Applying survival analysis to operations management: Analyzing the differences in donor classes in the blood donation process

Abstract

Much of the research in operations management deals with the study of events. These events take many different forms: the length of time until a job is completed, the time interval until a component breaks down or the time taken for the development and introduction of a new product. When studying events, we are often interested in examining how factors such as the introduction of a new dispatching rule or a change in the level of capacity utilization affect the resulting events (e.g., completion time for a job). Analyzing events is not an easy task. Events are often time‐varying (where the probability of the event taking place changes over the life of the event), seldom normally distributed (with the distributions often being highly skewed) and affected by censoring (observations lacking a beginning or ending point). Finally, information about events is frequently collected longitudinally. These traits create problems for the application of such commonly used procedures as ANOVA. To cope with such problems, new statistical tools are needed. This paper introduces such a statistical procedure, survival analysis. Survival analysis is a set of statistical techniques (non‐parametric, semi‐parametric and parametric) used to determine quantitatively the impact of independent variables on a dependent variable which represents the time interval between events. The usefulness of this procedure is illustrated by applying it to data taken from a study focusing on the blood donation process. The paper concludes by pointing out some of the problems encountered in operations management which can be readily analyzed using this procedure. Overall, the paper tries to make the reader aware of the capabilities offered by survival analysis.

Possible Applications of Survival Analysis in Franchising Research

“…Survival analysis examines the hazard that a certain event occurs (Elandt-Johnson & Johnson, 1980). Survival analysis possesses two main aims (Melnyk et al, 1995). On the one hand, we want to estimate the time period during which the event can happen.…”

“…Formally defined, an event is ''a change in state as defined by one or more qualitative variables within some observation period and within the relevant state space'' (Blossfeld, Hamerle & Mayer, 1989). It consists of some form of change in state (Melnyk et al, 1995). According to Allison (1984) The Measurement Window…”

“…We have said that censored data are data which are incomplete. But in order to be complete, the information must meet three conditions (Melnyk et al, 1995): 1. The time during which the subject is exposed to a particular risk type must be specified; 2.…”

“…This unknown date can be close, distant or even non-existent. When data are right censored, there is a starting time but no ending point, because the process has not yet ended (Melnyk et al, 1995). For instance, right-censored data are present when some firms are still present in the market, or some stores are still open at the end of the measurement window.…”

“…Left censorship indicates the fact that we cannot determine the initial date of the event studied, for instance the date of acquisition of a product for people asked about their purchase (Ray, 1988). We can have data for which there is an ending point, but no information about when the subject was first exposed to risk (Melnyk et al, 1995). Data are left censored if the initial date of the event is unknown or prior to the measurement window.…”