Now we're going to look at data management considerations for multi-center or network based studies. Specifically, we'll look at team roles and responsibilities, some extra logistical planning that's needed when we're doing multi-center studies and data hosting, quality assurance and management issues. First of all, it's important to think about the team roles and responsibilities. When you're dealing with multi-center or network based studies and, you know, I think you should at least, I think we should at least think about things in terms of study site. Study site might be enrolling patients, they might be gathering data, they might be generating the data that's going to be managed, but typically they're going to be one of multiple sites. Then in many cases there, there is a formal data coordinating center. This is the group that will be managing the collection, the storage, the, the quality assurance, the cleaning of data and dissemination of that after the study is complete. And then sometimes, there is a, or there always, there's a, there's a regular study operations coordinating center. And this group, if it's separate and apart from the data coordinating center, will be sort of, in charge of the study operations and, and, the coordination for the entire entire study or network based trial. Again, sometimes the data coordinating center and the coordinating center are the same, sometimes they're different. It usually just depends on how big the study or trial is. So first thinking about team roles and responsibilities. One thing that's different about multi-center work is that you know, you have bigger teams. With multiple institutions you have multiple IRBs. You have multiple contracts and, and even subcontracts to get a study launched and working. You got to think about data use agreements and even business associates agreements. And we've talked about good clinical practice and certifications and training. All of that has to happen in duplicate and triplicate. And, and, how many sites, sites that you have in your study. You have to sort of do that many, many times. There is an economy of scale that works in here and you can do many of the processes in parallel, like the IRB submissions, etcetera. But you do have to work it each time and so there's a lot of extra organizational work in these studies and trials. Need to consider data use and publication policies. With multiple sites come multiple investigators. These are all very bright people. They're all leaders in their institution and you have to work through a priori things like data ownership, scientific credit. remembers as well that in, in longitudinal studies particularly in network based studies, a lot of times sites can be added or even removed in the middle of a study. So, so this is not a one time deal. This is sort of a setting of the policies up front and working through the logistics even downstream of that, when multiple sites are added downstream. randomization of subjects across centers. We've talked about all of these concepts, you know, a little bit in, in the single site instances. But again if you're random, randomizing subjects across centers you've got to operate telephone services or, or call in services or maybe computerized web services so, so that as patients are consented in and enrolled they're, they are randomized in the appropriate way. same with assignment of IDs, making sure that that IDs are consistent, but, but they are unique across sites. Then there's logistics of equipment. we run a a fairly large data coordinating center here in, in one of my teams, and you know, there's a lot of work expended in terms of sending out laptops and sending out you know, equipment to run the studies and the trials. If it's a, if, if it's a sort of a normal normal type trial that does not require a lot of extra protections on the regulatory side, sometimes you can, you can use just equipment that's on hand. But if you're doing things like FISMA and some of the other things that we've talked about on the regula, regulatory side, where you've got extra security conditions, you know, you, you, you may well have to participate in, in coordinating the, the logistics of the equipment. And you know, that includes shipping, it includes hardware support, software support, all sorts of things there. training. centralized training can, can be very useful. sometimes we have to go out and we have to train face to face. one of the things that we try to do as a group is, we try to, try to as much as possible, centralize that training. And do more remote training with go to, go to webinar or you know, other webinar type training and teleconference calls. We've found that the combination of having you know, good solid platform that data platform and, and sort of common training mechanisms, training, training materials around that platform work out pretty well and can really reduce it. But it's very important that all key per, personnel working with the study be trained adequately. user management. So you know, just, just as in a single trial, you need to be very, very careful of adding and removing people from your project in, on the, on the data and operations side. This, this becomes even more complex on the multi-center studies where you really need to keep very, very strict control over the, the training of individuals before they're allowed into systems, any kind of data use agreements that are needed, etcetera. And, and just be very, very diligent about adding and pruning the, the, the record so that, so that just the right individuals have access to the data and can participate in the right ways at the right time. Data hosting. This is this is probably one of the first things that I think about when I, when I meet with an investigator team and we're looking at multi-center or network based trials and studies. It's, it's how we're going to deal with the data harmonization. So there's really a couple of different models that, that we think about. The first one is centralized. And, and here I've got a slide that, that, that has two sites, site A and B, just for example. But again, you might have ten, you might have 40 sites in, in this same sort of principle and paradigm. in a centralized model the, the, the, each of the sites would be you know, would be entering data either on paper and, and faxing it or, or sending it to a formal data coordinating center and having those guys enter the data into an electronic system. Or ideally they would be accessing the electronic data system remotely. And, and, and basically the idea here is that all of the data collection instruments, all of the forms and, and the things that we've talked about up to this point, it would be the same as in a centralized study but in this case we would sequester the data. So that site A would be able to see their data, site B would be able to see their data, and, and there would be no mixing of that. Again, the standardization of the forms, everything would be the same except for that sequestering. the other model to consider is a de-centralized model. Here, maybe site A and site B and the other sites would have their own sort of local electronic data capture system and process. And, and, and once those data are collected they would push them or the data coordinating center might pull them out, out of systems at the right times. Do whatever extract, transform and load pro, process needs to happen to get them into a harmonized dat, data data platform. So there are there, there are really advantages and disadvantages to both. the disadvantages to the centralized it's an easy decision for trials. It really is the right way to go if you can do it. sometimes in pilot collaboration, sort of network-based collaborations, we find that you know, if it's, if it's a very new project again, not a, not a formal clinical study or trial but more of a collaboration registry. Sometimes sites are a little bit nervous about sort of giving up control of their that you, you know, being able to sort of have their own data sets here until, until things kind of roll along. the pitfall on the, the decentralized model is you can do it but boy, it's really, really important to standardize thing. And not just once not, not make sure that we're collecting the data once the same way but make sure any kind of versioning that goes on downstream gets propagated out to all of the different sites. It's doable but it's really difficult. The only time that that, that I recommend doing that is if the politics or whatever the local policies are are just completely prohibitive of the centralized. Otherwise you're almost always going to be better in my opinion, with trying to to push for a centralized model of data collection and, and management for these multi-center studies and trials. So you know, other issues to think about special considerations in multi-site studies would be study monitor. Sorry, monitoring the study progress. monitoring for data completeness and monitoring for data quality. Including query formal query issuance and resolution and being able to stratify that across all of the sites. Lot of communication needed there to make sure that everybody's pulling their load in terms of the recruitment process, but also in terms of keeping their data clean. Keeping it pristine so that so that, so that basically the scientific benefit for the study is, is as strong as it can be once things are closed up. Again, it's these are all important when you're doing single site studies but it just sort of amplifies when you're doing multi-site studies in trial. important the data quality. making sure that if your the data coordinating center in these multi-center studies and trials you're going to be typically tasked with freezing the data. With you know after the quality checks, after the sort, sort of the the, you know, all the monitoring that we just mentioned, there are going to be points in time when you'll be expected to freeze the data. When you'll be, be expected to absolutely clean it, de-identify it, and then disseminate those data sets for analysis and publication. So those are just a few considerations. Again this is like many of our special considerations topics, we're going sort of wide across overview in, in terms of some of these topics rather than, than very deep which would, you know, maybe take a course in it, in it of itself. But I think we've done an adequate job here of covering team roles and responsibilities, logistics and as well as data hosting, quality assurance and management.
