Thousands of Lives, One Study: The Making of the Ontario Health Study
00:00
26:18
What can a 225,000-person study reveal about the environmental, lifestyle, and biological factors that shape health and disease over time? In this episode, Vicki Kirsh and Sandford Borins discuss how long-term health data, biological samples, and administrative health records are being used to advance discoveries in chronic disease, cancer, and health-system equity.
The Ontario Health Study is housed at and funded by the Ontario Institute for Cancer Research (OICR).
TRANSCRIPT
Misty Pratt
The most important discoveries in health research don’t happen overnight. Sometimes they take decades. Today, on In Our VoICES, we’re talking about the Ontario Health Study, a groundbreaking research initiative that follows 1000s of people over time to better understand how our genetics, environments, lifestyles, and life experiences influence our health. It’s a powerful reminder that every participant’s story is part of a much larger picture. Joining us today are Victoria Kirsh, Interim Director of the Ontario Health Study, and Sandford Borins, a participant in the study and Professor Emeritus of Public Management at the University of Toronto. Together, we’ll discuss the value of long-term research, the commitment behind it, and how insights gathered today could improve health outcomes for future generations. Welcome, Vicki and Sanford, to In Our VoICES.
Vicki Kirsh
Thanks so much.
Sandford Borins
Pleased to be here.
Misty Pratt
Vicki, I’ll start with you. What is the Ontario Health Study and how did it get started?
Vicki Kirsh
So the Ontario Health Study is a long-term health study of about 225,000 adults across Ontario. People joined between 2009 and 2017. They told us about their health, about their habits and their lives, and many of them gave a blood sample. We now follow them over time, both actively, so that means through questionnaires and passively through linkage to administrative health data, and we see who develops what and why, and then we try and associate those health outcomes with likely causes. This type of study to an epidemiologist is called a prospective or longitudinal cohort study. Ontario, I should add, wasn’t the only province doing this. There were several regional cohorts that were running in parallel, and over time we harmonized our questionnaires so that the data could be pooled, and that became known as CanPath, the Canadian Partnership for Tomorrow’s Health, which now comprises seven regional cohorts spanning all 10 provinces, and includes about 350,000 participants. Ontario, though, is the largest single piece of it.
Misty Pratt
You know, it sounds like what you’re looking at is not necessarily just health. You’re also looking at lifestyle too.
Vicki Kirsh
Exactly. I mean, the data that we collect is really comprehensive. Our questionnaires are, you know, long and comprehensive. They ask about health outcomes, but also social determinants and mental health and well-being, and so that allows us really to address a variety of outcomes, diseases, and and psychosocial factors as well. What underpins all of this work is really being able to answer the question: Why do some people get sick and others don’t? So we collect, you know, right across the range from genes to behavior to the neighborhood you live in, and that lets researchers work out which of those things actually drives health outcomes. And you really can’t answer questions like that with a short study of a few 100 people. You need many people and many years of follow up.
Misty Pratt
Right, and what are those types of diseases you’re looking at, the most common ones?
Vicki Kirsh
I mean, cancer is the main one, and it’s really sort of you know why the study was set up at the outset. We were we had funding from the Canadian Partnership Against Cancer, and cancer was a main interest. It’s also sort of made easier to study because we do have you know legislative reporting for cancer in Ontario, and we have the Ontario Cancer Registry, so that captures almost all new cancer diagnoses, but now we can study a variety of outcomes through these administrative health linkages, and also you know through through our questionnaires. So heart disease, asthma, stroke-the list really goes on.
Misty Pratt
And how common are these kinds of studies in the research world? Because when I think of research, and some people may be listening, thinking of clinical trials, right? That seems to be kind of like gold standard. How common is it to have these big studies with like hundreds of 1000s of people that you’re following over a really long period of time?
Vicki Kirsh
Well, I mean, CanPath has over 350,000 participants, and that really rivals the the biggest study that many people know about, which is the UK Biobank. They have about 500,000 participants, so you know we’re right up there with these you know an international study like that one. There are others that people might have heard of, the Nurses Health Study that’s you know run through Harvard. There’s the All of Us study in the in the U.S. You know most countries have you know a large prospective cohort study, and we’re now starting to work together and pool of all of our data so that we can look at some of the rarer health outcomes.
Sandford Borins
As I understand it, longitudinal studies have to be big because people are continually dropping out of the pool.
Vicki Kirsh
Exactly.
Sandford Borins
So they you want to have large samples so that you can come up with sort of statistically valid conclusions.
Misty Pratt
How hard is it to keep people engaged over such a long period of time?
Vicki Kirsh
It is difficult, and we do have people, you know, drop out over time. And sometimes people come in and out. There’s a certain questionnaire that might appeal to them more than than another. They’re interested in telling us about their diet, maybe, but maybe during. COVID, you know, they were preoccupied and less likely to fill out a questionnaire then. So, you know, people drop in and out. I would say of the study, and then there’s also this ability that we have in Ontario to follow people passively through these administrative linkages. So even if they don’t fill out the questionnaire, they’re not lost to us, and they’re still, you know, part of our analyzes and part of our database, and you know, researchers are very much interested in in the data that they have already provided.
Misty Pratt
Sanford, I just want to go to you first too before we talk about how the public was involved. You know, tell me about your involvement in the OHS.
Sandford Borins
I don’t remember exactly when, but I did hear about the initial survey, and I thought it was a good idea. As an academic, you could imagine that I always thought that research is a good thing. And when the initial survey became available to me, I completed it. And then during COVID, there were one or two COVID surveys, and I completed the COVID surveys. And then recently, there’s another survey that looks at people’s diet and people’s time allocation, and I just completed that. So you might say I have a thing about completing surveys.
Misty Pratt
Yes, I am with you. I love surveys.
Sandford Borins
In part because a lot of my research was based on surveys.
Misty Pratt
You understand the importance of surveys, and Vicki, how was the public involved in some of the other aspects of the study, like from the get-go?
Vicki Kirsh
So everybody volunteered to participate, right? So we already have, you know, a group of people who are interested in research, motivated, as Sanford said, and they they came to us through different means, like through their family health team, some through their employer, some came because a friend or family member told them about the study, and then some of them actually showed up to these community centers that we had at the outset. They were set up in different regions where people could come in. Some some of them provide physical measures, and they were able to give a blood sample close to home.
After that baseline enrolment, they consented to several things. So they consented to have their share their data with researchers, to have it linked to administrative health records, and then to be contacted again in the future. And it’s really that consent that makes everything possible. So after that baseline enrollment, participants were given their own login on our portal, and that’s really where they complete questionnaires. They keep their contact information current. They choose what they want to hear from us and how often, and that’s how we go back to them. Because anyone who consented to recontact can then be invited through that portal to complete a new questionnaire or even an ancillary study,
Misty Pratt
Like the Heal study, like the one that’s now asking about diet and…
Vicki Kirsh
Exactly.
Misty Pratt
Right.
Sandford Borins
Yeah, that’s that’s the one that I just completed. That was a complicated study. I remember I it was supposed to be done on a previous day, so at the beginning of the following day, I you know made notes on both diet and my activities to get ready to go in and complete the form.
Vicki Kirsh
Yeah, the dietary collection I have to say is probably the most cumbersome of our questionnaires, but it’s it’s so important and it’s one of the areas that we we missed at the outset. You know, and it was this this gap that we didn’t have dietary assessment, and it’s important for so many different health outcomes, even if we just need to adjust for you know total energy intake or or something, it’s they’re important variables that we need for all sorts of analyzes. And I’m glad that we’re finally getting around to collecting it, and and that we there was there was funding to do so.
Misty Pratt
And how is the OHS connected or being connected to ICES data? And you touched on this a little bit. So what is what is that about? How is how are we combining this data where I go in and answer a questionnaire with ICES data, which is very much what we call administrative? So it’s it’s our visits to the ED, it’s our visits to our family doctor, and much much more.
Vicki Kirsh
Right. So you can think of it really as as two halves of a person’s health story. We hold the half that only the person can tell you. So whether they smoke, what they eat, how much they earn, how they rate their own health, their family history of disease, for example, and ICES holds the half that only the health system can tell you, or you know, or that the health system can tell you best. So every hospital visit, every physician claim, every prescription for those over 65, every cancer diagnosis, and together there are particularly powerful sources of information.
Mechanically, you know, how does this partnership work? Well, we do both deterministic and probabilistic linkage between the data sources. So most participants gave us their health number, and that that’s really helpful because that match is then exact. We have one person, one number, and then the match is done. For people who didn’t, we can think probabilistically instead. So we match on a combination of things like name, date of birth, and postal code, and you know we can calculate the probability that two records belong to the same person.
Misty Pratt
Oh, that’s fascinating. So you’re you’re guessing for some of those. Well, I mean, it must be a very accurate guess.
Vicki Kirsh
Yeah, it is. It is. We’re pretty- we don’t use it unless we’re sure that we have the linkage right.
Misty Pratt
Of course.
Vicki Kirsh
And then what we’re linking to at ICES includes hospitalizations, emergency visits, physician billing, lab data, the cancer registry, and also the chronic disease cohorts that ICES has built and validated. And I think you know that last one is really worth explaining because a chronic disease cohort isn’t something that exists, in the data someone actually had to build it. And so every time someone receives an insured service, a physician visit, or a surgical procedure, for example, a record gets generated against their health card number.
So ICES takes those administrative data. You know the inpatient hospitalizations, the surgery records, the lab data, and they apply these what’s what are called case finding algorithms to them. The algorithm might say something like this many physician claims with this diagnostic code within this window, you know, or one hospitalization with this code that counts as a case of disease X, and then each algorithm gets validated against a reference standard, so a primary care chart, for instance. So you know, you know how good it is before it’s ever ever being used for research.
Sandford Borins
Doesn’t this happen just only because we have a single payer healthcare system? In other words, if there weren’t a single payer system, how would data like this be collected on a society-wide basis?
Vicki Kirsh
Right, 100% None of this would be possible without universal health care, and because the fact that everyone Ontario presents the same card for every insured visit, so you know the single identifier follows a person throughout the entire system. It’s not a research database that anyone set out to build. It’s sort of, as you allude to, a byproduct of how we pay for healthcare, but it means that we’ve got near complete population wide records that researchers in countries with fragmented private health insurance just can’t assemble. It’s really amazing.
Misty Pratt
Absolutely, I learn so much about my own organization every episode, and just the whole fact of the case finding, like of being able to then look at the data and just kind of say based on these few things, these visits we’re seeing, this could be a chronic disease of some kind, right?
Vicki Kirsh
Exactly. And ICES has, I mean, I think there are maybe 15 or more of these chronic disease databases, like diabetes, asthma, kidney disease, heart failure, and all of those are then used for research purposes.
Misty Pratt
So, what kinds of research questions can be explored using the OHS data?
Vicki Kirsh
The range is really as wide as what researchers bring to it. I mean, I can give you a few a few real examples that integrated both our questionnaire data and the ICES data. The first centers around heart disease risk. So, there was a team that used linked OHS and ICES data to show that the Framingham risk score was overestimating risk in Ontario, and then they recalibrated. So this is sort of information that’s used to determine whether somebody’s put on a statin, for example, and that’s the kind of finding that really changes what happens, you know, in an exam room between a physician and and a patient.
Misty Pratt
It was being overestimated, meaning more people were thought to have need a statin for…
Vicki Kirsh
Exactly, exactly.
Misty Pratt
Okay, okay.
Vicki Kirsh
And then that group went on to find some really striking sex differences so they found that as cardiovascular health declined women’s risk actually climbed more steeply than men’s and they found that self-rated health was actually a really strong predictor among women so that’s just one and then you know there there are many others.
Cancer etiology. I’ll just touch on that. One is another important area that we can explore. And this work is really closest to me as a cancer epidemiologist, and it’s a big part of what the cohort was built for. There’s a group using our work history data with job exposure matrices, for example, to look at occupational physical activity in lung cancer, and at endocrine disruptors in colorectal cancer, another group is developing polygenic risk scores for breast cancer. We’re looking for risk factors for early onset colorectal cancer, which has been increasing really dramatically over time. The common thread, though, in all of these different studies is that you need exposure measured before diagnosis, which is exactly what a cohort like ours provides.
There’s air pollution. One study found that even at Ontario’s relatively low pollution levels, long-term exposure was linked to higher cardiovascular and respiratory mortality. We can also study mental health outcomes. There was an international consortium that asked whether depression and anxiety raise cancer risk, and they found you know it was reassuring. They found that that no, it doesn’t, and that was reassuring for a lot of people. There are researchers looking at health system inequities, for example. The list goes on.
Misty Pratt
The list goes on. Yeah, it sounds like a lot. Yeah,
Sandford Borins
It’s probably important that participants not know too much, in other words, not be told sort of what the hypothesis is that that’s being tested, right? Because the researcher doesn’t want the sort of knowledge or the attitudes of the participants to influence what they say in the questionnaires. So we should just have a general idea of what we’re doing, fill up the questionnaires as honestly as we can, and then leave it to the researcher to make sense of the data that we provide.
Misty Pratt
It’s pretty common, I think, for like surveys and and data like that that participants do tend to underestimate or overestimate depending on how they would like people to think they’re living their lives. Like I might say, well I have four alcoholic drinks a week when in fact maybe I have 10, right.
Sandford Borins
So researchers, you know, sort of want to keep the participant somewhat out of the loop so that their you know their their answers won’t be biased. I’ll just mention the study in which I was asked to be a patient partner, and it’s a study using the administrative data. In the last year, I had heart valve repair surgery, and the surgeon who operated on me is also a faculty member at U of T, and he’s applying for a grant for a research project that compares heart valve replacement with heart valve repair in terms of which patients have which of the two, and whether the right decisions are are being made in in terms of the choice between sort of the two methodologies, and when I looked over the proposal, I was really quite impressed to see that he could say we have data on 15,000 mitral valve operations in Ontario that were conducted over the last 15 years, 10,000 which were repairs, 5000 replacements, and so on. That’s a database, and said, “Oh, this is this is really interesting, and I’m certainly happy to be part of this study.” It sort of puts my own experience, which turned out to be a good experience. Things went well. It puts my experience in a wider context.
Misty Pratt
How has it changed your view on health overall, your own and the rest of Ontarians?
Sandford Borins
I think that completing these surveys has emphasized for me the importance of environmental factors and also lifestyle.
Misty Pratt
Yeah, absolutely. I think environmental factors are something we don’t tend to think of when we think of cancer, we think of diabetes, especially right. Whether our neighborhood maybe affects whether we’re developing diabetes.
Sandford Borins
One of the lifestyle factors is is alcohol, right? And I thought that one drink a day was okay, and I had a conversation with my cardiologist, and he had a different view, and I was convinced.
Misty Pratt
So, what are the unique features of such a long-term, large-scale study like the OHS?
Vicki Kirsh
I mean, there are a couple that stand out. The first is is time, and that’s why a study like ours is considered to be the gold standard among epidemiologic study designs. You know, in epidemiology, we talk about establishing temporality, and that just means knowing for certain that the exposure came before the disease. That might sound obvious, but it’s actually the hardest thing to get right. So, as Sanford alluded to before, that you know, if I ask someone who already has cancer about their diet, well, their memory of it is colored by by having cancer, and their diet may have changed because of early symptoms that they didn’t even recognize at the time.
So I would have no way of telling whether what I’m seeing caused the disease or was caused by it, and that’s the fundamental weakness of asking people, you know, to look backwards. But a prospective cohort like the OHS solves this really by design. We asked everybody the questions first while they were well, and then we wait. Nobody knew who was going to get sick, and the same goes with the blood. You know, you can’t go back and determine in someone with cancer what their blood looked like in 2012. We have that blood though. It’s in a freezer. It was collected years before anyone got sick, and then finding early markers of disease is what it can be used for, and it’s one of the goals, in fact, of our study.
Misty Pratt
What do you do with those bio samples with the blood that you collected?
Vicki Kirsh
Right, so there were about 40,000 participants who gave us a blood sample, and you know I don’t think people appreciate sort of how unusual that is, and what a what a really precious resource that is. So the blood at the outset gets separated into serum plasma, red blood cells, buffy coat, and then it’s been stored in liquid nitrogen. We keep it at these two mirrored bio repositories.
We also have dried blood spot cards that were collected during COVID, and then over time, some of those participants have been diagnosed with cancer, diabetes, heart disease, which means that we can go back to those samples that were taken from a healthy person years before diagnosis, and then ask, you know, was there something already there? And that’s what a biomarker for early detection is. It’s something measurable in the blood that shows up way before symptoms do. You can’t find one by studying people who are already sick because then you have no way of telling you know whether what you’re seeing caused the disease or was caused by it. You need that sample collected first from somebody who looked perfectly well at the time, and that sequence is the thing that we have that almost nobody else does. And really, those samples they keep gaining value.
So a sample frozen in 2013 can now be tested with technology that was invented, you know, now or in the future in 2030. We’ve also genotyped about 20,000 participants, and we’re hoping to do whole genome sequencing on all of them. And none of that was remotely affordable when the study first started. And genomics is is just one layer, so those samples can be run now for proteomics and metabolomics. So you you can look at what proteins and metabolites were circulating in someone years before they were diagnosed, and then there’s also cell-free DNA. So that’s the basis of a lot of the early cancer detection work that’s happening right now. So it’s really a dynamic picture that can be generated for research.
Misty Pratt
Yeah, that’s fascinating. I can imagine in the future there’ll be even more that you hadn’t expected that you can do with those those samples.
Sandford Borins
I didn’t give a blood sample at first, I don’t think, but I know that I gave blood samples for the COVID studies. But the other thing I sort of think about is that what the study needs is a level of trust that someone sort of sends me a kit, and I have to trust that you know I’m going to cause myself a little bit of pain, but I know that I’m doing this. I’m not doing it face to face. I’m doing it at home, but I have to trust that the sort of the pain that I’m undergoing is contributing to something socially valuable,
Misty Pratt
And trust that it’s going to be kept safe.
Sandford Borins
Yeah, the pandemic was a very unusual time, and people were being called on to do things that were inconvenient and sometimes painful in the name of the greater good of society. And I guess I saw this type of participation as one way of sort of fulfilling my duty as a citizen.
Vicki Kirsh
I hope that we are building that trust with participants, and I think some of that comes as we generate new findings and we share that with researchers. We engage with them through newsletters, you know, we host webinars where participants can log in, town halls. So I think all of that helps to sort of build a relationship.
Misty Pratt
Yeah, it really is a relationship because you are, it’s a- it’s almost like a community that you’ve built that don’t know each other and that are fully anonymous within the the database, and yet, you know, you have to kind of maintain some contact with them over many years. So,
Vicki Kirsh
What we give up though in is by using a study design like this is time and and money, I’d say. Waiting over a decade is slow. It’s expensive. What we get in return is being able to say something about cause rather than just about association. And there’s really there’s no shortcut to that. And then the second main thing I say is scale. So Ontario is enormously diverse. Questions people actually care about tend to be specific. So, does this drug, you know, work in the same way in this group as in that one? Why does the incidence of disease vary by subgroups, by ethnicity, by location? You’re not answering those kinds of questions in a small study, but with 225,000 people and more than a decade of linked records, you can start to answer questions like that.
Sandford Borins
I guess the way I think of it is that you know you have a huge database, and then you have cells that focus on particular diseases. And if the database is big enough, all the cells will also be big enough. And you want big cells so that you can analyze out of the 250,000 people, you get, you know, 5000 with heart disease, and you can look at the ones with heart disease. You know, even though it’s a small subset of a very big data set, it’s a large enough subset to do the statistical analysis that you need to do.
Misty Pratt
Yeah, you don’t want a small cell of like five people because that’s not going to really tell you much, right?
Vicki Kirsh
Exactly. I mean, now I mean we we have about an average of 13 years of follow up, and that time, you know, we’ve accrued 1000s of cases of certain diseases, 2500 breast cancers, you know, 7000 strokes, 9000 cases of type two diabetes, and it’s it’s those kinds of numbers that are really powering the analyzes.
Misty Pratt
And ultimately, you know, giving us answers for the future, hopefully, right?
Vicki Kirsh
Exactly. Yes,
Misty Pratt
That this could potentially feed into our knowledge on types of interventions that we could maybe make down the road to try to change those outcomes.
Vicki Kirsh
Yes, absolutely.
Misty Pratt
Sanford, what do you hope the OHS study will accomplish with gathering all this data across our lifetimes?
Sandford Borins
I hope it will achieve its objectives in terms of us understanding more about the sort of environmental and behavioral causes of diseases. I mean, I think I think that’s very important, and you know that information can be made available to people, and hopefully will influence their behavior. There are lots of other things, but these two leap out to me as being really important.
Misty Pratt
Yeah, that people feel maybe a little bit more of a sense of control, perhaps.
Sandford Borins
Yeah.
Vicki Kirsh
Absolutely. Like I think a lot of what what will eventually answer is still ahead of us. You know, using samples and data from these people who signed up in 2012 and then went back to their lives, and it’s it’s a pretty remarkable thing for 225,000 people to have done for each other and for research, and all of this only works because of because of their involvement.
Misty Pratt
Yeah, it truly is remarkable. Thank you both so much for being here today. Thanks for joining me for this episode of In Our VoICES. Check out the show notes for links to research and any other information that we’ve referenced in this episode. A reminder that the opinions expressed in this podcast are not necessarily those of ICES. Please be sure to follow and rate us on your favorite podcast app. If you have feedback or questions about anything you’ve heard on In Our VoICES, please email us at [email protected], and we will get back to you. All of us at ICES wish you strong data and good health.
