 |
Cletis Earle, CHCIO, VP, CIO Penn State Health Image courtesy – Penn State Health |
Q. I want to talk about the interview you did for HIMSS and delve into a couple of points on that a little more deeply but before we start could you just give me a brief description of your current position as senior VP and CIO at Penn State Health
And as you can imagine as a CIO you come in with objectives around technology but unfortunately, a lot of that was thwarted because of the pandemic. For a lot of the last year- and-a-half at least, I've been really focused on the pandemic and working from home developing the systems you know for testing and then vaccine administration. Just as things began to ramp down and started getting back to the world of regular IT work the Delta variant came down the pike, I guess this is the new norm where down the road we're now going to be ramping back up. So it has been a roller coaster and that's where my career has been regarding in particular, the last year and change, dealing with pandemic related things and regular IT operations.
Q. Going back to an interview you did with HIMSS, you said, there are now countless scenarios to use a gold mine of data to provide people with the best results. Tell me first about that gold mine of data. What are you talking about there, and in your view what are one or two of the most likely scenarios for mining that data?
Yes. What I mean by the gold mine is that the healthcare information that we have is the most complete data sets that we have on the human. It is not only everything about you, and we’re not just talking about your demographic information, we're really talking about all of the elements associated to your being. That’s important in that we can go down to the genomic level, right? For organizations such as ours that have significant amount of data over a course of years, we are really talking about a scenario where we have more tangible information in a digital form about the human being than at any other time.
From that perspective, that's great. Then when you think about it even further, now you also have the data available that family members, your kids, your aunts, uncles, your niece, your grandmothers, and your mothers, which can inform your doctor about your family health history. All that information is also aggregated to form a clearer picture of a community health.
Again, at no time have we ever had this kind of repository available, and the key here is we now have the ability to work it and do something really good with this data. As we start to develop machine learning and other technology, we will be able to look at this information, be able to extract the details in correlations that we were not able to find on a manual basis.
We're really talking about, at Penn State University and Penn State Health, and our College of Medicine where we have a major research arm, we're really talking about the ability to have redacted or de-identified information that is truly a culmination of years and years of physiological and clinical content that can help predict and prevent disease in the future.
Q. So are clinicians or researchers using that data to ask questions that they never able to answer before?
Yes, researchers and clinicians are asking, and again this is de-identified data all right, so they are able to ask and look for access to all this information to pursue a wider range of research directions, at other times they are coming in with concrete ideas and concrete initiatives to pursue research. And then sometimes they're looking at this to say, hey how do we have better access to this information so that we can put it through some algorithms to be able to abstract information in different ways.
Q. Are we just talking about electronic clinical outcomes assessments or more broadly about all kinds of clinical research?
I think it's all kinds. Whether its imaging, whether it's electronic text, you're really talking about tapping into this information in different ways in order to have machine learning and artificial intelligence being able to identify some areas of opportunity to improve the health of our patients.
Q. Can you just give me a little example of a question a researcher came to you and said, how can we get this out of the data we have?
You know, whether they need the data to look at a program around say, endocrinology and diabetes studies they want to be able to look at data information that is deep and robust, so that they can actually start to see where there may be some opportunities to provide better care based on their research studies.
I can mention hundreds and hundreds of scenarios, but I think the endocrinology piece or the diabetic portion is probably one that comes to mind because there are also opportunities to not only connect to the data but then connect to patients’ mobile devices so that we can have a real-time data and really abstract information continuously so that we can provide real-time care.
Q. I am assuming that a lot of this is coming out of electronic medical records, including medical imaging data. Now what are some of the keys to getting access to that gold mine of data? What do you think are the obstacles into tapping into that?
First and foremost, we have to de-identify it because that information cannot be tied back to a specific person. So the biggest challenge is always making sure that you are making it available for the researchers so that it cannot be traced back to resources. I think that's the single most important challenge, and then providing access, general access for our researchers so that they have better access to this information. De-identifying and providing access is by far the biggest lift we have in order to provide the information as quickly as possible.
Q. A lot of clinical research is conducted by collaborations across multiple academic institutions. Do you have to be able to find ways to allow researchers to share data across different health systems.
Are you talking about health systems or research because universities tend to share data amongst themselves by default. That's what they tend to do. If we are talking about the Cure Act component of the data, we have Federal legislation out there that really talks about the fact that there's data blocking and that nobody should block data from others.
As a competitor we adopted the strategy that the data belongs to the patient. We were one of the first organizations that took up Open Notes, so we believe in pushing out the data to the patient, allowing the patient to take that data anywhere. We have patient portals, we have all of the things that allows the patient to be able to share that information with another provider regardless if it's with a competitor or not so that the outcome can be the best possible.
Depending on the scenario again, if it's clinical, then yes the answer is yes, we share data but it's more of the patient sharing his or her data, right? Because again, it's in their hands. If we are talking about researchers they tend to share data and their results of that data inherently in common.
Q. You mentioned machine learning a little bit ago. How does this machine learning help providers determine better pathways to care?
If you use a current examples, such as the IBM’s Watson solution, Memorial Sloan Kettering, has some really good information that indicates that machine learning has been extremely efficient in in being able to show providers that their data and their images have potential scenarios for cancer where artificial intelligence is screening through tens of thousands, if not millions of images and seeing where there are anomalies that can then be reviewed and provide feedback to clinicians in a much more comprehensive fashion.
I think inherently that's one of the most recognized ways where machine learning has made an immediate impact in being able to be able to say look, we know the technology can help facilitate the doctor in finding information faster because the speed of a computer can't be matched. So I think that's a great example on the imaging side.
Q. The ability to share data across or between different hospital systems, and I'm talking about large integrated delivery networks, such as Penn State Health, is that something that's becoming more common? Or is it still a problem or are there barriers, policy barriers, technical barriers to doing that?
Yes, it's becoming more common I can tell you at Penn State Health and State College of Medicine on the health system side, we've adopted that approach, we use CommonWell, we've been a proponent of HIE, we continue to extend ourselves in HIE, Health Information Exchange, areas so that we can share data with our colleagues. Our goal is to share as much information as possible in HIE so that information is accessible to anybody in the state and then potentially anybody who would connect from a national side.
We’ve found that having a robust data sharing program, particularly around clinical information and clinical content is essential to the best outcome for the patient. And we use these components like CommonWell, and our HIE as the catalyst to make that happen successfully.
Q. Has the interoperability problem been solved?
No, no no, that's not quite there yet. Again, from that interoperability side we are using the HIE as the clearing house, right? But there are times where, you know, EPIC needs to connect with Cerner and Cerner needs to connect to MediSite, and Allscripts and all of those others, so still there are some challenges. Again, it all boils down to sometimes the technical capacity and capabilities of the EMRs. Some are better than others around that, but it also boils down to the organization's willingness to share that information.
We're really talking about, at Penn State University and Penn State Health, and our College of Medicine where we have a major research arm, we're really talking about the ability to have redacted or de-identified information that is truly a culmination of years and years of physiological and clinical content that can help predict and prevent disease in the future.
Q. So are clinicians or researchers using that data to ask questions that they never able to answer before?
Yes, researchers and clinicians are asking, and again this is de-identified data all right, so they are able to ask and look for access to all this information to pursue a wider range of research directions, at other times they are coming in with concrete ideas and concrete initiatives to pursue research. And then sometimes they're looking at this to say, hey how do we have better access to this information so that we can put it through some algorithms to be able to abstract information in different ways.
Q. Are we just talking about electronic clinical outcomes assessments or more broadly about all kinds of clinical research?
I think it's all kinds. Whether its imaging, whether it's electronic text, you're really talking about tapping into this information in different ways in order to have machine learning and artificial intelligence being able to identify some areas of opportunity to improve the health of our patients.
Q. Can you just give me a little example of a question a researcher came to you and said, how can we get this out of the data we have?
You know, whether they need the data to look at a program around say, endocrinology and diabetes studies they want to be able to look at data information that is deep and robust, so that they can actually start to see where there may be some opportunities to provide better care based on their research studies.
I can mention hundreds and hundreds of scenarios, but I think the endocrinology piece or the diabetic portion is probably one that comes to mind because there are also opportunities to not only connect to the data but then connect to patients’ mobile devices so that we can have a real-time data and really abstract information continuously so that we can provide real-time care.
Q. I am assuming that a lot of this is coming out of electronic medical records, including medical imaging data. Now what are some of the keys to getting access to that gold mine of data? What do you think are the obstacles into tapping into that?
First and foremost, we have to de-identify it because that information cannot be tied back to a specific person. So the biggest challenge is always making sure that you are making it available for the researchers so that it cannot be traced back to resources. I think that's the single most important challenge, and then providing access, general access for our researchers so that they have better access to this information. De-identifying and providing access is by far the biggest lift we have in order to provide the information as quickly as possible.
Q. A lot of clinical research is conducted by collaborations across multiple academic institutions. Do you have to be able to find ways to allow researchers to share data across different health systems.
Are you talking about health systems or research because universities tend to share data amongst themselves by default. That's what they tend to do. If we are talking about the Cure Act component of the data, we have Federal legislation out there that really talks about the fact that there's data blocking and that nobody should block data from others.
As a competitor we adopted the strategy that the data belongs to the patient. We were one of the first organizations that took up Open Notes, so we believe in pushing out the data to the patient, allowing the patient to take that data anywhere. We have patient portals, we have all of the things that allows the patient to be able to share that information with another provider regardless if it's with a competitor or not so that the outcome can be the best possible.
Depending on the scenario again, if it's clinical, then yes the answer is yes, we share data but it's more of the patient sharing his or her data, right? Because again, it's in their hands. If we are talking about researchers they tend to share data and their results of that data inherently in common.
Q. You mentioned machine learning a little bit ago. How does this machine learning help providers determine better pathways to care?
If you use a current examples, such as the IBM’s Watson solution, Memorial Sloan Kettering, has some really good information that indicates that machine learning has been extremely efficient in in being able to show providers that their data and their images have potential scenarios for cancer where artificial intelligence is screening through tens of thousands, if not millions of images and seeing where there are anomalies that can then be reviewed and provide feedback to clinicians in a much more comprehensive fashion.
I think inherently that's one of the most recognized ways where machine learning has made an immediate impact in being able to be able to say look, we know the technology can help facilitate the doctor in finding information faster because the speed of a computer can't be matched. So I think that's a great example on the imaging side.
Q. The ability to share data across or between different hospital systems, and I'm talking about large integrated delivery networks, such as Penn State Health, is that something that's becoming more common? Or is it still a problem or are there barriers, policy barriers, technical barriers to doing that?
Yes, it's becoming more common I can tell you at Penn State Health and State College of Medicine on the health system side, we've adopted that approach, we use CommonWell, we've been a proponent of HIE, we continue to extend ourselves in HIE, Health Information Exchange, areas so that we can share data with our colleagues. Our goal is to share as much information as possible in HIE so that information is accessible to anybody in the state and then potentially anybody who would connect from a national side.
We’ve found that having a robust data sharing program, particularly around clinical information and clinical content is essential to the best outcome for the patient. And we use these components like CommonWell, and our HIE as the catalyst to make that happen successfully.
Q. Has the interoperability problem been solved?
No, no no, that's not quite there yet. Again, from that interoperability side we are using the HIE as the clearing house, right? But there are times where, you know, EPIC needs to connect with Cerner and Cerner needs to connect to MediSite, and Allscripts and all of those others, so still there are some challenges. Again, it all boils down to sometimes the technical capacity and capabilities of the EMRs. Some are better than others around that, but it also boils down to the organization's willingness to share that information.
Although we have decided to share everything to the HIE, not every organization feels the same way because they're, in my opinion, stuck on an old model. They may keep the data of their patients in the belief that in doing so they have more power. We’re choosing to compete on the quality of our care. So, we've looked at that as an approach and we have an amazing leader in Steve Massini (CEO of Penn State Health) who continues to push that narrative and we just continue to strive forward that way.
Q. So that's kind of a policy barrier that is up to each individual Institution?
Yes. Hopefully one day everybody starts to feel the same way. Right? It all comes down to who owns the record that's all it is, and again if you finally get back to this concept that it belongs to the patient, then I think that's the game changer, and then these technology barriers that these EMRs have will come down over time, and once they come along, and they're coming along slowly, I think we're going to be in a much better place when you get those two components done right.
No comments:
Post a Comment