Advance admission scheduling via resource satisficing

Abstract

We study the problem of advance scheduling of ward admission requests in a public hospital, which affects the usage of critical resources such as operating theaters and hospital beds. Given the stochastic arrivals of patients and uncertain usage of resources, it is often infeasible for the planner to devise a risk‐free schedule to meet these requests without violating resource capacity constraints and creating adverse effects that include healthcare overtime, long patient waiting times, and bed shortages. The difficulty of quantifying these costs and the need to safeguard against resource overutilization lead us to propose a resource satisficing framework that renders the violation of resource constraints less likely and also diminishes its impact whenever it occurs. The risk of resource overutilization is captured by our resource satisficing index (RSI), which is calibrated to reflect a risk‐adjusted utilization rate for a better interpretation to the healthcare planner. Unlike the expected utilization rate, RSI is risk‐sensitive and serves to mitigate the risks of overutilization better whenever overutilization can be avoided in expectation. Our satisficing approach aims to balance out the overutilization risks by minimizing the maximal RSI among all resources and periods. Under our proposed partial adaptive scheduling policy, the resource satisficing model can be formulated and solved via a converging sequence of mixed‐integer linear optimization problems. A computational study establishes that our approach reduces resource overutilization risks to a greater extent than the benchmark methods.

Feature‐driven robust surgery scheduling

“…In Subsection 3.3, inspired by the recent work of Zhou et al. (2022), we make the scheduling decisions of emergencies partially adaptive to some of the system state information. (4)( Computational intractability ) Note that even computing the objective function in () for some given allocation decisions $$\bm {x}$$ is #P‐hard in general (Hanasusanto et al., 2016). To solve model (), one usually resorts to SAA.…”

“…Our model naturally describes elective and future possible emergency patients, and it can find solutions that adapt to the feature information of emergencies. A similarly adaptive allocation approach is proposed by Zhou et al (2022) in a different setting.…”

“…Our adaptive allocation policy is inspired by that proposed in Zhou et al. (2022), in which the ward admission and scheduling decisions are adapted to patient types. We introduce their idea into our setting and enhance it by explicitly using patient features to define patient types.…”

“…A similarly adaptive allocation approach is proposed by Zhou et al. (2022) in a different setting. We introduce their idea into our setting and enhance the approach using patient features to define patient types. (iv) Tractability and numerical validation : Our approach can be reformulated as a mixed‐integer second‐order cone programming.…”

“…The state space is huge and grows exponentially with the surgery duration as well as with the quantity, arrival times, and types of emergency patients. It remains challenging, if not impossible, to establish a fully adaptive model that is tractable (Zhou et al., 2022). Because the elective patients are allocated to ORs daily and emergency patients are allocated in real‐time upon arrival, a trade‐off between model fidelity and computational tractability should be made.…”