If It's Avoidable, It's Not Risk Adjustable
If you've taken Vishnu Rachakonda's claims analytics course or used the Tuva Health data mart (yay Aaron Neiderhiser!), you're likely familiar with the NYU Avoidable ED Visit algorithm - a standard approach for identifying potential savings opportunities in claims analytics.
Recently, I had an interesting thought: what if we connected the NYU algorithm with CMS-HCC? In other words, is there a relationship between avoidable ED visits and risk adjustment factors?
The chart above reveals the answer. Using the 2017 NYU patched algorithm [1] and the V28 HCC diagnosis set, I analyzed diagnosis codes with Non-Emergent Scores above 0 (noner > 0) - essentially, visits that "could have been" avoided with some probability. The x-axis shows the Non-Emergent Score quartiles, while the y-axis displays diagnosis code counts, with eligible codes in blue and ineligible codes in yellow.
What stands out? As visits become more definitively non-emergent, we see increasingly tall yellow bars, indicating they're not eligible for Risk Adjustment. The correlation between Non-Emergent Scores and non-RA eligibility is striking. It's almost like a rule: if it's not emergent, it's not risk adjustable.
This insight offers interesting perspectives. These non-emergent visits represent pure cost - they don't even contribute to risk adjustment. You could reduce them significantly without affecting your benchmark at all.
The question becomes: what are the best strategies to reduce avoidable ED visits? It seems like such a clear-cut goal. Sometimes, combining seemingly unrelated datasets can provide surprisingly valuable insights.
[1] https://lnkd.in/dgyKRFEE