Dr. Molly Lauck: Welcome, everyone. Thank you for coming and joining us for this exciting, uh, presentation. As Dr. Lynn said, my name is Molly Lauck. I’m the director of, uh, faculty research and sponsored programs in the Center for Research Support. I also have the honor of chairing the committee that is charged with reviewing all of the wonderful dissertations in doctoral studies that are nominated by faculty for this award. We’re going to begin with Dr. Eckrode’s presentation of his dissertation research on Streptococcus pneumoniae infections.
Although, before he begins, I would like to share with you some of the review committee’s comments on his dissertation because it’s always interesting to know what reviewers find to be meritorious. These are some of the things that they said. They thought that the problem statement was concise and well-articulated. How many times have you heard that this week during your disser—during this residency, right? There it is. They said that the author did a nice job of justifying the rationale for the study he completed, drawing in results of similar kinds of studies, and established the clear, positive implications for his research. Again, things you’ve heard this week.
The study design and methodology were quite sophisticated, above and beyond what is typically expected of dissertation research. That’s a high accolade. The author carefully examined the uses of models in other areas outside the purview of his study, assessing the strengths and limitations of these models. And finally, they felt that the math medical—mathematical, excuse me—modeling approach was a novel application used to address this problem in this study. With that, please welcome Dr. Eckrode.
Dr. Carl Eckrode: Good afternoon. I’m struggling. I’m kind of at a loss looking for a word right now, um, the word that describes the phenomenon of being recognized for your writing by being asked to speak, and, uh, I think the word is probably “privileged,” and thank you for those comments. I’m going to talk, attempt to talk, in 20 minutes on my dissertation. My wife felt that I should have no problem filling 20 minutes. I’m going to try and get it very quickly, so let’s go.
Periodicity of epidemic of invasive disease due to infection with Streptococcus pneumoniae in the United States. Large title for something that you’ll find impacts all of you, and I will also like to say, this isn’t something I did by myself. I’d like to recognize my committee, Dr. Maria Ronhill, Dr. Shanna Morrill, Dr. Hadi Danawi, and my wife, for all the support and effort that they put into this study. Anything like this is a group effort.
What is invasive pneumococcal disease, then, if this is something so ubiquitous and common that we saw this great show of hands; what’s important about it? Well, sometimes when this organism gets out of its environment, it causes pneumonia, it causes meningitis, it causes bacteremia—infection in the blood, it causes sepsis, a whole-body systemic infection.
It can kill you. And it does that. And it kills roughly about 40,000 of us in the U.S. every year. It’s responsible for hundreds of thousands of hospital admissions. We know a lot about this organism. We know what causes it. We know how it’s transmitted. We know who’s vulnerable to it. It’s a normal human commensal. It lives with us always. It lives in our nose. If you live with a five year old who’s in day care, you’ve probably got it living in your nose. It’s vaccine-preventable, and it’s curable by antibiotics, okay. That’s the good news. Other things that we know about it: good news. We know that it comes for us in the Northern Hemisphere in the winters and in the fall. We know that it comes right along, trails right behind influenza.
If you see influenza, you see streptococcal infections. What we didn’t know, leading up to this study, was when the big epidemics came. It’s like going to the seaside—going to the coast, going to the beach, whatever you want to call it. We go to the coast in Oregon. And we have tide tables. And we know the tides are going to come in at a certain time, and they go out at a certain time, and we can predict it. When is the big wave coming? When’s that big wave going to crest, the one that the brave souls surf and the wise souls run from? When is that big wave coming to us? We don’t know this. We didn’t know this. So that’s what we set out to try and figure out.
When did the epidemic come? Is there a pattern to epidemics of invasive pneumococcal disease? What was the problem? What did we want to find out? Does invasive pneumococcal disease in the United States occur in a periodic, predictable pattern? Does it do that on the whole, does it do it by gender, does it do it by age, can we figure this out by geographic area, by census region? And that’s what we set out to do. So now that we’ve got this, we know this organism and we’ve got this problem, and everyone now is a little bit worried about the Streptococcus pneumoniae that’s living up your nose, right? Okay?
Get your vaccines. That’s my thing for vaccination. What did we do? Well, the first thing we have to do is a retrospective cohort study. For the non-epidemiologists, what’s a cohort? It’s a group of people who have some event in their lives in common. Retrospective means they had it in common in the past. After we gathered the data, we look and say, what did they have in common? Risk, pediatric population, very young, very old, adults that smoke, have chronic obstructive pulmonary disease, heart disease, asplenia, HIV, the very young and the very old, and you can see that spike there on the left? That’s 5 to 14 years of age. About the time we go to day care and school and all start exposing each other and sharing.
So here’s where we are. What did we do? We found a data set, a robust data set—the National Hospital Discharge Survey is a national continuous probability survey. It gathers from all over the nation, from large hospitals, the CDC. The Center for Health Statistics collects this, validates it, puts it out on CD-ROM, it’s downloadable from the Internet. It’s out there—decades of data, decades of data. It came to us in ASCII format, large strings of data, weeks and weeks of consolidating it to fit in the statistic program SPSS.
We won’t go into the depths of that, the study population. How do we pick a population out from this data set? We looked at anybody who had a discharge diagnosis consistent with invasive pneumococcal disease from 1979 to 2006. That’s 28 years of data. As my colleague back in Oregon, Mr. George Hicks, has told me, the best way to predict the future is to look at the past. We looked at the past. We extracted these cases by use of the International Classification of Disease, Revision 9, Clinical Modification Code Set. We went, found what’s the code commensurate with pneumococcal meningitis? What is the code? What’s the code? Extracted by the code. Why did we start in 1979? 1978, they used the ICD-8. No bridge between.
Why 2006? Last year of reported data, but it gave us a very large—very large—data set. Roughly 28 million people, seven million cases. Then we subjected it to analysis. Overall yearly incidence. How many people per 100,000 got this disease? We used this to compare to previous studies to help ensure the accuracy of the data set. That’s something that, if I had to leave a take-home message today, it would be check your data sets. I decoded them three times, then compared them to each other and found errors in the set, went back and re-decoded them again. We used SPSS 17 for Mac because that’s what I had and that’s what I’m familiar with. You can do it with SASS, you can do it with our SPSS commercially available software.
We came up with the means of the incidences for endemicity. What’s the average level of this disease in the population in the United States? And we looked at it by geographic region, by age and ranges, by gender. We didn’t look at it by race. I would have loved to have looked at it by race, but there’s an under-reporting problem in the NHDS. About 20 percent of the data set, nobody reported it. Left kind of a gap. Then we did a twist: frequency domain time series analysis. As I was explaining to someone earlier, imagine a Slinky. Wonderful toy; I wish I’d brought one with me. The time and disease in the cycle is a Slinky moving away from us. Each coil is one year. Divide that Slinky into 12 parts, that’s a month. We looked at month after month of the disease. Then look for regularities that repeat. Every four coils, if something repeats, well, hey, you’ve got a period, you’ve got a spectrum.
And we used time-series analysis in the frequency domain to attempt to determine if there were repeats in that great coil of disease over time. Hope that made sense. Okay? Used a command set called Forecasting. It’s used in business. It’s used in manufacturing to calculate mean time between failures. We used it to look at a disease pattern and then expressed it graphically as periodograms. That’s the periodogram for invasive pneumococcal disease in the United States from 1979 to 2006 in the aggregate. And what do you notice about it? You notice that there’s no spikes. The spikes in there, a spike in there would have shown a regular recurring event. And this is just noisy. This is random. This is chaotic. And that’s what we found out.
We have a finding that invasive pneumococcal disease, when you exclude pneumonia—take pneumonia away from it—when you look at meningitis, bacteremia, septicemia, sepsis—we find out that it’s a chaotic dynamic. It just happens at random. Had we identified a pattern, we could have modified an existing model. Sutton Banks, Castillo Chavez, came up with a beautiful mathematical model. We could have modified part of that model and used it to predict the epidemics. As it is, their rather nice, elegant model stands. The use of long-run probability calculations. Can’t do it with IPD. There’s—we did not find a period.
The good news? Vaccination, our current practice of mass vaccination, is validated by this study. It’s a sound practice. If you can’t predict when it’s going to come, if it’s just hitting when it hits, what do we need to do? Vaccinate. And this is something that I would like to bring up. Vaccinating the pediatric population reduced it in the adult population. Again, if you’re around five year olds in day cares, you’ll meet this bacterium. And we found that reduced incidence in the study. Relying on herd immunity? Not quite as defensible when it’s random, now, is it? Not quite as defensible for randomness. Thus, education, prevention, vaccines, vaccination become much more strongly supported as a result of our work, assessment of vaccine efficacy.
How well does the vaccine work? We always have an argument about that, don’t we? Does it work, does it not work? Does Pneumovax prevent pneumonia? Does it prevent IPD? One of the things that could have happened, if you give the vaccine at the bottom of the periodic cycle and then measure its effectiveness at the top of the period, you could be fooled. That increase in disease could have been just the period. If there’s no period, you can rule that out. You can look at other things to improve the vaccine, like the emergence of non-vaccine-addressed types, the use of time-series analysis as an analytic tool in epidemiology.
It’s been done for measles—very successfully in England; it was done for pertussis and actually helped us to know when pertussis changed its epidemiology and re-emerged. We should do this more. We can embrace this technique. This used off-the-shelf hardware and software. Having an understanding spouse and a UPS driver, I could get the devices that I needed to make this work. And I have a very understanding wife. All credit to her for working with all the twists and turns of the dissertation process and helping us to come up with this. It uses off-the-shelf hardware and software. We support evidence-based practice. The study, I believe, supports evidence-based practice. Our current surveillance, the active bacterial core surveillance system, our current method of prevention, vaccination.
We keep emphasizing those, we’ll keep reducing the incidence of this disease. We can reduce the morbidity and mortality of this disease. 1899, William Osler, the ever-quotable Osler, in his chapter on low-bar pneumonia, called this disease the captain of the men of death. Okay, 1899. We still see it affecting us today. We now have a bit better understanding of it, though. We know that the tide comes in and the tide goes out on this disease in our winter months in the Northern Hemisphere.
We know not to wait for the big cresting wave, to take precautions now. And that is where we see some social change and some impact as a result of this study and as of related studies. Thank you very much for your time, thank you very much for your patience, and I’d like to open this up to questions now.
Woman: Um, first and foremost, congratulations to both colleagues.
Dr. Carl Eckrode: Thank you.
Woman: Yes. Um, Dr. Eckrode, my question is, uh, knowing that you’re in the public health field, I’m curious to know what brought you to this particular study.
Dr. Carl Eckrode: What brought me to this particular study?
Dr. Carl Eckrode: 1998. If you’ll come to the poster presentation, you’ll see a graphic I didn’t use in this slide. There was an unusual spike in this in 1998. I was working as a fairly new respiratory therapist in an intensive care unit. We were getting hammered, and people were dying in their mid-30s of invasive pneumococcal disease. I asked an older therapist, what’s going on? He says, oh, this happens every once in a while. And I asked, did it? I wanted to know when this was going to happen again, and that’s why.
Woman: Okay. Thank you.
Dr. Carl Eckrode: Thank you. It came to me.
Emmanuel: Hi, my name is Emmanuel. I’m, uh, I’m with the School of Management in the Ph.D. of Management with a concentration in Information Systems Management, and I’m interested in, um, not a similar study but something in the medical field. And you mentioned the data set you used, the database. How readily available was it after you had the IRB approval, if you could briefly relate to us how you came across the database? And my second question is, what were some of your recommendations for future study, and have any of these recommendations, uh, been worked on or other continuous work in these areas? Some students may find those gaps and may contact you for, um, some ideas on how to follow up with the research that you’ve just completed. And what were some of the limitations you came across while completing your dissertation? Thank you.
Dr. Carl Eckrode: Those are great questions, and I’ll try to tackle them 1-2-3. The availability of the data set. It is a publicly available data set. Once I had IRB approval, it was a matter of going online to the National Technical Information Service, NTIS, and ordering them, uh, the data sets from 1979 to 2000, uh, came on CD-ROM. From 2001 to 2006 can be downloaded from a file transfer protocol server at the CDC, through CDC Wonder. Uh, the total cost of the CD-ROMs was $149.80. Okay, why that particular data set? Uh, I was taking the class in informatics and was searching through data sets and got, uh, assigned, we all got assigned to look in CDC Wonder, and it certainly is exactly what it says, a wonder. There were numerous data sets available, and when I found this one, I thought, oh, there it is, there’s the gift right there.
Uh, what recommendations do I have for future studies? I specifically recommended that in the future, somebody try a mathematical approach called wavelet analysis, and I please ask that nobody ask me to explain wavelet analysis; it’s quite complex. I believe that if there’s a hidden periodic component, if, uh, that wavelet analysis might have been able to extract it, but it’s very mathematically complex. Uh, I have not published on this work as of yet, so that recommendation has not made it out into the profession.
Emmanuel: Would—would your wavelength analysis include, uh, the frequency, and let’s say lambda?
Dr. Carl Eckrode: Uh, it would address the frequency. I don’t believe that it would necessarily look at lambda. It’s more of a spectrum density. It looks at the density of the spectrum. Uh, I’ll do a short version. Imagine you’re back at the coast again; we’re back at the beach. You’re standing there next to the water, and you see the waves coming toward you. Well, you see the wave in front of you that’s five foot tall, but it obscures the one on the horizon that’s 25 foot tall. Wavelet analysis allows us to look at the densities through that data set. And I would like, I would love somebody to tackle it.
Emmanuel: And the limitations?
Dr. Molly Lauck: I’m, I’m going to encourage you to continue this discussion afterwards, ‘cause it seems like you have some really wonderful questions to ask Dr. Eckrode, and I think you’re going to benefit from one-on-one conversation.
Dr. Carl Eckrode: Thank you.
Emmanuel: Thank you.
Dr. Molly Lauck: And he’ll be at the poster session as well.
Dr. Carl Eckrode: Definitely. Please come and see me at the poster session.
Dr. Molly Lauck: Okay. I think we have one more question over here.
Bob: Um, Bob Williams, student of, uh, public policy administration. Through the dissertation process, what was, uh, one of the most rewarding experiences you went through, and what was one of the most challenging or frustrating?
Dr. Carl Eckrode: Oh, the most rewarding experience, the defense. I’ll be quite, quite honest—the defense, it was a wonderful experience. You—everyone has had the stories of the defense gone out of shape? Well, no. It was wonderful. I got asked some very, very good questions, some thought-provoking questions and, uh, came away with feeling like I had really accomplished what I set out to do. And please, stay with it. Stay with it. You’re going to finish this process. Make the decision to finish it with a Ph.D., okay? What was the challenge and the frustration? The first time I tried to open the data sets.
Dr. Carl Eckrode: And I realized that four gigs of memory was not enough, so overnighted, and then I realized that 12 weeks was going to be about the minimum amount of time to open these data sets and then not being able to run the analysis, spending two weeks to find out that I had misnamed four variables. That’s okay because when you discover it and you fix it, you get to do a very discreet, out-of-the-way happy dance and get on with it.
Layton: Hi, I’m Layton Simon. I’m a D.B.A., uh, candidate from the School of Management. I’m inquiring about, uh, the data that you looked at in reference to what caused the, um, the patient’s reaction and/or were there any, um, underlying ailments? Um, particularly, I was wondering if there were any, uh, children with, uh, sickle cell or immunosuppressed systems that potentially got worse and potentially have died. I’ve had to deal with a child with that symptom over and over again since six months birth.
Dr. Carl Eckrode: When we pulled the data set, when we extracted from the data set, I did not look at comorbidities specifically. I looked at age, gender, and geographic area. But within the literature review, we do know that, uh, indigenous populations and African Americans are affected disproportionately by this disease. Those who are immunocompromised, including some people with a genetic variant that causes them to be quite vulnerable to this disease, uh, experience it at rates, in some cases in Alaskan Natives, 100 times the normal expected level. Uh, I think that if you were to go back and look at the data set, I don’t know if it would tie directly to sickling cell because this organism does not reside on the red blood cells. It affects the serum in the blood, and it causes hyper-permeability of the vasculature, to get technical. But I think you would also find those immunocompromised and immunosuppressed patients to be very, very vulnerable to this, even in the hospital. You can acquire this one in the hospital.
Layton: Thank you.
Dr. Molly Lauck: Great. Thank you very much.