Payment systems and hospital length of stay: a bunching-based evidence

Despite the huge attention on the long average hospital length of stay (LOS) in Japan, there are limited empirical studies on the impacts of the payment systems on LOS. In order to shed new light on this issue, we focus on the fact that reimbursement for hospital care is linked to the number of patient bed-days, where a "day" is defined as the period from one midnight to the next. This "midnight-to-midnight" definition may incentivize health care providers to manipulate hospital acceptance times in emergency patients, as patients admitted before midnight would have an additional day for reimbursement when compared with those admitted after midnight. We test this hypothesis using administrative data of emergency transportations in Japan from 2008 to 2011 (N = 2,146,498). The results indicate that there is a significant bunching in the number of acceptances at the emergency hospital around midnight; the number heaps a few minutes before midnight, but suddenly drops just after midnight. Given that the occurrence of emergency episode is random and the density is smooth during nighttime, bunching in the number of hospital acceptances around midnight suggests that hospital care providers shift the hospital acceptance times forward by hurrying-up to accept the patients. This manipulation clearly leads to longer LOS by one bed-day. In addition, the manipulation is observed in the prefectures where private hospitals mainly provide emergency medical services, suggesting hospital ownership is associated with the manipulation of hospital acceptance time.