Augmented Intelligence Tool Helps Ensure Timely Referrals to Palliative Care for Patients With Advanced Cancer


An intelligence-based system could help improve the timeliness of referrals to palliative care and hospice services in a community oncology environment.

In an interview with Pharmacy Times, Ajeet Gajra, MD, MBBS, FACP, discussed research into how an intelligence-based system could help improve the timeliness of referrals to palliative care and hospice services in a community oncology environment.

Q: Can you explain the augmented intelligence-based tool that you discuss in your presentation?

Ajeet Gajra, MD, MBBS, FACP: So, this was born out of the work and the tool is owned by Jvion, it’s called Jvion CORE. So, what it is, is a machine learning tool that is commercially available and we published other data ahead of this current study in the Journal of Clinical Oncology Practice, where this tool basically integrates clinical information, which is, you know, your tumor, your stage, comorbidities, polypharmacy, age of the patient, gender, all of that. So that's the clinical information. Then it has all of the claims data, you know, because patients may be getting oncology care at the center, but then they also have all this other care that they're receiving, so that gets pulled in from there. The novel piece of this AI tool, the Jvion CORE, is that they integrate data from social determinants of health. So, this is beyond just income, poverty, education. So, many people say, oh, yeah, just use the zip code. No, zip code is too broad. So, this goes much further than the zip code. And this is data that they are able to collect from publicly available databases, like the US Census Bureau, or the USDA, or the NOAA.

And so all of this is brought in, and what essentially, to put it very simply, what they're doing is they're creating these clusters of patients which are similar based on their disease and their social structure, and then trying to assess the trajectory of a patient. So essentially, a patient is being matched to other patients with similar clinical and social characteristics to determine what the outcome or the trajectory might be. So, the key component of this is that the risk of deterioration, risk of mortality, risk of doing poorly, you know, is what is identified. And so, once a patient like that is identified, and of course, in the context of an advanced cancer diagnosis, then these patients are the ones who are appropriate for a palliative care consult.

So, a clinician might say, “Oh, why do I need all of this? Because I know.” But the surprising part has been that there are patients who are obviously, unfortunately, not doing well clinically. And so, everyone knows they need palliative care. But then there are patients who may appear clinically sound or not at risk, but then based on all of these other factors are determined to be at risk. And then also, there's a component of patients who may appear as high risk, but they may have a modifiable risk factor or, you know, a latent infection or some other social issue, like lack of transportation, which can maybe put them in the high-risk bucket. So, it basically amalgamates a lot of information from various sources, and this has been previously validated. And like I said, it's a commercially available tool that we've used for this purpose.

Q: How can this tool be implemented in the real-world setting?

Ajeet Gajra, MD, MBBS, FACP: Yeah, so you know, there are some practices utilizing this tool already. And of course, we've worked with one of these practices for the data that we're presenting at ASCO 2022—and this was the Northwest Medical Practice, Dr. Sibel Blau and her team out of Seattle Tacoma area in Washington State. So, the way it works is that the AI system needs to be integrated, or rather, it sits on top of the EMR. So it is EMR [and] EHR agnostic, so it doesn't matter what type of electronic medical record the practice may be using. This AI tool can, let's just say, read off and communicate with the EHR and derive the information that is needed from the HER, and it will generate output and this is a continuous outcome. So, essentially, every week all of the patients that are in the practice or in the EHR are reviewed by the AI tool to identify those at highest risk. Now, there are also other vectors or other components to this AI so, you know, patients can be assessed not only for deterioration and mortality risks, but for pain, for depression, for mental health. Again, a slew of other possibilities and new vectors can be created. You essentially have to train the machine and the learning set and then a validation set, but the bottom line is in the mortality vector or at-risk factor, or risk of deterioration, the tool creates essentially a report for each week in terms of who are the patients who are highest risk. And this report is literally pulled off by the practice to identify the patients, you know, that have been identified with this tool. And then you can take corrective measures, which may include palliative care, or if that is not the case, say someone with adjuvant therapy who doesn't have active cancer, if they are at high risk, well, then they may have something preventable or you know, reversible, that is putting them at high risk, and so it can be addressed in that manner.

Q: What did you find about how the tool impacted timely referrals?

Ajeet Gajra, MD, MBBS, FACP: Yeah, so what we've demonstrated in our earlier work with this tool is that utilizing it in the context of palliative care increased the number of palliative care referrals, and also the referrals to hospitals within this practice. This is a large community-based practice, and we also have similar data actually from another practice that has piloted this tool.

But what we found in the current work is about the timeliness. So, we're looking at, okay, well the referrals increased but were they timely referrals? So, what we really found was that, and we chose because this is a real-world study, real world data. So, this is a retrospective look. So, all patients who unfortunately had a mortality event at this practice in the given timeframe, we reviewed their records to see if these patients received a palliative and hospice care referral. So, we did a before and after analysis. So, we did it a before analysis for the time period before the tool was implemented, before the Javion CORE was launched at the practice. And then we did a post period after the Javion CORE launch to see if there was a difference. We know we would like for patients to be referred as early as possible to palliative care, so we chose a 90- to 180-day window. So, again, it could even be longer, but we just have a time point.

So, we noticed that there was almost a 2-fold increase in patients being referred to palliative care in that window. And similarly, we use the shorter window with hospice, to assess if patients were being referred 14 days prior to their mortality to hospice. And we basically quantified it at various time points. And, of course, that's in the poster, I would have to probably look at those numbers quickly. So, the 90- to 180-day window for palliative care was doubled, like I said, [with an] increase by 111%. And then for hospice, actually, the 14- to 90-day pre-mortality referrals increased also by 179%. So, almost a tripling of those referrals in a timely manner. The other important piece was that before the launch of this tool in this practice, 86% of patients were not getting a palliative care referral at all. And so, after the tool was launched, that number decreased to 72%. Of course, you may argue there's still a lot of patients not receiving palliative care. But again, you know, we have the whole pandemic and all of that in this time period. But the bottom line is that it was a movement in the appropriate direction. And similarly, prior to the deployment of this tool, 83% of folks were not receiving a former hospice referral—that doesn't mean that they weren't getting hospice. And this number dropped to almost 52%, so a dramatic decrease in patients who were not being formally referred to hospice, and then we saw in the longer time points, a lot of patients being referred earlier in their course. So, that's the takeaway: Patients started to be referred earlier in the course of palliative care, earlier in the course of hospice, because the bottom line and the takeaway is, you know, the earlier in the course that we can do that, the better I think it is for patient outcomesand for quality of life.

Q: Is there anything you would like to add?

Ajeet Gajra, MD, MBBS, FACP: You know, I think this is a valuable tool. It's already shown utility in a community oncology setting and, you know, certainly larger studies would generate more data and greater confidence in its utility across various practice types.

Related Videos
pharmacy oncology, Image Credit: © Konstantin Yuganov -
Pharmacist holding medicine box in pharmacy drugstore. | Image Credit: I Viewfinder -
Pharmacy Drugstore Checkout Cashier Counter | Image Credit: Gorodenkoff -
Mayo Clinic oncology pharmacy
Therapy session -- Image credit: pressmaster |
Testicular cancer and prostate cancer concept. | Image Credit: kenchiro168 -
Medicine tablets on counting tray with counting spatula at pharmacy | Image Credit: sutlafk -
© 2024 MJH Life Sciences

All rights reserved.