Understanding mortality among hematology oncology patients identified from hospital administrative data.

Abstract

215 Background: Hospital administrative data has large ramifications for quality of care and quality improvement projects. Unexpected inpatient mortality among patients with a hematology/oncology (H/O) diagnostic related group (DRG) is not widely studied and reported. We sought to investigate the unexpected inpatient mortality rate among cancer patients who were admitted to a large, urban, tertiary care safety net teaching hospital. Methods: We obtained the hospital’s adjusted mortality rate and evaluated the subset of specific observed/expected deaths ratio (O/E) of patients with a primary H/O diagnosis. We reviewed each case that was identified as an unexpected inpatient mortality from 2016 to 2018. A chart abstraction tool was designed for data abstraction that included demographics, location of admission, comorbid conditions, if mortality was expected, if documentation was reflective of the severity of the illness. Two independent reviewers abstracted each chart. A third reviewer assessed each case to make a final determination regarding expected mortality based on medical complexity and if documentation was reflective of severity of illness. Results: On review of the inpatient mortality data, the O/E for the H/O diagnostic related group (DRG) from Vizient was 1.14. Twenty two cases were identified as having an unexpected death. Among those cases, 23% of patients were transferred from an outside hospital, and 23% of patients did not have a known cancer diagnosis on admission. In the majority of cases, initial documentation did not accurately reflect severity of illness and/or specialists in hematology or oncology were not consulted at time of diagnosis. We noted the majority of patients were not followed by a hematologist/oncologist within the hospital system (63.6%). In 55% of patients, the cause of death was acute respiratory failure, and 14% of patients had an Advance Directive/DNAR in the chart prior to admission. Delayed antibiotics in febrile neutropenia and a missed blood transfusion reaction were among causes of unexpected deaths. Conclusions: Hospital generated data reported that mortality among H/O patients was greater than projected, however careful chart review of each case demonstrated a significant number of cases that were expected but the severity of illness was not documented properly to account for the death. Templates for proper documentation using the Vizient mortality variables is a key area to lower unexpected inpatient mortality.