Background: Transitions of care, whether between or within institutions, are an important source of errors, inefficiency, and unnecessary costs. Inter-hospital transfers are complicated by incongruent information systems, indirect and asynchronous communication, and geographical distance all in settings of high patient complexity and acuity. We developed a large database of patients transferred between hospitals to identify novel predictors of risk, to estimate impact on US health care costs, and to determine hospital outcomes.
Methods: A large dataset of transferred patients was generated using Health Care Utilization Project’s State Inpatient and Emergency Department databases from 5 states (FL, IA, NY, UT, VT) between 2011-2013. Two methods to measure diagnostic concordance were used to investigate the impact of information agreement on subsequent outcomes. The first measure (Diagnostic Concordance Index, DCI) involved percent agreement of 29 validated chronic conditions based on ICD-9 coding known to impact inpatient outcomes and utilization. Another measure involved diagnosis gain and loss in 9 categories of ICD-9 coding. Diagnosis loss denotes a diagnosis present in the referring hospitalization and absent on admission to the accepting hospital while diagnosis gain represents the opposite. The primary outcome was adjusted inpatient mortality, with secondary measures of cost and length of stay (LOS).
Results: We identified 180,337 inter-hospital transfers where the referring and receiving hospital stay could be identified. When compared to a non-transferred population, transferred patients had higher adjusted inpatient mortality (OR 1.39, 95% CI 1.36–1.42 p < 0.001), longer length of stay LOS (8.2 days vs 4.7 days, p < 0.001), and higher cost ($59,314 vs $38,410). Across all transfers, the mean DCI was 52%. A higher DCI was an independent predictor of reduced inpatient mortality (OR 0.65, 95% CI 0.59–0.70, p < 0.001, LOS (p < 0.001), and reduced total charges (p < 0.001). Diagnosis gain was associated with increased mortality (OR 4.49 95% CI 3.74–5.37 p < 0.001), LOS (p <0.001), and cost (p < 0.001). Diagnosis loss was also associated with increased mortality (OR 3.28, 95% CI 2.5–4.3, p < 0.001), LOS (p < 0.001), and higher cost (p < 0.001).
Conclusions: Inter-hospital transfers contribute disproportionately to in-hospital mortality and cost. Discordance of diagnosis coding between referring and receiving hospitals occurred frequently. Greater diagnostic discordance between hospitals was associated with reduced survival and increased cost. Possible contributors to discordance include miscommunication during the transition of care, discovery of diagnostic errors, development or discovery of new conditions after transfer, and evolution of disease. While error or disease evolution may explain the impact of diagnosis gain on outcomes, diagnosis loss illustrates the profound impact of miscommunication on patient outcomes.