Incorporate Byline Order with airSlate SignNow

Byline zip

SREE VUTHALURU: Hello, everyone. Welcome to the "Analytics to Fight Opioid Abuse" use case at the SAS Global Forum 2020. My name is Sree Vuthaluru, and I am a systems engineer supporting our government business unit at SAS. I'm excited to talk to you all today and show you a demonstration of our analytical capabilities that have been tailored to help fight the opioid epidemic. Before we go into the demo, I will pass it to Dr. Steve Kearney to give us an overview of the problem at hand and the approach SAS has taken to help fight opioid abuse. Dr. Steve Kearney. STEVE KEARNEY: Thank you, Sree. Hi, everyone. I'm Steve Kearney, and I have the privilege to work with an exceptional team here at SAS and with all of you to solve some of the world's most challenging health care, life sciences, and government problems. Sree is one of those brilliant SAS colleagues that works tirelessly to make the world around her better every day, and I look forward to her demonstration of how to use analytics to fight against opioid abuse. As we deal with the current COVID-19 pandemic, there is also an epidemic that continues to rage across the US and other parts of the world, the opioid epidemic. At SAS, we have been working with all the key stakeholders in this fight for years. Traditional or base SAS is used by every department of health and federal agency involved in this daunting problem, but many times it is with incomplete, latent data that is often in silos. In recent years, we are proud of the innovative solutions and partnerships that we've developed in this battle and we had signs of improvement in 2019. However, many are concerned that the current state of our health system will erode the great strides that were made. The numbers before the 2019 improvements look something like this. More than 12,000 people were treated in ERs every day for opioid abuse and opioid overdoses. Now with COVID-19, those patients are struggling to access treatment and many that were doing well now feel isolated and without resources. Also, more than 130 people died every day in 2018 from an opioid overdose, and it was estimated it could have been as high as 180 people per day. There is grave concern now what the new trends look like. There was a decrease in prescription opioid deaths, but unfortunately, there was a 10% increase in synthetic opioid deaths, like fentanyl. The majority of those that were using fentanyl received it through their transition from prescription drugs to heroin that now has become laced with fentanyl. The SAS approach promotes interoperability and combining data sources. It allows for a common data language in an analytic environment that promotes data-driven decision-making. Through the use of visualization, network analysis, risk scoring, and real-time alerts, this analytic environment utilizes AI and machine learning, along with the ability to develop or compare any model including open source, to turn numerous disparate data sources through APIs and advanced data management into actionable data that can empower law enforcement and health officials to get the services to the people real time and most importantly, save lives. Now, I will turn it over to Sree for our demonstration. SREE VUTHALURU: Thank you, Steve. Let's jump into the demo. This demo uses some publicly-available data, along with some demo data, and represents information that may be collected by law enforcement and prescription-drug-monitoring programs. As Steve mentioned, combining data from multiple agencies helps us gain the most insight in tackling opioid abuse. I've already cleansed the data and prepared it in order to get it to the dashboard format you're seeing on the screen today. I've done this within the same SAS platform. And some of the steps I've taken include standardizing addresses and names, joining disparate data sets together, and removing duplicates across my data sets. We'll start by looking at an overview of the city, along with some key values. Here, you can see the number of fatal overdoses by county. I also have the number of drug-related arrests for the entire state and naloxone administrations. The line chart below shows me fentanyl-related deaths in the yellow line and oxycodone-related deaths in the orange line. Let's jump into the next page for more details. This Dashboard page is split in two. On the left-hand side, we'll have more law-enforcement-focused insights, and on the right-hand side, more health and PDMP-revolved insights. In the middle, we can see the same map broken down by county for fatal overdoses, where Essex County has the highest. The next map underneath shows naloxone deployments, where Camden has the highest. And the last two maps are drug possession arrests and drug distribution arrests, again, at a county format. The purpose behind both of these maps is to better understand if the areas where individuals are purchasing drugs are different than those where they're using drugs. Now, let's jump to the left-hand side. Here, we have a table where we've calculated risk of drug overdose at an individual level. Now as you can imagine, this consists of highly-sensitive information or PII. We would recommend and enable role-based access to allow only certain users or groups of users the ability to view and interact with this sensitive information. This can be done throughout the platform at a data-set level or even at a report level. The specific risk score here takes into account a few different metrics ranging from the number of drug offenses, previous convictions, total MME the individual has been prescribed, as well as many others. Now, the risk score can be altered to add other metrics or take out certain variables depending on what is most relevant for that specific state, municipality, or country. The model to create these risk scores is created within SAS and can be automatized to output results as the data is updated. Let's move to the byline chart below. I have two byline charts here, one for strict count and one that takes into account the population, and is standardized per 100,000 people. Looking at this chart, I can see the orange line is the number of drug-related arrests per 100,000 and the blue bars are the number of fatal overdoses. As you can see in Atlantic County, there are many more overdose deaths than arrests. The opposite is true for some other counties, like Monmouth County here. This small insight helps policymakers better tailor their approaches for different parts of the state. One of the counties may need more treatment centers and naloxone kits, whereas the other may need more law enforcement. Now on the right-hand side, we've once again risk scored, but now for providers. And we've put them into five different risk groups, which are the bars. The yellow line shows the risk score. The risk score here can take into account metrics such as the number of patients that have overdosed under the provider's care and the number of MMEs the provider has prescribed as compared to others in their specialty. Again, this is something that is highly customizable for any region, and we would work with local domain experts and incorporate their input into this model. The risk scores so far have been at a provider or individual basis. However, identifying hotspots and knowing where to focus efforts, especially with low resources, is another key insight. Let's go back to the map and focus on Essex County and drill into a zip-code-level view. Here, I can see the top few zip codes by both risk score as well as fatal overdoses and fatal overdoses per one million. So far, everything we've looked at is from a high-level situational awareness perspective. This approach can be taken for not only this opioid use case, but many others-- for instance, the current COVID-19 pandemic. It's important to understand where the problem is. However, this will take us only so far. So let's go back and investigate further at an individual perspective and see what actions we can take on this data. I'm back on my main dashboard page. And focusing on this table here, I can see an individual with an obviously high risk score. I'll double-click to take me into the individual scorecard page. Now, the scorecard gives me a lot of information about this individual, including their previous offenses, whether or not they went to jail, and if they were drug dependent at the time. Again, this is a page with very sensitive information, and we would assure that only certain individuals have access to it. Now, I would like to know more about John Doe, so I'm going to jump into our investigative piece of the platform by double-clicking and linking to Visual Investigator. The investigative side of this platform will allow me to investigate John Doe and any involvements that he's had, as listed down here. This takes into account multiple data sources and multiple data entries. However, this is all done through entity resolution behind the scenes, so I'm not actually making any connections. It looks like there's an open burglary investigation. So I'll go ahead and open that involvement and see some basic information about this entity. Here, I can see that the dwelling was burgled and prescription drugs were taken from the medicine cabinet. This happened back in March of 2019. I'd like a better picture about John Doe, so let me go back and open up a network view in a new analyst project. The network view will give me a better understanding of all of the entities that John Doe is connected with. Although these entities may come from disparate data sources, SAS is able to connect them. I'll go ahead and expand these links one level. I'm not manually drawing any of the lines here. Again, it's the entity resolution done in the backend. Here, I can see that John Doe is connected to multiple prescriptions, some arrests, and some naloxone deployments. I'd like to know more about the Vicodin and oxycodone prescriptions, so I'll expand these nodes one level. I'll start with the Vicodin. Doing so, I can see that all three prescriptions come from the same doctor and same pharmacy. In other cases, we have seen multiple prescriptions from multiple doctors, which has been indicative of doctor shopping. While investigating providers, we found cases where prescriptions were always sent to a specific pharmacy. This was an initial clue to understanding that the doctor was colluding with the pharmacy and further investigation was needed. In a similar way, alerts on high-risk-scored pharmacies can help uncover pill mills through similar types of investigations. Back to John Doe, I'm seeing many other types of entities, including arrests and naloxone administrations. I'm curious and want to know more about the timeline here, so let's take a look at that by opening the time slider. I'll start at the beginning, which is back in December of 2018. As I scroll through this timeline, we'll see the entities pop up in chronological order. So we see that first was the Vicodin prescription, followed by oxycodone, and a second oxycodone script the next month. Continuing to scroll, I'll start to see in March that there was a possession of prescription opioids arrest for John Doe. Continuing to scroll, I see that burglary entry, where he was a suspect, as well as a loitering charge, and then a naloxone deployment followed by an assault followed by a second naloxone deployment. I'll continue scrolling here, and we'll see that there was a third naloxone deployment just a few days later, and finally, a drugs overdose charge, and he has passed away according to the toxicology report. Now, all of these insights and any notes can be saved as part of an ongoing investigation. The platform itself allows for alerts on different entities, perhaps in this case, risk scores on patients, providers, and prescribers, as well as assigning tasks in a workflow to claim, investigate, and escalate cases. Throughout this demo, we've tried to give you a high-level illustration into one of the many ways SAS is using its technology to better understand the opioid epidemic, as well as make data-driven decisions and actions through investigations. Thank you for watching and please connect with Dr. Steve Kearney and myself on LinkedIn. And feel free to email us with any questions.