Apr 6, 2021
Qventus applies innovations in artificial intelligence, machine learning, behavioral science, and operations management to hospital operations. Our products include the first patient-flow automation solution for healthcare, and new tools developed in response to COVID-19 that help hospitals forecast and plan for their Personal Protective Equipment (PPE) needs and post-acute planning. We have helped our partners—large healthcare systems, academic medical centers, and community hospitals—lower exit delays and reduce in-patient length of stay.
COVID-19 disrupted the healthcare industry to an unparalleled degree. The challenges facing the industry include unprecedented financial pressures, capacity pressures, canceled surgeries, and increased hospitalizations for COVID-19 patients. With the increase in COVID-19 patients came more demand for PPE supplies, and capacity constraints that were difficult to plan for. With limited visibility into the demand coming their way, facilities didn’t know what to expect. This is where Qventus saw an opportunity to help.
Our goal was to help hospitals answer these questions:
These questions were aimed directly at the most pressing problems facing hospitals during the early stages of the COVID-19 health crisis.
As an analytics architect at Qventus, I used Looker to build and rapidly deploy custom data planning solutions for our hospital partners. We chose Looker because it worked well for our needs to:
We leveraged Looker’s big data analytics to curate hospital management solutions with embedded analytics during this crucial time. With the tools we developed at Qventus, hospitals can now leverage customized dashboards to see a more complete view of the impact of certain critical decisions, such as when to reopen surgery rooms and whether (and how) to make changes to supply usage policies. We arm hospitals with the tools and information they need to make these important decisions in a strategic and data-driven way.
Results so far have been great. We spent a lot of time building and working with clients to use the tools, and sometimes we forget to take a look back and see where we came from.
Our data model for supply usage is based on complex factors. We’ve continually iterated on this model by incorporating feedback from our hospital partners and by accounting for new factors as the pandemic evolved.
Our planning tool provides a full spectrum of assumptions that can be tailored according to the unique needs of each hospital. The categories include:
We built the interaction-based usage model in LookML, Looker’s Git-versioned modeling layer. We used some constants (based on known experience in hospitals), and also provided some default selections based on our localized SEIR (Susceptible, Exposed, Infection, Recovered) epidemiological model. The customizability of LookML and the ability to alter assumptions on the fly played a significant role in the model’s improvement as we continued to develop new and more robust versions of our tool. After building the initial metrics and measures, we could create any number of layers on top of the LookML model to handle these complex calculations and aggregations.
Our tool uses customized dashboards and visualizations to show hospitals how many days they have remaining for each type of supply and what category is expected to deplete first. They can now project the quantity of supplies that will be needed months in advance, based on their unique situation.
During the initial COVID-19 crisis, hospitals did not know what to expect in terms of capacity constraints. We set out to provide them with a robust model that would take into account our localized SEIR epidemiological model, third-party information, and published research in order to forecast how many patients might need to be discharged daily and where they could be routed.
We were easily able to pull from publicly available datasets through Google BigQuery. This gave us a robust model with multiple complex calculations in Looker. We could adjust each metric based on whether the data was available or not with simple SQL statements in Looker. Healthcare analysts can override our assumptions if they don’t feel they are applicable to their unique situations.
Our tool shows hospitals how many daily discharges there would be for COVID-19 and non-COVID-19 populations. It shows them which ones would be discharged to post-acute care and to which type of care—for example, to home healthcare or skilled nursing facilities. And it shows them how many days they have left until capacity is reached.
In this unprecedented time of crisis, we are so grateful that we have been able to help our partners address their most pressing issues, and that we could so quickly develop and deploy the data tools that made it happen. The feedback has been overwhelmingly positive. Facilities are using these tools as a core part of their COVID-19 planning strategy, and we are proud to serve as a reliable partner that is there when we are needed most.
For more on the solutions we built for hospitals, watch my JOIN@Home Talk, Hospital Planning for COVID-19.