Background: Inpatient mortality rates have been used as a quality measure in stroke. Some factors are within the control of healthcare providers and others, such as age or severity of illness are by nature non-modifiable. Various risk adjustment models (UHC, GWTG) attempt to account for severity of illness based on patient characteristics with the presumption that remaining differences in mortality are accounted by differences in health care quality. In this study we examine how patient mortality can be effected by a number of independent variables, including care provided prior to admission, involvement of palliative care team, and patient and family preference
Methods: We initially reviewed our mortality reports from GWTG-Stroke and UHC for 2014 at Dartmouth-Hitchcock Medical Center (DHMC), and identified a cohort of 49 patients deceased from stroke using the appropriate ICD-9 codes and the discharge status of “expired”. We compared cases reviewed by The Division of Quality, Safety, and Value (DQSV) at DHMC with independent chart review performed by neurologists from our stroke program, who focused on factors such as route of hospitalization, management prior to admission and end of life care. Deaths were categorized into three categories: 1. Non-preventable. 2. Possible iatrogenic mortality 3. Non-iatrogenic, but possibly preventable based on specific circumstances of the case.
Results: A significant number of preventable cases were found on clinician’s review compared to the case reviews based on the DQSV. Notably at least 7 cases were found to have complications prior to transfer to our facility. None of these preventable deaths were due to premature withdrawal of care.
Conclusions: These results suggest that institutions similar to ours would benefit from improving quality of care of stroke patients at the referral hospitals prior to transfer. This work also implies that as health care moves towards an incentivized and outcome-based system, using mortality scores alone to label hospitals as excellent or poor may have the unintended consequences of compromising patient care in order to reach benchmarks that do not take into account the unique variables of different hospital settings.