JMIR Publications

ABSTRACT

Background:

The average hospital length of stay (LOS) has increased in recent years, leading to negative outcomes for patients and increased costs for patients, hospitals, and insurers. Some of these prolonged stays are for non-medical reasons such as discharge planning. Electronic health record (EHR) audit logs provide an opportunity to measure healthcare professional EHR-based workload during extended hospital stays. Both EHR-based workload and prolonged hospital LOS are known to be associated with sociodemographic disparities, emphasizing the importance of ensuring equitable healthcare access and outcomes.

Objective:

The objective of the study is to identify trajectories of EHR-based workload, examine clinical and socio-demographic differences across trajectories, and explore relationships between trajectories and hospital discharge disposition.

Methods:

Inpatient medicine hospitalizations with LOS ≥ 3 days from January – April 2023 at 14 hospitals within a large healthcare system were included. EHR audit logs were used to compute EHR time per patient per hospital day, a proxy of EHR-based workload. Group-based trajectory modeling was used to identify latent patterns of EHR time during hospitalization. Patient demographics and encounter characteristics were compared across the identified trajectory groups. Associations between trajectories and discharge disposition were examined using multivariable mixed effects logistic regressions.

Results:

Four trajectories of EHR time were identified among 9537 hospitalizations (median LOS=5.23 days, IQR 3.95, 7.89). The two largest groups were “Moderate Early [EHR Time]/Low Overall [EHR Time]” (n=6812, 71.4%) and “Moderate Early/Moderate Overall” (n=2013, 21.1%). Among the two smaller groups, “High Early/Low Overall” (n=397, 4.2%) had high initial EHR time (>20 hours), short LOS (4.13 days, IQR 3.54, 4.89), and disproportionately included patients who were Black (n=202,50.9%) and had Medicaid insurance (n=105, 26.4%). “Moderate Early/High Overall” (n=315, 3.3%) had the longest LOS (21.7 days, IQR 17.8, 28.5) and the largest proportion discharged to skilled nursing facilities (42.5%). Adjusted odds ratios (aOR) of discharge to skilled nursing facility, other medical facility, or hospice differed between trajectory groups after controlling for demographics and clinical variables. Compared to the largest group, “Moderate Early/Low Overall”, both “Moderate Early/Moderate Overall” (aOR=2.76, 95% CI 2.45, 3.12) and “Moderate Early/High Overall” (aOR=4.91, 95% CI 3.73, 6.47) were significantly more likely to be discharged to a SNF, other medical facility, or hospice.

Conclusions:

There were distinct EHR time trajectories in inpatient medicine hospitalizations, and socio-demographic and clinical characteristics differed between these trajectories. Discharge location differed between groups after adjusting for confounders. While results of the current study are not predictive, these descriptive trajectories suggest potential future implications for patient throughput planning.