/adjudication-engine-ruby

Primary LanguageRubyMIT LicenseMIT

Beam Adjudication Engine

Beam is processing our own claims, and we need to make an adjudication engine, that judges whether or not we need to pay the claim, depending on many factors. One of your coworkers started the engine, but left for a vacation, so we need you to pick up the slack.

Getting started

bundle i

./bin/adjudication-engine ./spec/fixtures/claims.json

Instructions

There are several steps to this adjudicator, some of which your coworker roughed in for you (yay!). Here's a breakdown of the remaining work:

  1. We need to match each claim with a provider record from our network. You can get our network data in CSV format from an API endpoint. The data is available at http://provider-data.beam.dental/beam-network.csv . Luckily for you, there's only a few records.
  2. Unfortunately, our data sources sometimes aren't the best. So, after fetching this data, you should sort out ones with bad NPI's -- national provider ID's.
    1. A valid NPI should be of length 10.
    2. A valid NPI only has numeric (0-9) characters.
    3. Log bad NPI's to STDERR.
  3. We will have claims data provided via an argument to the command line interface. You can run ./bin/adjudication-engine FILE. For ease of ad hoc testing, we have a fixture in the spec folder. So, you could run ./bin/adjudication-engine spec/fixtures/claims.json to test the process as you go.
  4. Next, the claims data needs to be matched up to providers. Luckily, we import the claims with an NPI included, so just match those up to the appropriate provider record.
  5. Now that we have a claims matched up to valid providers in our network, let's adjudicate the claim. Here are the rules you'll adjudicate with:
    1. As you can see in the claims JSON fixture, a claim is an instance of a visit to the dentist. It has line items, which are individual services billed by the dentist. Each line item or service is adjudicated by itself, and can be rejected or paid by itself.
    2. Reject any duplicate claims. A duplicate claim is defined as having the same start date, patient SSN, and set of procedures codes as another claim. All but the first claim in a duplicate set should be rejected.
    3. If the provider was not successfully matched, reject the claim and all of its line items. This plan does not cover out of network at all. :(
    4. Fully pay any preventive and diagnostic codes. Luckily, your coworker already provided a helper method on the ClaimLineItem model to help you figure this out.
    5. Pay any orthodontic line item at 25% of the charged rate. Again, your coworker added a helper method on ClaimLineItem to help you with this.
    6. Reject anything else.
    7. Log any rejections to STDERR.

In order to allow the CLI (in the bin folder) to work properly, the Adjudication::Engine::run method should return processed claims.

Notes

This solution requires Ruby 2.3.4 or higher (see the .ruby-version file). There were some module changes made to existing code, so if this gem is being installed and used some client code may need to be updated accordingly. This solution was also fully Test-Driven.

Recommendations

  • Batch processing applications such as this one typically allow for checkpoints and the ability to restart processing at arbitrary points in the file. These are features that we might want to implement in the future.
  • SSN is typically not used as an identifier due to privacy concerns and HIPAA compliance. This field should be replaced with a unique but non-personally identifiable value. Alternately a combination of other fields can be used in lieu of the SSN.
  • The adjudication-engine script should be updated to allow the Beam providers file location to be passed in. This will allow the script to be run in a development environment.