 I just have to correct one thing that Peter said. There is one factor that will drive patients. It's parking that's well-known. Okay. There we go. So what I'm going to do is give an overview of some of the definitions of precision health and learning healthcare system. Actually not spend much time introducing the Geisinger My Code Community Health Initiative since Alana did that this morning, but then talk more about the Geisinger Genomics Screening and Counseling Program and then discuss some of the key processes and how we've actually utilized the tools of implementation science to get this large scale population sequencing program off the ground. So precision health, at least in the way I'm using it, emphasizes prevention while encompassing the interventions that are inherent in precision medicine, which is a more disease-focused as opposed to a health-focused entity. So in precision medicine, the definition that I use is one from Clay Christensen, the provision of care for diseases that can be precisely diagnosed, whose causes are understood, and which consequently can be treated with rules-based therapies that are predictively effective. That's from the innovators' prescription, a disruptive solution for healthcare. We do view our project as a precision population health initiative and as such we've renamed it as the My Code Community Health Initiative to distinguish it from where it started, which was as a biorepository. Now inherent in this are educational efforts that are directed at participants, providers, payers, administrators, and other stakeholders. And as we talked about earlier today, this is endorsed at the highest level of the organization as a strategic initiative. So it's explicitly represented in the strategic plan of the organization, both the clinical organization as well as the research organization. You saw the slide earlier, the Community Health Initiative. We're looking to do exome sequencing at a minimum of 250,000 patients. We're going to go through those exomes to look for medically actionable results and then return those results to patients and providers. And we'll support those patients and providers to the follow-up along with long-term management. And then we'll be operationalizing this as a scalable return of results and infrastructure in this integrated health care delivery system. So again, this is the biorepository was where we started 11 years ago. It was a very traditional biobank at that time, but it was informed by extensive consultation with our patients that informed the design. And at that time they were saying, you know, we would really, if there's information that you think is important to our health, we would like to have that back. So there was interest from our patients already about the possibility that we could return something to them. So the participants signed a broad consent to combine their EHR data, which can be prospective as well as retrospective, can be identified or deidentified as well as biospecimens. And a more recent version of the consent which we're trying to convert people over to includes the ability to re-contact participants for future projects and to communicate medically actionable results. So all the patients that are being consented now are being consented under this broad re-contact and return of results. To date, actually this is not to date, this is August 1st. This number is now 877, and we have, as Alana said, 216,000 consented at the present time. Our initial analysis indicated that we thought about 3.5% of the population would have an actionable genetic condition. I think as we've learned more, particularly about some of the genes that are associated with the inherited rhythm disorders and cardiomyopathy, we're revising that down a bit, maybe closer to 2, 2.5%. This has also changed since I turned in my deck here. We're not the guys in your 61, the Brugata paper came out and said, well, we really don't have any good associations with 20 of these 21 genes, and so we said, well, if that's the case, we shouldn't be returning them. So it illustrates the dynamic nature of implementation and trying to really use best evidence, but we do focus this on conditions that we consider to be actionable, meaning that there is a medical intervention that can be done. We recognize that actionability can go well beyond medical interventions, but we're using a medical model for this. It does build on the ACMG incidental findings list, which was recently updated in 2016 and continues to undergo revision, a lot of focus on cancer predisposition, cardiovascular disease. This again is from August, and I did update the numbers here. We've now returned 881 results to 877 patients, so four of our lucky patients have two disorders that we identified. We've put right up front here the CDC Tier 1 conditions, the ones that the Office of Public Health Genomics thinks that there's adequate evidence to consider public health interventions here, and so you can see that we've had over 250 hereditary breast ovarian cancer, 107 FH, and 85 Lynch syndrome results that we've returned. We're going to take a look under the hood and look at some of the challenges that we occasionally encounter when we lift up the hood. So this is the high-level process. We start with consenting and sample collection. We move to sequence interpretation, confirmation, and reporting. That's reporting back to our clinical return teams. That's reporting from the laboratory to the Genomics screening and counseling program. We then report results to participants and do cascade contact and testing for family members, and then we measure outcomes that are attributable to reporting. So I'm going to go through each of those at a high level. So consent and sample collection. Participants can consent in person. They can consent online through a my-guysing or portal, much as Peter was talking about, or through a smart device, also tethered to the patient portal. We consent between 800 and 1200 patients per week. We found, though, that in-person consenting is, by far, in a way the most efficient way to do it. We do sample collection as part of the routine blood draw. And then because we have a standing order, additional samples can be drawn over time to replenish biospecimens. So as opposed to many biorepositories where you get a specimen, and then when that specimen's gone, you have to recontact the patient, potentially reconsent them. Here I go in from my annual physical. I go to get my lipid study, and they said, oh, you're in my code. Can we draw a couple of extra tubes of blood? Sure, go ahead. And then we have that ability over time to replenish. I want to just make one point here about how we implement the consenting. We put concenters in clinics. What we recognize is after a concenter's been in a clinic for a certain period of time, they achieve saturation. So what we actually did was to design a system by which the scheduling software would communicate with our research software. This was a purpose-built iteration. What it allows us to do is to say for each of the patients that are being seen in this clinic, how many of them are already in my code, so we're not approaching people that are already in there and bothering them. How many people have said, I don't want to be in my code, so we're not really not bothering them? And then which patients have not been previously approached, and those are the ones that the concenters target in real time. And then we can also look to say, okay, what percentage of people that are coming into this particular clinic are eligible? And as we see that drop, then we can move that concenter into a different clinic that is more naive to my code consenting. And by this way, we've been able to kind of keep our numbers up and also achieve satisfaction because we're not constantly bothering people, particularly those that come into the clinic frequently. Sequencing confirmation and reporting. This is in theory. You have our eligible participants that are sent for exome sequencing. That sequencing is done in a research laboratory that is not CLIA certified, so we have to do a bioinformatic analysis against the genes that we're interested in reporting if we identify variants in those genes where there's a potentially reportable result and we send that for confirmation to a CLIA CAP certified laboratory that issues us a report. If there's none, then we just save that exome sequence for a feature bioinformatic analysis as Yogi Berra once said, in theory theory is better than practice and practice it ain't. This is what it actually looks like for just one laboratory in terms of actually trying to operationalize that system. I am not going to walk through this other than to say that if you don't understand the underlying workflow, you will miss opportunities. And even since this slide was collected, we had a process by which we had the sequence being analyzed by an external vendor. We now have an internal sequence analysis pipeline, so this is now excised for this particular process. But for each of our laboratories, including our internal laboratory, which is doing testing for C282Y hemochromatosis, we had to build a workflow of this type and then have monitors in place to make sure that the system stays up and running. For reporting results to participants and their family members, this is our workflow. When the result comes back to the laboratory, it does go to our genetic screening counseling program where we do a second review. A to say, did we call this pathogenic? Did the laboratory call it pathogenic? And is there new information or is there information within patient's chart that would lead us to believe that perhaps this is not correct? So we screen all of the results, but once we have met that threshold, then they are released to the primary care physician. Now, this is somewhat different than our usual process. It's not unusual when I go in from my lipid profile that when I'm on the hall going back to my office that I get my guising or alerts saying your lab results are ready. And they go, they're released simultaneously to the patient and to the provider. But in consultation with our providers and with our patients, they said these are kind of different. These are a little bit more complicated. We think we should build in a lag. So we were able to engineer the process such as a five day waiting period after the result goes to the PCP so that they can take advantage of getting up to speed. The PCP in some cases will just go ahead and contact the patient. Otherwise, we then send an electronic message for those patients that are on the patient portal or a letter to those patients that are not. And then we call the patient to make sure that they did receive the result. If we can't contact the patient or get no information that they've opened the message in the patient portal, after three attempts to contact, we do send a certified letter. And so they at least have the results. So they will have that result in a certified letter. And then sometimes they will contact us. And at that point, then we disclose the result using phone scripts that are specific for the different results. We begin to collect family history information and then we can send the result and support materials. So we send information for as many first-degree relatives as the patient says that they have so that they can distribute that to support cascade testing. The patient has the option they can follow with their primary care physician with the genomic screening counseling program or both. We also then take advantage of existing programs where we have conditions specific specialists for all the conditions that we're returning. We have a respiratory pressure-varying cancer clinic, high-risk colorectal cancer clinic, cardiomyopathy clinic, et cetera that we can utilize. And then we serve a care coordination role in this process to make sure that the patient can navigate the system. Outcomes are critically important to implementation. Many of you have seen this before. But when a result comes back, patients can fall into one of five categories. They may have a prior diagnosis through clinical testing where if it's a BRCA result, we say, hey, we found this variant. Yeah, I know I had clinical testing. Hopefully they found the same variant. However, we found there's still value there because there has been inconsistent communication of what exactly that patient should be doing and what their family members should be doing. So there are missed opportunities that we can fill in a care gap here. In the second group, this is a patient that has a condition like breast cancer but does not in fact know that it's due to a genetic variant. So we now provide a unifying diagnosis and provide opportunities for differential care as well as testing for family members. In this case, this is an individual that is not aware that they have a condition but on the basis of the intervention. Say breast imaging, they find a subclinical breast cancer which can be hopefully dealt with at an earlier stage. And then two groups have no phenotype at the time of the disclosure after the initial evaluation. In one of those groups, the phenotype will emerge over time but we do know that none of these conditions are completely penetrant. So we know that there's going to be a group for which we've returned a result that they will never have a benefit from personally because that for whatever reason their phenotype would never emerge. But of course we don't know how to distinguish those groups. Maybe at some point in the future as we understand the complexities of this we'll be able to understand that better. But again, here there are still our opportunities for cascade screening for family members that might in fact develop the condition. These are all the outcomes we look at. There are process outcomes. Did a person get a test? Intermediate outcomes, what is the actual lipid level? If we do an LDL, we ultimately want to get the health outcomes. Did we prevent a cancer? Did we treat a cancer in an earlier stage? Did we reduce heart attacks? What's the cost? There are behavioral outcomes or patient reported outcomes. We heard about some of the things that Peter's programs collecting. All of these are really important. There are also outcomes from the system perspective, costs incurred or avoided, utilization, visibility and reputation which actually is kind of a big deal. We've been told by our business people that we've had groups that have come to us because we have this program. I wasn't expecting that in central Pennsylvania. And then the patient experience, we actually get paid for that. If our patients have a good experience with us, we get more money. So ultimately what we want to do then is to create this precision health in a learning healthcare system. And again, you saw this PDSA cycle that Alana had presented earlier. So we plan this, we release it. But again, we guarantee it's going to be wrong when we send it out there. But we have to know how do we collect data and how do we fix it? We study it. We disseminate it. We might try it out in one clinic and then from what we learn we move it to others. And so there's a constant analysis but at each of those points you have to collect data to really know what's happening. And so that's what we're really trying to do. And you might ask the question, well, this is pretty complicated. Does it actually work? And I'm happy to present at least one piece of data that suggests that it might work. So this is our results returned to date. So these three are annual returns. So we returned 99 from May 1st of 2015 to May 31st of 2016. The next year we returned 209, 254. But we realized that this was really, we had a lot of people sitting out there that weren't getting results. So we in the process here took a look at all the different aspects of the process. And in three months from June 1st of this year to the end of August, we returned 315 results more than we'd done in any year previously. So over a third of our results were returned in the last three months due to extensive re-engineering of our sequence interpretation pipeline, the confirmatory testing and reporting and then how we can improve the timeline of reporting results to participants and family. So in conclusion, precision health is an emerging technology. We have to be able to demonstrate improved value in the healthcare delivery setting before it will be adopted. Outcomes have to be defined ahead of time and systems built to support measurement to determine which services actually add value. And implementation is complex, who knew. And requires a systematic approach. The learning healthcare, this was recorded wasn't it? I'm in big trouble. The learning healthcare system framework may represent a robust implementation model. And I would like to thank all of our participants, again over 200,000, the great team of the Genomic Screening and Counseling Program and then all of our collaborators at Geisinger and at Regeneron. Thank you. Thanks, Mark. Any quick question? Seems not, thank you. Okay, next to Lincoln Nadald on implementation of cancer genomic medicine.