Stanford Research Communication Program
  Home   Researchers Professionals  About
Archive by Major Area


Social Science

Natural Science

Archive by Year

Fall 1999 - Spring 2000

Fall 2000 - Summer 2001

Fall 2001 - Spring 2002

Fall 2002 - Summer 2003




Changing Work Behavior in Medical Organizations

Carol Cheng Cain
Department of Biomedical Informatics
Stanford University
December 2001

Medical care is a complex, team-based activity which requires coordination among the doctors, nurses, and support staff. The decisions each one makes can affect the work of the others in unpredictable ways. Hospital administrators have noticed that medical workers sometimes change their behavior based on factors which are unrelated to their patients. For example, they order fewer tests when their workload is heavy, and limit specialist referrals during the evening. These changes often have unanticipated effects on the care patients receive. I am creating a computational tool that models medical work. The model results predict how medical care workers will respond to the changing demands of their work, and how their response affects the coordination efforts and work completed. My model can help administrators design better work practices and anticipate how staffing and procedural changes will affect patients.

Existing computational tools help doctors modify the patient's treatment plan based on the patient's health. Consider the case where the doctor delays regular antibiotic treatment because the patient's health has deteriorated. My approach focuses instead on the medical worker's environment. The same doctor may also choose to delay therapy because he is off work the next day, and wishes to see the patient the day after starting antibiotic therapy. The decision results in a longer patient stay, which is undesirable to the hospital and the patient. Had the hospital organization been able to predict this behavior, they could have intervened by assigning the patient to another doctor, or allowing a nurse practitioner to monitor the patient. By addressing the work context of the medical team, my research helps administrators examine their intuitions about important contextual influences on medical care decisions.

My system, called the Context Aware Virtual Health Administration Team (CAVHAT), represents four work contexts:

-Time-based cyclical patterns such as day and night
-Schedule-based contexts such as length of time on shift
-Workload of clinicians and their backlogs
-Availability of colleagues for coordination and consultation

The model includes the workers in an organization, their work responsibilities, and the organization's policies for handling coordination and consultation. The system simulates the behavior of the organization as it tries to complete its work, and outputs potential scenarios. CAVHAT predicts outcomes of patient care such as the completion time, communication quality, the probability of failure, and adherence to pre-specified work processes. The simulation behavior is probabilistic: clinicians have a preference for working a certain way, but may not do so every time. Simulation experiments work well in studying organizations because empirical experiments are expensive and sometimes infeasible.

I use CAVHAT to run two kinds of tests: idealized and realistic. In idealized experiments, I create very simple models of prototypical organizations. By altering their policies, I can generate theories about how clinicians will respond contextually. For example, an organization whose policy is to refer all decisions to the head physician may find that the head physician becomes swamped with work, and clinicians begin to make decisions on their own, in order to speed up patient care. In realistic experiments, I observe medical organizations and create models which closely parallel the people and tasks. The model predictions in this case are very useful to the particular organization, but are not easily generalized. In both cases, the models are useful for sensitivity analyses to see which policies most strongly affect the results, and as a testbed for new policies and ways of organizing.

I am still developing CAVHAT and performing idealized experiments to test its capabilities. Over the next year, I will be observing and modeling work in the Intensive Care Unit (ICU), and then testing CAVHAT's ability to represent the ICU and make accurate predictions. CAVHAT will assist in our understanding of how contextual behavior affects organizational performance.