Background: Approximately 1.6 million patients are transferred between hospitals yearly, with disproportionately high cost and mortality. Despite the risk associated with inter-hospital transition of care, there is no established best practice in coordinating and triaging transfers. This gap in knowledge is partly driven by a dearth of studies which integrate clinical data before and after transfers, and a lack of high-quality interventional studies toward improving this transition. The transfer is a process across several hours to several days where physician communication is incongruent from the actual transition of responsibilities. While errors and miscommunication have been documented, the value of structured communication has not been established in prospective studies.
Methods: We evaluated a convenience sample of 1000 consecutive prospectively collected inter-hospital transfers with a templated transfer note to a representative tertiary referral center. Each note contained structured elements which were extracted and merged to patient outcomes. The primary outcome was in-hospital mortality, with secondary measures including length of stay, and escalation of care within 24 hours. Finally, we evaluated the rate of hand-off error: that patients were reported to be stable but were transferred to the ICU within 6 hours of arrival. We performed multivariate logistic regression to evaluate individual elements prediction of outcome adjusting for age, race, payer, and comorbidities. C-statistic was used to determine how effective hand-off elements predicted inpatient mortality and early transfer to the ICU.
Results: Information contained in the transfer note provided fair prediction of in-hospital mortality. In fact, relying on information in the dot phrase and age alone provided good prediction of in-hospital hospital mortality (AUROC: 0.71) and fair prediction of transfer to the ICU within 24 hours (AUROC 0.66). When adjusting for age, demographics, and Charlson comorbidity score, individual transfer elements predicted subsequent mortality including whether the patient would be stable throughout transfer (OR 0.34 p <0.001), whether recommendations were given (OR 0.47 p = 0.016), and whether the reason for transfer was to evaluate for transplant (OR 3.59 p = 0.018). Errors occurred in 6% of cases, where patients were documented to be stable, but were transferred within 6 hours of arrival. This was associated with higher rates of inpatient mortality (OR 2.99 p = 0.003).
Conclusions: In this prospective, observational study of inter-hospital transfers, we show that a structured hand-off supported by documentation provides valuable information. Both documentation of predicted stability and feedback for care where shown to be associated with lower risk of death. Additionally, in this context, error rates were rare (6%) but associated with higher adjusted mortality. This study supports a growing body of evidence that structured hand-offs and supporting documentation likely improves care and should be applied to high-risk transitions which includes inter-hospital transfers.