MHE Talks: Efficient Use of Infusion Centers, Other Facilities Improves Access to Service for Patients

Here are some excerpts of a transcript of our conversation with Mohan Giridharadas, founder and CEO of LeanTaaS. The transcript has been edited for clarity and length.

Some background

I have a background in computer science and in business, and I started my business career by joining McKinsey in ’91. I spent 18 years as a management consultant, most of it focused on operational improvement at large scale customers.

I started LeanTaaS in 2010, with the idea being to deliver operational excellence, but delivering it on software as a service platform.

We initially thought that we will be building operations on analytics using Lean principles and delivering customized applications that will very problem specific. Our initial clients were across a variety of industries. Google was a client, Flextronics was a client, Clorox was a client. So we were in banking, technology, insurance, and son Even Home Depot was a client.

We perfected the model of sophisticated data, operational analytics, developing machine learning recommendations and delivering it through a SaaS application But we felt no single solution was scalable. And so our idea was to build something that could really scale. And in that model, we ended up solving the infusion scheduling problem for the first time at Stanford Healthcare. And we realize that every infusion center in the country and on the planet, probably face that issue. And so we realized we had found something and had solved a very hard problem that needed to be solved, and therefore, pivoted the company and put all our energy behind it. Up until that point, we had only raised about five or $6 million. Once we pivoted [about five years ago]. So that's an order of magnitude different in terms of scaling. The company was only about 20 or 25 people when we pivoted, we're over 200 people now. And so the decision to pivot was exactly right. And we followed it up with being able to deliver behind it.

Matching demand and supply

What we do is very sophisticated demand side analytics to predict the volume of patients, the type of patient, the duration of their treatment, when they will show up, the propensity for them to run late, etcetera. So we get a very analytically sophisticated read on the upcoming demand signal for the rest of today for tomorrow, for this week, for next week, for next month.

So while we're building out the demand forecast, we are also understanding the supply side, because to deliver a healthcare service, you need the staff, you need the equipment, and you need the room.

So. for instance, if you're delivering infusion or chemotherapy, the pump the drug, the nurse, and the chair have to all be ready, willing and able at that moment in time. If the pump is broken, the drugs are late for the nurse called in sick or the chair has not been cleaned, that infusion isn't happening. So in the supply side, you've got predictive analytics that are constraint based, because there are a finite number of each of these asset classes that all need to come together.

You need to make the demand and the supply match in very tight windows of time, which is an incredibly hard math problem. It's got predictions and prescriptions and optimization and simulation, etc. So we do that. And based on that, we can recommend how the health system should use its assets.

At the end of the day, we know exactly how the day went, because we've got continuous data feeds. So our algorithms continuously look at what did we think would happen and what actually happened. And was that a one- off deviation or is there a systemic problem that causes us to tweak our algorithms and refit our models?

In our operating room product… we've got very sophisticated categorization methods. So we don't say a spine surgery is a spine surgery. We know if it's a two-level fusion versus a four-level fusion. So we understand the distinctions between the surgeries, we categorize and classify correctly. And we look at the accuracy by surgeon by type of case. And based on that, the next time they're scheduling, we can offer them guidance. “I know you think this is a two-hour case. I'd suggest you leave two and a half hours for it.” So our algorithms are learning in that way.

Relationship to EHR

We are EHR agnostic. EHRs are an excellent repositories of data. They've got every time stamp, they put everything there. So we built data pipelines that are at varying levels of granularity. Once we get going, we get an automated feed once a night after their database warehouses synchronized typically, that's sometime in the middle of the night, we get an automated data push again.

The data belongs to the hospital. And so the hospital IT teams know how to pull the data out of their systems. And so the hospital IT will pull the data out and push it to us. So the automated push every night is after the warehouses are synchronized. We've understood exactly the fields we want; we've helped the IT team write the script that gets pushed to us. And then in most cases, we also get a real time feed.

Relationship to Improving patient access

What we do is we unlock capacity by using the magic of supply demand, algorithmic balancing, and the ability to link disparate services into a seamless flow of a patient encounter. Because, remember, a patient encounter is a series of connected events—they go to the lab, they go to the doctor, they get a procedure done, they recuperate, etcetera.

We can systematically unlock 10 or 15% capacity in most of the assets we work with. OK, so if we work with infusion centers, they should be ablecomplete 10 to 15% more treatments within the same cost envelope. meaning the same hours of operation in the same nursing staff.

With ORs, we can systematically improve the utilization of each and every OR by five or six percentage points during primetime hours.So what happens when you unlock capacity? Unlocking capacity is the key to access, meaning I get have more opportunities to create appointments today, tomorrow, this week, next week. Instead of saying, sorry, I'm full up, the earliest I can give you is next month. So access is directly related to capacity. The more capacity you have, the more access you create. So in that way, we directly help our health systems unlock capacity.

On interfering with clinical judgment

We don't want to opine on clinical matters. Clinical matters are very complicated. They're multifactorial, they require deep, deep specialist with 30 years of clinical experience, we are not going to pretend our algorithms can do that. But algorithms can make very good operational recommendations. And so they do. So what we do is we offer up an intelligent recommendation, we don't jam it down anyone's throat. So think of it like Netflix. It's understood your viewing patterns. It's understood the kinds of movies you like the kinds of actors and actresses you like. And it offers recommendations. It's not forcing you to see a movie, but its recommendations are pretty good. Because they understand you quite well.

Responding to the pandemic

Very early on, we created free tools that we made available to everybody, not just customers, to help calculate the backlog on surgeries because as surgeries are getting canceled, the backlog is growing. But someone who needed a heart surgery still needs a heart surgery, even if you push it out a few weeks —that's not going away. Someone who needs a knee replacement needs a replacement. Given that the OR is quite stressed, how do you bring that backlog back? And so we created backlog calculator tools.

With infusion, how do you alter your templates in a world where every other chair needs to be left empty for social distancing? And so we came up with all of these templates and backlog tools and made them generally available.

We conducted a series of webinars where we offered up our perspectives. But more important than our perspective we brought in our customers. Our customers are the leading health systems in the country. We brought them in to share what they were doing on monitoring, on testing, on managing shifts, on remote work, on patient isolation,And those who are very well attended, we'd have hundreds of people from dozens of health systems attend these webinars. So we shared all of those best practices. Finally, what we were able to do is take some of those tools and incorporate them into our product. So for customers who have our products, we were able to leverage the backlog creation in the backlog work because we had all the data.

Success with Novant Health

To give you an example of one particular health system that responded very quickly during COVID Novant health in North Carolina, it's a massive health system. We've been talking to them off and on for many months. But then when COVID hit, they suddenly had this urgent need to fix the utilization problem because they could see the backlogs building up. And so in a very short period like four or five weeks, they agreed to launch our product across all 120 ORs at Novant Health.

Up until then our deployment required our team to be on site — not a lot but on site two or three days at a time, for five or six weeks. But no one's going on site. And so we innovated with them to make the whole launch remote and the whole launch accelerated. We went live in a six- or seven-week period. All virtual — no one from our team set foot on property. One hundred and twenty-plus ORs scattered across 20-plus surgery sites with 300 to 400 staff members needing to be trained.

And it went off superbly. They published about it. They also use our infusion product.

The big picture

Across the US we have $2 trillion of healthcare assets, not counting real estate. There are 5,000 hospitals. Let’s assume they each $300 to $400 million worth of stuff. An OR is $15 million. An imaging machine is a million bucks. So collectively, we've got one and a half or $2 trillion of assets. The utilization in healthcare is worse than any other asset-intensive industry: airlines, transportation, shipping, etcetera. So if you could unlock 10 points of utilization by smarter supply-demand matching, you can create 200 to 300 billion dollars of value a year, which is massive for the healthcare system. So that's our North Star.

We focus entirely on the United States. There are 5,000 hospitals. Three hundred of them work with us. So we got only 4,700 to go, so that's kind of how we think about it.

Resistance

I think this is all new to health systems, relatively speaking, because health systems digitize their records less than 10 years ago, right. Deploying applications that are in the cloud is relatively new to them. So they're learning as they go. So because it's relatively new, it's hard to tell the level of investment and effort it takes.

So a health system might reasonably say, “I've got three smart data scientists, and I've got a couple software engineers.” So when I think about what stops us, I don't spend any time worrying about what other people are doing. It's just how can we persuade a health system that actually believes, “Hey,I've got this.” And so we've got to be able to point out, yes, that's great. Those are great resources. But let me tell you the level of investment it takes to solve even one of these problems.