The health industry’s shift from a physician-centric view to a organization-centric approach, has quietly rippled across allied sectors. Sales used to be hyper-focused on the physician, but not anymore. With value-based contracting, decisions on purchasing are becoming more and more centralized within Integrated Delivery Networks (IDNs). Knowing the locus of control within an organization is critical, but the complexity of IDNs makes this a challenge.


This is where the right data can help.

A “data-driven approach” isn’t a buzzword anymore. Accurate affiliation data that shows the connections between healthcare professionals (HCPs) and healthcare organizations (HCOs), can help commercial teams identify decision-making structures within IDNs. But as with any network, these affiliations are never static. Multiple factors lead to change in the degree to which any two entities may be related to each other - a physician changing jobs, changes in referral patterns, billing pattern changes - all of these can result in HCP-HCO affiliations changing over time.

With 2M active HCPs and more than 8M relationships, affiliations are constantly in flux.

To make sense of the size and nature of changes, we queried our database to look at affiliation relationships over a 2 month period. The analysis was done on snapshots of data two months apart; we bucketed each affiliation by examining the score between a HCP and a HCO. Scores are categorized as Very High, High, Medium and Low, based on the strength of the relationship. Being able to quantify the degree to which two entities are linked to each other gives a lot of flexibility. Depending on the use case, the universe of entities that we are interested in changes - and score is an excellent way of expressing our expectations.

By tracking when a particular pair of entities - an HCP and an HCO - changed the strength of their affiliation, we were able to plot some neat charts.

Chart 1: Share of affiliations that changed

Indicates the number of HCP-HCO affiliations that experienced a change in relationship; includes new affiliations, some that got removed and others which had a change in the affiliation strength

Result: About 3% of the total volume of affiliations changed during this period, most of the changes being changes in strength (69.48%). This is followed by removals (19.71%) and closely followed by additions (10.81%).

If 3% by volume seems a small number, consider the fact that it is 3% of 8.58M or 460K HCP-HCO affiliations. Extrapolating this to three months, almost 5% of physician affiliations experience a change every quarter!

Chart 2: Variation in HCP-HCO affiliations by type of facility

Indicates the share of HCP-HCO affiliations by facility that experience a change in relationshipResult: Facilities like Mental health centers (10.3%) and Clinical labs (9.9%) show a bigger relative change in affiliations compared to HCOs like Hospices (6%), Home health (4.7%) and Physician groups (3.8%).

Chart 3: Changes in strength of HCP-HCO affiliations

Indicates affiliation changes through the lens of strength; The flow diagram shows how affiliations evolve over the time period in consideration

Given the score of an affiliation, we can segment them into the four score buckets - Very Strong, Strong, Medium and Low. On the left we find buckets representing affiliations from 2 months ago, on the right we have current data. The lines flowing from a bucket on the left to the bucket on the right indicate the volume of affiliation changes between the two.

Result: We find a significant amount of flow between the Weak and Very strong buckets - these are mostly affiliations that have been added or removed. A larger proportion of Strong relationships move to the Very Strong bucket than it does to the Medium and Weak buckets. The data tells us that Strong relationships generally get stronger over time.

Chart 4: Changes in strength of HCP-HCO by region

Indicates relative change in affiliations per CBSA county; affiliation changes are normalized to each county to remove effect of large population regions

Result: Some regions in the mid-west show zero or little change. Coastal areas in the west have more variations in affiliations compared to the rest of the country.

Conclusion

Throughout this post, we have sliced and diced the data in different ways. We found interesting trends when examining the data based on affiliation strength, facility type and location. But we have merely scratched the surface here. This exercise also demonstrates the importance of powerful meta-attributes of the data. Hence when considering datasets for commercial healthcare, it is important to ensure that your data ticks the following buckets:

  1. The data solution accurately captures affiliations between entities
  2. Affiliation connections are categorized by type and strength
  3. All entity types possible (physicians, institutions, payers and manufacturers) are present and linked
  4. All entities can be matched to a customer data master with a unique identifier
  5. The data is refreshed frequently

An opportunity exists to examine trends from the data we have. Keep coming back to this blog for more such analysis, or better yet, sign up for updates to the blog with the form on the right.