COVID-19 analytics: Back to the future

Nirav Dalal:

Hello, and thanks for joining us at the State Government Solutions Track Focused on COVID analytics. My name is Nirav Dalal, and I'm a vice president and regional general manager. I'm joined today by two thought leaders, Jayce Johnson, and Tanuj Patel. Collectively, we have over 40 years of experience in data warehousing and analytics, and more than 50 years of Medicaid experience. Our focus today is going to be on COVID analytics, and specifically how we can take advantage of these challenging times to make the type of progress that we've been pushing for. And since we all wish we had a time machine, we'll some of the memories of a great '80s movie, Back To The Future.

Nirav Dalal:

Marty McFly is letting us borrow his DeLorean for this meeting. So we're going to use it as a time machine and move through the space time continuum to see how the environment is ripe to make some real advancements in analytics. Doc Brown has the DeLorean ready for us. So let's get in and see if we can get this thing moving to 88 miles an hour.

Nirav Dalal:

And we've arrived in November, 2019. Remember November? Let's talk about what the environment looked like for analytics at that time. The following were some of the roadblocks that really prevented us from building some models to drive population health. They included silos of data between agencies that were tough to navigate, few governors were engaged in driving data sharing, social determinants of health analysis was a consistent theme, but resources were tough to come by. And the focus was primarily on driving down constantly services and impacting the houseless population. This was just the tip of the iceberg.

Nirav Dalal:

Even with those roadblocks, we aspired to push organizations to use new data and in new ways to impact our communities. Okay, I've got the DeLorean filled with plutonium and the flux capacitor is, well, fluxing. Let's jump in and move forward in time to see where we are today.

Nirav Dalal:

The images on this slide depict how much our lives have changed due to COVID. In-person work, sporting events, concerts, and shopping malls are not the current norm. Instead, people are using this opportunity to explore new activities that they previously didn't have time for. One bike manufacturers cited that they sold more bikes in the month of June than they did in all of 2019. Sourdough bread has suddenly become a thing. So the question is, how can we use this time and this current situation to address the challenges that we previously faced in bringing data together to drive real insights.

Nirav Dalal:

The current state is that there is tremendous support from the governor's office for data to shared, and for analytic models to be built rapidly. In addition, the health disparities have driven more and more social determinants of health analysis as it pertains to COVID-19 spread, as well as the treatments for those infected. With massive populations of data hounds that are clamoring for data, and trusted data, this has really shined a light on the issue of bringing lab data into the Department Of Health and getting it ready for analysis. There has never been a better time to strike while the iron was hot to create an environment where data is shared to drive actions. I will now turn over the keys to Jayce Johnson, to discuss what we recommend you do with this data to facilitate better insights. Jayce.

Jayce C. Johnson:

Thanks, Nirav. Hi, my name is Jayce Johnson. I have the pleasure of leading our state data warehouse and analytics practice for Optum State Government Solutions. I've been doing this line of work for over 20 years, working with over 30 States to bring solutions to their data and analytic needs. Before we get started, I'd like to talk about where we've been. And in some cases, in most cases, still are in terms of data availability to support the analytic needs during this pandemic response.

Jayce C. Johnson:

First off our US healthcare system data is fragmented and siloed, as Nirav mentioned earlier. There's a lack of access to comprehensive payer data, health plans have their patient encounter data specific to their own plans. Medicare has data separate from Medicaid, chip and Medicaid expansion data, qualified health insurance plans have their own information stored separately. Vital statistic information is collected by States Departments Of Health, including causes of death and death certificates that talk about the reason people died in certain locations. They store that in separate locations.

Jayce C. Johnson:

And coronavirus test results come from many different lab sources, consolidated yet into separate locations; and that's just the start. Even having that data in separate locations, some data is not current or timely. Providers need to submit their claims data in some cases, some period of time, seven, 10, 30, 90 days, dependent from the time of service, dependent on the payer. And the data is then consolidated at the payer level. Test results are sent to yet another place, consolidated in yet another. There's confusion on the test result, data reporting requirements. The CDC has requested certain data sets over time for analysis of COVID-19 test results. The format and content has changed over time. And The Department Of Health And Human Services has also stepped in to receive information separately.

Jayce C. Johnson:

Data does not always include healthcare encounter claims information, especially when we were talking about test results that could be used for simulations, models, to predict the impact of COVID-19 conditions on specific individuals. This includes derived or enriched group data, or data that identifies those patients at highest risk, and those with existing pre-existing conditions, existing disease conditions that contribute to adverse effects with COVID-19.

Jayce C. Johnson:

And then we have the privacy and security concerns that need to be addressed. People are oftentimes adverse to providing their location information, where they've been, with whom they've been in contact for how long. Contact tracing information via cell phone triangulation or credit card transactions are basically off the table in the United States due to privacy concerns. Laws sometimes protect individuals from sharing such information without mandates at the state and federal level. And then for lab test data, the submission standards have not always been adhered to. Large labs do adhere to standardize submissions, but smaller testing labs submit data as fast as they can, in whatever format they have available to them. There's a lack of testing results altogether. And this is the main piece of information that's needed in this pandemic response.

Jayce C. Johnson:

So testing and test results are pivotal to the COVID-19 analytics. To quote Dr. Claire Stanley from Georgetown University back in April, "Testing is critically important for pandemic response as a tool for gathering information on who is infected, where, and when. Accurate estimates of the number of people infected within a population are the cornerstone of evidence-based public health decision-making during an epidemic, and can help hospitals, and other frontline services plan, prepare, and more effectively respond." Universal testing would be preferred with consolidated results, but short of that, we need to use the results that we have as a sample in which to extrapolate our specific analytics.

Jayce C. Johnson:

By combining traditional demographic information, with new COVID-19 related source information, we're able to provide analysis, model risk, and stimulate or predict what will come next. With looking at individuals comorbidities or pre-existing conditions, age and gender, the type of job that a person possesses, living conditions, and other demographic information, we can make simple predictions. By combining that with test results, COVID-19 test results, and symptom information to determine infectiousness, proximity and exposure time with others, from contact tracing data acquisition. And even self-reported behavioral information like mask wearing compliance, we can increase the capabilities and accuracies of our statistical models. Our hope, all of our hope, is that the COVID-19 pandemic is short-lived. Even if so of though, this data and this process knowledge can help with our next contagious outbreak, what can follow, what will follow and our analytic response at that time.

Jayce C. Johnson:

So How can we address these obstacles and create an environment to support COVID-19 analytics? Well, to start, we suggest establishing a cloud-based data lake or data warehouse using cloud native technologies, as often as possible, to reduce costs and deployment times. Or, where data that already exists well formatted, and existing sources, we can use a federated or virtual database approach. Why? Well, these approaches tend to be less costly to stand up. And we all hope that this analytic need will be short-lived. And the speed to put the analytic environment in place is much, much quicker than to build a centralized repository.

Jayce C. Johnson:

Or, if the data already exists in different sources, putting together a solution to access the data in a federated manner or virtual environment, directly at the source of where the data exists using APIs, is also preferred. But we also need to partner with the data providers, all of the silo data that exists today needs to be accessed, and it needs to be granted and allowed for the specific purpose of saving lives. We could use demographic information from one or several, all payer claims, databases, APCDs, that exist in some States today. That would be a jumpstart.

Jayce C. Johnson:

But on top of that, we can use claims grouping technology to take the administrative claim's information, group it together to find out disease conditions, and to quickly find out those who are most at risk, if contracting COVID-19. And to quickly code for the comorbid conditions, the comorbid comparison groups, that have the existing contributing conditions, that have related severity to comorbid situation of contracting COVID-19. We can use contact tracing information when available to add to the richness of such analytics, giving us the data input to get in the way to avoid transmission, but the data must be timely. We can consider integrating longterm care data to predict downstream impacts of COVID-19 on our healthcare system.

Jayce C. Johnson:

For example, what we know today, dementia leads to people not following self protective protocols, such as mask wearing adherence or personal hygiene. People that have renal disease that can impact what sort of potential treatments we would use in treating COVID-19. And then other existing conditions such as blood clotting, are more likely to lead to adverse outcomes if we don't know that people have those conditions upfront.

Jayce C. Johnson:

And as Nirav mentioned earlier, we can combine the data with social determinants of health information from people's job or living conditions. We know that they have a correlation to outbreak advancement. And then, we can consider working directly with labs. Quest and LabCorp are the two biggest labs. We can get information directly from them instead of having them send the data to a clearing house, to get that data quicker and have it more timely. This all helps to create a data environment, and to perform the needed data preparation that is lacking today.

Jayce C. Johnson:

So with the access to this information, we can perform basic and starting analysis, answering questions such as how many members have been exposed to the COVID-19 virus? How many of the exposed members also have a high fever? How many of the exposed members have had a recent telehealth visit? How many members have been diagnosed with the coronavirus? Meaning they fit the clinical symptom criteria. Or how many of the diagnosed members have also had a lab test confirming COVID-19 infection?

Jayce C. Johnson:

How many of those have underlying conditions related to adverse outcomes? And where are they located? That's the basic information. But then we can use predictive analytics and machine learning to expand the analysis and dig even further down. Which geographic areas, zip codes, for example, have the highest prevalence of exposure or members showing severe symptoms today? Which cohort of members or retirees for our plan, are most at risk of contracting the COVID-19 virus? And where are they?

Jayce C. Johnson:

And let's not forget, with any novelty or change in the healthcare industry, there's an opportunity for those to defraud the healthcare system. We need analytics to ensure proper safeguards, like fraud and abuse detection algorithms, are in place to protect from things like telehealth abuses. Yes, we're seeing this. Providers submitting COVID-19 related telehealth visit increases that in some cases, yield unreasonable amounts of numbers of visits in a single day; that's abuse. That's fraud. And that's the vision.

Jayce C. Johnson:

This is just the start of the analytic possibilities that can be achieved with the integration of, or the access to, relevant data being available for COVID-19 analytics. Now, I'll hand out the presentation to Tanuj Patel, who will discuss actual analytic methods and the maturation of analytic possibilities. Tanuj.

Tanuj Patel:

Thank you, Jayce. Hello everyone. My name is Tanuj Patel. I'm an Analytics Product Director at Optum, where our primary focus is to collaborate with customers and help them discover actionable insight from data, to drive their organization strategic business decisions. I have been in healthcare industry for more than 20 years, serving in different capacities from providing direct patient care as a medical doctor, a clinical researcher, health informaticist, and most recently applied data analytics promoter.

Tanuj Patel:

In this part of presentation, I'm going to walk through some of the relevant analytical platform and capabilities, which may serve as your flux capacitor towards more holistic data center future. We are too familiar with the phrase, "Drowning in data versus [inaudible 00:14:30] for knowledge." Just like any other industry, we have been generating, collecting and storing all different sorts of data from [inaudible 00:14:38] system at a monumental pace. However, integrating such data and deriving strategic insight, is the key value proposition that analytical platforms provide.

Tanuj Patel:

Since the start of this pandemic, we have realized that in order to come out ahead, we need data sharing and data integration in a way that we never had before. As Nirav and Jayce mentioned earlier, systemically applying analytics across integrated data platform is becoming even more essential in order to improve health outcomes, gain operational efficiencies, and reduce cost of care. There are many opportunities to integrate and factor in data outside traditional health system walls, which hold great potential for improving overall health.

Tanuj Patel:

One said source can be social determinants of health data. Studies have shown that social and behavioral propensity account for 60 to 80% of health outcome and utilization. And during this acute times, it can be even amplified by understanding, integrating, and using social and behavioral data in an ethical, safe, and transparent way. Healthcare stakeholders can help change behaviors, gain insight to deliver more personalized care, that results in a better consumer experience, as well as improves health outcome.

Tanuj Patel:

With technological advancing and improving infrastructure, data driven organizations have been transitioning their data analytics capability to more of a self-service in real time, at point of care models. With organizations who are still in the process of gaining such maturity, the ongoing COVID crisis, along with the work from home models, is forcing them to quickly reassess and reform current capable.

Tanuj Patel:

This is where self-service analytics, or the platform that operates on, comes in. It plays a key role in analytics preparedness during such mercurial times where there is heavy emphasis on emerging data analysis, critical data driven decision-making, and et cetera. So here at Optum, we have been partnering with various healthcare agencies and organizations, to deliver and develop their self-service analytics capabilities, and discussing how we can collaborate, innovate, and focus on providing timely, yet value-based care.

Tanuj Patel:

What are some of the advantages of self-service analytic platform? So it makes data easily available at the fingertips of defined stakeholders, according to their individual roles. It eradicates barriers to access data, and makes it easier to derive insights. So users can do analysis when they want, how they want, and what they want. Simple to use BI tools, which are part of this platforms make creating, publishing, and sharing reports, and dashboard more streamlined. It allows for a faster, quicker, data discovery and insight.

Tanuj Patel:

This will help in a greater control, agility, and speed for the end users. This unprecedented times are great fit for flexible self-service analytics platform, when use cases are more or less nebulous, data is coming from all directions, and it's constantly changing. Such platform and capabilities, encourage, enable, and empowers community of users with analytical mindset to efficiently [inaudible 00:18:32] from data, to information, to knowledge. And such journey is no longer result for only data scientists, and other elite groups of professionals.

Tanuj Patel:

With such a single source of truth platform, we can foster collaborative business intelligence, and reporting environments. Where stakeholders are able to derive not only retrospective, but prospective insights, which we'll talk about shortly, on population characteristics, and identify members who are at high risk for a condition such as COVID-19. And if they are going to have an adverse event related to it. Using such platform, our customers are already identifying the members who are at high risk of catching coronavirus, as Jayce, mentioned.

Tanuj Patel:

With such a timely and shareable analytical insight, leaders are making evidence-based decisions, whether it's making policy changes, allocating resources, prioritizing testing, and other treatment facilities, developing care programs. Or even developing novel contact tracing, or telemedicine applications where they can streamline their operations. Such platform, in addition, can act as a repository for all analytical assets within agencies, and for futuristic actionable analytics. So data driven decisions are made at the right time, for the right people, for the right purpose.

Tanuj Patel:

As we gain more analytical maturity and travel further in the future, such self-service analytic platforms can be enhanced by integrating comprehensive sets of artificial intelligence, and predictive analytics capability, for more augmented intelligence, which can help further in resource planning, financial planning, closing the gaps in care. And most importantly, delivering preventative measures for better health outcome.

Tanuj Patel:

There are many examples that come in mind, but mainly we can apply machine learning techniques to speed up and automate data preparations, analysis insight discovery, which are the fundamental components that Jayce talked about. We have configured and built the pipelines in our analytical offerings, which take advantage of open source tools and packages that are available within global data science community. So the data scientists can reuse, refine, test, and deploy a variety of predictive models with ease, and they can be embedded within end user workflow for high value cases.

Tanuj Patel:

One example would be to incorporating models, which quantify and predict disease transmissibility for COVID-19, predict next COVID hotspot. And we can tie all those predictions with the health system capacity in surrounding area, for better crisis preparedness. We can even leverage social and behavioral data, as Jayce and Nirav are mentioning to study and predict impact of social isolation, and loneliness on the mental health of the population, in context of COVID, or outside the context of COVID. Which can even help us develop Strategies to prevent the development, or exacerbations of such mental disorders.

Tanuj Patel:

Healthcare model that considers social and behavioral factors in my opinion, holds potential to guide us towards more improved quality of life, and sustainable healthcare ecosystem. In another use case, Optum has a partnered with Medicaid health plan for analyzing their social and determinants health data, to address critical healthcare needs of the houseless population, whose vulnerable position has been linked to more utilization overall. We can also use natural language processes to speed up findings and insights from unstructured data; such as clinical notes, survey data, case investigation, and contact tracing conversations.

Tanuj Patel:

So the advantage of self-service platform is that it allows to deploy and operationalize such predictive models in a user workflow. Along with [inaudible 00:23:00] to business intelligence, to drive more value, and provide opportunities to take more preventative and proactive approach, rather than the reactive one. So with the right people, right processing, and technology, we can align any agency's analytical goal towards a platform and use it for their flux capacitors to travel towards data centric future.

Tanuj Patel:

In conclusion, COVID-19 has been making our data analytics trends more relevant and urgent than we had previously realized. This pandemic, and the next one, will keep producing data. And will require data integration and analytical preparedness to better understand fluctuating crisis, and meet population healthcare delivery needs. As a result, your organization's oral analytic strategy should be adaptable to right people, right processes, and right technology.

Tanuj Patel:

This is the opportunity to double down on updating data sharing agreements, data integration, and data governance practices, policies. And last but not least, rethinking the person that is around data and analytics. And analytics is going to be a journey. It's not a destination. And it's going to be your bridge between data and evidence, which will support your decisions. Self-service analytics can be a vital step in this journey from data silos, towards data sufficient analytics. Hopefully the analytical maturity and preparedness we developed during this crisis, will continue to be fruitful, even after the pandemic subsides. So with that Nirav, please take back the control for the DeLorean and show us the future. Thank you.

Nirav Dalal:

Thanks, Tanuj. Okay. It's now time to hop back in the DeLorean and get a sneak peek of what the future might look like, if we take advantage of the current situation to bring data together and fundamentally change the way that we use that data to make the decisions. And we've arrived, in 2022, the memory of the pandemic is still very fresh, and we still mourn the loss of life, and the impact of the wellbeing of our communities.

Nirav Dalal:

However, we are able to make real measurable strides towards models that address the social determinants of health. We have changed how decision-makers look at data, and we've created a culture of sharing data in a secured way. The appetite to consume data has shifted from a select few to the masses, thanks to the self-service tools. We hope that this quick trip through the past, present, and future, have provided you with some ideas of how to take advantage of these challenging times. Our team welcomes an opportunity to continue the conversation with you, and as Doc Brown stated, "Where we're going, we don't need any roads." So let's blaze the path together.