1. Data
  2. Health
  3. Startups

Clinical Concepts on a Limited IT Budget : Tuva Health

Recently I came across the Tuva Health Project – an open-source collection of data-engineering processes and clinical concepts you can apply to your existing healthcare data. Tuva’s stated goal is to commoditize building your core health data infrastructure and clinical concepts so you can focus on building deeper population insights.

I see an opportunity in the Tuva project for smaller healthcare practices, which may be limited in budget, to extract more insights from their raw data while leveraging their existing team’s skillsets. Many of these organizations are faced with an open ended problem of how to start transforming their raw data into clinical concepts.

Several challenges exist when transforming health data into clinical concepts:

  1. Transformation – The schema of raw claims and EHR data serves a specific purpose (clinical operations, billing). Transforming that data into population insights requires moving into a new conceptual model for analysis.

  2. Limited Expertise – Your team may be new to building data engineering processes or lack expertise in building certain clinical concepts.

  3. Data Quality – Upstream data quality issues can bleed into your model, causing more distrust of the insights across the organization. Data quality issues can range from from incomplete data to domain specific integrity problems. An example would be an ESRD patient with no history of dialysis.

  4. Cost– A limited budget may drive who you can hire or what third-party systems you can buy.

  5. Open-Ended Outcomes – You may not be immediately trying to solve a specific problem, such as better understanding your potential to take on risk-based contracts. In this scenario, you may be looking to build core clinical concepts (such as readmission rates or categorizing chronic conditions) as a building block for future data projects.

Note: My goal is to highlight why the Tuva Project may be a useful starting point for your organization. Even if your long-term plan is to purchase a third-party population health system, implementing Tuva can assist your organization in building a conceptual understanding of data and engineering processes used in healthcare.

The Tuva Health Project

Coco Zuloaga and Aaron Neiderhiser built Tuva as an open-source framework to map raw health data into a standard model for building clinical concepts. Their goal is to commoditize the way you structure and transform health data so you can focus on what important. The basic process it to map your raw clinical and claims data into a standard input format. That standardized format is then transformed into the Tuva Common Data Model. The Common Data Model provides the based set of concepts, tables and lookup values to then build clinical concepts – such as readmissions.

Transforming data from the common model to clinical concepts is done through dbt (data build tool). The dbt is a popular data engineering tool for transforming data with SQL select statements.

You can checkout the Tuva roadmap here.

Why you may consider Tuva Health

  1. A Common Data Model – You’ll end up transforming your raw data into a common data model that was based off the FHIR R4 (https://hl7.org/fhir/R4/) standard. As noted in the Tuva docs, R4 was used as the starting point. FHIR data is nested JSON vs. a relational model so there is not a 1:1 map between FHIR and Tuva’s CDM. For your team however, this may help align concepts you see in RIM and FHIR into an analytics domain and allow you to reason through the clinical domain. The common data model will also provide a starting point for relationships between patients, encounters, conditions, procedures, etc.

  2. Based on dbt (data build tool) – DBT is a command line and cloud based tool for transforming data through standard SQL select statements. It’s the T in ETL. DBT makes it easy to connect to existing data warehouse. Your team can then build their transformation in SQL and run those transformations through the dbt command line tool or dbt cloud. (Note: There is a cost associated with dbt cloud and connecting to stores with PHI requires setup of an Enterprise plan with the sales team).

    If you or your team is new to data engineering pipelines, dbt provides a framework, best practices, open source tooling and a core building block for building repeatable data pipelines. There is a learning curve to using the tool, however your team can utilize their existing SQL expertise to get going quickly.
  1. Data Quality Checks – Quality issues in healthcare data sources can cause analytics problems downstream. For example, you may want to ensure that each patient has an attributed provider. Tuva’s tooling, based on dbt, allows you to write quality tests in SQL to validate the state of your model. Data quality tests will help protect against future quality regressions and build more trust in the output concepts and insights..

  2. Open Source Clinical Concepts – You can read through how Tuva readmission models are built. Often times, clinical concepts we build on top of our data are relatively complicated. You can reduce your time to valuable insights by utilizing the Tuva models that already exist, or read through documentation and source code to better understand how a model works. I expect these concepts will continue to grow as the project evolves. Leveraging existing, tested concepts reduces the potential for errors in your model and acts as an authoritative source for your team.

  3. Community – GitHub, Slack, Open office hours with the founders. You and your team has the opportunity to collaborate and ask questions. This may be incredibly valuable in terms of knowledge transfer and building a shared understanding of your data model in the organization.

If you are just starting out transforming data sources into population insights, budget and team expertise may be a limiting factor in how quickly you can move. The Tuva Project looks promising for launching a new healthcare analytics project while providing the rails needed to build a production ready data-engineering pipeline.

Comments to: Clinical Concepts on a Limited IT Budget : Tuva Health

Your email address will not be published.