Open Collections

Strategic resource planning in healthcare under uncertainty : British Columbia Cancer Agency and Vancouver's Downtown Eastside

University of British Columbia

In two very different healthcare settings we demonstrate the benefits of long term planning using operations research (OR) tools. We present models that handle considerable variability using solutions based on relatively simple approximations. In the first setting we present a mixed integer program (MIP) with a goal programming (GP) formulation for strategic workforce planning at the British Columbia Cancer Agency (BCCA). Our model considers experience, minimum and maximum durations, and redundancy in staffing to guard against unanticipated employee leaves. We evaluate the model parameters using simulation, and analyze the simulation output with logistic and Poisson regression. The core model can be generalized to other workforce planning applications in healthcare or to other human resource intensive industries; the full BCCA model illustrates a real-world implementation. This research introduces to the workforce planning literature a technique for building robustness into the plan, together with experience and duration constraints. In the second setting we study a marginalized population for which myriad organizations provide healthcare and other services in the absence of system-level quantitative planning. We use a queueing network to model clients with complex concurrent disorders (CCD) flowing through services in Vancouver’s Downtown Eastside (DTES). We perform sensitivity analysis on the input parameters, validate our solution against a simulation model, and conduct scenario comparisons to evaluate potential procedural and policy changes to the system. To analyze this network we present a novel approximation technique—-called a linearized closed queueing network (LCQN)—-for solving closed queueing networks. By using an open queueing network with the fixed population mean (FPM) approach, and by including a trick for dealing with capacitated stations, we create a network representation that is solved with a linear program (LP). This method scales to much larger systems. We derive the approximation ratio between this approximation and the exact solution for a small network, and use simulation to show that this gap is of no practical significance for the full network.

Thesis/Dissertation

eng

Vancouver : University of British Columbia Library

10.14288/1.0073451

Doctor of Philosophy - PhD

Business, Sauder School of

2013-05

Graduate

DSpace