Introduction
SNF E/M codes are the staples of SNFist billing. CPT codes 99304, 99305, and 99306 are used for initial consultations, and CPT codes 99307, 99308, 99309, and 99310 are used for subsequent consultations, i.e. follow-ups. The greater the last two digits of the codes are, the higher the complexity of the consults, and the higher the reimbursement.
To estimate the average billing rate of each CPT code (at the provider level), we did the following:
- We bifurcated the FY2018 SNF E/M codes (the latest data available) into initial and subsequent categories and used the ratio within each category, rather than across both categories. This made the analysis more intuitive to understand and the results are more consistent with how practitioners consider code-mix.
- We built a model that estimated the mean for each CPT code within initial/subsequent categories at the national and the state level. The model looked at individual providers rates, controlled for the number of encounters (i.e. was more confident about providers with a larger number of encounters and less confident about providers with fewer encounters), and partially pooled information to arrive at the state and national level estimates that were more robust than the traditional MLE (Maximum Likelihood). The model was also more confident about states that had more providers, and less confident about states that had fewer providers. The model also provided credibility intervals for its estimates at the state level.
Initial SNF E/M Codes
The initial SNF E/M CPT codes are 99304, 99305, and 99306. These are often referred to as initial Level 1, Level 2, and Level 3, respectively. Below we have forest plots for each CPT code. Each plot shows the estimated national mean as a red line, and whether a state is above and below the national mean, with credibility intervals indicating uncertainty around the estimate. For each of the Levels, we see significant variance from state to state. For example, we can see that the average provider nationally bills approximately 47.7% of their initial SNF E/M’s as 99306, whereas the average Snfist in Arizona bills 73.9% and the average Snfist in Iowa bills 20% of their initial SNF E/M at 99306.
The same data as above is represented below as faceted color maps. We can see some regionality, for example, the West tends to bill 99306 at above the national mean.
Subsequent SNF E/M Codes
The initial SNF E/M CPT codes are 99307, 99308, 99309, and 99310. These are often referred to as subsequent Level 1, Level 2, Level 3, and Level 4, respectively. As in the case of the initial SNF E/M codes above, there is a significant variance from state to state. For example, we can see that the average provider nationally bills approximately 8.4% of their subsequent SNF E/M’s as 99310, whereas the average Snfist in Minnesota bills 22.4% and the average Snfist in Kentucky bills 3% of their initial SNF E/M at 99310.
The same data as above is represented below as faceted color maps. Again, we see regionality, with the West, Great Lakes, the Carolinas, and parts of New England are more likely to bill above the national average for the higher level subsequent E/M codes (99309 and 99310).
Conclusion
This brief look at SNF E/M codes billing code-mix shows significant variance between states for both categories of SNF codes. Additionally, we see some geographical clustering as it relates to the level of E/M billings. This begs some interesting questions as they relate to causality, for example, are these patients more complex, are SNFists allocating more time, or rendering different levels of clinical decision-making in a systemic way?
Further, this analysis makes me wonder about how CMS’ audit partners gauge over-billing. Are they using naive algorithms based on MLE, which can give startingly misleading estimates? Are they benchmarking relative to a national benchmark or applying a regional benchmark?
Should these algorithms be made public, so that they can be critiqued?
