عنوان مقاله [English]
Health services research (HSR) is of great importance to communities because decision-makers and public consider HSR as the primary source of information to determine how well health systems are meeting their specifications. Nowadays, in healthcare, there are many therapeutic processes whose results are obtained by some related stages. For studying these kinds of processes, commonly referred to as multi-stage processes, two concepts are important: one is risk adjustment, and the other is considering the cascade property. An example of a
multi-stage therapeutic process is thyroid cancer surgery, which is usually performed on patients in two stages and part of the cancerous tumors are removed at each stage. In the two-stage thyroid cancer surgery, the quality of second surgery will be affected by the results of the first stage operation. For monitoring these processes, various control charts are used including model-based control charts. To design such charts, an appropriate model should be identified at first; then, control charts could be proposed based on the identified model. In this research, a risk-adjusted time-varying linear state space model is introduced for analyzing the multi-stage therapeutic processes. The state space models are statistical models that many researchers have used to analyze multi-stage processes. These models are based on engineering
knowledge and the physical laws of real systems. Then, the model order and its parameters are estimated by Hankel singular value decomposition (HSVD) and prediction error minimization (PEM) methods, respectively. This is called input-output identification. The model performance is evaluated using numerical simulation and a real world two-stage thyroid surgery dataset. Based on the satisfactory results, one can use the model while simultaneously considering risk adjustment, cascade property, transmission error, and test error to forecast and monitor multi-stage therapeutic processes.