Increase productivity, improve processing, and gain higher customer satisfaction with Machine Learning
How a leading managed healthcare provider used Machine Learning to improve the physician’s authorization process and improve the quality of the healthcare provided to members.

INDUSTRY HEALTHCARE

The client

The client is a leading managed healthcare provider, issuing Medicaid, Medicare Advantage and Medicare Prescription Drug plans to over 3 million Americans. The client engaged Pluto7 to help identify practical applications of Machine Learning (ML) to solve key business challenges for its products, offerings and services.

Overview

This client looked to Google Cloud and Pluto7 to help them identify the right use cases for driving transformational change. Pluto7, using their deep Machine Learning expertise, ran workshops with the customer to establish the root of their needs. The client chose to focus on routing classification to achieve the transformation they needed: emergency vs. routine authorizations vs. retroactive changes.

Challenges

Currently, most of the client’s decisions are made by people and based on the information they gather and assess. This includes specific care provided to customers, claims management, monitoring the patient’s lifestyle, determining their risk levels and also the physician’s authorization processing.

Although the client was using Optical Character Recognition (OCR) software, the physician’s authorization processing was manual, requiring high levels of human intervention. This caused delays in handling priority authorization requests, including patients needing emergency or urgent care.

Strategy

Based on Pluto7’s expert knowledge of Google Cloud Platform, advanced analytics including Machine Learning and Artificial Intelligence, and applications development, the following approach was taken:

  • Utilize Google Cloud Platform to collect and centralize all related physician’s authorization information.
  • Leverage Machine Learning and Artificial Intelligence to predict the classification for requests meeting the emergency criteria.
  • Modify the process to leverage emergency/non-emergency classification, and subsequently reduce the manual processing involved.

Results

Using the Machine Learning model to drive the processing of emergency authorizations reduced the delays significantly, with no hourly delays recorded.

The team was able to scale and process more physician’s authorizations with reduced headcount.

“I know that technology can help us become more productive, however we need to make sure we apply the right technology to help, considering that OCR-based processing has not worked as expected. We are looking forward to having Machine Learning help us improve processing physician’s authorizations and use that learning to improve other processes.”

- EVP and Leader, Managed Healthcare provider

This managed healthcare provider has experienced rapid growth in recent years, with their products and services used by over 3 million members. This growth meant that certain manual processes, such as physician’s authorizations processing, were slowing things down. They needed to be reassessed from the perspective of achieving breakthrough productivity.

The client received faxed documents which were either printed or handwritten. The fax contained various forms, some relevant and some non-relevant. These could include Authorization Requests, Expedited Authorization Requests, Retro, Ad Hoc Ads, or more. From the fax documents, the client needed to identify relevant information, then classify whether to expedite the document.

What does it take to automate physician’s authorizations processing?

The client’s current OCR based system unfortunately failed as it was unable to process whether the authorization notes from physicians were expedited or not. This is because the notes were often handwritten.

By using Machine Learning, human decision making was reduced. ML and AI was able to determine the content of the faxed documents autonomously, being able to more accurately identify the contents of handwritten notes. This in turn helped drive faster processes and reduce lead time to action on the content of the fax. Ultimately, the ability to classify expedites became faster.

Pluto7, with its GCP and Machine Learning expertise, determined that image processing, combined with natural language processing, could meet the challenge by classifying emergency and non-emergency requests correctly and transferring them to the right processing queues.

“Once we understood the power of Machine Learning and its ability to detect images and also process text we felt comfortable that the solution will work.”

- Claim Handling Specialist, Managed Healthcare Provider

Proactively identify expedited physician’s authorizations for immediate processing

With these objectives in mind, this managed healthcare provider partnered with Pluto7 to transform their physician’s authorization processing. Faxed document data, fax volume, and the labor hours involved were analyzed to finalize the solution.

Using Cloud Storage on Google Cloud Platform, Datalab for manipulating data, and Natural Language Processing (NLP), the proof of concept demonstrated that the solution met the desired requirement. This laid the foundation to enact the plan and roadmap for production implementation. These improvements lead to perpetual savings, year over year.

“Google Cloud Platform’s immense processing power and ability to leverage easy to morph pre-built vision and natural language processing models along with Tensorflow to model custom prediction was the key enabler in developing a productivity platform for the managed healthcare provider.”

- Salil Amonkar, Global Head ML/AI Practice and Delivery, Pluto7

Stronger through innovation

Working with Pluto7, this client was able to take their innovative mindset and apply it to their business, improving internal productivity and changing their members’ experience. By leveraging Google Cloud Platform and its components (Cloud Storage, AutoML Vision, , TensorFlow, Data Studio and Machine Learning Engine) they were able to drive innovative process improvement, seeing tangible business benefits.

Why Google

This client chose Google Cloud Platform because it:

  • Offers a flexible, scalable, ready-made infrastructure in the cloud.
  • Delivers powerful data processing, data warehousing, and state-of-the-art Machine Learning and Artificial intelligence capabilities.
  • Provides a cost-effective platform that’s easy for business team members to use.
  • Delivers speed, security, reliability and flexible pricing.

Products used

Google Cloud Platform

AutoML Vision

Cloud Natural Language

Google Cloud Storage

Google Cloud Dataflow

Google Machine Learning

Tensorflow