Over the past 5 years ORBIS - a big Dutch medical and healthcare concern - has developed a data driven approach to optimize their complete annual ‘production’ planning. This has given them an extensive operational data set regarding patient, staff and bedding capacity planning. Their ambition is to enhance their operational management around these assets in terms of:
- Real time insights (incl. root cause analysis) in their current utilization of these resources
- Predictive analysis with at least a 2-3 weeks time outlook, enabling pro-active management
- More transparency in performance and variance down to the required level of detail (e.g. process, specialism, disease, individual staff and/or patient
- Discovery of (present or developing) patterns that currently remain unknown but who might be of serious influence
- Integral view across interdependent planning entities providing ‘what if’ type of decision making
Cutting-edge technology
The technology to realise these ambitions is a combination of RapidMiner and QlikView. Although simple linear regression models for forecasting are already available in QlikView, the Orbis case needs more sophisticated models, e.g. non-linear regression, that potentially provide better prediction accuracy than simpler models.
The two-way QlikView RapidMiner integration that Xomnia built made this possible, and lead to some first (early stage) results:
- The supply of patients appears to be very predictable, both in number and diversity of diseases. As long as the health insurance system and the regulations do not drastically change, the pattern repeats itself year-on-year to 96%.
- Emergency patient volumes can be scheduled, influenced and reduced
Significant cost reduction
Hospital beds, operating rooms and diagnostic systems remain unused 11-20 % of the time. Even a fractional reduction of those rates implies a significant cost reduction. Insights provided by sophisticated forecasting models that are being processed by domain experts, will support Orbis to achieve this. To be continued!