skip to Main Content
artificial intelligence

IT Briefcase Exclusive Interview: Understanding How Big Data Is Transforming Healthcare

Sanjeev Agrawal

Sanjeev Agrawal

President and COO, LeanTaaS

A common challenge with healthcare in general, and cancer care in particular, is the peaks and valleys in patient appointment scheduling, which often results in inefficient use of scarce and expensive resources, unpredictable wait times for patients and overtime hours for staff. Sanjeev Argawal, president and COO of LeanTaaS Healthcare explains how data-science can optimize the scheduling of patient appointments based on a holistic view of healthcare operations.

A: “Big data” is a big word, but I can tell you how data is transforming healthcare. There are three very exciting things happening right now:

– First, a lot of patient data is going online; EHR adoption is happening at an unprecedented pace. And then, there’s a lot of public data that’s opening up — data about diseases, trends, clinical research, etc.

– Second, the technology to analyze this vast amount of data and make sense of it is becoming more advanced and accessible. From natural language processing to distributed parallel processing to image recognition and things like that. These technologies were not quite there yet a few years ago. Today, they can do a lot fairly quickly and affordably.

– Third, there’s immense pressure from payers to be accountable and responsible. The shift to value-based care can only happen if we have ways to measure, monitor and optimize. That’s forcing providers to rethink everything they do and see how they can leverage data to improve accountability and reduce costs.

Q: How are hospitals embracing these big data technologies?

A: Pretty much every healthcare CXO I talk to is aware that they need these technologies sooner than later as part of their migration to value-based care. Now that they have implemented the EHR, which is a big step, they’re asking: How can we now leverage all this data to make better decisions? How can we add layers of intelligence on top of this data to provide contextual awareness to our staff and providers?

Unlike other industries such as retail, healthcare takes a cautious approach to adopting new technologies — for a good reason. Before a piece of technology can be adopted, it must go through a number of rigorous checks with respect to accuracy, stability, performance, usability and compliance. But the biggest question is: What’s the ROI?

That’s where LeanTaaS and other advanced technologies that are focusing on operational improvements (as opposed to clinical improvements) have an advantage. The highest and best utility of precious operating room time can be ensured by mining OR case data, understanding utilization patterns, forecasting trends, and matching the right surgeons with the right OR blocks based on operating constraints and surgeon preferences. We look at non-sensitive data and our algorithms don’t have to be 99.999 percent accurate. Our ROI is pretty high, even with 90 percent accuracy.

Q: Let’s talk about ROI of predictive analytics in a bit; but first, what exactly can predictive analytics do for healthcare?

A: Sure. Predictive analytics is a broad term. People know analytics. Analytics is about “what happened.” Call them dashboards, reports or analytics; they were designed to tell us what is going on. Predictive analytics goes a step further. It can tell us “what might happen.” More importantly, it can also tell us “what we can do about it.” Some people call the latter “prescriptive analytics,” but there’s really not much difference.

Fundamentally, it’s about learning from data. If we have a “ground truth” — a set of facts about something — then we can use various techniques to find insights from those facts that could be valuable. With those insights, we can try to predict the future to some extent. For example, if we know what people are buying on Amazon, we can analyze that data and come up with insights such as “young females who buy brands X and Y also tend to buy brand Z” and then recommend brand Z products to them. This is a broad generalization; we can do a lot more with AI and machine learning.

Let’s take a cancer center example. Improving chair utilization for an infusion treatment center is complex largely because of the wide range of treatment durations; some treatments require less than 45 minutes while others require more than eight hours. If a large majority of treatments required approximately the same length of treatment, the problem would be a lot simpler.

The key is to mine the pattern of prior appointments in order to develop a realistic estimate of the volume and mix of appointment types for each day of the week. Armed with that information, it is possible to create an optimal pattern of “slots,” which reflect the number of appointment starts of each duration (i.e., one-hour duration, three-hour duration, etc.) at each appointment time (i.e., 9:10 a.m., 9:20 a.m., etc.). The slots have to take as many of the real operational constraints into account. These include the hours of operation, the number of chairs, various rules that depend on clinic schedules as well as patient-centric policies (e.g., treatments longer than four hours should be assigned to a bed and not a chair).

The Holy Grail is to move the chair utilization graph from the typical triangle that peaks somewhere between 10 a.m. and 2 p.m. each day to a trapezoid that ramps up smoothly between 7 and 9 a.m., stays flat from 9 a.m. until 4 p.m. and then ramps down smoothly from 4 until 7 p.m. Mathematical optimization makes this possible.

iQueue for Infusion Centers ensures that infusion chairs are optimally utilized by mining historical appointment data and generating algorithmically optimized scheduling templates that ensure smooth patient flow throughout the day.

Q: Okay, so what is the ROI with predictive analytics? How do you measure it?

A: Let me give you two examples.

If you run an infusion center, you are well aware of the 10–2 mid-day peak problem. The day starts slow, and then there’s a sudden mid-day rush when lots of patients show up and your staff is scrambling to take care of them, often skipping lunch. Then as the day winds down, things slow again. Pretty much every infusion center sees this pattern. Our infusion product uses predictive analytics to smooth out patient flow throughout the day. As a result, two things happen: 1) the utilization increases during mornings and evenings, and 2) the mid-day chaos reduces.

This translates into a clear ROI. On the operational side, the center sees lower patient wait times, more patients and happier staff. On the financial side, the increased utilization results in up to $20,000 per chair per year of additional revenue. When you have 10, 20 or 50 centers, that number adds up to significant savings.

If you are an OR manager, you know how hard it is to balance your revenue and utilization targets while keeping your surgeons happy. You spend a lot of time just looking for data — pulling reports from multiple systems — and then dealing with a whole bunch of subjective conversations around who should get more block time and why. It can take months to make block schedule changes to meet your targets. Our OR product simplifies all of that. It shows all the OR metrics in one place, explains the WHY behind them, and then makes fact-based recommendations about who should get more time, where and why.

That again translates into a clear ROI. On the operational side, OR managers see significantly increased productivity, surgeons are happier and more productive, and patients get faster surgical procedure appointments. On the financial side, every minute of OR time could run into hundreds of dollars in revenue, so the ROI is pretty clear: $100,000 per OR per year or more.

With predictive analytics, the ROI takes some time. It takes time to analyze the data, come up with insights, implement changes and see the results. It’s not something that you plug in and see positive results right away. It took us a while to consistently measure it and confidently define it for new customers.

Q: So how do you see this evolving? How will healthcare look 10 years from now with all these technologies in the picture?

A: Given the pace of technology evolution and healthcare adoption across the board, such as mobile, internet of things, data science and machine learning, I think healthcare will look very different 10 years from now.

On the preventive care side, we’ve already seen massive changes in the past decade. Mobile and the Fitbits of the world have transformed how we monitor health and stay fit. With sensors and the internet of things, we’ll see even more in the next decade.

On the primary care side, the biggest shift in the past decade has been telemedicine. The fact that you can see a doctor anytime, anywhere with the push of a button or that you can have a doctor visit you whenever, wherever is unprecedented. This will continue and become even more valuable and massive in the next decade.

On the acute care side where our focus is, we see an “air traffic control” analogy in the next decade. An interesting metaphor for the future of hospital operations is how air traffic control and sophisticated scheduling and airport operations have transformed air travel for passengers. They too have enormous complexity and the mission-critical requirement of passenger safety in the face of challenging external conditions. Three direct parallels:

– In order for a single flight to transport passengers safely from point A to point B, it requires “above the wing” services (boarding, food, crew) and “below the wing” services (baggage, fuel, tire check and other inspections) to come together seamlessly. Similarly, in order to perform even a routine surgery, services like labs, pharmacy, the clinician, the surgeon and the supporting team all need to come together to be able to safely and successfully treat the patient.

– At any busy airport, tens of thousands of passengers each day navigate their personal journey across connecting flights while relying on “invisible supporting services” such as their bag transfers, rebookings in the case of delays, weather systems, etc. Similarly, in a busy hospital on any given day, thousands of patients navigate their personal journey across a continuum of care while relying on the supporting services of labs, pharmacy, etc. to be timely and accurate.

– The volume of airline passengers has grown from a few thousand to a few million per day, and airports and airlines have been forced to do “more with less.” Similarly, the Affordable Care Act and a growing and aging population combined with the increased incidence of chronic disease will require hospitals to do “more with less.”

The aviation industry has diligently invested in the required technology, systems and processes to monitor, measure, collaborate and orchestrate. Similarly, hospitals are also beginning to invest in the technology, systems and processes to maximize patient access at each “node” and to streamline the linkages across nodes.

Just as the advent of air traffic control and fine-grained scheduling transformed airports like JFK from handling only a few hundred flights per day in the 1960s to managing thousands of takeoffs and landings per day within the same airspace, modern technologies and predictive analytics will lead to the creation of a similar “air traffic control” capability for hospitals. Assets like the OR, inpatient beds, clinics, infusion chairs and MRI machines will be far better utilized throughout the day; many more patients will be treated within the same facilities; and they will need to wait far less between “legs of their flight” across the continuum of care.

Originally published in IT Briefcase.

Back To Top