Background: It has been nearly a decade since the National Academy of Engineering and the Institute of Medicine recommended the application of systems engineering approaches in order to deliver healthcare that is efficient, effective, and patient-centered, yet large academic medical centers remain highly complex systems that are rife with inefficiencies. One area that has been shown to contribute to inefficient, ineffective patient care and poor patient and provider satisfaction is the geographic dispersion of traditional inpatient care teams. Simulation modeling has been used successfully to identify flow issues in the emergency department setting and has the potential to improve geographic assignment of admitted patients in the setting of variables such as high occupancy and the need for appropriate workload balance. In order to do this, the simulation technique must first be validated using historical distribution data.
Methods: Admissions data for the Internal Medicine housestaff teams for calendar year 2013, including time/date of admission, initial bed assignment and discharge time/date, was extracted from the Electronic Health Records (EHR). Length of stay was calculated by the differential between admission and discharge times. Utilization of team patient load capacity for EHR data was calculated using Markovian queuing theory. Applying systems engineering techniques with SIMIO simulation software, a current state model was developed to represent team admission rate and volume, bed location distribution and any queuing processes. For validation purposes, 50 iterations of the current state model were performed to decrease the variance of: 1) the mean of the number of patients in each unit, 2) the mean team census; and 3) the percentage utilization of each team’s patient load capacity. The validated estimate was compared with the 12 month EHR data from 2013. A linear regression model and goodness of fit test was performed between the validated simulation data and EHR data.
Results: A total of 3,953 patients were admitted to the teams. These patients were distributed across 17 nursing units throughout the hospital. The mean number of patients admitted to different nursing units was estimated from the simulation model then compared to the EHR data from 2013 respectively; the differences were not statistically significant (p >0.05). Average team census based on EHR data was noted to be 12.0 compared to the validated simulation model of 11.5. EHR derived percentage utilization of each team’s patient load capacity was 72.0% compared to the validated simulation model of 72.2%. Goodness of fit for the validated model as compared to the EHR data resulted in a coefficient of determination of r2= 0.99, signifying the validated simulation model is similar to the EHR data from 2013 and therefore reliable.
Conclusions: Systems engineering principles such as simulation provide scientific and data driven methods to model highly complex processes, an example of which is the assignment of patients to hospital beds and care teams in a large academic medical center. Once validated by historical data to accurately reflect a particular health system’s constraints, simulated distribution models can be developed and modified based on current state variables such as hospital occupancy or changes in provider-specific characteristics such as workload, allowing for targeted interventions that decrease waste both in time and effort over traditional methods of operational improvement.