This course presents critical concepts and practical methods to support planning, collection, storage, and dissemination of data in clinical research. Understanding and implementing solid data management principles is critical for any scientific domain. Regardless of your current (or anticipated) role in the research enterprise, a strong working knowledge and skill set in data management principles and practice will increase your productivity and improve your science. Our goal is to use these modules to help you learn and practice this skill set. This course assumes very little current knowledge of technology other than how to operate a web browser. We will focus on practical lessons, short quizzes, and hands-on exercises as we explore together best practices for data management.
Research Data Planning 1
Loading...
From the course by Vanderbilt University
Data Management for Clinical Research
357 оценки
From the lesson
Research Data Collection Strategy
This introductory module reviews the course structure and basic concepts in clinical research. We also discuss best practices for designing your clinical research data collection.
Meet the Instructors
Stephany Duda, PhDAssociate Professor
Department of Biomedical Informatics
Paul Harris, PhDProfessor
Biomedical Informatics and Biomedical Engineering
So in this section we'll be covering research data planning.
This is a very important topic for any
clinical and translational research study that you're considering.
Specifically we'll cover basic concepts behind
the planning and collection of research data.
We'll talk about the importance of thinking things through in
a real world practice before starting the collection of data.
And finally we'll talk about the importance of
good recordkeeping and documentation of records for ongoing and shared research.
The way we'll structure this section is we'll
talk through another of important concepts and best
practices that we've picked up over the years
supporting and, and investing in many, many research teams.
These are very common things that, that we see and
by, by conveying and teaching them in our own local environment,
we find that we increase the capacity as
well as the value of the research done here.
So we'll run through a number of examples as we go.
First example that I always give is the importance of, of good
data planning up front taking a lot of time up front to
make sure that you are collecting, that you are planning appropriately, so,
so that at the end of the study, you get good results out.
we have a phrase," Garbage In Equals Garbage Out."
And I've seen this play out many times
in research teams that are eager to get started.
They want to, they want to move ahead because they've got a
great idea, they want to move ahead with the study very quickly.
And in many cases, they'll forget important
things because they didn't, plan up front, that
they didn't take the time required up
front to make sure they weren't missing anything.
A lot of times this really doesn't
become evident until late in the study when you're analyzing the data.
And so, so I always stress, even when you think you've got it right, run it through
another iteration and make sure that everybody on
the team thinks it right before you move forward.
This really becomes an ethics issue. No matter what type of study you're doing,
you're always spending resources, you're always putting individuals
at some inconvenience, or even risk, by
doing a Clinical and translational type studies.
And so if you don't get the the details right from the front
end and you compromise the ability to leverage the data that you've collected.
And you're not able to publish, and, or make definitive statements on that data at
the end of the trial or study, as you would if you had done it right.
Then you've really done the people that working
with, the individuals that are contributing to your
study as volunteers, you've done them a great
disservice, because you haven't gotten the scientific benefit.
That goes along with the risk or
the inconvenience that they put into the study.
So I'd stress it's, it's an ethics thing as
well as a data and a study reliability thing.
So what are the things that, that I see a lot
here is we'll, we'll have a, someone who has a great idea.
We will meet with that individual and, and sort of start fleshing out that idea.
And a lot of times they, they haven't really though things through.
What they really had an idea on is hey, let's just start collecting a lot of data.
And this maybe, might come down from a department chair or maybe You know,
well meaning individual that, that's really
kind of getting their getting their interest peaked
around data and being able to sort of collect and manage large data sets.
And sort of the promise that if we do this then great things are going to happen.
A lot of times if you don't put the time up front really sort of
dial things down and think very hard about
a primary hypoethesis for what you are doing.
May be a secondary outcome measurements.
But instead he's just trying to do everything at ones, bad things happen.
You know good way that this happen is practice I think i'll meet with resource
teams from time to time and rather than
individual research team I'm meeting with the department.
And, and, and the department chair will say, hey, I've got this great idea.
And what we want to do is we want to create this great big registry.
And, and I'll meet with the group and say,
well, you know, what, what are, what are you going to do with that registry?
And, and sort of, the, the, the answer to that is very short.
And it's just, well you know, if we had a, if
we had a registry and we were able to collect data.
From all of our individuals that are seeing patients and the patients might
be entering data themselves through surveys,
you know, we can just do everything.
And I'll usually say, well, you know, that's, that's great,
but, but, you know, who is going to be entering
the data?
Well, everyone is going to be entering the data.
And, and we'll get it from all of the, all of the systems around campus as well.
And, and then I'll ask well what are you going to do with that data?
What kind of reporting are you going to do with that data
and, and you know the answer will be you know, everything.
You know, we, if we had this, we could do recruitment, we
could do quality improvement studies all sorts of things are possible here.
And usually what I try to do next is say well you know let's,
let's think about that because one thing that we've
learned about time is that by caring about everything.
You really care about nothing, when it, when it comes to data.
Because there's no way you're going to be able to invest all of
the resources it's going to take to have really solid good quality data.
unless you really, really think it through.
Unless you quantify, sort, sort of the benefits
and the return on investment on these things.
So, so that you're really getting good quality,
oversight of the data.
So rather than thinking in, in terms of everything,
why don't we think about it in terms of you
know what would be thing that we could do
in two months, that would really change the world here.
And if we could come up with something small, and sort of
build up and around that, then we've got something to build from.
Or, in the case where we're doing a clinical study or trial, I would
typically say" Hey, this is, this is
bad policy, we really don't want to go fishing''
when were doing a clinical study or trial.
We really need to sort of reign things in
and you really need to think about that primary hypothesis.
You need to be able to think about what figure one
in the study is going to look like when you publish it.
Or table one and you need to think, think very
carefully about collecting data around one or two things that
you're going to be testing definitively rather than coming at
it with sort of a full-fledged, full, just collect everything and
decide later.
So, so it sounds simple, but in practice, it's
really hard to, to practice self-control and go through
the process, making sure that you're collecting the primary
data that you need for your clinical study or trial.
So sort of going the other way, I've seen a lot
of studies that, that have come to me that, that basically say,
you know this is a really simple study. All we need to do is collect a few things.
Then we run the analysis.
And then we publish in the New England Journal of Medicine, etcetera.
I remember specifically a study team coming to me a
number of years ago, and, and they used that phrase.
You know Paul, this is a real simple study.
We're going to have people come in for two visits.
And we're really just going to
do two things.
And we're going to do blood draw, and we're do from that a metabolic panel.
And we're going to shoot an MRI image. And, from that.
Well we're going to, we're going to do the study over, over two visits maybe it was
a randomization of, you know, you know a
little bit of randomization with the patients etcetera.
But, but from the, the standpoint of the data it was very, very simple.
Two visits, we're going to measuring a couple of things and then we're done.
So,
I started asking questions at that point.
I said, well, you know, I understand a little bit about imaging.
So, I also understand enough about it that I know that you
can't just sort of feed an MRI image into a statistics package.
So, maybe let's think about it, a little bit, what you're planning on doing there.
Is it the tumor size, is it a diameter, is it a volume type measurement?
Think about those things that are going to be coming out of that high-density
image data that you're going to be using for your quantitative analysis.
And from that, you know, maybe we came up with five or six
measurements that were going to be important at each of those imaging events.
And then we started talking about metabolic panel.
You know, that's not just one thing.
That's really a number of things, you know,
that might be the cholesterol, the glucose level.
And so we started thinking about that and
particularly putting things in, in terms of units.
It really helped to study the study team start thinking very discretely.
Oh yeah, yeah, I guess we do need a field for
glucose, I guess we do need one for, fo-, for, blood pressure.
Oh, and, and while we're collecting blood pressure, maybe we
could split that up into Systolic and Diastolic blood pressure.
So, so I'd say by the time that, study team left my office that day.
We probably had them up
to about 60 measurements that they were going to be collecting.
So the things that I just mentioned, as well
as things like, you know, do you really need
to know, you know, some sort of identifier for
the patient, or do you need to know a name?
An address, phone number, maybe gender or ethnicity.
Those things that might, might really play a
factor in helping you analyze this data later on.
So they're left with about 60 measurements
that they were going to be collecting rather
than two. But also left them with this exercise.
But I said, you know, go home and think about your primary, secondary hypothesis.
[INAUDIBLE].
Let's look at those outcome measurements.
Think about how you're going to organize that, and then come back to me, and come
back with an even richer set of
information and data that you're going to be collecting.
And I stressed, as I always do, the thinking
about this in terms of what exactly would be
stored in a you know in, in an Excel type field or a, a data table type field.
And what units would belong
to that particular measurement and if you, if, if
they do that, what I've found is you know,
really, it really is hard to start thinking about
MRI images in one of those Excel spreadsheet type cells.
And so I say well go, go do this exercise and really sort of
think about these things in a discrete way and then get back to me.
By the time we launched that study I think we had
about 400 variables that we were collecting on, on every patient.
And so, I started with two and
ended with 400 and, and I would say that, you know,
that, that's pretty common when you start thinking things are, are simple.
Lot of times they become more complex later on.
And that's good because we really want to make sure that
we are collecting the right things that we're going to analyze later.
So, another, another thing I always try to stress in, in,
when individual are looking at putting a data collection strategy together.
[COUGH]
is in addition to thinking about each measurement, think about the
right type of measurement for, for, for each of those concepts.
And so and you know, in basic statistics
classes, individuals learn about the, the differences between
nominal, ordinal, and continuous variables nominal being things
like sex and race where there's no ordinal effect.
Ordinal maybe where there's some clustering or cata-,
cata-, categorization, of data that has an ordering effect.
That, but it's not, ta-, tiny and discreet like a continuous measurement may be like
a blood pressure which, which would have, you
know, one increment for every millimetre of mercury.
So, so I always stress, that you know
again while we're thinking about, the collection of
this let's think about the analysis phase.
And let's make sure that we're choosing the right type of variable
for each of the, each of the different entities that we're collecting.
as well, I always stress that unless there's some really good
reason on the front end don't ever collapse data before you need to.
So continuous variables.
You know, if you got a temperature measurement
you can always calculate later whether they had a fever by some rule but
you can't ever go the other way just by knowing yes, no and fever.
We can't go backwards and say let's calculate the continuous variable.
So, you know, don't, don't collapse variables
before they before they need to be collapsed.
The other thing that comes out a lot in the discussion around
measurement types is what do I do with that high density data?
Things like
an ECG waveform, or an MRI or a CT or an X-Ray scan.
And again, we talked about that in the last last slide.
really I, I always stress that, you know, keep the high density data intact.
You know, you don't want to throw away the MRI image.
You don't want to throw away the ECG waveform.
But typically when you get ready to analyze those data.
let's, let's choose the ECG as, as an example in this, in this case.
You know typically you, your not going to be analysing, you
know, the whole ECG and comparing that against, everyone else's ECG.
What your probably going to be doing is looking for the average heart rate.
Or maybe the, the longest QT interval during
this segment of, of an exercise profile, etcetera.
So, so I always recommend that we keep. The high density data, keep it on file.
But, but in terms of the, the structured data that we're
going to be collecting for the study and
analysing later, think about the post process values.
That, that long Q T interval.
The, the tumor diameter as measured on that MRI.
Ознакомьтесь с нашим каталогом
Присоединяйтесь бесплатно и получайте персонализированные рекомендации, обновления и предложения.
Coursera делает лучшее в мире образование доступным каждому, предлагая онлайн-курсы от ведущих университетов и организаций.
© Coursera Inc., 2018 Все права защищены.
