Snowfish works with a significant amount of health care-related data supplied by multiple vendors including the industry’s leading companies. Purchased data has multiple uses from identifying the appropriate physicians for an offering to understanding usage and adoption patterns. Strategic decisions are often based on single sourced data purchases.
The bad news is that the data is often wrong, inaccurate, incomplete or some combination thereof. The challenge for any company is how to inform business decisions with wrong, incomplete, or inaccurate information.
The good news is that there are ways to get around this. Before we propose a solution we will address some of our “interesting” data acquisition experiences.
What’s the link between hospital executives and golf courses?
Most would think this is quite the absurd question. I would customarily agree except for an experience I had when acquiring data on executive-level hospital professionals. The data provider assured us that they could provide thousands of professionals fitting our criteria. When we received the results while the first 100 were appropriate, beyond that was completely irrelevant. They included dentists, golf courses, funeral directors, and even comedians. For that data provider they clearly felt that any data was good enough.
You really want all the data you requested?
One of the more interesting challenges we commonly confront is that data providers fail to point out the limitations of their data. For example, we recently requested data on oncologists for a certain type of procedure. We thought by going to the largest and oldest data provider we would receive a very robust data set. After receiving the file we noted that less than 50% of oncologists were in their data set when we compare the data to a separate file we use. We asked the logical question, where is the rest of the data? We were told initially this in the complete data set. Doing some more digging and questioning the data provider finally acknowledged that they capture less than 50% of the market and a significant amount of volume goes through “specialty pharma.” Failure to acknowledge the limitations of the data is a common and omnipotent problem we confront.
Every clinician in our database is phone verified…NOT!
Working with another well-known data vendor we were very happy to hear that we would be provided accurate demographic and profile information based on phone verification for each and every clinician in their database. Not sure how this truly can be done, we still took them at their word that we would get the information that we needed to effectively inform our project. What we found was a fairly large number of discrepancies in the data. For example, the address for a particular physician said “Massachusetts” but their current institution was listed as “Cleveland Clinic”. With a little detective work we found that the physician did indeed live in Massachusetts but went to medical school at the Cleveland Clinic. It was clear on multiple counts that every physician is not being phone verified. In actuality they are using the national provider identification (NPI) and bouncing that against another file and that qualifies as phone verification. This clearly, is a unique definition of phone verification.
Oh you wanted those physicians?
When requesting certain physician specialties who coded for specific CPT codes we received a dataset with a large number of records that appeared inappropriate. When questioning the data vendor they attempted to tell us that these were indeed appropriate and that their title was just another name for the specialists we were requesting. We ended up spending several hours and finally identifying that they individual did not code the data request properly. Quite simply, you need to check both the input and the output.
Crediting certain physicians for multiple prescriptions by NPs/PAs they work with.
A lot of decisions in the industry are based on Rx volume by a clinician. One of the greatest misperceptions out there is that only physicians write prescriptions. A while back we decided to get to the heart of the issue. We learned that some leading data providers default to the physicians due the ease from a data perspective. For example, a nurse midwife practice could have five nurse midwives and one physician as part of the practice. The data provider would credit the physician with all the Rxs even though at most one-sixth of the actual Rxs were written by the physician. We then reached out to over 500 NPs/PAs and did our own independent research which is available in a white paper. It clearly pointed out that NPs/PAs involved in the diagnosing and treating patients is being significantly under reported. But that still is a dirty little secret.
So now that we shared some of the common challenges we thought we would provide you some solutions to address the Data Land challenges:
- Make sure you understand the limitations of the data upfront. The limitations are often never disclosed it will often take multiple telephone calls to really uncover the limitations.
- Make sure you understand how data is being captured. For example, there are three “switches” that capture medical claims data. No data source purchases data from all three they buy from one and sometimes two “switches.” So realize the inherent limitations of the data.
- Make sure you never rely on one data source from decision-making. The more robust and varied your data sources the greater your insights will be.
- Work with a company that knows the limitations of the data to help you develop a more sophisticated targeting strategy.
Data is a key element for developing strategic and tactical plans. Working with a company that integrates and value adds the data into a business context can create a more robust and comprehensive picture for informing decisions.
Dave Fishman is President of Snowfish, a strategy consulting firm serving the pharmaceutical, medical device and biotechnology industries.