Editor’s note: In this guest blog, we look at how healthcare startup Vida Health built a virtual platform on Google Cloud that cut costs and overhead, saves healthcare providers valuable time, and delivers machine learning capabilities that operationalize their data for better patient health outcomes.
At Vida Health, our virtual healthcare platform is designed to deliver whole-person healthcare by treating multiple conditions and integrating both mind and body medicine. In choosing Google Cloud to help us with our digital transformation, we were able to reduce costs 60% by switching from a managed platform to Google Kubernetes Engine (GKE), and are using Google solutions like BigQuery ML to innovate new products that help our patients and empower our clinicians.
Accelerating the heartbeat of digital transformations
Traditionally, healthcare has been a slow-moving industry with a bias toward risk aversion and maintaining the status quo. The COVID-19 pandemic challenged this mindset and encouraged many healthcare organizations to accelerate their plans for digital transformation. At the forefront of this transformation is virtual care/telehealth and the ability for providers to offer the same high-quality patient experience over the web and mobile as they do in person.
During the pandemic, Vida Health faced challenges scaling our original infrastructure on another cloud provider to meet the growing demand. We also felt that this CSP’s suite of machine learning (ML) services didn’t provide the value add we were seeking. After performing research into competitive cloud technologies, we chose Google Cloud for their flexible, secure, and scalable solutions that integrated seamlessly, reduced our operational overhead, and gave us the tools to build innovative products powered by ML.
A key differentiator of Vida in the healthcare marketplace is our platform. Where many competitors have solutions targeting single conditions, we took a horizontal approach, with a platform designed to treat multiple conditions and to integrate both mind and body. Nearly half of Americans have more than one chronic medical condition, and we want to help them with whole-person health solutions that acknowledge the reality of their situation.
Our platform is powered by a spectrum of Google solutions, including Looker, an enterprise platform for business intelligence, data applications, and embedded analytics. With a unified dashboard experience, Looker helps us aggregate all of our data and gives us a holistic view of each patient. To take advantage of artificial intelligence (AI) and ML technologies, we were well situated by using BigQuery, Google’s serverless data warehouse, to store all of our data in one place. Even as our datasets in BigQuery grow more comprehensive, it remains easy for our ML engineers and data scientists to use and experiment on that data. We can then take that data into production with BigQuery ML, which allows us to build ML models with only SQL skills.
Prescribing ML for new use cases
In our use and exploration of AI/ML in our platform, we go beyond pure AI tools by including human-in-the-loop programs and treatments. For example, we provide coaches, therapists, and dieticians that work with each individual patient, providing tips, strategies, and accountability. Our patient-provider interactions are digitized and stored, giving us a robust training dataset that we can now operationalize using all of the Google tools available. Using these provider interactions, we can track a patient’s progress to ensure they’ve improved their health outcomes, whether it’s weight loss, stress reduction, blood sugar management or beyond.
We want to endow our providers with superhuman powers, which means using AI/ML to manage and automate all of the tasks that aren’t member-facing, freeing up the providers to focus their time and energy on their patients. We’re currently experimenting with our Google tools around transcribing the provider’s consultation notes and then applying data analysis to uncover insights that will lead to better health outcomes. Other time-saving solutions on our roadmap for providers include pre-filling standard fields in the chat function and managing end-of-day approvals.
We’re currently using BigQuery ML for our “next action recommender,” a member-facing feature on our mobile app that recommends the next step a patient can take in their treatment, based on past datasets of information provided by the patient. At the start of their journey, the steps might be basic, such as scheduling a consultation, adding a health tracker, or watching a health video. But the longer a patient uses our platform, the more sophisticated the recommendation system gets.
On the provider side, we have our Vidapedia, a comprehensive list of protocols for treatments that providers can follow. In the past year we’ve invested in Vidapedia cards, which are distinct sets of clinical protocols that have been codified. We’re up to 150 cards, and instead of providers needing to keep all of that information in their heads, we’re working on using BigQuery ML to extract the actions a patient has taken so far in their treatment. Using that data, we’ll then recommend to the provider the most relevant cards that apply to the specific conditions. Having that information at their fingertips reduces the amount of time they need to spend on each member offline, which helps us build efficiency and lower the cost of delivering care.
We’ve also used ML in our customer acquisition process, which has traditionally been a costly endeavor for healthcare startups. A company first needs to market and sell to payers and providers, and then understand the total addressable market (TAM) for their patient base before convincing that segment that their platform is the best decision. We’ve successfully applied ML to this process, sifting through hundreds of different data inputs to better predict who is likely to use our platform, saving us time and money.
Invigorating virtual healthcare with Google Cloud solutions
The rest of our current Google Cloud stack is robust, featuring BigQuery Slot Autoscaling, a preview feature that optimizes costs and scales for traffic spikes without a sacrifice in performance. We use Looker for data reporting and dashboarding, and Data Studio for quick, ad hoc data visualization. Our relational database is Cloud SQL for PostgreSQL, and we use Data Catalog for data discovery and search. Other Google services in our stack include GKE, Dataflow, Data Fusion, Cloud Scheduler, and AI Platform.
The seamless integration between Google products and services has been impressive and time-saving. Many of our clinical protocols were originally written in Google Docs, and the ability to import that data directly into BigQuery has saved us so much time and effort. Using Looker to then democratize access to that data internally across our organization, and BigQuery ML to build ML applications upon that data, feels like a secret weapon that puts us ahead of the competition.
As the healthcare industry continues to adjust to the demands of a changing world, we’ll be working with Google Cloud to deliver cutting-edge solutions that exceed the needs of our patients and providers.
Learn more about Vida Health, then apply for our Startup Program to get financial, business, and technical support for your startup. You can also read more about other organizations using Looker and BigQuery to modernize business intelligence.
Cloud BlogRead More