 Hello everyone, thank you so much to the Our Medicine Conference for having me and thank you all for being here. I'm Asma Toomey and I'm the Director of Analytics and Research at PursuitCare. Today I want to talk to you about how we're leveraging open source and professional tools to deliver care to patients actively suffering from mental health conditions or in recovery. For over two decades now we've faced worsening rates of mental illness which have had devastating medical, social and economic consequences at the individual and the societal level. We're in the middle of an unrelenting wave of drug overdose deaths now, the majority due to the misuse and addiction to a class of drugs called opioids. In fact, since 1999, opioid drug overdose deaths have quadrupled and over half a million have died. To put this into perspective, that's 136 people dying every day from an opioid overdose and one life every five minutes. This has only gotten worse during the ongoing COVID-19 pandemic. After an already catastrophic increase in 2020, deaths have actually risen to a new record in the year of 2021. Mortality data from the National Vital Statistics System show that we've gone from a new record last April of 75,000 to now 108,000 in the last year. The rise in overdose deaths can be seen across the country. With some states experiencing as much as 50, 60, 70% increases from last year. COVID-19 disruptions alone cannot explain trends that were already there before and continue to worsen. The reasons for the steep increases are complex and vary in nature from region to region. It is true that the pandemic has exacerbated inequalities across racial, ethnic and socioeconomic class. It has also worsened access to care as many in-person clinics closed or significantly decreased operations. A recent study looking at county level differences have shown that among 3,200 counties in America, almost 80% have no opioid treatment programs and nearly 30% have no providers that can dispense life-saving medications for opioid misuse disorders. These disparities in opioid treatment access and opioid medications are also seen by gender and race and ethnicity. At the heart of what we do at Pursuit Care is eliminating every barrier that we can. This is why we leverage telehealth technology to meet patients truly wherever they are. It allows us to meet patients who do not have adequate local resources and who need treatment fast. The flexibility of telehealth technology not only allows for same-day treatment but it also allows us to implement functionality through our app that keeps our patients engaged in their treatment plan and their recovery. We also offer in-person care in select locations. So in-person and virtually, we're able to treat all substance use disorders and other mental illnesses. We deliver our care primarily using our application, which has support for the dispensing of medication assisted treatment, digital therapeutics, individual and group therapy, psychiatric care, and medications from our own pharmacy. Our goals, broadly speaking in the data science team, starts with detecting trends which then inform outcomes which are monitored and given those results procedural changes are established and enacted and those processes are also monitored with the lens of whether those processes have made a significant difference in the outcomes that we care about. First, starting with detecting trends, our internal tooling is built entirely to fulfill that goal and others. Being a startup in which it was just me a year ago, I had to make sure that our backend and our data engineers that were brought on later built it in such a way to optimize for efficiency, automation, and reproducibility of our outputs for several reasons, namely reducing overhead costs and using our team and resources in the most efficient and productive way possible. With this in mind, I looked to some inspiration from others who've had success building a truly data-driven organization. Emily Readerer from Capital One had a fantastic blog post about the importance of investing in internal tools. The gist of it is that there may be a lot of public packages out there that are built to fit as many use cases as possible, but they may not be the best suited for your specific use case. In doing so, significant advantages can be gained like improving quote quality and increasing the knowledge base of your company. In my first year, we've focused on building our backend primarily as naturally the first step in any data science department is to obtain data and structure it in a way that makes analysis possible and as streamlined as possible. We've built packages that streamline database connections and querying, ETL processes and API requests. This was crucial for our organization to do, given how our own health care data was fragmented and distributed across different platforms and different vendors, such as our electronic health record and our laboratory data and our telehealth app, all of which have their own way of structuring data. So it was important to consolidate it all together to get a full picture of our business. Our next consideration was taking this proprietary patient data and augmenting it with summarized public data. So with a few lines of code, we can marry our patient data with insightful spatial demographic and socioeconomic data from the American Census Bureau, the CDC and others. With a data warehouse housing these highly structured data from our patients, application and public data were able to detect trends spanning different outcomes of interest, such as demographics, socioeconomic indicators, healthcare resource utilization, clinical characteristics and treatment plan journey. We deliver these insights using an analytics publishing platform called RStudioConnect, soon to be known as PositConnect, to the different stakeholders in the company, such as providers and executives, who are then able to guide decision making in the company. This continual monitoring of trends also allows them to weigh in on what our important outcomes or key performance indicators are and how they ought to be measured. Our studio Connect as a deployment environment allows us for the deployment of highly customizable lightweight and portable reports, all in a version controlled manner. Stakeholders and the data science team work together to establish and monitor outcomes that are important for the delivery and the improvement of patient care. The four main outcomes we monitor involve engagement of the patient with our product and our care, their retention in our care, their drug use or lack thereof, and adherence to the treatment plan set out by the providers. While some outcomes are derived from simple counting and grouping, the majority do involve some additional lift due to the nature of the question and the complexity of the data. The tidyverse, coupled with modeling packages like tidy models in BRMS, allow us to tackle the analysis and modeling of longitudinal, multi-level data. Our knowledge base, built through the detection and monitoring of outcomes, allows us to then establish and implement process changes that are either operational in nature to improve our internal processes or to help personalize the treatment of patients based on their acuity and their desired goals for recovery. And with that, we've closed our loop here for the data science team. With this loop, we're able to fulfill our main missions, which are early detection, continuous improvement and hopefully better outcomes. Thank you so much to our medicine for having me and thank you all for listening. If you have any questions or comments, please reach out on my email or on Twitter. Thank you.