»As part of the quick check with Fraunhofer IPA, we were able to see the potential that AI offers for our appointment scheduling – especially for complex rescheduling. The developed prototype provides interesting impetus. Now, it's important for us to enable practical integration into our hospital information system.«
Sarah Grant – Purchasing Officer Hamm Kliniken GmbH & Co. KG
Hamm-Kliniken GmbH & Co. KG
Contact at the AI Innovation Center
Christof Nitsche
AI agent system for scheduling
Initial situation
Hamm Klinik Nahetal in Bad Kreuznach is a specialist clinic for oncological rehabilitation and follow-up rehabilitation. Around 60 new patients start their three-week stay every week. At the beginning, an individual treatment plan is drawn up based on medical guidelines and initial medical consultations. The staff then take over the scheduling and allocate the treatments to the therapists' available time slots.
This task is complicated by numerous constraints: applications should be evenly distributed throughout the stay and certain therapies must be carried out in a fixed order. Although there is a planning algorithm in the HIS system used, this does not provide optimized plans. Manual intervention is necessary, especially when there is a lack of availability: In order to create new capacities, existing appointments have to be laboriously reorganized.
Solution idea
Constraint programming was used in the Quick Check to optimize treatment planning. This approach made it possible to formally map all relevant conditions, such as sequence specifications, even distribution of applications and limited resources. The planning can be optimized across several patients, resulting in consistent and valid assignments that meet all requirements. The approach has a modular structure and can therefore be easily expanded. However, as the modelling and adaptation of such algorithms requires specialized expert knowledge, the feasibility of an agent-based system was also examined, with which the planning algorithm can be configured and further developed using natural language.
Implementation of the AI application
The implementation was carried out in close cooperation with Hamm Kliniken in order to develop an understanding of the professional and technical requirements. The planning problem was modeled as a constraint program based on real planning data from the clinic. At the same time, a simple user interface was created to test planning scenarios as a prototype. To evaluate the expandability, an agent system was also developed that incorporates new requirements in natural language and supplements them as modular constraints without changing the existing model. A prototype demonstrator was presented to Hamm Kliniken and the results are promising.