Background: Unpredictable patient census is a common problem amongst hospitalist groups, especially in large academic and community hospitals. Traditionally, hospitalist groups do not have a formalized staffing mechanism to deal with census variability, which can create physician work overload and dissatisfaction, delays in appropriate staffing, and compromise patient safety. In our group, dealing with spikes in census resulted in multiple phone calls and emails with variable reliability, poor management of clinical burden, and delays in care.

Purpose: Develop and implement a service model with an explicitly flexible and reliable activation scheme to manage high census variability. This model would help management of clinical burden and also help alleviate administrative burden by creating a predictable and reliable pathway for staffing.

Description: In the formative years of our program, our academic hospitalist group did not have a formalized mechanism for dealing with spikes in patient census or admission rate. Our release valve was ad hoc: hospitalists were called in last minute or asked to care for patients beyond the intended model. With a continually expanding daily census, this informal process became increasingly ineffective and burdensome.

We then moved to a model of pre-scheduling moonlighters on a monthly basis to work when needed. However, we were unable to reliably fill our moonlighting shifts, and the model still required constant coordination of the schedule.

After several iterations, we developed a specific shift to be utilized for high census situations. These “jeopardy” shifts were assigned to the entire group according to clinical full-time equivalent (FTE) percentage. The shifts were released in advance for the entire academic year as part of our normal scheduling process. We also developed an algorithm by which the shift would be activated.

Data collected on the usage of this shift over the past 5 months shows that it was activated in a standardized fashion by our hospitalists an average of 2.4 times per month, without the involvement of any external administrative guidance.

Conclusions: Implementing an explicit shift designated to manage census needs along with an activation algorithm resulted in minimal but effective usage during times of high census. This process has created a more reliable system for our hospitalist group to deal with variable census, with a goal of distributing work load more evenly and providing safer patient care.