Contextually Changing Behavior in Medical Organizations

Agent-based simulation is a new vehicle for researching and designing effective protocols in organizations. We have extended an existing framework for modeling engineering organizations to represent the less routine, dynamic tasks facing medical organizations. We model changes in a decision maker's behavior based on their temporal context and workload, and measure the effects on organizational communication and coordination. Our initial results indicate that such behaviors can have unintended effects on protocol compliance. Recent studies"2 have documented systematic differences in healthcare activities based on nonclinical factors such as the time of day or the month of the year. These differences arise despite organizational goals for uniformity of care. Organizational policies often fail to consider the work context in which clinicians make decisions, an important component of the quality of decisionmaking.3 As medical care becomes more comnplex, the communication and coordination tasks among health care teams becomes non-trivial. A simulation model of the work process can assist in diagnosing areas of unintended variability, and yield insight into the most effective interventions. We have been developing a model of service organizations involved in non-routine, diagnose-andrepair tasks4'5 such as medical care. Because this is often a continuous process, we have extended the model to include non-clinical work contexts such as diurnal or weekly patterns (Fig.l). We also include the contextual effects of concurrent care processes. Here we present on-going work on the representation of such contexts and predictions of their effects on decision making behavior and process outcomes. Our model represents the actors in an organization, their roles, skills, and experience levels, as well as their tasks and task alternatives. We derive contexts including cyclical temporal patterns, task delays, concurrent workload, and colleague availability. In these contexts, an actor's decisionmaking behavior deviates predictably from protocol. We focus on contexts which have had a documented effect on decision-making behavior in the medical domain. We model the effects of decisions made by lower-skilled actors, and of actors delaying work or hurrying through tasks. Our goals are two-fold: to describe the heuristics used by decision makers to respond to recurring contexts, and Fiure 1. Contextual changes In behaIorn Diumal exmple Alhogh ending physiians mae tremnt decisions durng th day, in IheeYeung$uch decisions may fall t resident. to measure the effects of these local behaviors on organizational-level outcomes such as: task conpletion time; communications sent, replied to, or ignored; errors corrected or ignored. We validate the model by representing situations described in the literature. Our initial results indicate that, although local contextual optimization may decrease the time individual actors spend on direct work tasks, it can also reduce coordination efforts and increase total task time because of subsequent rework. Thus the aggregation of small decisions can lead to significant changes in the "macro-behavior" of the organization. Contextual changes in decision nialing are an understudied phenomenon with potentially large effects on process execution. Simulation models are a controlled, cheap, succinct and quick alternative to experimental organization studies As system-level approaches to quality improvement gain importance, these models can illuminate the desired and undesired effects of proposed clinical practice protocols and guidelines.