Using DMW To Optimize The Management Of Clinical
Data From InForm and Other Sources Hello everyone. Thank you for your patience
and welcome to the webinar titled” Using Oracle’s Data Management Workbench To Optimize The
Management Of Clinical Data From InForm and Other Sources” presented by BioPharm’s Mike
Grossman, vice-president of clinical data warehousing and analytics. I’m Eugene Sefanov, the marketing manager
at BioPharm and I’ll be going over some housekeeping items before I turn it over to Mike. During the presentation, all participants
will be in listen-only mode. However, you may submit questions to the speaker at any
time today by typing them in the chat feature which is located on the left side of your
screen. Please make sure you state your questions clearly. Keep in mind, other webinar participants
will not hear questions or comments. Then the list of questions to the speaker will
be addressed as time allows. If you still have unanswered questions after the webinar
or would like to request information from BioPharm, feel free to visit the company’s
website for contact information. As a reminder, today’s webinar is being recorded
and will be posted on BioPharm’s website within 24 hours. WE will also be emailing you links
to the recording as well as a PDF version of the presentation. This concludes our housekeeping
items. I would now like to turn the call over to Mike Grossman. Thank you very much Eugene. And thank you
everybody for joining the call. I really appreciate it. Today, we’re going to go through just
a few slides and I’ll do some introductions, but the vast majority of the webinar will
be a live demonstration. So, it should be interesting. First of all, my name is Mike Grossman. I
run the data warehouse and analytics group at BioPharm. I’ve been at this for about 4
years. Prior to that, I spent 10 years at Oracle and I’ve spent quite a long time in
the industry, both in data management and in the IT side as well as doing warehousing. So, let’s just briefly review the agenda.
First thing I’m going to talk about a little bit is just a very short discussion about
why DMW. Why does it exist and then I’m going to go into how do you set up to the data management
workbench? And then how do you conduct trials with the data management workbench. So, basically
a set up phase and then the actual conduct phase. So, a little bit about why we want to use
a tool like the data management workbench. Really, it comes down to ease of data, such
as we collected. Inform collects much of the clinical that we need, but it doesn’t collect
all the clinical data that we need for our clinical trial. And, we’d really like to be
able to clean the data and review the data ongoing, not just the EDC data, but the (?) data. So, various companies are dealing with this
in different ways. Some people are data such as central lab data and they’re importing
it into the EDC system. Others are pulling the data out of EDC system and doing manual
compares. And, really, a data management workbench is really trying to address this whole data
review, data cleaning and data reconciliation process to encompass not just the EDC data,
but all the data involved with a clinical trial. In addition, it’s also trying to handle the
ability to transform the data to a common standard, so that it can be pooled with other
data. WE can do a cross study analysis, an ad hoc analysis. My, why does this exist? Well, we’re really
just trying to increase productivity. It’s just too labor intensive to do study specific
approaches here and to handle multiple systems and manual processes. We’d like to have an
end and traceable process for all the clinical data from the quick capture independent of
the source of data we capture right through to reporting and analysis. Architecturally, the data management workbench
is built on top of the life sciences hub. The data put in there, the life sciences data
hub is really designed to get a holistic view of all your clinical and metrics data in a
single, scalable environment that’s regulatory compliant. So, the data management workbench
is essentially a business specific layer that sits on top of the life sciences data hub,
that supports this data capture, data standardization, and data validation step. But, because it’s
such a, it’s built on top of an enterprise architecture and application, a designed phased
architecture, all the downstream motions, such as data pooling, analysis, data review,
warehousing, metrics warehouses, all of that data that’s coming in through data management
workbench is part of an organized architecture, so you really don’t need a different solution
for that architecture. Okay, so what, what’s the philosophy behind
the data management workbench? Well, it’s looking at a process where you load data from
your EDC System and this case, this release it’s really focused on Oracle’s InForm product.
So, you load data from InForm and metadata from InForm, and then you load other data.
Could be central lab data, in these cases, examples I’m going to be walking through,
its central lab data. But it could be deroginomics data; it could be, um, you know, adjudication
data. It could be any other kind of data. But you don’t necessarily want to enter into
your EDC system because the sites themselves are not directly involved with that data capture. So, that data is, comes in and you have an
opportunity to look at the raw data model or you can have other data models. So you
can transform that data to review data model, data cleaning model, you could even have an
analysis model that, that is multi-stage processing. So, all of this is available conceptually
inside life sciences data hub. But, DMW gives an alternate interface that’s a little bit
more user friendly, helps you write the transformations as you will soon see. Okay, so, one of the basic concepts of this
data management workbench is there is something called a study. And this study really correlates
to an InForm study. So, what we really want to do is: A study consists of one or more
clinical data models. It concludes mappings and transformation programs to move from one
data model to the next data model. It includes edit check programs that validate
the data. Now you will edit check programs already in InForm, this is really to do cross
edit checks between the central lab data, the loaded data, and the InForm data. And
then there are also custom listings as well as default listings that actually help you
review the data and track the review of the data. Just a, hold on, one more. Okay, so I’m going
to stop the slides right now and then I’m going to share my desktop, so hopefully this
works very well. And we will take a look at the applications themselves. Okay, so hopefully you can see my desktop.
I basically have a browser. Currently looking at the InForm 6-0 application. And we have
an example study called Cardio and I’m just going to do a quick review of this study. So, in this study, we’ve got lots of sites.
Looking at site number 14 here. Got patients within the sites. If I click on a patient,
you could see the normal InForm matrix. If I go to the InForm view, we can look at things
like the ECG strip. And then, this is just a normal interface that you would expect from
InForm 6-0. So, there’s nothing special here. I just wanted to kind of set the stage for
study that we’re going to be working with in this example. So, you can see we’ve got
lots of visits, lots of forms. You can do data entry, normal study. The second application I want to look at,
in the main application, we’re going to be looking at is the data management workbench.
And this is the interface for the data management workbench and I’m going to go through this
in some detail. And then, the third supplication that we’re
going to be looking at is the life sciences data hub. So, we’ll get an opportunity to
see how this, the data management workbench and the life sciences data hub interact with
each other. So, that’s really, what we’re going to be taking a look at this morning
and afternoon. So InForm. InForm feeds data into the DMW, which really is an alternate
fund for the life sciences data hub. So, first we’ll start with a little bit with
studio setup. So, when you come into the data management workbench, you come into your home
tab. And in your home tab, you basically have a list of all the studies that you’ve set
up. Now, this is a development environment, so there’s not a great, a great amount of
information here. And then, you pick the study you’re going to work on. Now, in the home
tab, when you select a study, you actually can see what’s the run-time, so loading of
data, transforming data and validation checks. We’re going to cover that in the second part
of the presentation, after we discuss how to set up a study. So, let’s go into setting up a study. So,
if, the first thing I want to show you is, when I’m working with a study, what the study
properties are. So, I’m on the home tab again, and if I click on study properties or I create
a new study, it has a name; it has an indicator whether the study is a template study. So,
you can create example, studies as templates so that when you create new studies, you can
create them from a template rather than just creating them from the, uh, scratch. Whether
it’s active or not, list of therapeutic areas. This is a list that’s driven, customer driven
and you can set that up as well as part of the administration. And then, study size.
The study size is really there for performance reasons and I’m not going to go into the technical
reason why, of how you would set up partitioning. But, that’s what the study size is for. In addition, as you’re building out this study,
it can go through life cycles. So, development, quality control and production. We’re really
going to focus on the development mode for now. So, you create the study and then after
setting up some administrative steps, you go into study configuration. And in study
configuration, you have your clinical data models, your data transformations and your
validation checks. So, in this particular study, we have three
data models. We have InForm and this is the data model that’s been pulled across of InForm.
We have a central lab data model. In this case, this is just some central lab files
that’s being loaded to data set and then we have a downstream data model that really combines
the lab data and it combines the InForm data together. Now, I’m going to, when I go into some of
the transformations, I’m going to show you that typically you would have a much larger
downstream model for a particular study. This is just for this specific use case example. So, let’s take a look at the InForm data model,
The InForm data mode is directly extracted from the InForm study itself. So, if you see
the tab here that talks about InForm configuration, you actually can go to this remote location
which is defined under administration. The focus of this talk will not be an administration.
Then there is some encounter information, life cycles, the InForm study, where web services
are, so this is essentially MBURL. So, this just essentially where you say hey, what’s
the study that I’m going to grab? Right now I have it suspended. If I had it
suspended, I can then press load metadata and then it goes into the InForm study, pulls
out all the extracts to table structures, RD tables and it creates them in here. So,
there’s no need to manually synchronize the Inform study structure with the DMW structure.
So, if I click these, so here’s con-meds (?), it’s here. If I were to look at the ECG area here
that actually would correlate with ECG area that you would see here in InForm. So, there’s,
it automatically synchronizes between the InForm metadata, between the DMW metadata. So, once you have the actual model in place,
you can load the data just by pressing this button. The tool manually loads the data or
once you move the study into production, you can set up a frequency. And the frequency
can be minutes, every 30 minutes. The way that Oracle has designed the frequency of
loading the data into DMW is an incremental load. And it’s incremental really down to
the patient level and easy to (?). So that you’re only pulling small amounts of data
rather than pulling the data into your DMW environment once a week or once every two
weeks and doing the reviews then, you literally can do this every few minutes and pull it
across so it’s like a trickle feed. And, I think that’s a very important differentiator
to how most people are doing their work today. That’s very interesting. Okay, so we have our InForm data model. We
have the central lab data model and the central lab data model is a variation of C disc in
this particular case. Now, because this is a file base approach, we have this concept
of file watchers. So, depending on the validation, you know we have datasets, production data,
test data, there are launchers in set up and directories sitting on a server where you
can just drop files and it’s expecting a fast file. So, if it’s that dataset, it’s just
dropped into this directory. It’s using the underlying LSH technology to pull that data
in. So, part of setting up the metadata is where are these files going to be. And it
can handle fast data sets, it can handle fast transport files, it can handle fast C-port
files, then it can also handle text files that include CSZ, the limited files, et cetera.
It really uses the underlying specifications. And, as you can see right now, there’ve been
some, it runs a bunch of checks. So, we’re going to go into some of the tasks when we
do the run time. So, once you’ve defined the models that you’re interested in, you could
always keep adding more models. You install the model. In the case of the manual, or like
central lab models, for InForm when you bring the metadata across it takes the model automatically. And then we have a combined model. So, it’s
very important to understand that you have these multi-step models and the models interact
with each other. And we’ll talk in a minute about how the models interact with each other. So, I’m going to switch to another study to
talk a little bit about how models interact with each other. So, just go over here, pick
a different study. It just happens to have STTM as one of its models, so if we go to
study configuration, we can see not only does it have the InForm structures and the central
lab structure, but we’ve got a STTM model. So, this STTM model typically would come from
the library. The library basically is a library of models. So you have two ways of standardizing
in DMW. You can either have models have libraries and you can have studies that have templates.
So, the library models are really metadata structure for tables or data models and then
templates that include validation checks. Would include transformations between models. So, if we actually look through this, it’s
kind of your standard STTM, if you look demography dataset, which we’re going to take a look
at. Here, it’s the standard STTM model. So, what we actually want to do here, is how do
we get the state out of InForm there into this model, into the STTM model. So, that’s
the first example a transformation. So, this accordion tab here, transformation, and we
have two target models: 1 is the cross analysis and 1 is STTM. So, this STTM model is under
version control of the mapping. So, there’s mapping from source to target so all of this
sources are InForm and then here’s our target table. So, we have all our target tables here
for STTM. And to look at one of the target tables, such as DMW, how do I get the data
from InForm to the STTM. So, the first thing you do is you pick what
your sources are and then when you have your sources, and in this case, it’s coming from
DM, and death. So, the Death record and the DM record on the InForm side are going to
be used to produce DM on the target. And, the type of transformation we want is join.
So, we’re going to join the death record to the DM record to produce an output. But, you
can see from this drop down, you can union things together. Direct just basically means
it’s a straight 1:1 type mapping. There’s a pivot and there’s an unpivot mapping, so
this really allows you to do things like go from a finding or a structure, if you’ve collected
vital signs in horizontal structure, and then you want to make it normalized, you can use
the pivot-unpivot. And then there’s a custom mapping. So, we’re not going to go through
all of these in this particular session. We’re just going to look at the join, it’s the simplest
one of the site’s direct mapping. So, once you’ve picked yourself, you’ve mapped,
okay, I want this and you press map and then it shows up down below, and then you find
another one and you press map and it would show down below. You can then begin to build
out your join clause. So, you can basically go okay, for table 1, I want to join it to
table 2 what we’re basically saying is we want all the demography records and we want
to join to death records, but we don’t, if, if there’s no death record, we still want
to keep the demography record. And that is really what outer join does, these outer join
flags. And then you have to put the criteria, in this case, it’s by subject ID. Now, if
you had a custom transformation, for mapping, it would just, it would show here. But we’re
not going to review that right now. So, now at a table level, you’ve said, okay
I’ve got, picked my sources and targets, now I want to go and add a column level and I
actually want to build out the map. So, if I select the column mapping icon, I then brought
to this screen. Now, the simplest type of maps are just 1:1 maps. Now, just, so for
example, the subject ID here is really mapped to the subject ID, let me just scroll down.
UMM , down here, you just click them both and say map. And you can even do a multi-pick
and say map. And, in that case, you actually have to build an expression. So, if I were,
for example, to look at the race, you can see that for the race I’ve already got several
different columns all combined here as sources. But, the output is a single field. So, when
you have multiple inputs that you want to do some sort of manipulation across multiple
inputs with a single output, you use the expression builder. So the expression builder here, I’ll click
on that, will allow you to concatenate or really use any function and we’re going to
look at more advanced function in a minute, to actually bring together data at a column
level on a function base. So, you can do 1:1 mapping, you can do a combined mapping, function
based mapping that basically has single output with multiple inputs with any function in
between them. And then you do a completely custom mapping. Target table by target table.
So, once you have these mappings, you are then able to produce in a very traceable and
a repeatable way and we’ll look at the traceability a little bit later, exactly the source the
target relationship from one model to another. There’s also some efficiency functions, such
as auto-map, install, you can validate them, you can also say, okay, what’s missing, what’s
not missing in case you’ve left out certain fields. So, you have quite a lot of different
approaches on how you can do these mappings. For a 1 hour teleconference, I really can’t
go through all the features of the mapping utilities, but I think that you can see that
it’s fairly flexible. So, that’s how you set up the mappings. So, you can go from one model
to another. So, the next function I’m going to look at
is the validation checks and then I’ll look and see if there are any questions that have
been collected up and takes some time to answer them.
So, there lots of different types of validation checks, but the ones that are most interesting
to me are the ones that you can actually compare the data collected in your EDC system, in
this case InForm, with the data that’s collected outside of InForm. So, I’m going to switch
back to another study, not the old one, InForm Cardio, and take a look at its study configuration.
So, we’re still just doing the setup here. So, validation checks are a new object in
DMW, and do not exist in LSH up until this point. So, this is interesting if you have
been following the history of life sciences data hubs, it’s a new object. So, again, we
have central lab and then we have InForm, those are the raw data and then we have kind
of a combined. Now, the way that this has been combined,
we’re going to take a look at very briefly, actually skipped a step in data transformation.
So, if you will just bear with me, I’m going to step away from data validation and go back
to transformations and look at how these were combined. So, in fact, for the P-Lab table, we’re taking
the central lab data, which is just a dataset that has been loaded and we’re taking the
InForm data and we’re combining them. And how are we combining them? Well, if I look
at the map column that is combined with a join, but it is an outer join. Oops, if I
do that at the table level, here’s adjoin, okay click on this, right. So, if the subject
ID and the visit ID are the same, we’re actually saying ECG record should have triggered according
to the protocol that there was also a lab draw. So, for every ECG record, we want to
make sure that we’ve got a lab table. So, we’re combining the two, and we’re doing an
outer join. Now, if we look at the column level mapping,
it’s actually as you would expect. It’s mostly 1:1 mapping. I just wanted to point out one
particular case. The actual high and low flag is a little bit artificial, are coming from
a function look-up. So, a custom function look-up. So, if we actually look at the expression,
we see that it’s lab utiles get low and this lab utils get high is the other one. So, these
functions, lab utils get low and lab utils get high, those are custom functions that
have been built that are available to DMW. Now, Oracle can’t think of all the functions
you want, as a result, they really haven’t given you a lot of functions you want. But
within the life sciences data hub, now I’m going to switch to the LSH interface. Within
the life sciences data hub, you can actually create any type of custom function you want.
So, there is a domain that gets created within the life sciences data hub, domains are just
like folders for those of you who are not familiar with life sciences data hub. And within there, there’s a DMW utils domain
or folder, and then any shared function that has been moved to production, is available.
So, in this case, the lab utils PL=program, has been moved to production, and it is shared.
So all the functions that are available in this lab utils program. So, if I actually
go to the source code and look at it, all of these functions get low get high, whatever,
are available within DMW. So, you really have the ability to build up a library of mapping
functions that are completely unique to your environment, but still are available to the
general team to actually grab and work with. So, it’s smart to mention that about data
transformation before I moved into validations checks. So, setting up validation checks. So, InForm
has its own validation checks, but when we bring that data together, we’re want to run
some validation checks on the combined data. Some of them are very, very simple. So, these
are just normal checks to say, okay, did someone mark this as abnormal reading but it’s within
the range. So, let’s just take a quick look at that. Very simple check. Okay. So, you basically can say, okay, what
are the sources for your validation check? In this case, there is only one source that
you have, and then, these are your variables. And you can actually pick the columns that
you want. There’s no joins in this case, it’s just from a single table that’s already been
combined. And then you put in the criteria and there’s a language for this criteria that
basically says that the lab and our indicator is abnormal and that the, the result is actually
correct. So, to, the key here is that we’re actually running an edit check on a table
that’s a derived table and it’s not a table that’s a raw data table. So, we’re actually
able to combine checks across sources. We’ll take another, a look at one more of
these. Cancel that. So, if we look for no lab data exists, so let’s look at that one
because we’re going to follow up on that one in a bit. On no lab data exists, we actually
are looking at date of visit and visit information that happens to be from, from InForm, and
we’re looking at the test code. So, if we look at the actual criteria, it basically
says, the test code is none. So, one of the things we setup when we combine the data is
that the test code, if we didn’t have a lab reading from the central lab, but we did have
an ECG reading, the test code was marked just none. Just part of the conditions of adjoin.
So, we can raise the discrepancy back to the lab that basically says, you know, where’s,
where’s this test. And then potentially we can raise a discrepancy back to the site as
well. So, you can basically can an arbitrary set
of complex validations. Those validations can exist on any of the data models. And then,
you can back trace those in the original source. So, you don’t always have to write your edit
checks directly on the source. So, that’s a brief discussion of the setup
of the data models, transformations and validation checks before I go into the tun time and the
listings and the discrepancy management, et cetera. I’m just going to pause for a moment
and check, check to see what questions are out there, if any. So, hopefully this unshares
my desktop. Here we go. Okay looks like I have a number of questions.
If you’ll just give me a moment, I’ll read through them. Okay. Okay, so the first question is: How is DMW
different from other CTMS and CDS such as ClinTrial and Oracle Clinical Okay, that’s
an excellent question. So, in some ways there is overlapping to clinical data management
system. The difference is, underneath the architecture of these clinical data management
functions, is an architecture for warehousing analytics and reporting. So, because the life
sciences data hub is under this architecture, the data coming from InForm is actually coming
into a clinical warehouse that is fully integrated with staffs, has version control, has blinding,
has all these functions that don’t have a pre-defined data model. The data models extracting
and multi-stage. So, if you’re working with a ClinTrial or Oracle Clinical, you basically
have one layer of data model and one layer of extracting. And while you can do the validation
check ClinTrial or Oracle Clinical, what you really can’t do is multi-stage processing,
(?) days reporting, for different consumers have different architectures. So, that’s a
very brief summary. I think it, there’ a much longer discussion. How do DMW metadata changes mid-study both
from EDC and external data? So, the way that it handles it is basically everything is under
version control and as new versions come through in development, then they go through a life
cycle of development QC at production. So, while you have your production study running,
as the changes come through, the metadata changes, et cetera, those get worked through
the development and only as you promote them through to Quality Control and production,
after you’ve had some changes, actually you can always apply them to the production data
and the production flow itself. The next question I’m seeing is: Can this
tool handles 100% STDM mappings for very complicated objects? Yes, it can. And I think the way
that you would handle for very complicated objects does come down to custom mapping.
So, the situation is, with the kind of mapping that I showed you, where you are basically
doing 1:1 mapping or function look-up mapping, you could probably cover about 90-95% depending
on how rich your function library is. But, there’s always special cases. So very similar
to how I showed you the custom functions. You can actually create completely custom
programs and validate them. And then those custom programs can be used so that few percentage
that the graphical mapper or the interactive mapper doesn’t cover. So, yes it can cover
100%, but it’s not magic. At the end, I , we, we, what I find, in my experience, is that
there’s always a few little tweaks, few little special programs that need to be handled in
order to move forward. And then: How do you maintain traceability
with custom mapping? So, we can maintain traceability with function based mapping. Variable level
traceability across a pure custom mapping is not really possible because underneath,
inside that code, you don’t know if aggregations is occurring or something like that, But,
if you don’t use the custom map base, then in fact, you get full traceability and we’re
about to take a look at some of those features. And a bunch more questions are going to pop
up, but since I’m only about half way done with the presentation and I have 20 minutes
left, what I ‘m going to do it go back to the presentation and then I’ll, if, if any
time is left at the end, I will go through and answer the remaining questions. But, thank
you for all the questions. I really appreciate it. All right, so I’ll share my desktop again.
And, now we’re back into the data management workbench. Okay, so I’ve got my study setup,
I’m working through it and I want to start loading data. So, I’ve set up, let’s say,
I manually loaded data through the screen I showed you previously under the data model,
and I’ve dropped some files into, for the central lab into this directory. So, here’s
an example on the 23rd of January, someone dropped a file into the directory and I’ll
get something like a result. So, for example, I can output files, I can
log files. This is the same LSH functionality for those of you who are familiar with LSH.
It was a fast dataset, so you end up with a SAS log. And, basically, you can see this
is a SAS log that basically talks about this is the output, let me close that one. And,
then if I actually look at the results here, one of these two, aahhh. And this is the LSH
or DMW log. So, you end up with a SAS log and DMW log, it’s just a text file you would
end up with, you would end up with a sequel loader file there. Okay. So, basically, this is just a history of loads
that have occurred. So this studies the mode once is the metadata and twice we loaded the
data. If you have it loading every few minutes, you would get lots and lots of lists here.
Studies currently in development, so it’s not using automatic load set. On the transformation side, you get a history
of the transformations and when they’ve been run. And it’s really by model. And here, we
can see we ran the P-lab model once. So, basically, this, this is all the run time information
and you can, if you have the right to schedule the job, this is just a place where you monitor
the job. And, similarly, for the, let’s say the InForm cross analysis. You get the record
of your, the runs here. So, that’s really the first thing. Second thing I want to look at is data listings.
So, I click on listings and depending on which data model you want to look at, you get those
listings. So, all the tables here are here for the listings. So, for example, if I want
to look at a particular listing or a particular default table, just click on it and I ‘m actually
looking at the data that’s been loaded. So, in this case, you can see there’s 28, you
can go through it. Ummm, it’s pretty straight forward for sorting with that. If I wanted
to sort the subject ID (? –garbled and double talked) you just use the little triangles
and it sorts positive, I mean it sorts ascending or descending. If you want to actually search
for a particular records, you click the find button and you get all the fields and you
can search on any of those fields. So, for example, if I was interested in site 14, I
just put site 14 in here and oops, say search. And that should filter by site 14. Okay, that’s
not what I expected. Let’s try that again. I must have made a mistake. Ahh, cause it
is the wrong field. It’s actually called site pneumonic if I remember correctly. Okay, and
there you can see the records for site 14. So, it’s a pretty simple browser. This is
not Spots Fire, this is not OVIE, but it really does give you information. If there are discrepancies on a particular
record, you get to see those discrepancies. So, if I tab through here, I can see the little
yellow flag there. And, it’s saying, okay congestive heart failure, I can tell there’s
a discrepancy. I can say show discrepancy. So, it goes right down to the cell level and
it says please complete the comment form prior to current medications. So, this actually
came from InForm. We’re going to look at discrepancies in a moment. And there’s also a constant,
a concept of flags. Currently, there are no flags on these records. I have to find some
records with flags. So, you don’t need to define any of these.
This all comes from the data model. Similarly, we can look at our join data, so that the
data that came from both central lab combined with, combined with the InForm data, where
the date of visit is coming from InForm and the, if the lab data is actually loaded , it’s
coming from there. And, you can see that we’ve got various discrepancies here. If I say show
discrepancies on this one, you can say, see look no lab data exists. And this really is
the validation check that we saw two minutes ago. So, it comes right down to the cell level. Now, one of the questions you might ask is
about traceability, because I think this is very important. So, here, I’ve got this combined
dataset that has date of visit and a few other fields from InForm, but the actual lab data
is coming from the central lab data. So, in this case, if I’m going to like click on this
cell, and I say trace data lineage, it goes back through the functions and we just expand
on that. And, you can see that it’s coming, it’s in the field itself is this P-Lab1.dov,
that’s the field we’ve highlighted. But, it’s actually traced back to InForm EGsent.dov.
So, this comes from the date of visit in the EG sent field. So, you do have this full traceability
to see where it came from. So, that, I think, is a very important feature because similarly,
if I went and looked at one of the others that actually have lab data, let’s just say
this protein, and I say trace data lineage, this now as you can see comes from the central
lab file. So, even though we’ve joined data, we actually can see the data lineage of where
exactly the data came from. Did it come from a central lab file or did it come from the
InForm file, in this case. So, the InForm, there is no file in the InForm database. In addition to these types of simple listings,
you also can create custom listing. Now, I haven’t created any custom listings, but it’s
a fairly simple concept. You just press the plus sign. I’m not going to take too much
time here. Give it a name and you’re going to have to actually pick what models, oops,
okay. So, I’m not going to go through that now, but I’ve seen some of these custom listings
work where basically you pick the source, and in this case, the source could be PLab1,
but I might be working in another model. Let’s just cancel on this one and move over to the
InForm model. Cancel. Are you sure you want to cancel? Okay. Greta, do it this way. Let’s go to the InForm
model and create a custom listing there. Pick my sources. Ahh, seems to be a bit confused
see source. All right, we’re going to skip this for now, and we’ll come back to it we
have time later. So, basically, you, you, can go through and
you can create join and you can create pivots, all the things that you really saw in the
maps, where you can basically create, centrally just a path review. And then, the last type of listing really
is the, what’s called the validation listings. So, you may actually want to see a listing
of only data that came, that failed a particular validation check. So, for example, if I just
want to see only the stuff where no lab data exists, I can say, okay, that’s it. So, rather
than going to through everything, I’m really just focusing on this. The advantage of this
is when we go to export this stuff, you can just send all this stuff straight to the lab.
It’s all ready pre-filtered and it’s ready to go. And we know that the lab can get this
stuff. Now, one of the other things I want to show
you is the ability to create discrepancies on the source systems. So, let me go back
out and do a little bit of a reset there. Click on listings, select InForm. Let’s go
look at AE table. And, let me filter it. I know I’ve got to, all right let’s go to a
site here. One-four, scroll down do search so I can restrict my records here. Did I do
the same mistake I did last time? I believe I did. It’s site pneumonic. Do a search. So
here’s zero zero one four. Here’s the adverse events that exists for this particular site.
Now, I’m going to keep scrolling through. Let’s say I’m doing my data review, and I’m
like oh-oh, this is wrong. I mean this is fine, there’s no medical problem with this,
I’m just using that as an example. But, I can actually do. I can create a discrepancy
as part of my review process here. And, I can say something like, please rephrase. So
this is particularly useful if you have a coding system that’s outside of InForm and
you just want to code the data in DMW and you want to send some things back to the site.
So let’s say they want you to split terms or something like that. Okay, just going to
say okay. All right. So, in now there’s a discrepancy in DMW and if I actually say,
go to discrepancy, let me highlight it first. Go to discrepancies, it will jump me across
to this discrepancy itself. So, now I can add comments to this discrepancy.
I can do whatever, you know, whatever I want. But, what’s interesting is that I can actually
click on this set to InForms. So, let me edit the discrepancy. I want to make an active
discrepancy. Sent to InForm. State. No, I want to change the state. Okay, let’s see
what happens. Oh, reason is mandatory. Okay, it’s thinking. Oh-oh. And, of course we get
an error. Wo we’re actually not going to do that right now. This kind of thing that works
extremely well when you’re practicing. So, um, let’s refresh this. Here. Double click
it. All right. Now I’ve upset it. Let’s continue back through the discrepancies
and we’ll come back to that in a moment. So, on the discrepancy tab, you get basically
all the discrepancies that exists both in DMW and in InForm. So, if I click on from
InForm, you get all the discrepancies that come from InForm. So, part of the synchronization
when data comes across is that you can look at the discrepancies from InForm along with
the discrepancies that are not in InForm. So, weight, not within specified range. You
can go here and see this. This is an InForm discrepancy. It came from InForm. You can
say display full record and it actually gives you all the data associated with that record
for that discrepancy. And, then you can actually add comments to that discrepancy. So, those
are InForm discrepancies. But, then if we actually look at the manual
discrepancies, these are the discrepancies that primarily have been created inside DMW,
you can see those together. So, you have lots of ways to filter discrepancies and mange
those discrepancies. So, you can restrict by, you can say what model you want, et cetera.
And you can actually manage those discrepancies just like you would in a clinical data management
system. So, you really have the ability to filter and manage those discrepancies together. So, unfortunately, I’m not sure why that discrepancy
didn’t go across. I was going to actually walk you through in InForm the actual crossing
of those discrepancies. I have another subject that I did that on the other day and I can
show you that in a moment. But, before I do that, let’s just finish this discussion. So, one of the things that you can and is
particularly useful for listings and custom listings and non-custom listings, is you can
export the data. Now, you can do a very simple export, so let’s just say I want to take this
medical history or physical exam table and I want to export it. So, there’s a very simple
way to export it. I can say export it all to excel and then it just takes that file
and brings it up in excel. So, this is an example. So, you can work very quickly. Similarly, you can export it to CSV. Basically,
you know, works, works the same way. But, what’s, what’s much more important is that
if you go over to the life sciences data hub, you have the ability to work with this data
downstream. So, you can do things like create a business area. So, what happens is, I’m
now switching over life sciences data hub user interface. And in life sciences data
hub user interface all your therapeutic areas are there that you’ve set up. In this case,
we’ll look at cardiology. Then, each one of your studies are there, and in this case,
we’ll look at the InForm cardio study. And then, for each one of your life cycles, you’ve
got death, production, QC and we’re really focusing on death. These names are not really
meaningful. In fact, Oracle has chosen to make them unique by giving them some internal
ID, but the description helps give you an idea of where the data is. So, the InForm data itself is here. The metrics
data is going to be in here. For each one of your data models, you always have a container
as well. And, then you’ve got some stuff about validation programs and some other programs.
So, let’s take a look at some InForm work area. So, I click on InForm. What you actually
see is that all the tables are actually tables inside the life sciences data hub as well
as the query information, flag information. So, if I were to look at that same data here,
inside the life sciences data hub, this is the same data that we were seeing through
the listing. This is very important because the life sciences data hub, downstream from
the life sciences data hub, any of these work areas, you can create any program. It can
be QOC sequel programs, SAS programs, informatic programs, anything that you want. On top of that, you can create what is called
a business area. And this really is not a discussion about life sciences data hub, but
the business area allows you to use tools like Spot Fire, Business Objects, J-Review,
right on top of this data. And it’s not a separate copy of the data. So, as your data
is flowing through from InForm through the (?) models, you can actually let, take your
standard data model and then layer it to like a Spot Fire or J-Review on top of that and
actually use that as your review process. And you’re not making extra copies of this
data. The same traceability applies across the board, whether it be inside LSH or whether
it be inside DMW. So, I am seeing that I have approximately
4 minutes left on my clock, so I’m going to switch over to questions. Obviously, this
is a very deep topic and we can talk about it much more than an hour. So, with that,
I’ll switch back to questions and see if I can answer some of them. Okay. Data cleaning (?) DMW, how do you flow
back to the InForm transaction base? Okay, transaction database. So, that’s what didn’t
work in the example. Although I definitely had it working this morning. Basically, it’s
done through web services. So, as soon as you say send to InForm, it calls the InForm
web service and it actually synchronizes back to InForm. Next time, data comes, you load
the data from InForm, it also loads all the discrepancy information. So, you have the
full control loop. So, the data cleaning goes back and forth. Basically you go through and
you mark the discrepancies as sent to InForm and it’s tagged against the variable that
you’re working against. The next question asks: Can we use DMW when
using Oracle Clinical as EDC tool? So, this current release really only accepts files
and InForm. So, we do not, Oracle does not have the connector for OC-RDC or RAVE or anything
else. It really is focused on InForm. My understanding is that future release they are planning on
adding some of that functionality. But, since we are BioPharm, we can’t really state what
Oracle is going to do. But, current it really only supports InForm and files as its inputs. Next question is: Do you start building your
trial DMW after your InForm trial has been built or is done in parallel? In general,
DMW will lag behind InForm’s, so while you’re still basically constructing it, you wouldn’t,
you wouldn’t (skip) immediately. There’s no reason it couldn’t lag, say a few days or
a week behind. It’s certainly not well behind. But, you will actually want to get some stability
in your structure in InForm before you pull it into DMW. It’s really the best practice
on how to handle that. The next question I have is: When creating
discrepancy on listings, can (skip) directly (skip) to InForm or do have to go to discrepancies?
My experience is that you have to go to the discrepancy and send it to InForm, but I think
that in the user documentation it says that you can send it directly to InForm. I haven’t
actually gotten that to work. I think, in theory, you can send it directly to InForm,
but, usually, my experience to make it faster, you go to review the discrepancy and then
send it to InForm. So, that may be a bogger it may be just my lack of knowledge. The next question I have is: Is there any
option for graphs in DMW in addition to reports? So, this functionality is limited as you seen
it. If you want more advanced graphing with a tool like OBIE or Spot Fire or J-Review,
those same listings through the LSH interface are available to all of those tools. So, you
do lose (skip) in Spot Fire, J-Review the ability to back propagate discrepancies. But
for really nice interface, right now I would say use a tool like Oracle Business Intelligence
or (? skip). I have a couple more questions. I do want
to take a minute to wrap up though. When discrepancy is created in DMW, how is it routed back to
InForm? So, I think I answered that already. It is done through the InForm web services/ And then, the next one is: Can we consider
DMW the metadata management tool for updated cleaning and discrepancy management tool?
It primarily is the purpose. LSH underneath is a metadata management tool. This is an
alternate interface for that. because you can have models (skip) various and their kept
under version control, you can actually use this as a metadata management tool. I think
the actual (?) of metadata management. It’s not explicitly handling the whole approval
life cycle, the new variable being added to our model as a workflow, but it does offer
all the underlying version control and metadata management. And, in the LSH side, you actually can write
comparative reports that say how is my standard model compared with this other standard model. Um, we’re kind of running out of time now.
So, I do want to wrap up. I wanted to thank everybody for joining. I really do appreciate
it. BioPharm is knowledgeable on DMW debut. We work very closely with Oracle to really
become experts in this area. And, please feel free to reach out to us and
ask questions. If you think you’re interested, if you have questions about this, LSH or any
clinical warehousing analytics applications. We’d be more than happy to answer your questions.
So, with that, I think I’ll hand back to you Eugene and he can wrap up the webinar. Thank
you again. (Eugene) Great, thanks Mike. So, as a reminder,
we will be sending out a recording of the webinar as well as a PDF version of the slides
within the next 24 hours. You can also go to biopharm.com to access the recording and
presentation. We want to thank you for your participation and hope that information that
Mike provided today was helpful. And, as always, feel free to register for our other upcoming
webinars which you can find on our website. Thank you so much and have a great rest of
the day and evening.