Barriers to Early Hospital Discharge: A Cross-Sectional Study at Five Academic Hospitals

Hospital discharges frequently occur in the afternoon or evening hours.^1-5 Late discharges can adversely affect patient flow throughout the hospital,^3,6-9 which, in turn, can result in delays in care,^10-16 more medication errors,¹⁷ increased mortality,^18-20 longer lengths of stay,^20-22 higher costs,²³ and lower patient satisfaction.²⁴

Various interventions have been employed in the attempts to find ways of moving discharge times to earlier in the day, including preparing the discharge paperwork and medications the previous night,²⁵ using checklists,^1,25 team huddles,² providing real-time feedback to unit staff,¹ and employing multidisciplinary teamwork.^1,2,6,25,26

The purpose of this study was to identify and determine the relative frequency of barriers to writing discharge orders in the hopes of identifying issues that might be addressed by targeted interventions. We also assessed the effects of daily team census, patients being on teaching versus nonteaching services, and how daily rounds were structured at the time that the discharge orders were written.

Study Design, Setting, and Participants

We conducted a prospective, cross-sectional survey of house-staff and attending physicians on general medicine teaching and nonteaching services from November 13, 2014, through May 31, 2016. The study was conducted at the following five hospitals: Denver Health Medical Center (DHMC) and Presbyterian/Saint Luke’s Medical Center (PSL) in Denver, Colorado; Ronald Reagan University (UCLA) and Los Angeles County/University of Southern California Medical Center (LAC+USC) in Los Angeles, California; and Harborview Medical Center (HMC) in Seattle, Washington. The study was approved by the Colorado Multi-Institutional Review Board as well as by the review boards of the other participating sites.

Data Collection

The results of the focus groups composed of attending physicians at DHMC were used to develop our initial data collection template. Additional sites joining the study provided feedback, leading to modifications (Appendix 1).

Physicians were surveyed at three different time points on study days that were selected according to the convenience of the investigators. The sampling occurred only on weekdays and was done based on the investigators’ availability. Investigators would attempt to survey as many teams as they were able to but, secondary to feasibility, not all teams could be surveyed on study days. The specific time points varied as a function of physician workflows but were standardized as much as possible to occur in the early morning, around noon, and midafternoon on weekdays. Physicians were contacted either in person or by telephone for verbal consent prior to administering the first survey. All general medicine teams were eligible. For teaching teams, the order of contact was resident, intern, and then attending based on which physician was available at the time of the survey and on which member of the team was thought to know the patients the best. For the nonteaching services, the attending physicians were contacted.

During the initial survey, the investigators assessed the provider role (ie, attending or housestaff), whether the service was a teaching or a nonteaching service, and the starting patient census on that service primarily based on interviewing the provider of record for the team and looking at team census lists. Physicians were asked about their rounding style (ie, sickest patients first, patients likely to be discharged first, room-by-room, most recently admitted patients first, patients on the team the longest, or other) and then to identify all patients they thought would be definite discharges sometime during the day of the survey. Definite discharges were defined as patients whom the provider thought were either currently ready for discharge or who had only minor barriers that, if unresolved, would not prevent same-day discharge. They were asked if the discharge order had been entered and, if not, what was preventing them from doing so, if the discharge could in their opinion have occurred the day prior and, if so, why this did not occur. We also obtained the date and time of the admission and discharge orders, the actual discharge time, as well as the length of stay either through chart review (majority of sites) or from data warehouses (Denver Health and Presbyterian St. Lukes had length of stay data retrieved from their data warehouse).

Physicians were also asked to identify all patients whom they thought might possibly be discharged that day. Possible discharges were defined as patients with barriers to discharge that, if unresolved, would prevent same-day discharge. For each of these, the physicians were asked to list whatever issues needed to be resolved prior to placing the discharge order (Appendix 1).

The second survey was administered late morning on the same day, typically between 11 am and 12 pm. In this survey, the physicians were asked to reassess the patients previously classified as definite and possible discharges for changes in status and/or barriers and to identify patients who had become definite or possible discharges since the earlier survey. Newly identified possible or definite discharges were evaluated in a similar manner as the initial survey.

The third survey was administered midafternoon, typically around 3 PM similar to the first two surveys, with the exception that the third survey did not attempt to identify new definite or possible discharges.