Breathe in … breathe out … let the data flow

How many of us in compliance sit mired in email, policies, audit and investigation reports, and a myriad of other data? It’s not that we lack enough data points, it’s trying to decide which ones are valuable. We have disparate systems that don’t talk to each other and any number of structured and unstructured data sets.  Layer all that in with the June 1 announcement by the Department of Justice that data analytics is a critical plank in any effective compliance program, and we are definitely feeling the pressure.

This is one exercise where taking a monumental step back to see the proverbial forest rather than the trees can be an extraordinarily helpful exercise. The first step is to determine which are the most critical controls in our compliance framework, and how do we test that they are operating effectively? This is so important, because we will often see a client who has built beautiful dashboards that provide information to little or no effect. Each company is unique, and each needs to go through the exercise of determining which are the vital few as opposed to the useful many.

In this article, we will walk through a fictitious fact pattern and use it to concretely demonstrate how to integrate data analytics and process flows into a program.

Clinvio – Project Pericles

“Clinvio” is a global pharmaceutical company that manufactures products in several competitive therapeutic classes, including cardiovascular, cancer and neurology, that are managed through three separate and corresponding divisions. They have global revenue of approximately US$3 billion and a market cap of US$19 billion. Clinvio sells its products across various global markets, including North America, EMEA and Asia Pacific, although the indications for certain products may vary across those markets, and some are sold in certain markets in conjunction with other partners. Some older products face generic competition, and most face some kind of therapeutic competition. Clinvio’s cancer pipeline contains some particularly promising candidates, including one in Phase III.

Clinvio started out as a single-product cardiovascular company. The company has grown through seven acquisitions in the last five years. Some of these purchases were of entire entities, and others were structured as asset purchase agreements to facilitate pipeline growth. As a result, Clinvio grew from a US$300 million biotech company in 2014, primarily in the United States, to a now-US$6 billion multinational. Its largest acquisition was of Neurocaple out of Singapore in December 2019, which effectively doubled the size of the company from US$1.5 billion to US$3 billion in revenue. 

In each of the seven transactions, the company decided to allow the resulting entity to operate at least partially autonomously; as a result, there was no integration of the financial or other systems, except in rare instances. In order to report out financial performance, there is a combination of automated pulls of data as well as manual integration of spreadsheets.

The acquisition of Neurocaple resulted in significant integration issues, as they had an existing compliance program and personnel that the CCO has been asked to integrate but with an admonition that it should be “light touch” and not interrupt the business, as the acquisition white paper called for significant growth and returns immediately to justify the purchase. 

On January 10, a hotline call comes in that alerts the organization that it has a problem with its advisory boards. Specifically, the allegation is that ad boards are being used to “reward” the top prescribing physicians. After doing some preliminary interviews, it is determined that Clinvio definitely has exposure on this front. 

Over the course of the next several months, Clinvio’s compliance team conducts an internal investigation that ultimately leads them to make a disclosure to DOJ and HHS, which results in a substantial fine and a Corporate Integrity Agreement.

Now that we have a fact pattern to work with, let us walk through all the points where data comes into play.

Viewing the landscape

When a company has grown through acquisition as Clinvio has, it is not uncommon to end up with a Frankenstein’s monster of a computer network topology map. See, Figure 1. This picture is actually not that bad in the sense that at least all the nodes are connected. Blue represents Clinvio, and red is all the acquisitions.  It is not a pretty picture, but it would not be uncommon to see groups of nodes that are not connected to the nucleus at all. In that case, it would require manual transference of data, which often results in data leakage where mission critical data is lost, unusable or difficult to access.

Figure 1

Now that we understand that we have a fragmented computer network, it is now incumbent upon us to find out what the data sources are and how are they viewed. Structured data is anything that can be displayed in rows, columns and relational databases. Gartner estimates that 20 percent of the enterprise data fits this definition. Unstructured data is everything else. For example, unstructured data would be emails, Word documents, images, audio, video, etc. The role of “Big Data” is to bring the two together. See, Figure 2.

Figure 2

Although in Clinvio’s case they are chasing the data under time constraints, the overall goals should be to set up a system that regularly monitors and tracks the data. An example of such architecture is shown in Figure 3. What the picture is trying to show is how to transform the data using SQL servers to a point where the data can be interrogated. SQL stands for Structured Query Language. A SQL server is a relational database management system. 

What this requires is taking the unstructured data and organizing it in some way so that it can be stored in rows and columns. For example, this can be accomplished by scoring emails for relevance to specific search terms. The score would be in the relational database and, if relevant, it is possible to trace back to the original unstructured data to see how it is impacting the results.

Figure 3

However, data visualization is not just ones and zeros. Data visualization can take the form of process maps. In Clinvio’s case, we know there is a problem with how advisory boards are approved. We learn that each one seems to be okay, but when looking at all of them in the aggregate, there are doctors that are attending too many, and there were no controls to detect the issue. 

Let us first examine current state. See, Figure 4.

Figure 4

As can be seen, there are no controls in place to determine if an HCP has participated in multiple advisory boards or speaker events. In an investigation like Clinvio’s, it is critical to stop the bleeding as soon as possible. The facts still need to be investigated, but the current process can be modified immediately. 

Often when just reading a policy, it is difficult to see where the gap occurs. The process map makes it clear how things move through the system. In this case, preventative controls are inserted in four places to enhance the controls architecture. See the orange shapes in Figure 5

Figure 5

By the same token, when managing any large initiative, project management is essential. This is a type of data management in that it is important to understand what is being done when and the results they will deliver. By taking the time to plan out all the workstreams it can be determined where the gaps in resources are, from both a human and capital point of view.

To be clear, in compliance, we almost always use Agile project management. Traditional project management is used when there is a defined endpoint.  For example, if one were building a home there is a defined endpoint – a finished home. It may be hugely complex when considering all the choices to be made, but the endpoint is still relatively fixed.

In contrast, Agile project management is an iterative approach to managing a project throughout its life cycle. In a case like Clinvio’s, we do not know what the result will be. Will they disclose the issue to the government? How many processes need to be examined and changed? We just do not know. Therefore, the iterative flexibility is key.

The project plan is also strategic and has multiple purposes. By scoping the project, we can explain it to the project participants and stakeholders: what has occurred, what is planned, and what are the remaining unknowns? In this way, the same plan can be repurposed for different audiences. It becomes an important messaging vehicle, particularly we later need to explain to the government what was done, when and for what purpose. 

In Figure 6, we illustrate a simple Gantt chart and project plan for the document review element of the investigation. The red line represents the “Critical Path.”  This tells us those projects that have dependencies and cannot slip if the project is to be delivered on time.

Figure 6

Finally, it is time to get to the real data and answer the primary question: was the advisory program being misused, and could that have resulted in kickbacks to doctors? As seen in Figure 7, we have all the data. Now, if one is a data maven, the story is in the spreadsheet and is right there for all to see. However, if one is a data novice, a picture would be helpful.

Figure 7

Transforming data into compelling pictures has never been easier. The two most popular data visualization programs are Tableau and Microsoft Power BI. For Clinvio, we used Microsoft Power BI. As with anything one has not done before, it can be scary and intimidating to wander into a new computer program. We swear that the water is fine in the data lake, and you should take the plunge.

With one click, the Excel spreadsheet in Figure 7 can be imported into Power BI. Once it’s there, pretty pictures abound. Essentially, you pick the graphic that you want, eg, pie chart, bar graph, map, etc. You then select which column in your data will sit on which axis. That’s it. The program does the rest and provides the picture.

What becomes dangerous is, just because data can be displayed a certain pretty way, it does not mean that it is appropriate. Think through what the real question is – was Clinvio’s ad program being abused, and was it tied to how much a doctor was prescribing? Let’s walk through Figure 8 and see what it’s telling us.

Figure 8

We will start with the picture of the map and rotate clockwise. The picture of the map is very impressive. Who doesn’t like a geographical representation?  However, it tells us nothing. Yes, we can see where the doctors are from, and the size of the bubble corresponds to the amount of spend, but so what?

The second picture shows us the amount of sales by the doctors, and we can see that Dr. Alex Karev sold a lot. Again, it does not answer the question directly.  We do note that these five named doctors are well above the norm of other doctors in their region. This is not dispositive though. These five could be world-renowned doctors and therefore treat the sickest patients. They would consequently prescribe more than most.

The third picture shows the amount of product each of them sold and breaks it out. This is an interesting question, in that we want to see if there is a big difference between these two blockbusters. Surprisingly, it is pretty consistent, and we are not that concerned.

Now we come to the fourth picture. This one is definitely interesting. We would want to break down the years and look at the individual doctors, but the spend line directly corresponds to the sales line. It looks like Clinvio may have a problem.

Things get worse when we look at the final picture. The Spend and Sales lines, when viewed over the years, track identically. Even the little dip in Sales has the identical dip in Spend. Clinvio definitely has a problem.

There are plenty of other graphics that will show some serious issues. In particular, if looking at the spreadsheet and Dr. Zhivago’s sales in 2018 are significantly lower, then in 2019 it looks like he is penalized, and the amount the company uses him also drops. This does not look like it will end well.

All of this is shown in order to drive home the point that just because the data can be asked a question, it doesn’t mean it is a good one. There is a skill to interrogating the data while always keeping in mind the questions we are trying to answer.

Figure 9 gives us a sample of what good data starts to look like and how it can progress the investigation. In the top left picture, we see an anomaly in the data. Dr. Zhivago was a top paid doctor by Clinvio, but then it tailed off significantly in 2019.

Proceeding in a clockwise fashion, in 2018 we see that the amount the Dr. Zhivago prescribed dipped significantly in 2018.

Figure 9

In the next chart we see the number of events Dr. Zhivago did in 2018 was high, but then dropped off in 2019. And finally, in the last picture we see the two plotted against each other. When his sales tailed off, it looks like he was “punished” by having fewer events.

This is not an open and shut case. It is necessary to do additional analytics and ask for the story of what occurred here and if there is a possible reasonable explanation. It certainly looks very bad, but more importantly the data picture gives you permission to ask the question and further interrogate the data to see if there are additional instances that mirror this pattern.

Key takeaways

This is, obviously, not even the tip of the iceberg, but what we hope it drives home is that conducting an email search and some interviews is no longer sufficient. It is critical that compliance understand the data architecture and leverages it to support the controls work that they are tasked with administrating.

The tips for succeeding on this data journey are as follows:

• Develop a data strategy for your compliance program. If you do not, the next plaintiff or government action will design it for you
• Determine your critical controls and how to effectively connect disparate systems that will provide “answers”
• Electronic database of control evidence is a “must have”
• Project management is critical for every aspect of your program
• Make data fluency a key objective for each member of your team