How to Manage Denials – AI Style

How to Manage Denials – AI Style

In our last post, we discussed what our ideal version of an effective denial management process would look like. In this post, we are excited to discuss how to execute those strategies!

To execute the strategies discussed before, providers need niche technical capabilities. While a base model based on logistic regression or random forest could be a good starting point, incremental improvements are possible with a more intelligent analytical engine which back-propagates the error and tune dynamically. Newton AI is adequately equipped to do so and handle your voluminous data in quick time. In a way, it consolidates all intelligence from your denials data into one single platform and repeatedly pushes the workflow for the most optimal denial management. The automated denial categorization and follow up list helps providers improve their first pass payment rate and maintain a healthy days sales outstanding (DSO). The engagement model guarantees a minimum reduction in denial rate and assures continual improvements, making Newton AI a niche tool worth your time for exploration.

How do you currently manage denials at your end? What strategies have worked in your setting and have you been able to automate those? How do you see analytics playing a role in your initiative? We would love to hear your views!

An AI Based Strategy to Manage Denials

With MACRA’s payment reforms underway and Alternative Payment Models (APMs) within striking distance, claim denial management are bound to get renewed focus by providers. A report from HIMSS Analytics estimates that 66% of those providers currently not having a denial management solution will be looking for such a solution over next 6 months period. This calls to revisit the question: “Are older ways of manually reconciling the denials and managing A/R resolution efforts adequate?” In this blog post, we present an automated and arguably more efficient means to handle denials.

At Apexon Health, we believe that an effective denial management should have 3 components working together:

Firstly, the provider should be adequately forewarned of a likely denial. Working re-actively to denials is a sub-optimal strategy. An automated denial prediction mechanism suggesting optimal denial avoidance strategies and incorporating them into the daily workflow of providers is needed. Pre-bill claims data provides ample data fields which capture such intelligence and help build denial prevention strategies around them. Building the predictive model requires a good understanding of data science and revenue cycle processes. Our experience also suggests that a single predictive model alone may not good result in a multi-specialty setting. A way to overcome this challenge is imbibing self-tuning models which operates at optimal performance and then having an ‘ensemble’ which weeds out the poor performers and dynamically puts into place the best performer. Such ensembles can accurately quantify the risk of denials.

Secondly, the provider should be able to quickly process the denials and read through the patterns. It should help them answer:  Is it coding on claims from a particular locality for a specific CPT posted to a particular payer leading to more than usual denials? Essentially, what all different claim attributes, when present together, have been leading to denials? Can we quickly identify them and put them as part of claim edits or scrubber to stop any further leakage of revenue? The human brain has limited capability to recognize such patterns. For example, if there are just 10 claim attributes of a denied claim, it calls for interpretation of 2^10 i.e. 1024 patterns!! Sophisticated algorithms which can scale up fast and can do mammoth calculations in quick time are needed. That’s where our Newton AI comes in; it rapidly captures millions of patterns, sorts them by their frequency of occurrence and confidence, and enables HIM apply his expertise to pick the patterns to be pushed to claim edits.

Lastly, effective denial management calls for intelligently handling the claims after their denial. Strategies based on FIFO (First In First Out), high dollar denials, and payers with best resolution rates can be good but may not always be optimal. These are just few dimensions and a payer’s decision to deny the claim is typically an aftermath of several considerations. A time based model which predicts payments over several horizons can enable providers adequately manage their AR. ApexonHealth’s Newton AI has in-built modules which create multiple scenarios for AR handling. It essentially takes historical denial resolution, uses appeals and production data as the feed, does likely collections with user defined look-ahead periods, and suggests most optimal AR inventory available for the dent.

Keep looking out for our posts to see how to most effectively execute these strategies. Let us know how an ideal and effective denial management process looks like to you.