The Cost of System Congestion: Evidence from the Healthcare Sector

Abstract

We investigate the cost of system congestion using data from 225,473 maternity admissions at 30 hospitals in the state of Washington. We identify congested days using the distribution of patient admissions for each hospital-year combination and use the rate of Caesarian sections (C-sections) as the proxy for the cost of congestion. We use two separate logistic regressions to investigate the relation between congestion and the decision to operate. The first regression includes data from the full sample of patients. Contrary to expectations, we find that congestion does not increase C-section rates. We estimate the second regression using data only from those patients classified as being at risk for a C-section and find that congestion does lead to increased C-section rates. The difference obtains because our full sample analyses do not control for physician incentives to classify patients as being “at risk,” and whether the patient is “at risk” is included as a control in the regression. The nature of our findings also suggests that the cost of congestion manifests itself in subtle ways, underscoring the importance of understanding the specifics of the process when we examine issues such as the cost of congestion.