The use of artificial intelligence and data analytics in life insurance
The emergence of data analytics and machine learning is providing insurers and reinsurers with new insights into how they drive and monitor their business.
Announcer: This podcast is intended solely for educational purposes and presents information of a general nature. It is not intended to guide or determine any specific individual situation and persons should consult qualified professionals before taking specific action. The views expressed in this podcast are those of the speakers and not those of Milliman.
Jeremy Engdahl-Johnson: Hello, and welcome to Critical Point, a podcast brought to you by Milliman. I'm Jeremy Engdahl-Johnson, from Milliman's Media Relations team, and I'll be your host today. In this episode of Critical Point we're going to be talking about COVID-19 and the implications for healthcare costs. Joining us today are four of our healthcare professionals. Pamela Pelizzari is a principal and senior healthcare consultant from our New York office. Hi, Pamela.
Pamela Pelizzari: Hello.
Jeremy Engdahl-Johnson: Doug Norris is a principal consulting actuary in our Denver office. Hello, Doug.
Doug Norris: Hi, Jeremy.
Jeremy Engdahl-Johnson: Matt Kramer is a consulting actuary in our Milwaukee office. Hello, Matt.
Matt Kramer: Good afternoon.
Jeremy Engdahl-Johnson: And Stoddard Davenport is a healthcare management consultant also from our Denver office. Hello, Stoddard.
Stoddard Davenport: Hi, Jeremy.
Jeremy Engdahl-Johnson: So, this is quite the dream team we've got assembled here. Everybody has been involved in figuring out the perplexing question of how much COVID-19 is going to cost for the healthcare system. We've got people on this podcast who have been working on building models. We've got people on this podcast who've been looking at what the impact is for consumers and out-of-pocket costs. So, we've got a lot to cover and we're going to go ahead and just dive right in. One production note: with everyone working from home these days, we're recording this podcast via Zoom. So, we apologize if this isn't the crystal clear audio you're used to. We promise the content is the sort of high-quality information you expect from Critical Point. So, with that, let's go ahead and talk COVID-19 and healthcare costs. This first question I want to address to Matt and it's a doozy. What are the biggest challenges when it comes to modeling COVID-19 healthcare costs?
Matt Kramer: Well, of course, the biggest challenges are the unknowns. And, so, what are the unknowns? When we're looking at modeling this, we're looking at, I think, two main components: what is the cost of treatment and then what is the lack of cost that is caused by deferred care? By the fear factor? By people being afraid to go in for face-to-face health visits? So, I think we've got a pretty good handle on the cost per-service if we're talking about people who go to the hospital. We can model that with the data that we have. It gets a little bit fuzzier when you get to the question of what proportion of people are going to the hospital, because there's been such a wide variation in the proportion of people who are tested, the testing rates, the testing availability. So, the denominator of these percentages when we talk about severity distributions, about proportions of people who are going to the hospital, proportions of people who are going to the emergency room get a bit fuzzy. But the real fuzzy thing is the question of what proportion of people are avoiding services? What services are being avoided? We're doing some research into that, but I think the data is still very sketchy at this point here on April 7th when we're recording. There was a really good New York Times article on this just yesterday. It was called, "Where have all the heart attacks gone?" So, I'd refer people to that, if they want to start to get their arms around that question.
Doug Norris: One thing I'll add to that, there's a common joke about actuaries driving a car down the road while looking out the rear-view window. Data is our life-blood. And, in this case, the data's being continually re-stated. There's a lot of underreporting of cases, just because not everyone's getting tested. And I even heard a story in Italy this morning that many deaths aren't being recorded as COVID costs, just because of the capacity needs and the situation. So, for those of us that rely on data to make projections, it's quite challenging.
Jeremy Engdahl-Johnson: Yeah, I heard today that there have been a disproportionate number of deaths at home in New York and they're thinking a lot of them are unreported. Let's talk a little bit more about that data and just the-- I mean, I'm glad that you threw the timestamp out there, Matt, because this is changing seemingly by the day. I mean, how volatile are these assumptions? And I'm curious to get Pamela's take on this just kind of from the public health perspective, but, like, this sure seems like a very fast-moving river of information.
Pamela Pelizzari: Yeah. I totally agree with that assessment. It is a continuously changing landscape that we're working with. I would point to two primary issues that we have with the ongoing changes in the data. One is the infection rate: How many people are going to get the disease? That is changing based on a lot of things, but one of them being the actual changes that we're making as a society. How much is staying home going to affect whether or not we have 10% of the population or 20% or 60% or 80% eventually infected, because we've heard all of those assumptions made in various settings. The second thing I would point to as a real challenge as we're trying to estimate the impact of COVID-19 is how severe those cases are. And I would point out that that's a particularly challenging assumption, because we're looking at data from different countries that have different testing methodologies and different levels of availability for testing. So, any data we have about how severe the cases are, how many people will actually end up in the hospital out of the people who have infections, are all based on whom that country was able to confirm as having an infection. And different countries have taken completely different perspectives on this. So, it's a little too soon to look at the U.S. and we can't really be sure where else to look for that information.
Jeremy Engdahl-Johnson: So, the infection rate is one big open question. I want to go back to something that Matt mentioned, which is this question of care being deferred. What are the dynamics of that? I mean, what do we know? Again, totally understanding that it's early days in all of this, but are we already seeing some patterns emerge in terms of the sorts of things that people aren't doing? I think the question was posed, "Where have all the heart attacks gone?"
Matt Kramer: Yeah, in that article-- and this is extremely un-scientific-- but I'll quote from the article: "In an informal Twitter poll by angioplasty.org, an online community of cardiologists, almost half the respondents reported that they are seeing a 40% to 60% reduction in admissions for heart attacks. About 20% reported more than a 60% reduction." So, that's, obviously, a service where you can't defer it like a knee replacement. Maybe you just say, "I'm going to suffer through that until I feel safer going to the hospital." It's something where you need that care and it's being deferred. And, so, the article goes on to talk about concerns that physicians have about people having their conditions worsen, because they're not getting the care that they need. So, that's something we could see over time as well. We don't really have our arms around this just yet in terms of what services are being deferred. We are hearing from providers that this is potentially very, very bad for their revenue in the short term, because there are just so many services being deferred. And then the longer-term question then, I think, is what impact is that going to have when-- we call this "pent-up demand" is released and people are going and getting those services that have been deferred? And will it have an adverse effect on their health status?
Pamela Pelizzari: Yeah, I'll follow on that. I'm based in New York City where we're having a lot of activity in the hospitals in this area. And you're sort of seeing it at all levels. What we're hearing is that at the really intensive levels, ICU-level care, inpatient care, almost no one is coming in for anything other than COVID-19 at this point. So, even things you might not have thought would be deferrable services, we hear oncology practices saying they're going to defer surgeries for certain types of cancers, because it puts the patient at more risk to have them be an inpatient than to just treat that a little bit later. Services, like maternity, that you really can't defer, they're reducing people's visit counts, making sure that they only come to the office when they truly need to because of the risk to the patients. So, I think that we're seeing this impact as really enormous. And how much it impacts different stakeholders will vary substantially. From a hospital perspective, they're losing a lot of revenue from those services that might not be made up from COVID-19 patients because of the clinical presentation. From an insurer perspective, it really depends on what population you're covering. Are we talking about patients who are older, patients who are younger?
Jeremy Engdahl-Johnson: Well, staying with the hospital topic, they're certainly in an odd situation. On the one hand, all this care that they're relying on to hit their revenue goals is no longer coming through and, yet, they also have big issues with potential overcapacity. What do we know about that issue? We've certainly seen overcapacity issues in other countries. And the U.S. is butting up against it in New York, at least. What do we know and what do we not know about overcapacity?
Pamela Pelizzari: One of the things that's really interesting about COVID-19 is that the patients tend to stay in the hospital for a very long time. So, the impacts on hospital capacity are slow, but they build up. So, you'll have patients, maybe they’re being admitted, but the impact that those admitted patients has on the system lasts for two weeks, three weeks sometimes. So, one thing we've seen is that individual patients are having a huge impact, because every single one who gets admitted is taking up a bed for much longer than, say, the average admission in the U.S. That said, I also think we're seeing hospitals doing really creative things with their beds. They're finding ways to put beds in rooms that didn't typically see clinical care. They're finding ways to convert surgical beds to regular inpatient beds or ICU beds. And, so, we're seeing a lot of flexibility in the system, which will be interesting as the environment changes, because we don't know how that will convert to the regular situation. There's a lot of extra flexibility right now, which regulators could walk back, but we don't know how quickly or how much they will.
Matt Kramer: Another interesting question is what will happen with the field hospitals that we're seeing set up in Chicago at the McCormick Place and in New York? I heard that the Cathedral of St. John the Divine is going to become potentially a field hospital. So, these-- how will care that occurs in these field hospitals, wherever they are set up, how will that be compensated? Or does-- will the payers, the insurers or self-funded employer plans, or Medicare Advantage organizations, will they be on the hook for the care that-- the compensation for the care for the individuals who are being cared for in these very non-traditional settings? Jeremy Engdahl-Johnson: And who even gets paid in that environment? I mean, is it the federal government or--?
Matt Kramer: It becomes a question of who is performing the care, perhaps. Is it going to be a public-private partnership where there would be some private parties who would be seeking to be compensated for that care? I think that's still very much unknown. Again, that'll probably evolve over time. And, in a month, we'll know a lot more.
Pamela Pelizzari: And I think there isn't a blanket answer, right? You're seeing a lot of different things happening. In some cases, private hospitals are expanding their capacity into other spaces: hotels, dormitories. In some cases, you have the federal government or the state government assisting with the setup of additional beds, but then those are run by an existing hospital. And, in some cases, you have hospitals that are being fully set up by a governmental entity. And the answer might be different for all of those in terms of who has to pay for that care and who they're actually paying and how much.
Doug Norris: So, back to Pamela's point about capacity issues, I would say, how well we're able to ramp up in limited time to meet this particular challenge, things like field hospitals, 3-D printing of materials, jerry-rigging ventilators-- that will impact the capacity in a good way, of course. A lot of the capacity modeling I've seen so far is at a statewide level, which implicitly assumes that a bed in Buffalo is equivalent to a bed in Manhattan. And I think that's something that people should be thinking about a little bit more when modeling. We talk a lot about flattening the curve and the main way folks are addressing that right now is to get the peak capacity or to get the peak need lower. And we should continue to do that, but another way to do that is to get the capacity line hire.
Jeremy Engdahl-Johnson: Yeah, in one of his press conferences this weekend, Governor Andrew Cuomo was talking about having an inventory of every ventilator in the state and the capacity to kind of move them around. So, maybe they can get that ventilator in Buffalo to where it needs to be in a hospital in Manhattan. It certainly sounds like a major logistical task. All right. Well, I'd like to shift. We've been talking about these kind of big picture costs to the system. What does all this mean for the patient? Stoddard, you've done some research on out-of-pocket costs for people who have COVID-19. What did you find?
Stoddard Davenport: Yeah, so, the short answer on what does it cost out-of-pocket to be treated for hospitalization for COVID-19 is that it really depends. The US health system is a mix of different payment structures and different types of insurance and each have their own unique implementations for patients. And we have Medicaid that covers a lot of folks on the lower end of the income spectrum, Medicare that covers many of those that are aged or with certain disabilities. We have the private insurance market that covers folks that have employer-sponsored insurance or go out and purchase their own insurance in the individual markets. And we still have a sizable population that doesn't have insurance at all right now. The latest round of economic stimulus included some dollars earmarked to cover hospital bills for those that are uninsured. And the list of private health insurers that have pledged to waive out-of-pocket cost-sharing for patients seems to be increasing by the day. But, you know, for everybody else where there's not some pledge or program to assist with costs, this is going to vary quite a bit depending on the type of insurance they have, potentially where they live, how old they are, how long they need to stay in a hospital, the types of complications they experience and more. You know, to try to wrap our heads around what this might look like, we looked at past hospitalizations for influenza and pneumonia as a proxy for what COVID-19 hospitalizations might look like, and, of course, there are differences in the severity and the disease course and things like that, but the experience of these patients might help shed some light on how this could play out for COVID-19. On average, we saw very low costs for Medicaid beneficiaries. Most of those folks didn't have any out-of-pocket costs at all. For those on Medicare, the bulk of the costs was-- came from their Part A deductible, which is just over $1,400 now in 2020, although those costs are lower for those with Medicare supplemental plans or on Medicare Advantage plans that can sort of enhance the standard Medicare benefits. In the employer-sponsored market, which covers the largest share of those in the US, we saw costs averaging just over $1,000. But for those purchasing plans on their own, through the individual market, we saw costs north of $1,500 in the historical data. That said, those averages obscure a lot of detail. For example, 44% in the individual market actually had no out-of-pocket costs, potentially, because of the cost-sharing reductions that are available as one of the Affordable Care Act features. But for the other 56% that did have costs, those costs averaged north of $2,700 and 25% of those folks actually had costs in excess of $4,300. So, you know, long story short, it really depends quite a bit and we've tried to dig to kind of some of the different layers that come into play here. There's kind of some interesting and surprising, non-intuitive findings in some of what we've looked at as well, where, depending on what else a patient has experienced already over the course of the year, they may have already been more likely than others to have already had hospitalizations and met some of their deductibles and out-of-pocket limits, and things like that, earlier in the year, which could lower what they end up paying for a COVID-19 hospitalization. But, yeah, it's really kind of all over the place, depending on the unique situation of each patient. And, you know, digging into the details behind those averages, I think, is really helpful.
Jeremy Engdahl-Johnson: So, in this modeling work and in analyzing COVID-19, what kind of co-morbidities are we seeing? Are there certain things that are more prevalent than others?
Doug Norris: The CDC actually just released some data on this a couple days ago and the sample size is not large. There's about 7,000 people involved, which may or may not be large, depending on your point of view. But lung issues, chronic lung disease, cardiovascular disease-- diabetes is a big one. A lot of things, like liver disease or smoking status, things like that, I guess it makes sense on some level-- this is a thing that attacks the lungs-- and, so, things that are related to the lungs are going to make it worse.
Matt Kramer: Related to that, there was a New York Times article the other day talking about a study that had shown that people who live in areas with poor air quality have been shown to have significantly worse outcomes from COVID-19 as well.
Jeremy Engdahl-Johnson: Of course, the air quality is getting better everywhere now that no one's driving. So, maybe we can reverse that.
Jeremy Engdahl-Johnson: Okay, so, we've covered a lot of ground here in not a whole lot of time and appreciate the diversity of perspectives on this call. I'm curious, because we've got two actuaries and we've got two of you who are public health experts, kind of the unique challenges that you're seeing from your particular vantage and maybe we'll start with the public health folks, since the actuaries always get to go first at Milliman. So, Pamela and Stoddard, what are you seeing and going through-- what we're six weeks into this exercise of trying to model this strange situation?
Pamela Pelizzari: I think that one of the challenges that I often see-- one of the challenges I've seen over the past several weeks is the balance of using epidemiologic modeling techniques versus whatever modeling technique seems to work. I think that when we talk about modeling from a public health perspective, typically we're actually taking into account assumptions about disease spread and they're incorporated in particular ways. And where I see a lot of research right now, that sometimes does recognize that nuance and sometimes doesn't -- and that's something that's been really challenging is different models with different assumptions are coming up with very different assumed infection rates. And I don't think anyone is necessarily doing their modeling wrong. They're just doing it from very different underlying constructs.
Stoddard Davenport: Right. And, for my part, I think it's been a good experience to work alongside the actuaries in these efforts. You know, I think we're all personally impacted in one fashion or another. We're all anxious to see where we can help make a difference. I think we just sort of have entry points to where we plug in and contribute our skills and experience. You know, within our group I'm happy to let Doug handle anything that comes up in terms of payer operations and things like that. But I try to keep my eye on how this is impacting individuals, both in terms of their health and the health and financial crisis that can be created by an encounter with COVID-19, as well as in terms of the mental health ramifications for all of us.
Doug Norris: I'll turn that around and say that I mentioned this in a past Critical Point podcast episode, but the reason I joined Milliman was the opportunity to solve very challenging problems. And we're blessed at Milliman to have very good clinicians, very good epidemiologists, very good public health experts and very good actuaries and it's been fun to see everyone work together to tackle a problem as large as COVID-19.
Jeremy Engdahl-Johnson: Great. So, let's do a little crystal-ball time here. Clearly, things are moving fast and a week feels like an eternity. But look forward a little bit. Look forward, let's say, a month. What are the big questions or the big one question that you have that you want to get answered as more information presents itself, just given this totally fluid very unprecedented situation. And let's go around the horn. Pamela?
Pamela Pelizzari: I think that I'm very interested in seeing how we can learn more about the age relationship of the disease and how things like age and gender and socio-demographics impact patient outcomes. Right now, there isn't enough data for us to really draw conclusions, especially in an environment where testing isn't sufficient to identify all cases. But early evidence suggests that there is a big difference by age in severity and even by gender, but we don't have enough information to really run that down yet. So, I'm very interested in seeing how that plays out as we learn more.
Jeremy Engdahl-Johnson: Great. Matt, how about you?
Matt Kramer: I'm really interested in the question of how do we know when this is over? Will there be an antibody test? When will that come out? Will there be home testing kits, whether it be for antibody or for having COVID-19? And, so, I think we need to figure that out as a society. Scott Gottlieb had a really good article on that a few days ago that sort of laid out a game plan for that. So, I think there's certainly some ideas out there that'll help us move forward. But I think the question of how we move forward and does there-- is there going to be second hump of this thing? You know, when we look at this flatten-the-curve graphs, there's a single hump. Well, is there going to actually be a lull and then, maybe, let's say, later this year-- is it going to be November, December? Something like that? Is this going to recur? So, that's what I'd like to know.
Matt Kramer: Doug?
Doug Norris: There's about a two-week lag between the strictness of social distancing and social enforcement measures, and I'm curious to see how well it's going to play out when we do things and it seems onerous, but it's not quite looking like it's working yet. How will society respond and will it vacillate back and forth? And I guess similar to Matt's question, but more in the near term how this is going to play out over the next several weeks.
Jeremy Engdahl-Johnson: Great. And Stoddard?
Stoddard Davenport: Yeah, I'll pile on with Matt's comments there. I think the biggest X factor for me is whether or not we're going to see a second hump in this. Are we going to have the intelligence in place to know when it's safe to sort of relax some of the social distancing that's been going on? Are we going to have the capacity for the increased testing and contact tracing and things like that that might be needed? Or, you know, when are we going to get a vaccine? I think, you know, we're all looking at these curves for the current spike and there's a lot of modeling around those. And many of those models are really just not even prepared yet to tackle the question of what would a resurgence look like. So, I'll be keeping my eye on that.
Jeremy Engdahl-Johnson: All right. Well, as of April 7th, there's still seemingly more questions than answers, but we're doing what we can to sift through the noise and sift through all the uncertainty to provide some information that's useful to people as this pandemic moves forward. Thank you, Pamela, Doug, Matt, and Stoddard for joining us. You've been listening to Critical Point, a podcast from Milliman. To listen to other episodes of our podcast, visit us at Milliman.com. You'll also find our substantial and growing library of COVID-19 publishing there, easy to get to from our homepage at Milliman.com. You can subscribe to Critical Point on iTunes, Google Play, Spotify, Stitcher, or wherever else you get your podcasts. We'll see you next time.
Critical Point Episode 21: COVID-19 and the implications on healthcare costs
The COVID-19 pandemic will have far-reaching implications for both short- and long-term healthcare costs in the U.S., as Milliman consultants Pamela Pelizzari, Stoddard Davenport, Doug Norris, and Matt Kramer discuss on this episode of Critical Point.