Healthcare Operations Management is concerned with decision support planning and analysis of (service, production and information) processes in the health care with quantitative, empirical and qualitative methods.
The following areas of Healthcare Operations / Health Information Management are the focus of our activity:
Over the past 20 years, hospitals have seen a sharp increase in internal costs. These are the result of steadily growing patient numbers and the resulting increased personnel requirements. Since approximately half of the costs incurred are generated by the hospital staff, there is a need for efficient personnel staffing and scheduling to minimize personnel costs while maintaining a targeted quality of service.
In practice, shift plans are usually created by experienced physicians. This is, on the one hand, very time-consuming and cost-intensive, and on the other hand often only produces a result of medium quality since it is usually not possible for a human planner to consider all restrictions equally during the planning process. These restrictions and regulations include, for example, labor law provisions that are legally mandatory, but also aspects that relate to fairness and equal rights within the personnel department. Manually created plans often violate many of these conditions and lead to dissatisfaction. In this context, software-based mathematical optimization offers the possibility to generate plans that take account all of the rules to be considered within a short time.
The complexity of training increases in highly technological industries. Additionally, job training is often times not only time-related but also task-related. Predicting the exact number of procedures that a trainee can perform in these periods is not always possible. Accordingly, a trainee might not be able to perform all of the needed procedures resulting in an extension of the apprenticeship.
An important aspect of the quality of professional development is that the trainees must know in advance what kind of tasks will be performed in the following periods and in which department they will take place. This is not only relevant for trainees but also for personnel management. By means of a given curriculum, the management not only knows in advance who will join the department in the future, but can also assess the knowledge of the new trainee. In addition, it is desirable for organizations to ensure compliance with the promised duration of the overall training. Particularly within the framework of the work-life / family balance, a fixed duration of a training program is important. By using the same curriculum for more than one trainee, the systematics of a teaching can be increased. While the trainees are able to compare their performance level with other apprentices and share their experiences with younger trainees, the instructor is able to improve the skills of the training due to the repeating structure. Examples that include job-related training are training programs of physicians, as presented in our experimental study, and police training, but also some smaller complex programs such as driving license training.
Nurse Planning and Scheduling
Our research in this area focuses on mathematical procedures that allow us to quantify the benefits of flexibility provided by the deployment of cross-trained nurses. Key performance indicators include, for example, fairness issues such as the number of assigned weekend shifts and satisfied employee preferences, the average workload of staff, and adherence to given patient-to-nurse ratios. Flexible nurses allow you to respond to fluctuations in demand at different wards in a hospital by being able to quickly assign them to different medical units. If this flexibility is already taken into account in the medium-term resource planning, shift schedules can be generated that are better suited to deal with unforeseeable fluctuations in care requirements.
In practice, we support the Augsburg Hospital in the process of introducing Automated Employee Planning (AEP). The AEP uses rule-based heuristics in the HR management software to provide schedulers with (near) optimal monthly rosters. The quality of automatically generated rosters depends heavily on the initial configuration of priorities, employee skills and availability, and weighting of different objectives. Therefore, it requires a precise adjustment of the data and input factors that are used in the daily use of AEP.
Furthermore, we deal with the nurse planning in the emergency department of the Augsburg Hospital. It is important to optimize the shift design and personnel planning for the seasonal (hourly, daily and monthly) fluctuations in patient arrivals. An inflexible shift design with few shift types, such as "early shift", "late shift" and "night shift", can, for example, lead to serious under- and overstaffing, because the distribution of patient arrivals cannot be adequately matched. This results in systematically distributed times during the day, in which nurses are too busy or idle. With mathematical optimization, additional shift types can be determined to ensure a more balanced staff utilization.
Participating Team Members:
Health care always deals with the compromise between providing better health for all and long-term financial viability. In order to be able to carry out appropriate treatments at moderate costs, an efficient and effective use of involved resources is indispensable. Capacity planning includes dimensioning, allocation, long-term and short-term planning, as well as control and management of resources. These include premises (e.g., operating rooms, normal and intensive care units, emergency rooms), equipment (e.g., sterile instruments, medical consumables), and personnel (physicians, nurses, others).
Case Mix Planning
In most western countries, an aging society and a corresponding change in disease patterns can be identified. It is essential to ensure adequate availability of resources by anticipating such developments in the long term in order to ensure adequate health care. The long-term planning of the patient spectrum is referred to as Case Mix Planning. This problem is closely intertwined with the long-term allocation of resources to different entities. The resources to be considered include, for example, operating rooms, beds on intensive and normal stations as well as personnel. We analyze effects of, e.g., scale effects and random events, on the strategic planning of suitable patient spectrums using real life data.
Intensive care unit capacity planning
The intensive care unit represents a crucial and expensive resource for the hospital, which is strongly affected by uncertainty and variability in demand. As a result of the limited capacities, it is often necessary to find a compromise between high capacity utilization and a high level of service. When dealing with an intensive care unit that often reaches its limits, both the admission control of the new arrivals and the demand for premature discharges of existing patients should be taken into account. Obviously, the rejection of new patients as well as earlier discharges can lead to an increase in mortality rates and monetary risks. Therefore, it is important to find optimal admission and discharge policies to minimize the aforementioned negative consequences. This decision-making problem can, for example, be modeled as a discrete Markov decision-making process. Strategies for a wide variety of scenarios can thus be derived and evaluated using medical and monetary criteria.
Within a hospital, the operating room is not only one of the most frequented departments with 60-70% of all stays including at least one visit at the OR, it is also one of the most expensive departments, accounting for more than 40% of the total expenses of a hospital. Many factors influence the scheduling process of the department, making it difficult to reach a high utilization. When scheduling surgeries, not only resources needed within the department - e.g. physicians, anesthetists, nurses, and available ORs with the desired equipment - have to be considered, but also the resources of up- and downstream units. Connected departments, e.g. the patient wards, the post-anesthesia care unit, and the intensive care unit, have also limited resources and are directly affected by the planning decisions of the OR department. We use mathematical programming and develop linear programs to define scheduling rules within the OR department, which reduce peaks in demand of connected departments. Moreover, we apply machine learning methods to consider the impact of surgeries on downstream units. This allows to utilize the available capacity to the fullest and removes potential bottlenecks.
By assigning appointments, a clinic can directly influence the waiting times of their patients and the utilization of their staff. A reduction in waiting times is often accompanied by a reduction in capacity utilization and vice versa. In the literature these factors are therefore usually regarded as contrasting costs. The minimization of these costs is additionally made difficult by several stochastic influences. Patients look for their doctor because of different illnesses and have to be treated differently. Some of them visit a clinic without first making an appointment, are unpunctual or have an appointment and still do not appear. We develop models that help the scheduler to decide when which patient is to be scheduled.
Participating Team Members:
Real or simulated data, their description but also induction represent an integral part of performance measurement of systems. The Chair for Healthcare Operations / Health Information Management is concerned in particular with the following sub-areas of performance measurement:
Data Envelopment Analysis
Data Envelopment Analysis (DEA) is the most frequently used method for the efficiency measurement of non-profit organizations. It can be used, for example, to compare the performance of hospitals or universities. Best practice examples can be identified and learned from. Since the invention of the DEA method, a large variety of models has been developed. However, which of these models provides the most accurate and robust results is still unclear. With our research, we try to answer these questions in order to enable a reliable efficiency measurement in the health care system. Crucial for the benchmarking of different DEA models is the generation of artificial data. By the use of Monte-Carlo-Simulation, we generate a variety of scenarios where, among others, different efficiency settings, amounts of inputs and scale elasticities are utilized. Several instances of the same scenario are used, to guarantee the robustness of the results. Finally, the accuracy of the models is determined by different performance indicators.
Apart from the benchmarking of models, we published a literature review in the Data Envelopment Field, which is concerned with applications in the healthcare sector. Besides various useful descriptive statistics, we provide a roadmap to important methodological literature and publications, which provide crucial information on the setup of DEA studies. Further information on the literature review can be found here.
Allergic rhinoconjunctivitis, also referred to as hay fever, is among the most common allergic diseases in Germany and is associated with a high loss of quality of life in the affected population. The treatment of the pollen allergy is often in the hands of the affected themselves. In order for allergy sufferers to be able to react in a targeted way and in advance to the allergen exposure, they need information on the concentration of allergenic pollen in the air. Since 2015, the University Center for Health Sciences at the Augsburg Clinic (UNIKA-T), initiated by the Chair of Environmental Medicine (Prof. Claudia Traidl-Hoffmann) and the Chair of Healthcare Operations / Health Information Management (Prof. Dr. Jens O. Brunner), operates an innovative fully automatic pollen monitor from the company Hund at the Bavarian State Office for Environment in Augsburg. The device recognizes the majority of allergy-sensitive pollen taxa and transmits measured data to the UNIKA-T server on a 3-hour rhythm. We apply machine learning to identify and classify pollen. The project envisages the development of an innovative pollen application to provide this information to pollen allergists.
Testing more than one hypothesis based upon the same data set can lead to an inflation of type I errors respectively false positive research results. Therefore, statisticians and especially physicians have developed an abundance of methods for such multiple testing problems. Although highly relevant, these methods are not applied in sophisticated management research as our analyses show. The medical context of current multiple testing research, the abundance of methods and the two step method selection process seem to provoke this situation. We address the first points by distinct method recommendations for business researchers based on Monte Carlo simulations. In addition, our new multiple tests (SiMaFlex and SteMaFlex) are able to transform the two stage selection to a one stage selection procedure. Performance analyses show that these methods dominate well known multiple tests. By focusing these application questions, we hope to transfer simultaneous inference from medicine to management research.