Case Study

Mosaic Medical Uses Alteryx to Optimize Physician Resources

Mosaic Medical deploys the power of Alteryx predictive analytics to manage resources considering the relationship between patient health and demand for providers.

Mosaic Medical is a community health provider based in Central Oregon. Mosaic operates primary care clinics with family practice physicians and pediatric services; school-based clinics in elementary and high schools; and a mobile clinic that provides primary care to the homeless. As a community health center, Mosaic primarily sees low-income Medicaid and uninsured patients, often with high medical need. Over the course of a year, Mosaic treats about 19,000 patients, which works out to about 60,000 patient visits in a year. Marshall Greene is the Senior Healthcare Data Analyst at Mosaic Medical. His team develops all clinical, operational, and quality improvement reports for the organization.

 

Deeper Insights
Deeper Insights: Mosaic used Alteryx to see clear patterns correlating risk scores and patient panel size.
Minutes vs. Hours
Hours vs. Weeks: Before Alteryx, a patient risk score model took 75 hours to build. Using Alteryx, he built the same model in only three hours.
Intuitive Workflow
Repeatable Workflow: Updating the risk score model is just a matter of clicking a button, and the workflow is scheduled to send automated updates once a week.

 

Challenge

Mosaic Medical needed to understand their doctors' patient load, called a 'provider panel', in order to set target panel sizes and manage physician resources. The provider panel is the roster of patients that a given provider is responsible for keeping healthy. The number of patients a provider has on their panel can vary depending on whether the doctor works full-time or part-time, how many patients a given provider sees in a day, and how healthy those patients are. If the average patient is unhealthy, that patient will have more visits over the course of a year. The doctor will see that single patient more often and, hence, will require a smaller panel size. Patient panel size can vary anywhere from 800 to 2,400 patients.

Mosaic needed to set target panel sizes, but Greene's team didn't have a good way to adjust panel size targets based on the patient severity. The team created initial risk scores, but level loading was difficult to manage because Mosaic has a mix of providers who see chronically ill patients requiring more visits, and pediatricians who mostly see healthy kids only once a year for their well-check exam.

Greene needed to create a patient risk score to identify factors within Mosaic's electronic medical record associated with the number of visits to Mosaic clinics. He needed to predict the demand for a given provider to balance physician cases.

 

Solution

Before discovering Alteryx, Greene made a first version of the risk score and used a labor-intensive process. He downloaded reports from a tool provided by his Electronic Medical Record (EMR) vendor, exported the reports as CSV and used Access queries for data prep, and finally he would export data as a CSV into R. No one from Green's team was an R programmer. He estimates that it took about 75 hours to build that first model, and updating it monthly was a painful process.

With Alteryx, Greene built a risk score model in only three hours, and updating the model using Alteryx is just a matter of clicking a button. He even has it scheduled to update once a week.

For risk modeling before Alteryx, Greene used two SQL queries – one to produce a problem list of chronic diseases by patient, and another to gather basic demographics. The work required data prep and joining those two data sets to apply the risk model. With Alteryx this process is much easier and the risk score model is much cleaner. The workflow is annotated, so anyone can see what is happening at each step in the process. Plus, the models are more accurate because he is able to apply more modern predictive tools.

 

"That first risk score in Alteryx took me three hours to build. And that was while we were in our Alteryx trial period. We were trying to make a business case to the C-level staff to show the real value in this tool."

Marshall Greene - Senior Healthcare Data Analyst - Mosaic MedicalMarshall Greene
Senior Healthcare Data Analyst
Mosaic Medical

 

Watch Marshall Greene's Presentation from Inspire 2016

Watch now

 

Results

Before Alteryx, Greene implemented Tableau. He says, "We loved Tableau, but we quickly realized that there was a need for a lot more pre-processing for the data." Using Alteryx, he can clean and analyze data and then easily export patient risk scores to Tableau to identify the average patient risk score for each given provider. For the first time, Mosaic can identify clear patterns. For instance, Greene can show that pediatricians have lower average risk scores, and are able to handle larger patient panels. And internal medicine providers, who see more middle-aged and chronically ill patients, have higher average risk scores and smaller patient panels.

The speed with which Greene was able to create risk scores using Alteryx made his business case for implementation to the Mosaic executive team easy. He says, "That first risk score in Alteryx took me three hours to build, and I did it while we were in our Alteryx trial period. We were trying to make a business case to the C-level staff to show the real value in this tool, so I was still learning the interface, but was basically able to create the model in three hours."

Greene's use of Alteryx has evolved to allow him to analyze patient activity for proactive outreach. The risk scores allow him to identify target patient groups in need of outreach, especially "underutilizers." These are patients who have chronic conditions, but who haven't been seen recently or regularly. In many cases, if Mosaic clinics can reach these patients and schedule them for overdue check-ups, physicians can prevent poor health outcomes.

Greene can also use Alteryx to forecast provider needs, such as staffing and exam room usage. He can use claims data from Mosaic's local Affordable Care organization to perform patient claims forecasting, and estimate cost of care.

For future improvement, Greene wants to use Alteryx to predict a patient's likelihood of having a health event in the next year, or in 10 years, so that Mosaic Medical can be more proactive in medical care.

 

^Top