with the dashboards.
We’re now looking to take things one step further and implement continuous monitoring. For example, if you are trying to do capacity management and bed utilization, you would want to see how your rooms are utilized at that point in time so you can actually make decisions in real time. The same would apply to use cases like helping leaders predict staffing needs for an upcoming shift. It’s no longer enough to know who was in a bed a day ago – you need to know who is there now so you can staff accordingly. With near real-time information coming in to the clinical data warehouse, we now have the ability to surface that information continuously.
Having access to near real time data opens the possibilities of running complex algorithms such as artificial intelligence and machine learning to predict future state. This requires a whole new platform, able to provide system-wide and end-to-end visibility, layered with the predictive ability to say for example, “Your current ICU patient in bed 1 is likely going to require a SNF stay, so start planning now,” or even look beyond the four walls of hospitals to better manage occupancy based on forecasted demand. Another application is to leverage trended data to identify patients who are at significant risk of decompensation. That’s what we are building toward now.
We are working to make it easy for physicians to get information they need without overwhelming them with extraneous data. The whole point of this new platform is to surface the right information, to the right person, at the right time. We understand that managing a patient is a team sport and coordination is key. For all of this to work seamlessly, you need to be able to show the relevant information to the right person but at the same time ensure that we are not leading to alert fatigue or overwhelming our clinicians with excessive data. We’re building out a visualization platform whereby everyone sees information about the same patient and the same unit, room and bed, but in a role specific and contextualized manner.
Another important piece of this is that we are able to generate optimized work lists. For example, If you are an infection preventionist, you’re able to see all of your patients who have a urinary catheter and rank ordered by who is most likely to benefit from having it pulled. And you manage your cases based on that to minimize risk of CAUTI. Or, you can look at a specific cohort of patients and say, here are my patients that are most likely due for discharge today based on a predictive algorithm. Those are things that we’re building into the new platform which is primarily about optimization and creating an abstraction layer to present the underlying data.
Predictive analytics is a powerful tool for advancing personalized medicine and tailoring treatment plans to individual patients. By analyzing vast amounts of patient data, including medical history, lifestyle factors and treatment outcomes, predictive analytics can identify patterns and associations that are not readily apparent to humans.
This wealth of data allows providers to better predict the likelihood of treatment success or failure for a particular patient, helping them make informed decisions about the most appropriate treatment options. Predictive analytics can be used to identify patients who are likely to respond unfavorably to a specific medication or treatment plan, reducing the trial and error approach often associated with treatment selection.
Predictive analytics can also help identify patients at high risk of adverse events or complications, allowing healthcare providers to implement preventive measures or adjust treatment plans accordingly. This proactive approach improves patient safety and outcomes.
Bottom line: predictive analytics enables healthcare professionals to practice personalized medicine, providing individualized treatment plans based on a patient’s unique characteristics and predicted response to interventions.
This is about providing physicians and other members of the care team with useful information to aid in their clinical decision making in the treatment of patients. In providing this intelligence, CHS is committed to ensuring the following principles when applying AI:
We look forward to providing physicians with information and education about the new platform and the additional functionality it will provide in the coming months. Our current roadmap takes this work out to late 2024.
On the Lighter Side:
We have a two-year old, so we don’t have a lot of spare time these days, but when we do, we enjoy traveling. I’m also a bit of a history buff, so I like reading about history and watching documentaries.
As a physician, I knew I could make a positive impact on my patients, but the scientist in me realized that I could help deliver care on a much larger scale by focusing on bringing medical technology to the bedside. This was purely a case of connecting the “dots” in hindsight. When I got my PhD in in Computational Genomics, the field of AI was also making progress and I was able to use some of the nascent natural language processing (NLP) algorithms to decipher the language of DNA and RNA. It is amazing that today those same NLP algorithms have received a lot of attention with Chat GPT and other large language models.