An Inside Look and What It Means for Your Growth Strategy
As you may know, Stratasan recently updated our Physician Patterns application with referral relationships based on full-year 2016 Medicare data. This long-awaited update means you will have complete access to the most current patient data, giving your team a major competitive edge over other physician relations outreach programs in your service area.
- Analysts who regularly work with physician data will enjoy the behind-the-scenes look this post will give into all that was involved in this data update.
- Those of you who focus on business development and strategic planning will be excited to discover how this data update can positively impact your growth strategy.
- Physician Relations Reps, data is the foundation upon which your entire outreach strategy lives or dies. You will appreciate how this update will give you the most current patient data available.
First, a little background: Physician Patterns is Stratasan’s tool for planning and tracking a physician relations strategy, mapping service line opportunities and accurately measuring inbound and outbound patient trends. The Physician Patterns application provides information that recruiters, physician liaisons, and business development professionals can use to have more informed conversations with their physicians, thereby building and strengthening those relationships.
Physician Patterns was originally built using data that the Centers for Medicare & Medicaid Services (CMS) published as part of a Freedom of Information Act release. That release, however, only went through the third quarter of 2015. In order to update Physician Patterns with the latest data, Stratasan purchased access to CMS’s full claims data warehouse. Known as the Virtual Research Data Center (VRDC), the platform allows users to access a remote desktop client, where they can perform analysis on Medicare data and export appropriately masked results.
How will this data update impact your physician relations growth strategy?
As we’ve shared in the past, using the best available data can provide a clear roadmap for health providers focused on strategic growth. Historically, the most current Medicare claims data has been unavailable to commercial parties and healthcare providers; only hospital claims data was accessible, Now, with this update of our Physician Patterns application, you will be able to access all sites of care: hospitals, physician offices, imaging, labs, emergency and urgent care data.
You will be able to:
- Gain access to the most current patient data: 4½ months post-close of the previous quarter
- Acquire referral volume data for physicians in your service area
- Understand the most important referral patterns in your market
Access to the VRDC: Is it open to the public?
Researchers and entrepreneurs may apply to ResDAC to gain access to the VRDC. The application process involves submitting a research protocol, signing a data usage agreement, and paying for a seat license per user per year.
While anyone with an approved research protocol and the funds can apply for access, as we will share in detail through this post, the process of working with the raw data to create referral files is very time-intensive. It also requires a high degree of proficiency with the SAS programming language. For this reason, we feel it is a better value for Stratasan's clients to allow us to handle the time-consuming data work on our end. This will leave you more time for engaging with the data in a way that keeps you focused on growth initiatives.
The intricacies of updating the Physician Patterns application data using the VRDC
The end-to-end process of exploring the data, writing the code, running the code, testing the results (and repeating, if necessary!), then outputting the results, took about three months in total.
You may ask, “what took so long?”
Once we officially received access to the data, a thorough background check process and federal security training was required. This was because the data in the VRDC is the full claims data, which may contain highly confidential patient information.
After passing security training and the background check, the data updating process could officially begin. Since the VRDC houses all manner of Medicare data, there were literally hundreds of tables to peruse, just to ascertain which ones were the ones we needed. After determining which claims tables we needed, the next job was to perform some exploratory data analysis. This involved answering questions like, “do the records need to be deduplicated?” and “what is the frequency of missing values?”
It is worth mentioning here that there are seven different types of tables that needed to be queried: inpatient, outpatient, skilled nursing facility, home health agency, hospice, non-institutional carrier base claims, and non-institutional carrier line claims. In performing the exploratory data analysis, it became clear that the structure of each of these tables was either slightly different or, in some cases, very different, from each other. For example, one table type might only have claim date, while another file type might have both claim date and visit date. As another example, one table type may only have attending physician, while another table type might have both attending and operating physicians. So, it was necessary to transform the claims tables in a way that made them stackable with each other. And, given that the CMS documentation from the FOIA request is somewhat limited, we had to make some methodological decisions regarding anything on which the CMS documentation was agnostic.
After that, it was time to write the code. At a high level, the code had to accomplish the following:
- First, use the raw claims data (which has one episode of care per record) to create a table of referrals (which has two episodes of care per record, no more than 60 days apart from each other, for the same patient).
- And second, take that table of referrals across all of 2016, and count the unique patients for each pair of physicians.
That second step is quite simple, but that first step is surprisingly cumbersome given the enormous size of the tables we were working with, relative to the limited computational resources on CMS’s virtual desktop client. Remember, since the data are highly confidential, we could not simply download the data and analyze it on our machines.
It is always a best practice to write code efficiently, but when you are working with tables as large as half a terabyte on a shared workspace environment of extremely limited space, it becomes truly critical. So, a lot of time was spent writing a macro that created a series of smaller tables (each for a given 60-day window in 2016) on which a Cartesian join was performed, to determine the physician-to-physician links for that given time frame. The macro then recursively did this for each progressive 60-day window (moving forward one day each time), and stacked and de-duplicated the resultant referral table to the referral table from the prior iteration of the macro. The necessity of this was to capture all possible relationships; if we had simply run the code on one month of claims at a time, for example, we would have failed to capture referrals between two months.
As mentioned above, the biggest hurdle in this process was the workspace constraints in the VRDC. As is common practice in many organizations, CMS has its VRDC users set up using a shared server for their SAS workspace. However, the claims tables are so large, and certain parts of the code are so resource intensive, that it was a common occurrence to have a job fail if too many other VRDC users were active at the time. In other words, even if Stratasan had been the only VRDC user, the code still would have taken about a week to run, but since we were far from the only VRDC user, there were times of peak user activity when we would bump up against the workspace limitations, causing the code to crash.
What this meant was that, while running the code, it was necessary to keep a very close eye on it, day and night, so as to lose as little time as possible re-starting it whenever it did crash. Throughout the entire process of working in the VRDC to create the new Physician Patterns referral file, I became very acquainted with the habit of being most active on nights and weekends, when the other VRDC users were (presumably) not logged in, so as to have as much of the shared workspace as possible.
In the end, it took a handful of versions of code to get the results we wanted to capture, and along the way, a great deal of care was taken to ensure that we were truly capturing the physician referral relationships we wanted to capture. After about eight days of running, the final version of the code finally finished, and we were able to export the results (masked to counts greater than ten, of course!) and update Physician Patterns.
Through the VRDC we now have access to not only institutional claims—everything that happens under a acute hospital setting—but also non-institutional claims, which are labs, imaging, ASC's, and more. This provides a much clearer picture of the spectrum of care that physicians provide to patients in your markets.
From a referral perspective, we can now provide an unfiltered look into Medicare claims to build physician strategies, uncover growth opportunities, and increase patients into your facilities.
Interested in gaining complete access to the most current patient data? Request a discovery call with Sean Conway and find out how Stratasan’s Physician Patterns can support your physician relations efforts and help you uncover valuable opportunities for growth within your market.
Article by Ashley Graham, Senior Data Analyst for Stratasan