Decision support for hospital bed management using adaptable individual length of stay estimations and shared resources

摘要: Elective patient admission and assignment planning is an important task of the strategic operational management a hospital early on became central topic clinical operations research. The beds subtask. Various approaches have been proposed, involving computation efficient assignments with regard to patients’ condition, necessity treatment, preferences. However, these are mostly based static, unadaptable estimates length stay and, thus, do not take into account uncertainty patient’s recovery. Furthermore, effect aggregated bed capacities investigated in this context. Computer supported management, combining adaptable estimation treatment shared resources (aggregated capacities) has yet sufficiently investigated. aim our work is: 1) define cost function for taking estimations resources, 2) mathematical program formally modeling problem architecture decision support, 3) investigate four algorithmic methodologies addressing one base-line approach, 4) evaluate w.r.t. outcome, performance, dismissal ratio. expected free ward capacity calculated individual estimates, introducing Bernoulli distributed random variables occupation states approximating probability densities. represented as binary integer program. Four strategies solving applied compared: exact using mixed programming solver SCIP; three heuristic strategies, namely longest processing time, shortest choice. A baseline approach serves compare optimization simple model status quo. All evaluated by realistic discrete event simulation: outcomes ratio successful dismissals, model’s factors. simulation 226,000 cases shows reduction rate compared more than 30 percentage points (from mean 74.7% 40.06% comparing quo strategies). Each leads improved assignment. only marginal advantage over factors (≤3%). Moreover,this was achieved at price computational time fifty times that models (an average computing 141 s method, vs. 2.6 strategy). In terms its performance quality solution, strategy RAND preferred method case resources. Future research needed whether equally marked improvement can be large scale application study, ideally comprising all departments involved planning.