Scheduling chemotherapy appointments under uncertainty by considering different nursing care delivery schemes

TL;DR

This paper presents a stochastic optimization model for scheduling chemotherapy appointments that accounts for nursing care flexibility and infusion duration uncertainty, improving patient wait times and nurse overtime.

Contribution

It introduces a versatile two-stage stochastic mixed-integer programming model applicable to various nursing care models and develops novel scenario grouping algorithms for efficient solution.

Findings

Near-optimal schedules outperform current clinic policies.

Partially flexible care models significantly reduce waiting time and overtime.

Scenario grouping methods impact solution quality and computational efficiency.

Abstract

The flexibility level allowed in nursing care delivery and uncertainty in infusion durations are very important factors to be considered during the chemotherapy schedule generation task. The nursing care delivery scheme employed in an outpatient chemotherapy clinic (OCC) determines the strictness of the patient-to-nurse assignment policies, while the estimation of infusion durations affects the trade-off between patient waiting time and nurse overtime. We study the problem of daily scheduling of patients, assignment of patients to nurses and chairs under uncertainty in infusion durations for an OCC that functions according to any of the three commonly used nursing care delivery models representing fully flexible, partially flexible, and inflexible care models, respectively. We develop a two-stage stochastic mixed-integer programming model that is valid for the three care delivery models to minimize expected weighted cost of patient waiting time and nurse overtime. We propose multiple variants of a scenario grouping-based decomposition algorithm to solve the model using data of a major university oncology hospital. The variants of the algorithm differ from each other according to the method used to group scenarios. We compare input-based, solution-based and random scenario grouping methods within the decomposition algorithm. We obtain near-optimal schedules that are also significantly better than the schedules generated based on the policy used in the clinic. We analyze the impact of nursing care flexibility to determine whether partial or fully flexible delivery system is necessary to adequately improve waiting time and overtime. We examine the sensitivity of the performance measures to the cost coefficients and the number of nurses and chairs. Finally, we provide an estimation of the value of stochastic solution.