Episode 4: Tracking the Impact of State Policy on COVID-19 Hospitalizations

Featuring Pinar Karaca-Mandic

Kurt enjoys the company of Pinar Karaca-Mandic, a professor in healthcare risk management at the University of Minnesota’s Carlson School of Management. Pinar, among other accolades, is a research associate at the National Bureau of Economic Research and an editorial board member for the International Journal of Health Economics and Management. Kurt and Pinar discuss her research on the impact of state policy regulations on COVID-19 hospitalization rates. Join them to hear about the results and the implications for the next several months. 

Kurt 

Well Pinar, welcome, today, to the Carrot Shtick. I thank you for joining us here.

Pinar 

Thank you so much for having me.

Kurt 

I’d love to hear a little bit about, maybe share a little bit about your background and what you’ve been working on.

Pinar 

Oh, thanks, Kurt. So, I’m the C. Arthur Williams, Jr. Professor in Healthcare Risk Management in the Department of Finance at the Carlson School of Management. I’m also the Academic Director of the Medical Industry Leadership Institute, or what we call MILI at the Carlson School and I’m an economist. More specifically, I am a health economist. And in a very brief nutshell, my vision in my research is to improve value in healthcare.

Kurt 

That’s wonderful. Here at Carrot Health, we’ve always been impressed with the work done at MILI and it’s a great source of recruiting employees for us, so we love the work at the University of Minnesota.

Pinar 

Thank you so much. MILI at Carlson is just a wonderful community that brings together academics, students and industry partners, you know, with our goal to really shape the future medical industry through education, research, leadership and important conversations.

Kurt 

That’s great. Well, you know, you and I connected earlier this year, as Carrot Health was getting started on doing some research in the COVID-19 space. And I understand you’ve gone on since then to publish a paper and do some additional work there. I wonder if you could share a little bit about what that work is?

Pinar 

Absolutely. So, actually, MILI overall, we had a major response to COVID. You know, in addition to the paper that you mentioned, and I’m very, very looking forward to talking to you about it, we actually launched an important project, which we call University of Minnesota COVID-19 Hospitalization Tracking Project and it’s a collaboration with MILI with another Institute at the Carlson School, MISRC, and with the goal to really collect and disseminate daily information on COVID-19 hospitalizations, ICU numbers from official state Department of Health websites. So, when we started that project, mid-March, only 23 states were reporting accurate, any kind of data on hospitalizations, and it was all over the map completely inconsistent with each other. And, you know, we started collecting the data daily, reached out to state Department of Health media and communication leads to call for action to get more standardized reporting on COVID-19 hospitalizations. And I’m really happy to report today, all states are reporting some sort of hospitalization data. And, you know, as you mentioned, we use this data in terms of sharing with the public, with the media and other researchers through our project website that is housed at the Carlson School of Management. And we also analyze this data and answer some important research questions as well. You know, in addition to sort of more position papers calling for more standardized data, we have done a lot of work in showing preliminary estimates and trends. But also, we looked in a JAMA publication on the association of state stay-at-home orders and how they reflected on trajectories of hospitalizations. And one of the papers that we’re also very proud that we worked on is that we looked at the racial/ethnic disparities in COVID-19 hospitalizations.

Kurt 

Yeah, that’s a wonderful place to start. Because I know, you know, just from our work in the data world, if you don’t collect good data, it’s really hard to do anything about it. So that standardization of hospitalization and reporting at the state level is super important. I’m curious, what did you find on the stay-at-home orders when you did that analysis?

Pinar

Yeah, that was actually one of our earliest papers. And believe it or not, we had enough data just from four states. So, we looked at four states and the analysis was limited to that we basically looked at states with stay-at-home orders. And it was a very simple analysis that we actually did in Excel. We projected out, if there is no stay-at-home order, what would the growth rate look like in COVID-19 hospitalizations, and then we actually then compare that projected growth rate with data from after the stay-at-home order and we saw an incredible deviation from the projected exponential growth in hospitalizations in these states with the stay-at-home orders.

Kurt 

So clearly the stay-at-home orders helped. But now we’ve sort of slipped back into another growth phase, do we need to go back and do another stay-at-home order?

Pinar 

Very interesting question. And actually, you pinpoint another work we have, which is in the works. So, in which we’re actually looking at reopening of the states. And, you know, you can imagine, like closure and reopening, and until we find the vaccine is fully in, and everyone has access to it or some solid treatments, we are going to be going back and forth in these open and closed modes of the economy. And yeah, the closures and the reopenings are not the reverse image of each other. So, I can tell you that. But stay tuned for some of those results coming up in the future. But we can talk about race and ethnicity paper, because I know you really wanted to hear more about that.

Kurt 

Absolutely. Yeah. I’m very curious to learn a little bit more about how racial and ethnic data has been impacted at the hospitalization level.

Pinar 

Yeah, so as a background for that study, several other studies have highlighted disparities in the impact of COVID-19 in terms of infection rates, as well as mortality, but less is known about disparities and hospitalizations. And some earlier studies from the CDC were just looking at 1,500 hospitalizations where race and ethnicity data, unfortunately, was available in only 580 of the patients. Also, very limited studies were also giving us hints as to some of the important disparities when it comes to racial ethnic disparities and hospitalization. So, what we did is we extended that work, we, based on the collection of the data that we have at the Carlson School, around the April 30, several states started reporting not only the hospitalizations, but also their breakdown by race and ethnicity. So, we started collecting that data as soon as it became available. So, then we did a study during the study period of April 30 to June 24. And one caveat, only 12 states reported cumulative hospitalizations by race and ethnicity. So that’s another call for action that we need data from many more states. But really, what we wanted to do is we wanted to extend this work that kind of gave us hints and to the disparities, and do a state-by-state analysis of race and ethnic prevalence of cumulative COVID-19 hospitalizations, and then compare that, this prevalence in the hospitalization data, to ethnic and racial composition of each state’s population. So, what we found is remarkable. In all the states that we looked at, for hospitalizations, proportion of hospitalizations for blacks was higher than the group’s representation in the state’s population, with the exact opposite for whites. And this consistency across, yes, 12 states, but consistency across the 12 states really supports the hypothesis that there are some important underlying systemic issues that are negatively affecting the black communities. And we found very similar disparities for the Hispanic population in 10 of the 11 states where we had data on ethnic composition of the hospitalizations. We found that the proportion of hospitalizations by Hispanics exceeded the representative population in the state. And finally, in select states we saw similar disparities for American Indian/Alaska Native population, for example, in Arizona and Utah.

Kurt 

So, across the board for almost every population, other than the white population, you saw over-representation in severe cases of COVID-19 infection.

Pinar 

Absolutely. And a little bit of a highlight, Kurt, is, of course, this was a study that was completed, you know, looking at April to the end of June, like a two-month period, but looking at 50,000 hospitalizations at that time period. We keep an eye on these data, you know, this is unpublished analysis, but I can tell you that as of end of October, these disparities largely remain in terms of the, you know, the number of states with that are reporting higher hospitalizations in terms of proportion compared to state population for blacks and Hispanics. Well the magnitudes of the disparity seem to be shrinking, which is promising. But again, that’s sort of like our unpublished analysis and our kind of keeping an eye on the disparities work that’s going on.

Kurt 

Can you give us an idea of some of the magnitude of differences that you saw between these populations that were over-represented?

Pinar 

So, another highlight from this paper, actually, is that the disparities exist across the board, but they’re highlighted for black communities in some states. They’re highlighted for Hispanic communities in another set of states. So, if you look at the black individuals, the disparities were the greatest in Ohio. So, in Ohio, almost 32% of the hospitalizations were by black individuals compared to black individuals representing only 13% of the state population. And the other greatest disparity state was unfortunately Minnesota, where we saw that about 25% of the hospitalizations were by black individuals, whereas black individuals represent about 7% of the state population.

Kurt 

So that’s, that’s more than three times the baseline population that were hospitalized. That’s incredible.

Pinar 

It is very incredible. It’s a very large disparity, and if you wanted to discuss the Hispanic disparities, there, the Hispanic disparity was most pronounced in Virginia, where 36% of the hospitalizations included Hispanic patients, compared to 9.6% of the population being Hispanic. Similarly, Utah. 35.3% versus 14.2%.

Kurt 

Wow. So, in that, you know, sort of 200 to 300% range for over-representation. Did you have any clues that you found in the data for why this disparity might be occurring?

Pinar 

That’s a great question. You know, we did not directly look at link to cases and infections. But as I mentioned, there is a growing body of literature that by now establishes that infection rates and cases are also disproportionately represented in minority populations. So, you know, that’s not a particular question that we looked at, but there’s definitely a link between cases and infections than to hospitalizations. But really, you know, when we think more big picture, right, in the case of COVID-19, we also know from a growing body of literature that just these adverse economic, living and working conditions, social determinants of health, that Carrot Health, you know, looks at, you know, and takes to heart. These expose minorities to higher rates of infection, you know, we know that a larger share of blacks and Hispanics have frontline service jobs and production jobs, we know that they face higher socio-economic pressures to keep working, they’re more likely to lack reliable or safe transportation. We know from some studies that they’re more likely to be in more crowded housing situations, over-representation in institutional settings, all of these things contribute. And of course, there is the other piece of the big puzzle is that underlying conditions, which we know clinically now, that are related to more severe symptoms of hospitalizations, like these are cancer, chronic kidney disease, obesity, diabetes, more serious cardiovascular disease. So, all of these we know, disproportionately impact minority populations. So, to answer your question, you know, while we did not look at the exact link, in our opinion, this disparity that we find largely points to some of these long-standing structural inequities in our system. And, you know, like social determinants of health are living in working conditions, access to care, and just these continued long-standing issues, prevention, the lack of access to prevention, or more limited access to prevention, all of these are resulting in what we see today.

Kurt 

Yeah, I think that matches very closely what Carrot Health sees in our own data. When we look at incidence of disease states, for example, like diabetes and cardiovascular disease between, just to pick on Minnesota, between the black and the white population, you know, they’re also in that 200 to 300% range. Similarly, the underlying barriers to health those social determinants also fall into those categories. I was just looking this morning actually at homelessness data from Minnesota and the black population from a housing instability perspective faces that same sort of 200 to 300% gap compared to the white population for risk of homelessness, and those sorts of disparities tend to compound the clinical diagnosis, which puts them more at risk for a critical infection when it comes down to COVID-19. So that’s exactly what we would have expected to see in the data.

Pinar 

Yeah, I agree with you.

Kurt 

So, you made an interesting comment on the standardization of the hospitalization data and talking a little bit about the challenges in collecting the data. What were some of the challenges in trying to get data, accurate data, to be able to make these measurements?

Pinar 

So, as I mentioned, when we started this project, there were just a handful of states that were reporting some sort of data. And what I mean by that is, there was a handful of states that were reporting the number of COVID-19 patients currently in hospitals. Some of the other set of states would report total number of COVID-19 hospitalizations to date, so like, very difficult to make any apples-to-apples comparison. So, part of the challenge was to really try to get a good sense and a good stable set of data that was, you know, reported for the same metrics, like either new admits, or cumulative admits, or current admits. And then a limited number of states report another very severe measure, which we think is really important, which is the number of patients in the ICU or a number of ventilated patients, because that shows the really, you know, the most severe cases that is constraining some of the hospital’s capacity, ICU being very difficult to flex around and increase the ICU beds, for example. So those are some of the challenges. Then the other set of challenges are about when we collect data. What is that really measuring? Some states say that they’re measuring only confirmed cases of COVID-19 hospitalizations, while similar states report a combined measure of confirmed plus suspected. So, like, what does that suspected mean? It means right, it’s not very clear. Whereas the majority of the states are actually kind of agnostic about what they’re reporting, whether the numbers that we pull are confirmed cases or suspected cases. And another very interesting challenge that I didn’t appreciate until there was a period, you may remember, where Minnesota paused their data release, and then for about a week or two weeks, and then they released it again. So based on my conversations with some of the MDH staff, another important challenge for a State Department of Health is how you collect data, will you collect the data by asking each hospital like report our number of patients, reports the number of patients in your hospital that have COVID-19, for example, or report the number of patients in the ICU, or what Minnesota was doing in the earlier days of the data collection was actually they were contact tracing. Once the patient was hospitalized with a positive confirmed COVID-19, they would trace this patient, and then there were all these difficulties afterwards about, has this patient transitions from the regular hospital bed to an ICU bed, or did get discharged or unfortunately did they die, and all of these transitions and tracking these transitions and the timing of transitions and when to release these data, right? These are all important challenges that make it very difficult. You know, when we collect data at 9 am versus midnight, we are collecting data from the entire day or the previous day, and all of these metrics vary by state.

Kurt 

That’s so challenging the fact that we do this on a state-by-state basis, it sort of begs the question of why don’t we have a single national approach that could iron this out?

Pinar 

That’s an excellent question. And I think I tend to be more on the camp of being optimistic on that end. As you know, the CDC was releasing Sunday the weekly data. And then mid-July, there was this other sort of issue that, you know, created a lot of question marks for a lot of people when HHS took over from CDC, and our group took the approach of contacting HHS, and then getting involved in an informal working group with HHS, together with some of the other major groups that are collecting these data, for example, COVID Tracking Project, you know, their leads, is part of the HHS working group as well, and really working with HHS, to keep comparing the data that we’re collecting from states and the data that they are receiving from hospitals, and really understanding and reconciling the differences. Which, the reason why I’m staying hopeful, is that over the months now, I guess, working with these data and comparing it to HHS, that we are converging towards a case where the HHS data collection and the data that we collect from states are telling the same story, they may not still be the exact same numbers, but in terms of qualitative magnitudes, in terms of qualitative trends, that we are moving along the same ways. And I have to say some of our work has contributed to this federal data collection efforts. For example, I mentioned to you the difference between suspected cases versus confirmed cases. And you know, so HHS, based on our conversations with them, started asking hospitals to separately report suspected and confirmed cases and also in their own data release, they differentiate the two so that people looking at these data can make more apples-to-apples comparisons.

Kurt 

Yeah, that’s really interesting, because I remember about that time in the news we were hearing about some states maybe influencing how they reported information to seem like their state was behaving more or less favorably, even to the point of reclassifying cases from, instead of reporting as COVID-19, reporting them as respiratory infections or some other challenges. Did any of that show up in your data? Or did you have enough from other sources to be able to triangulate directionally on the accurate answers?

Pinar 

Yes. So basically, we track our data from the states directly, as I mentioned, some states are certified to report to HHS. So, some states may be reporting it the same way. But there’s definitely variations. I can’t speak to what goes on if a hospital or a state, you know, reports, respiratory illness as COVID or not, I can’t speak to that. We don’t have any data. No data is really is along those lines from the states for us to be able to say that. But as I mentioned, we are now able to mostly differentiate between these suspected versus confirmed cases. The other thing that is also what increases my optimism is that there is more and more transparency in the data. In that, for example, there is a new tool that HHS provides that, kind of, not daily, but on a weekly basis gives an indication as to the completeness of the data that they’re reporting, like how many hospitals or what percent of the hospitals in the state are reporting on X, Y, and Z? Actually, that’s another data challenge, right? Different hospitals may be reporting at different dates and keeping track of that for the sake of completeness of data is yet another challenge.

Kurt 

Wonderful, thank you for that. Let me let me ask, you know, so speculation now. So, we’re seeing, you know, particularly here in the upper Midwest, but sort of expanding out across the nation, infections are ramping up. Our own state is evaluating, you know, sort of additional shutdown measures, as we head into the holiday season here. What do you see for the winter? You know, as you look at growth patterns in the data and sort of speculate, nationally or regionally, where are we headed?

Pinar 

It’s hard to speculate. Maybe I’ll say, maybe it’s too easy or too hard to speculate. Somewhere in between it’s difficult. But let me just share with you some of the data that we’re seeing, we’re looking at these data all the time, especially more recently as there are important concerns of trends and peaks. So, for example, if you look at growth and hospitalizations, what we see in our data is, as summary I can say that, so currently, we are looking at sort of the first week of November, right? All states see positive daily growth in the last two weeks and 45 states see positive daily growth in the last week. So, this is pretty concerning on a nationwide, right. But trends are especially concerning in the Midwest, in some Midwest states. So, what I did, is we looked at our data, looked at hospitalizations per 100,000 adults. So, it’s a very good sort of rate measure that normalizes the hospitalizations in a state with the state’s population, right? And if you look at that, and then this is during the first week of November, and if you then rank these states, okay, the top 10 nationwide states in terms of their hospitalization per capita, eight of these states are Midwest states, South Dakota, Nebraska, Wisconsin, North Dakota, Indiana, Illinois, Missouri, and Iowa. Minnesota is sort of in the midway, Minnesota is ranked 24th in that measure, right. And then we can do the same analysis for the ICU numbers. So, if you look at ICU rates for 100,000 population, so ICU rates per capita, we have this information only from 34 states, not all states. And then top 10. Five of the top 10 are Midwest states, Indiana, Missouri, Wisconsin, Illinois, and Iowa. And note that South Dakota doesn’t report this number. South Dakota if they report, probably much difficult, right. And deeper dive into Midwest, we see daily growth rates of 2.5%, 3.5% over the last two weeks in per capita hospitalizations. And the other thing that you mentioned is overall capacity, right? And where we’re headed in terms of overall capacity. If you look at the Midwest, and this is now data we don’t look at the overall capacity, we only look at sort of percent of beds occupied by COVID patients, both for the hospital beds and the ICU beds. But if you look at the HHS numbers and, kind of, do a triangulation with our data as well, what you see is in the Midwest, on average, 70% of the staff beds are occupied, and 76% of the ICU beds are occupied. On average, again, 30% of the ICU beds are occupied by COVID. And 12% of the hospital beds are occupied by COVID. And you may say, well, what’s a 10%, 12%, like, what does that mean, right? So, there’s not to my understanding, to the best of my knowledge, there is not like a gold standard or a particular threshold that says, okay, after this, the capacity is overwhelmed. But typically, my read of the literature is that when COVID-19 hospitalizations exceed about 10% of the hospital beds, that’s like big alarm. And that’s really concerning. And if Midwest, I told you 12%. Nationally, 14 states exceed that number, 14 states exceed the 10% threshold. In fact, unfortunately, South Dakota, North Dakota, Montana, and Wisconsin, they exceed 15%.

Kurt 

That’s pretty discouraging. I think it echoes what we’re hearing. In fact, I think we heard from one of your colleagues at the University, Michael Osterholm, who’s been appointed to the COVID Task Force, that we’re looking at a pretty tough winter over the next couple of months. Obviously, the end isn’t in sight until we get a vaccine that’s deployed across the majority of the country. But how, you know, if we kind of think that that’s many months away, what does it look like as we get through this winter and beyond? What do we do to bring those numbers back down again?

Pinar 

I think we do, what we do to bring those back numbers back down again, is we go back to what we were doing in May, at the end of May, because if you look at these states, Kurt, like they actually like Minnesota, for example, very much the current numbers of Minnesota both in terms of hospitalizations and ICU, they match our May peak. If you remember, at the end of May, we were going to our big peak, we’re back to that peak. So whatever we did, to get rid of that peak is what we need to do now and if you remember, that was masks and prevention and trying to — Yes, we’re all COVID-19 fatigued and just sort of getting over that and realizing again that there is much more work to do to push through and not give into some of the temptations and just being very socially responsible on how our actions matter to other people. The typical case of like externalities as an economist, I’m sorry, I’m using a jargon, but that’s exactly it, right? My decision to not wear a mask is going to really hurt someone. And we just all have to be very cognizant of that and take accountability.

Kurt 

That’s great, great thoughts. And I this sounds like it touches on your research that’s forthcoming around what happens when we come out of a lockdown. And as states come out, what does that look like? So, we’ll have to stay tuned. I’m very eager to see the results of your next paper.

Pinar 

Absolutely. I would love to keep in touch and I would love to keep you posted on our work.

Kurt 

Wonderful. Well, thanks, Pinar. Thanks for joining us today and I look forward to hearing more from you in the future.

Pinar 

Thank you so much for having me and for giving me this opportunity to share our work. Thank you.