We are passionate about patient access and believe in the power of sharing our ideas with others who may be interested in collaborating with us. Here are a few examples of things that have gotten us excited recently.
Predicting No Shows
Patient no-shows have a serious and substantial impact on medical groups and health systems so we wanted to see if we could do something about it. Using data that is already available in most systems we created a machine learning framework to predict patient no-shows and to intelligently overbook appointment slots. (Nerd alert: after testing many models we decided to use a number of demographic attributes using a random forest “simplified” model.) This is now embedded in DASHcentral and allows our clients to intelligently overbook patients that are likely to no-show while avoiding overbooking for patients that have a high likelihood of keeping their appointment. Learn more in the Henry J. Austin Case Study.
Unused provider inventory directly correlates with poor capacity utilization. However, most practices block off slots based on physician preferences, not demand. We wanted to see if we could account for patient demand fluctuations while simultaneously respecting provider preferences to increase capacity utilization, reduce burnout, and importantly - allow for long-term planning.
We used drivers for patient demand such as seasonal trends, holidays, and promotions. Using linear programming our models can create optimal schedules that align with forecasted demand for the year. This is a computational challenge because a 100-provider practice can have approximately 6M variables so it takes a very fast machine about two hours to compute a year’s schedule for a mid-sized practice! But this exercise can help executives conduct long-range planning in a way they hadn’t been able to before. This service is available upon request to Radix Health customers only.
Computer Machine Vision
Insurance cards are notoriously difficult for a machine to read given the heterogeneity of design - unlike credit cards, the name of the covered individual, the plan, and the different IDs may be anywhere on the card. This means that standard OCR is not helpful in translating an image of the card to text that will save the patient time in scheduling an appointment and it adds considerable friction in the patient journey; we’d like to fix that. As a “Labs” project we are experimenting with methods to make the patient experience during this process seamless by not only reading the text, but also identifying what the information represents.
Home Visit Scheduling Optimization
Several studies have highlighted the positive impact of home visits by nurses in helping patients with chronic illnesses. Using data on patients, routes, and nurses, we are using machine learning and AI to optimize mobile care delivery that increases patient access, improves patient outcomes, and reduces provider burnout. This allows more patients to receive coverage from an existing pool of care providers, prioritize care continuity, and quickly add last-minute home-visit requests without causing chaos to the schedule or avoidable overtime. We are looking for care teams with large mobile nurse programs that would be interested in piloting this technology with us.
Seeking Alpha Customers
We’re excited about partnering with innovating health care provider organizations on these. Let us know if you’d like to be an alpha client.