My time at KPMG, LLP

What I learned in the world of Government Pharma.

I spent my time at KPMG learning the whole way through. I immersed myself in the world  of Government Pharmaceuticals which revealed the interesting way we receive the drugs that we use to treat our ailments. Specifically, the Government Rebate program for Medicaid and Medicare programs. So, before I tell you about what I did, let me tell you about this program.

When the poor or elderly need drug, the U.S. Government programs that cover healthcare costs normally covered by private insurance. For the manufacturer to gain access to the population that uses Government healthcare, they are required to participate in a drug rebate program. explains it here:

The Medicaid Drug Rebate Program (MDRP) is a program that includes Centers for Medicare & Medicaid Services (CMS), state Medicaid agencies, and participating drug manufacturers that helps to offset the Federal and state costs of most outpatient prescription drugs dispensed to Medicaid patients.

One of the responsibilities of the manufacturer is to calculate a Unit Rebate Amount (URA) for each drug that participates in the program. The local States are owed the rebate amount back for that drug.

So what did I do?

I supported the Medicaid Claims processing. All of those states with up to 30 programs each, submit an invoice that contains hundreds of drugs claimed in quarters reaching back t0 15 years ago. This is no small data processing challenge. An early assumption I made was that they implemented a system that would allow for a paperless process - this the last time I made that assumption.

I spent a month trying to identify ways to improve the existing process for the team responsible for ensuring the accounting and allocations line up for the clients we serviced. Afterwards, came my Python automations. With a rise in remote work due to COVID-19, most paper invoices were digitized. We used OCR and PDF tools in tandem with my quick Python scripting, and I developed 3 key algorithms that helped me do my job.

  1. An algorithm to simply classify documents. Prior to my entrance, this was a purely manual process to help us to classify key data associated with a file. The criteria was not complex, but it was rather lengthy. So to circumvent this, my script was equipped with our criteria in a CSV format. My algorithm did my work for me, taking into account the nuances of human readable text and the positioning of words in a sentence. Using text processing techniques such as Levenshtein Distance and Tokenizing, I never had to open a document to know what was in it. Literally saved me hours because of the sheer number of documents we could process in one week.
  2. A Data-Load Validation algorithm. We generally relied on a few variables from the states for us to properly classify this information, but since we are using accounting methods, there were some edge cases we identified. I was able to utilize great Data Governance practices to find mistakes in processing, and SQL breaking errors in the file, before the information was uploaded. Most importantly, it prevented accounting errors from making it into the database.
  3. An algorithm to automate emails to my team! Honestly, I did this because sending the same reports every day is not the most convenient, so automating this was fun. I don't believe my co-workers caught on so I'm very proud of that!

In short, I learned a ton about the world that is Government Healthcare Programs. Specifically how people get care, and how the Pharma business makes money. In the end, this experience has helped me identify the value I bring:

I'm a little lazy, so I'll automate it just to avoid doing it