Robots with a Social Memory

Abstract

More than occasionally we hear the protest that “People are not robots!” when trying to bring consistency to medical practice. We certainly agree, although the era of absolute clinical autonomy in medicine is long over. Ledolter et al.1 are likely in for a round of “not robots!” protests with their contribution in this month's issue of Anesthesia & Analgesia. People will take exception to being compared to machines that process jobs, and the implication that jobs are kept waiting because of poor decisions by the machines. Tough! The server/job analogy describes any system in which a limited, high-value resource (e.g., a physician) sequentially serves a series of clients (patients). This simple observation in no way devalues the unique role anesthesiologists play in patient care. Ledolter et al. make a simple observation: at night, when 3 operating rooms are allotted to be available, the on-call anesthesiology team frequently runs only 2 rooms, delaying service for patients, surgeons, and the rest of the hospital system. Several important questions arise. Is waiting a function of which anesthesiologist is on-call and making the choices about activating the second-call anesthesiologist, or are all alike in their decisions to let everyone else wait? Does the amount of waiting change over time? Do policy changes or implementation of organizational structure influence performance? These are important issues for hospital and anesthesia practice leadership. Ledolter et al. demonstrate that a simple method for doing so is as good as more complex techniques. Statistical Process Control (SPC) methodologies provide powerful, useful tools for monitoring waiting as a function of time (i.e., to answer the question, “Is my system performance changing or is it the same over time?). Despite the intimidating name, SPC methodologies need not be intimidating. Ledolter et al. demonstrate that a simple SPC technique developed by Shewhart, meant to be implemented with pencil and paper, is just as good as more sophisticated, less accessible methods. This conclusion that simple methods will serve is refreshing. The going gets heavier when trying to decide whether all anesthesiologists are alike in their decision making. Here, it is important to know whether all of the anesthesiologists (or more specifically, their observed behaviors) come from the same population, or whether some of the anesthesiologists are different from the others. Picking the correct distribution with which to model the population of anesthesiologist decisions is the critical step. Here, Ledolter et al. opt for the negative binomial distribution. What is it, and why use it? The negative binomial distribution accurately describes the probability, in successive tosses of a fair coin, of getting 1, 2, 3, …, n heads (or tails) in a row, or of having a third case waiting to start once 2 cases are already under way. Through hypothesis testing with reference to the negative binomial distribution, Ledolter et al. conclude that all anesthesiologists are the same. They (or “we”) make poor resource utilization decisions, letting a third case wait (including patients, surgeons, nurses, and staff) instead of calling in the second-call anesthesiologist. Poor resource utilization is one potential lens through which the anesthesiologists' decisions can be examined. Decisions involving calling in additional providers are inevitably influenced by more factors. The group of physicians will all share roles as the “caller” or the “one called in” over the course of time. Social psychology gives insight into how people make decisions given these constraints. Game theory, in the form of the prisoner's dilemma, is a useful construct to explain cooperative behavior (defined as cooperation with a second “prisoner,” in this case the second anesthesiologist) and uncooperative behavior. Two prisoners are presented with a choice: “cooperate” with each other, meaning that they remain silent during interrogation,