/podcast-ep17.png)
Oncology Analytics & Delivering Insights from Messy Data
Summary
Transcript
Data plays a vital role in cancer treatment. In oncology analytics, data analytics can help identify promising treatment strategies, offer better access to affordable care options, and provide critical feedback to medical teams.
In this episode, Susan Hoang, VP Oncology Intelligence & Analytics at McKesson, shares how her team overcomes the inherent challenges of messy healthcare data to deliver insights that can help save lives. Plus, she shares the unique journey that took her from economics and marketing to data science.
We discuss:
- Susan’s unique path to becoming a data science leader
- Sifting through messy healthcare data
- How to define measurable data science outcomes and gain buy-in
DAVE COLE
Hello! Welcome to another episode of the Data Science Leaders podcast. I'm your host, Dave Cole, and today our guest is Susan Hoang. She is the Vice President of Oncology Analytics at McKesson. Welcome, Susan.
SUSAN HOANG
Thank you.
DAVE COLE
Thanks for being here. First of all, what is oncology analytics? Maybe we can just start there.
SUSAN HOANG
Yeah. Oncology is cancer: cancer care. Oncology analytics is really around all the data and insights around cancer care. One of McKesson's biggest programs and strategies is around cancer care. I work for The US Oncology Network, one of the largest community oncology networks in the country. We see one in five cancer patients at one of our centers.
DAVE COLE
Wow. That's impressive. It's the treatment of cancer patients that you're focused on more so than anything from a pharmaceutical standpoint in terms of coming up with a cure, or is there part of that as well?
SUSAN HOANG
Yeah, all of the above. We have different arms of our company that do all of the above but, specifically for the analytics that we do, we provide the analytics that supports all the data and insights around the care delivery. Think about how treatment is given, whether that's surgery, radiation, chemotherapy or drugs; support for cancer patients, their family members and their caregivers; social work; how do you think about the multidisciplinary team (not just oncologists, but nurses, pharmacists, dieticians, social workers) that surround the patient and the caregivers that we support—the insights around that.
The other parts we think about in the healthcare ecosystem is, in addition to how care is given, how it's paid. Our team does a lot of the analytics around the reimbursement or working with payers, thinking about the reimbursement for the treatments and services provided for our patients, because cancer care is quite expensive in our country. I don't know if you know that the average cost of cancer treatment is $150,000 per patient.
Usually, adult cancer is a disease of the elderly, so these patients are typically over 65. Most of those patients you think about are on Medicare. Those are fixed-income folks. They're retired. They're on Medicare and it's 20% out of pocket. $150,000, 20% of that is out of pocket. That's pretty costly. We have to think about access from not just treatment, but the affordability of treatment and making sure it's appropriate and good quality as well.
DAVE COLE
Well, it's truly God's work. I mean, what you all are doing. It's got to be very fulfilling. I want to dive into it. I'm going to come back to this. I want to talk a little bit about the agenda. What are we going to be talking about in today's episode? One of the things that I think makes you very unique, Susan, is your background. You did not start with a traditional stats or math background. You were a pharmacist before you became an analytics leader. Is that correct?
SUSAN HOANG
Yes. Pharmacist. Marketer. All of the above.
DAVE COLE
Great. We're going to dive into that, and I think it's the provocative question of what background do you need to have as a data science leader? I think there can be varied backgrounds. You can be extremely successful coming from different walks of life, and you need not have a PhD in statistics.
The second thing that I think should be near and dear to every data science leader's heart is driving measurable outcomes. How should a data science leader be measured? What does success look like? We're going to talk a bit about some of the metrics that Susan uses as a data science leader. Hopefully, it will be universal and good knowledge for the rest of the audience here. First of all, get back into what we were talking about. Help me understand a little bit about the analytics that you're doing today.
You want to keep care manageable from a financial standpoint. It is extremely expensive. Also, it sounds like there's quality of life, there's improvements in terms of treatment that you're also focused on. What metrics are you looking at there? What are you driving?
SUSAN HOANG
I think our team is the kitchen sink for The US Oncology Network, but if you think about the entire ecosystem to support a cancer patient, we touch the analytics behind all of that. Let's start with how we think about treatment decisions. Our team actually does a lot of the analytics to identify what are the treatment strategies that are most optimal by cancer type. We actually provide the insights and the feedback to the caregivers on how they're performing, how they're benchmarking against other providers, as well as to the pharmacist and the other quality leaders who have to think about the population health approach.
Now, in addition to that, we have to think about reimbursement. How does a provider get paid for cancer care? What you see in the healthcare world is the idea that we have to bend the cost curve. It's just the pace that it's at is just unsustainable.
DAVE COLE
Yes.
SUSAN HOANG
In order to do that, our team actually does a lot of the analytics behind the payer contracting. It's not just what you think about as working with the payer contracting around, “Here's the model. Here's the performer. Here's what we forecast.”
We actually take on a lot of value-based care, which is really paying for performance. What is the ping on the outcomes of the patient? What's the quality of life of the patient? How well are we managing side effects? How well are we managing unintended hospitalization, emergency room visits? How are we managing your pain? And really, how are we managing the total cost of care for the patient, as well, and the patient satisfaction?
If you think about all those aspects, there's a lot of different dimensions. We have to think about everything from modeling and forecasting. What's the likelihood of payback in six months’ time? What's the likelihood we even actually have to think about reserving finance? We do the financial analysis behind that and say, “For different aspects of care, how might we improve or find the insights that can improve where we are and where the opportunities are to improve care?”
So it runs the gamut, but we think about it as the holistic insights that support the ecosystem of cancer care for the network.
DAVE COLE
There's a lot of things I heard there: quality of life; pain management; total cost of care; customer satisfaction, so that'll be patient satisfaction?
SUSAN HOANG
Yeah, patient satisfaction.
DAVE COLE
Yeah, patient satisfaction. If I had cancer, patient satisfaction is, "Hey, am I doing better?"
I would think the typical score would be more outcome-based. I imagine also there's the relationship that you have with the provider itself? Do you feel like you're getting the highest quality care? Is it a number? Is it as a survey that gets sent out? I'm just curious how that works.
SUSAN HOANG
Yes to all, and there's a qualitative component. It's interesting. The definition of quality depends on the stakeholder. To an employer or to a payer who's paying it, there's a different definition than what a patient defines as outcomes and good satisfaction. We know, in our industry, what matters to the patient and their caregivers is how attentive a provider is. Do they have the time? Do they have the support of the oncology care team? Not just the oncologist in supporting them through their care—not just on the care—but surrounding their entire socioeconomic situation.
For a payer or employer, paying the check for the treatment, it's a little different. They look more at things like, “What assurance do I have that there is a reduction in variability of services I will get, from your providers and your network, as well as the quality of care that you're given so that it's a balance of quality and cost?”
That's really the equation for value. What guarantees can we have that we have good value and predictable value for the care we see for our members, as they say in the payer world.
DAVE COLE
That makes a lot of sense. If we can turn back time a little bit. I want to hear a little bit about your journey in becoming a data science leader. Help us understand how you got there. Starting in the more traditional pharmaceutical world and even marketing, and now to where you are today. Give us a high level of how that all happened.
SUSAN HOANG
I would say my career has been a winding path. What unites my path is really a true belief in patient care and access to care. I started off not thinking I was going to go into healthcare. I was an international relations major in undergraduate. I actually did a lot of economics. My training was in economics, so I did really well there thinking I'm going to travel the world after I got my undergraduate degree, but being a dutiful Asian daughter, it must be healthcare. I was able to find a program at UCSF where they focused on pharmacoeconomics: health policy and pharmaceutical economics. That's the discipline of economics, but looking at the treatments and drugs specifically, and that's where I got my doctorate of pharmacy.
From there, I did a lot of clinical programs where we focused on thinking about treatment formularies. How do you think about the cost from a population health perspective? How do you make sure that patients have access to care, but making sure it's thinking about the balancing of cost and affordability and access to treatment. I did that for some time and had really fun jobs in all of that, but it was one of those things in life where you don't know where you're going next. My husband and I had a moment where we were deciding. I wanted to stay on the West Coast. He wanted to go to the East Coast for his training. Mutual assured destruction was to go to Texas.
DAVE COLE
Just go to the middle.
SUSAN HOANG
Here we are decades later, we're in Texas, and it was interesting because at that time I was introduced to a whole new career doing global strategic marketing at GE. I thought, “Perfect.”
I can go back to some of my global aspirations back when I was an undergrad. I was afforded the opportunity to work at GE and look at cancer care across the globe and how care is delivered differently, and the needs are slightly different, even though the unifying theme is cancer care. It was a great time in thinking about cancer care needs and what other portfolios GE had, to be able to put together surrounding cancer care.
What was interesting is, after seven years, I realized maybe traveling does lose its luster. We were on the road all the time in a suitcase. I was actually tapped on the shoulder at McKesson and asked, "Hey, we have a turnaround situation in pharmacy. We were wondering if somebody with your background could help us."
I joined the organization with the intention of going back to some of my pharmacy roots. I did share with the leadership team there that I would love to explore something new and different.
I would say that's where I've landed today, doing analytics. One of the key things that I recognized in my tenure at McKesson was how important it was to have data to drive the decisions behind it, so it's not just based on experience or conjecture; how might we elevate the decision-making through the data and insights. That's where I'm at, but I would say what's neat about my entire career is every job I've been in, I've always been the first at it. I've been able to define it as I go, so it's been fun, and I'm enjoying where I'm at now.
DAVE COLE
That's great. At GE, you were doing marketing, and then you had this opportunity at McKesson. Did you go to McKesson and start as a data science leader, as an analytics leader, or did you go there setting up some program and then say, "Hey, I really want to lead an analytics team."
I want to understand, first of all, why did you want to do that? Then second of all, how did you convince someone that given your background that you were the right person for the job?
SUSAN HOANG
One nice thing about McKesson is they're very open with folks who are willing to take calculated risks. One of the opportunities they saw, and I saw, was the ideas. We had pockets of analysts all over our business unit. The question was: “Is there an opportunity to combine different disparate teams to be able to do more by seeing more? How might we consolidate various previously siloed teams, come together as one unit to support The US Oncology Network?”
DAVE COLE
There were a bunch of siloed teams that were focused on what you said at the outset, which was all these oncology care, payer-related challenges. You're like, "Hey, if we consolidate them, I think we can make them more productive by working together than them all being out in these various silos."
I'm going to put you on the spot here, Susan. This was your idea. You saw this. You pitched it. Then somebody said, "Yes, I see what you see too, Susan. I think you should bring these folks together and herd the cats and manage them and have them work more collaboratively."
Did I put words in your mouth, or is that how it went?
SUSAN HOANG
That's a good summary of the journey so far. Yeah.
DAVE COLE
Great. Sometimes when you look back on your career, you skip over these risks that you take. I think that's fantastic: seeing inefficiency, raising it up. I'm curious: was it a coffee conversation or did you put together a deck? Do you recall how that happened?
SUSAN HOANG
It's probably a series of various different conversations, thoughts, jotting it down, having other conversations, pulling it together, and really having good support from leaders who say, "Okay, here's a risk. Go for it."
It takes a little bit of both and also being lucky; being set up in a situation where around the organization there were parallel conversations of the commit around analytics being a big part of the larger company's organization strategy too. It was a combination of luck and timing, and willing to challenge myself in a different way but also the idea that you can't do it alone. The team has to want to do it too. That's the piece that is interesting about a turnaround realignment situation.
DAVE COLE
Great. I'm of the opinion that, yes, there's luck, but I think a lot of times you make your own luck. The fact that you saw this problem, saw this opportunity, had those conversations, aligned the various folks: that takes skill. That's what great managers do. They see these inefficiencies and they take chances. Sounds like it's worked out great. I'm curious. I imagine your team has a mixture of data scientists and engagement managers, things of that nature. How has it been managing them? Have there been any challenges for you, or do you feel like there's a lot you learned throughout your career that has been very helpful?
SUSAN HOANG
All the above. I would not be telling you the truth if I said it was, what do they say, ‘bluebirds and lemonade’, although somebody said that bluebirds are very vicious. I would say it's a bit of both. One of the things I was able to take from my past experience is the idea that sometimes it's okay to start from new. It's okay to start with a challenge because sometimes, within resource constraints, you might find the most brilliant idea to innovate. I have a track record of building teams, turning around teams, so I use that experience. In addition to that, with the team I had, coming under one leader when you were in different groups, it must come with some skepticism, right?
DAVE COLE
Right.
SUSAN HOANG
I mean, is your job on the line? What else is going on? But the picture you have to paint for everybody and even if you are a leader, I always say this to my team is: just because you have a title, doesn't mean you can't influence without it. You must influence without it, actually. The higher up you go actually, I believe the better skilled you must be at influencing without authority. There is no way you can get folks to do anything unless they believe it too: that shared vision of what needs to be done and why it needs to be done. That was the picture I was able to paint and still am painting for the team. The transformation is not done. We're still working on it.
I'm not a classically trained data science leader, but you surround yourself with folks who buy into this vision of a team, have the right skill set and different skill sets. I think of it this way. I always explain it to my team as: “Think of a superhero team. Everybody has their own special superpower, but when you come together, that's when you really can do a lot of amazing things. The Avengers, the Incredibles, they each have their own superpower, but when you come together that's the difference-maker.”
That's the philosophy we have taken with our team coming together.
DAVE COLE
Diversity of skills, in terms of that Avengers analogy, makes a lot of sense. What is coming through just in our conversation here is the passion that you have for solving a very gnarly problem. Cancer is obviously one of the deadliest diseases that we know of today. Then you have healthcare, which is perhaps the most complicated industry on the planet, at least in the US and then you combine those two things and there's just challenges everywhere.
The fact that you have that passion, you understand the problems acutely, and you know what the various business counterparts, whether it be a provider, the payers etc., are really looking for, can help drive and focus a team that may be well versed in certain parts of the problem, but may not see the sum of the parts.
SUSAN HOANG
Yeah. Our team's makeup is a mix of both clinicians with a background in informatics who understand the data as well as data engineers, data managers who understand the data piece. To your point: you're right. Healthcare is complex so the data is very messy. We call it the swamp because you’ve got to mix the electronic medical record data with claims data, financial data and third-party data. It's not even the data itself that's messy, having to think about our multi-cloud, multi-platform strategies. Some things are on-prem. Some things are not. Our team must be able to put their feet in two canoes.
In our world there's no simple clean platform and that's the way it goes. We've had many team members come from other industries, like finance. I think healthcare is not as sophisticated as other industries when it comes to analytics, but others that have come in with that kind of background even admit that healthcare data is very messy. What's unique is oncology data, particularly, is very messy because of the complexity of the cancer types. Cancer is not just one disease. It's actually many, many diseases, so that's the challenge with it.
DAVE COLE
Why is that? Help me understand. What do you mean by cancer is actually many diseases?
SUSAN HOANG
Good question. The etiology (biology) of the disease is different. When one says, "I have lung cancer," or, "I have breast cancer," the tumor markers behind it are very different. How you treat a patient is very different. Even how it manifests when the disease advances typically goes to different organs.
One must have to think about how you treat it very differently and even the cancer doctors subspecialize. Some doctors might exclusively only see a certain type of cancer patient. Brain cancer doctors are very different from breast cancer doctors, or even the types of doctors. A cancer patient may see a surgeon, a radiation oncologist and a medical oncologist just to treat one cancer. It's quite complex compared to, I would say, even other chronic diseases.
DAVE COLE
Yeah. That makes total sense. Let's segue into the measuring outcomes. That is very near and dear to all of our data science leaders. What are some tips and tricks that you have for how you want to be measured and how you define success?
SUSAN HOANG
At the end of the day, our philosophy is how do we measure our impact for patient and our patient care, and how do we support the clinicians that deliver the care? Oftentimes before the beginning of a fiscal year, and throughout a fiscal year, we partner very closely with other departments and divisions and interlock roles. So if there is a key target to improve hospice or advanced care planning, we tie ourselves to the ultimate outcome and not just to deliver a particular report or a particular project; it's to the end target. Dollars saved, dollars generated, what's the shared savings? We tie it to that measure. We don't measure ourselves on incremental deliverables. We share it with our business partners.
DAVE COLE
What does that interlock look like? Somebody is putting together high-level targets of dollars saved or increase of patient satisfaction score. How do you say, “We're going to focus on these four key initiatives, but not these other two, and here's how we're going to help etc.” What does that look like?
SUSAN HOANG
It usually is an interactive process. It's an ongoing process, and we'll start off or begin a year with a whiteboarding session where we say, "Okay, what are the big goals? What are the aspirational things that we need to go after?"
When we get that shared target, we work with the other departments and say, "Okay, what would it take? Let's work backward. If that's our end game by the end of this fiscal year, let's walk backward and see what it would take to get there."
A lot of times, even to feed that discussion of what should be our big picture aspirations for this fiscal year, we come in there with some ideas and opportunities. We say, “Here's what we look like today. Here is where we think the opportunities might be,” and then go from there.
Sometimes that helps inform the decisions of what should be prioritized as well. What we also push for is we follow really two key philosophies in our principles as a team. The idea of design thinking is you start with the biggest pain, the biggest need that needs to be solved for, then figure out how to build prototypes. I'll call them quick prototypes. What's the fastest and cheapest way to test whether it's working or not, so that we can measure and learn, then improve on it? We do not take the approach of build, and then they will come, as best as we can. We will challenge even our fellow colleagues in different departments: “How do we know? How do we know this works? What's the signal?” We'll ask for that upfront, and we agree to it as a team, and we work on it that way.
A great example of that might be the opportunity we worked on around end-of-life care. We found an opportunity that we could really improve on advanced care planning and patient care at the end of life. With that, we found the opportunities that the biggest need was when it came to care delivery from a multidisciplinary approach, sometimes you needed numbers and not just the qualitative aspects of a patient's care, to have that conversation. That's where we applied machine learning and built a model and said, “Can we help figure out and predict the mortality risk within that last 90 days of life,” not to replace decision, but actually just to help supplement a care team's decision-making.
DAVE COLE
Was this predicting outcomes based on various treatments at end-of-life care?
SUSAN HOANG
It's a combination of patients' symptoms, the patient's treatment, the patient's disease. All of the above, so all the clinical parameters that might be behind to feed into the model. What was interesting about it is with this data science team that we work with, we pushed and said, "Actually, before you go there, start with interviewing the care teams: flying out to see the practices, seeing how the care is given, interviewing patients, interviewing care teams," so that they understand all the different aspects of needs first. Then we made sure that we had oncologists, who are invested in this, actually work with us so that we were co-creating with the customers in it. Along the way we pushed and said, “Can we go faster,” meaning what would be a good enough prototype?
Also having those tough conversations about what's a good enough sensitivity and specificity rate to be able to say, good enough, and what's the tolerance for that? Those kinds of decisions need clinicians and oncologists to buy in on it. What's neat about it is the speed at which we were able to build the model, test the model and how fast it was bought in. Sometimes the customer can be your biggest advocate. We made sure that machine learning was not a black box: it was defensible and transparent. That goes a long way to getting buy-in from oncologists and the care team. You want to know what's inside it. You're not going to trust a model that you're not sure how it was created, how it's made, and how it’s maintained.
DAVE COLE
Yeah, I think explainability is certainly a hot topic, but explainability in healthcare and certainly in oncology treatment is paramount. It'd be tough to get oncologists and clinicians to jump on board if they didn't trust and understand the model. One thing that, clearly, your background brought to the forefront was that you have a passion for actually getting out in the field. I talk about business users and the business side all the time on the Data Science Leaders podcast. In this case, it's talking to the clinicians, the oncologists, understanding what their pain is, what their frustrations are, and making sure that this model that you're building is actually going to fit a need.
When you say they bought in, well, of course they did, because you brought them in. Your team brought them in from the outset. They were part of the journey. They understood your goal, even how you were going to go about it and then, when the model was built, understood a little bit about data science and the fact that it's not a black box. When you said that it's not a black box to them, it made me think that you taught them what machine learning is.
SUSAN HOANG
Oh, yeah. It was really fun because when it came to the field piloting phase. The other thing with data science is you can tinker forever, but at what point do you call it, get out in the field and see if we can get practices to pilot it: “What's the level of fidelity that you need at each stage in order to get to the next stage?”
Having the oncologist, upfront, be a part of it, they were the best spokespeople to even recruit the next round of pilot sites so that they can speak the language of an oncologist, and they can speak firsthand from their own experience, piloting it and then that's how we expand. I think that's the biggest piece in order to make sure that you reduce risk in any innovation, with a design thinking approach. Start small. Think about what is the prototype, the fastest cheapest way to test, and get it into the field.
DAVE COLE
Right. A couple things I learned here, Susan, on measuring outcomes. First one is if you have a company culture where you have annual or quarterly goals, bolt onto those and show how you can be of assistance and how you can help achieve those goals. That's easy. That's a layup.
The second thing I heard is: when you're working on your own projects, get that buy-in early. Go to a proof of concept, proof of value type approach before you just build it and assume it's all going to go to plan and build consensus. Then you grow from there. Then you can really expand if it's working. If it's not working then, hey, you didn't waste more time and resources than was needed, and you tried a different approach. I think that helps at the end of the quarter or at the end of the year, when you're looking back and trying to explain how your team performed, you're not the only person singing the praises of your team. You have folks out in the field doing the same. I think that is critically important.
SUSAN HOANG
I agree. I think the idea is who's going to advocate for you when you're not in the room. That's the piece that I think is really important because it's new, and to many business leaders it's boring. They're not sure how to leverage it and that's the piece that I think our team is doing differently.
DAVE COLE
I think you're showing how to leverage it with each project and with each oncologist or provider, what have you, that you went over with the work that your team is doing. I think that's fantastic. Great advice.
DAVE COLE
Hey, I've really enjoyed this. This has been really interesting.
SUSAN HOANG
Same.
DAVE COLE
I think you certainly have shown that you do not need to have 20 years of statistics and coding and machine learning. If you truly understand what your business users want and need, and you have a passion for understanding their problems and then reflecting that back and working with your team to deliver solutions to those problems, you can be fantastically successful, and clearly, you have been.
SUSAN HOANG
Thank you. This was fun.
DAVE COLE
Yeah, this was a lot of fun. If people want to get in touch with you, Susan, can they reach out to you on LinkedIn or do you have any other social media?
SUSAN HOANG
LinkedIn.
DAVE COLE
LinkedIn. All right. Great. Well, Susan, thanks so much for being on the Data Science Leaders podcast. I had a blast.
SUSAN HOANG
Thank you, Dave.
Popular episodes
What It Takes to Productize Next-Gen AI on a Global Scale
Help Me Help You: Forging Productive Partnerships with Business...
Change Management Strategies for Data & Analytics Transformations
Listen how you want
Use another app? Just search for Data Science Leaders to subscribe.About the show
Data Science Leaders is a podcast for data science teams that are pushing the limits of what machine learning models can do at the world’s most impactful companies.
In each episode, host Dave Cole interviews a leader in data science. We’ll discuss how to build and enable data science teams, create scalable processes, collaborate cross-functionality, communicate with business stakeholders, and more.
Our conversations will be full of real stories, breakthrough strategies, and critical insights—all data points to build your own model for enterprise data science success.