Predictor Looks to Reduce Patient No-Shows

Among the reasons for skyrocketing U.S. healthcare costs is the relatively mundane problem of patient no-shows, which end up costing the healthcare sector billions of dollars in lost revenues each year. In response, healthcare analytics developers have come up with predictive models that help hospitals and clinics gauge when appointments might not be kept.

Using a data science platform developed by New York-based startup Dataiku (as in, “data-haiku”), healthcare analytics specialist Intermedix Corp. built a tool for predicting which patients are most likely to miss scheduled appointments. The partners said the tool is now being used in more than 50 U.S. private clinics.

The problem of patient no-shows is common and growing. The analytics partners cited estimate that between 5 and 10 percent of patients missed scheduled appointments. In response, some clinics have begun charging fees if patients do not confirm 24 hours before a scheduled visit to the doctor’s office.

The partners also cited studies estimating that primary care physicians lose about $228 for every no-show, resulting in lower reimbursements from insurers. Reduced revenues can have a cascading effect on clinical outcomes, and losses for specialist are estimated to be even higher.

In response, more clinics are said to be turning to analytics to reduce the probability of no-shows and the resulting lost revenues. One approach is using raw, or “heterogeneous,” patient data to fine-tune patient scheduling.

Enter Intermedix, which formed a data science team that used the Dataiku platform to develop a “no-show predictor.” The platform was used to ingest historical and demographic patient data. Based on the results, a predictive model was built to score individual patients based on the probability they would miss an appointment.

Office managers and schedulers are then alerted so they can target reminders to patients deemed most likely to miss their appointment.

The partners claimed such a predictive analytics model would normally take more than three months to develop. Their collaboration reduced development time required to deliver a more accurate prediction tool to about one month.

Dataiku promotes its data science studio as squeezing more results from raw data through a collaborative interface that links expert and beginner analysts with data scientists. The platform is intended to capture different skill sets in the development of analytics tools that can transform raw data into predicted outcomes.

Along with the healthcare sector, Dataiku said its more than 80 customers range from e-commerce and industrial to financial and insurance companies using its tools for application such as fraud detection, demand forecasting and predictive maintenance.

The startup has so far raised $3.7 million from a pair of investors as it seeks to expand internationally.

Recent items:

Data Startup Targets Machine Learning For Healthcare

Big-Data Backlash: Medical Database Raises Privacy Concerns