Throughout the whole project, the relevant data is gathered and anonymized from the cooperation partner to form a building block for the planning process of residents. To identify criteria of fairness and stability as well as potential weaknesses experts are interviewed and the results are used to generate mathematical models. Additionally, also interviews with residents are done to evaluate their satisfaction and receive feedback for the project. The insights of these interviews are used to point out and react to the strengths and weaknesses of the project and thus support the achievement of intended goals.

In the first step, it is important to clearly define demands and capacities for surgeries, not only for shifts but also regarding the different types of physicians. This is necessary since the number of surgeries that a resident can perform within a time period and a certain clinic is an elementary component of fulfilling the different stages of education. Therefore realistic predictions help to plan the duration of the different stages.

Key figures are an essential operational instrument for comparison and regulation as well as for the measurement of target achievement. Therefore it is tried to derive key figures from other service areas, use them for the modeling of objectives and evaluate interdependencies. In this context, thresholds should help with evaluating the fitness of the key figures. In collaboration with the project partner, a method for tracking these key figures within the daily routine is developed.

In order to implement the insights of the project to hospitals, a suitable interface for personnel planning software is necessary. Therefore the correlations between education and personnel planning are analyzed and mapped to an interdependency matrix. This also helps with identifying and avoiding potential problems and intends to simplify future research.
Additionally, a guideline for researchers and practitioners sums up the arising problems in resident planning to help them understand the complexity of planning.

A clear educational structure and reliable planning improves a hospitals reputation for attractive and efficient education and thus providing a competitive advantage in employing physicians. However, hospitals need to consider high service levels and the effective use of human resources. Paying attention to all of these aspects means in a first step to determine the optimal number of education appointments with consideration of resource and capacity limitations. Due to the planning complexity, mathematical models and methods are used to first determine the optimal number of resident jobs along with the appropriate education schedules. Therefore a mixed integer problem is formulated and solved with OR/MS algorithms, like column generation, branch, price and cut or Benders decomposition. The developed methods are tested within simulations as well as case studies at the cooperation partner.

After developing the mathematical model for resident scheduling, the problem is expanded by stochastic factors that occur in reality and affect demands and capacities. Therefore on the one hand, Markov decision processes are used and on the other hand, a two-stage stochastic programming approach, which intends to provide more ex-ante information. The formerly developed methods need to be adapted to these topics.
However, uncertainties occur not only within the strategical planning but also within the tactical planning level, where the educational schedule is transformed into a concrete working roster. In this context, certainty measurements help to avoid unfairness in terms of steady progress or number of clinic transitions.
Overall, the uncertainties regarding the number of surgeries are intended to be met by a robustness concept in order to improve planning reliability and lower the number of roster changes.

Since fairness directly affects employees’ satisfaction and thus the quality of work, different fairness concepts are defined and their influence on personnel planning is evaluated. In this context, especially the number of clinic transitions and the intended influence on the continuity of care is considered. Therefore the developed models are adapted and solved heuristically, for example with local search heuristics.