Hurry Up and Wait: Differential Impacts of Congestion, Bottleneck Pressure, and Predictability on Patient Length of Stay
Executive Summary — This paper quantifies and analyzes trends related to the effects of increased workload on processing time across more than 250 hospitals. Hospitals are useful settings because they have varying levels of workload. In addition, these settings have high worker autonomy, which enables workers to more easily adjust their processing times in response to workload. Findings show that heavy load plays a significant role in processing times. Congestion is associated with longer lengths of stay. More surprisingly, when there is a high load of incoming patients from a low pressure area (emergency medical patients), current hospital inpatients' stays are longer compared to when incoming patients are from a high pressure area (emergency surgical patients). Furthermore, high predictability of the incoming patients (e.g. scheduled surgical patients) is associated with shorter lengths of stays for the current inpatients than when the incoming patients are less predictable (emergency surgical patients). In this study, there was no decrease in quality of care for patients with shorter lengths of stay. Key concepts include:
- High congestion increases patients' length of stay by up to 0.81 days, which indicates inefficiency due to overloading of resources.
- Incoming inventory load with high predictability reduces patients' length of stay by up to 0.45 days, which is enabled by the ability of a worker to plan in advance for a new work assignment by discharging a patient to make room for the incoming one.
- With highly predictable incoming patients and no congestion on the day before expected discharge, there is a shift toward discharging patients currently in the hospital one day earlier than expected.
- A hospital would benefit from adding or allocating additional resources to the inpatient hospital units, and counter-intuitively, targeting a lower occupancy level to increase productivity.
- To further improve productivity, the allocated inpatient hospital resources could include adding a nurse on the hospital floors who is solely responsible for discharges and admissions.
High workload, from high inventory levels, impacts unit processing times, but prior operations management studies have found conflicting results regarding direction. Thus, it is difficult to predict inventory's effects on productivity a priori, inhibiting effective capacity management in high load systems. We categorize load into in-process inventory (congestion) and incoming inventory, decomposing the latter into its levels of bottleneck (BN) pressure and predictability, and quantify the magnitudes and directions of change on processing times. Using data from 283 hospitals, we find (1) high congestion increases a patient's hospital stay up to 28%, indicating inefficiencies from overloaded resources; (2) a patient stays up to 11.7% longer if there is a high load of incoming low BN pressure patients, consistent with the slowdown associated with "social loafing"; (3) a patient's stay is up to 10.2% shorter when there is a high incoming load of predictable patients, consistent with workload smoothing.