Avalon.AI

Avalon.AI is the US data analytics company providing the best outcomes for payors, providers, and medical device and pharma companies. Their innovative AI model combines value-based data analytics with clinical treatment and protocols to provide best-in-class healthcare. Avalon.AI’s proprietary technology is delivered via advanced dashboards and data sets that enable users to make more informed decisions – from the micro level of individual prescriptions up to doctors, hospitals, and entire healthcare systems – all to create better patient outcomes.

In response to the escalating Opioid Epidemic in the United States, Avalon.AI undertook the task to modify and enhance opioid prescription practices. Their goals included:

- Conducting a thorough analysis of doctors' prescription patterns and comparing them with patients’ common procedures and outcomes.

- Developing an all-encompassing analytics platform that enables healthcare providers to integrate data from various systems and convert this data into comprehensive reports.

- Providing valuable insights to healthcare institutions about prescription practices, thereby urging them to reevaluate their approach to prescribing pain medication.

Avalon.AI had a clear vision and profound understanding of the healthcare sector, yet the complexity of their goals demanded a partner with demonstrable technical expertise. 

Based on Svitla's multi-year experience in the healthcare industry, as well as the successful joint implementation of several previous projects, including the Hospitals Administration Platform and Healthcare Management Platform, Avalon.AI chose Svitla as a reliable technology partner.

The timeline was tight due to commitments the client made to the medical community and required Svitla to launch the initial application within 3 months.

Upon taking on the project, Svitla held several strategic sessions with Avalon.AI  management to understand the goals, construct a development plan, and suggest the best-suited technologies stack and relevant team composition.  

The primary challenge was to swiftly develop a Minimum Viable Product (MVP) capable of effectively handling and organizing vast amounts of unstructured and underutilized data from diverse sources. The goal was to convert this data into structured analytics and user-friendly dashboards, which would aid users in identifying trends in opioid prescriptions and subsequently offer means to regulate overprescribing.

Cloud Infrastructure Architecture

Given the need for rapid deployment of the platform, Svitla utilized AWS to build out the scalable infrastructure quickly and take advantage of AWS's robust security and compliance tools. This choice ensured HIPAA compliance, a critical requirement in handling sensitive healthcare data.

The architecture of this system consists of three parts: data upload, ETL, and visualization.

Firstly, medical data from various sources is uploaded to an Amazon S3 bucket. This action triggers a Lambda function that parses the data and generates corresponding records in a relational database.

Next, the data undergoes an ETL process using AWS Glue Data Catalog, as well as third-party tools such as Fivetran and Matillion. The processed data is then stored in Amazon Redshift.

In the final stage, we leverage Machine Learning algorithms executed by AWS SageMaker to uncover medical insights and correlations. 

The results and findings are presented to users via the Tableau Visualization Dashboard. This enables users to see the rankings, make comparisons, and modify pain medication approaches based on the data insights.

Choosing between Power BI and Tableau, our team stopped on the latter as it is able to efficiently handle significantly larger data sets.

Data Processing Through Advanced AI and RPA

The Svitla development team combined Python, SageMaker, and UiPath to create a multi-faceted tech stack for data processing. Python and SageMaker were utilized to construct an efficient analytics engine and develop and integrate an AI model, while UiPath was used to automate data-cleaning processes.

As part of the data cleanup process, the team designed a UiPath bot to transform disorganized and complex data formats into a more structured format. The bot was programmed to extract and reorganize the data into specific columns that were conducive to analysis, thereby eliminating the need for manual intervention and making real-time dashboard updates.

Moreover, considering the restricted data extraction capabilities of Ambulatory Surgical Centers (ASCs) from their Electronic Medical Record (EMR) systems, the team utilized the RPA bot to generate import files. These files were crucial in feeding our analytics engine with the data it needed for effective processing and interpretation.

Tableau and Machine Learning Integration

Once we cleaned the data, we built the first dashboards using Tableau. This helped to turn the data into easy-to-understand visuals for end-users. At the same time, our Data Science team used Python and SageMaker to create models that could predict how much of an opioid a doctor might prescribe. 

This was the pioneering effort in handling such a unique predictive modeling task. The insights gleaned from these models were subsequently incorporated into our Tableau dashboards.

UI Design for Final Dashboards

The dashboards were then handed over to UX/UI designer tasked with presenting the medical data effectively. We emphasized critical trends and behaviors, ensuring the dashboards were easy to understand and operate.

Expert Team Composition

To ensure the project's success, Svitla assembled Senior-level specialists to address the following directions:

-Cloud Solutions: AWS Lead experts to oversee the implementation of the cloud-based infrastructure and ensure a scalable calculation engine for data metrics.
-Data Science and Analytics: specific data science engineers to construct the analytics engines and calculations for categorizing patient prescription levels. 
-Robotic Process Automation: RPA Developer to create a process for ASCs with limited ability to extract data from their EMR systems. This involved utilizing a UiPath bot to generate import files for the analytics engine.
-UX/UI: Senior designer to visualize medical data effectively and highlight critical trends and behaviors.