In BMC emergency medicine
BACKGROUND : There is limited research on individual patient characteristics, alone or in combination, that contribute to the higher levels of mortality in post-transfer patients. The purpose of this work is to identify significant combinations of diagnoses that identify subgroups of post-interhospital transfer patients experiencing the highest levels of mortality.
METHODS : This was a retrospective cross-sectional study using structured electronic health record data from a regional health system between 2010-2017. We employed a machine learning approach, association rules mining using the Apriori algorithm to identify diagnosis combinations. The study population includes all patients aged 21 and older that were transferred within our health system from a community hospital to one of three main receiving hospitals.
RESULTS : Overall, 8893 patients were included in the analysis. Patients experiencing mortality post-transfer were on average older (70.5 vs 62.6 years) and on average had more diagnoses in 5 of the 6 diagnostic subcategories. Within the diagnostic subcategories, most diagnoses were comorbidities and active medical problems, with hypertension, atrial fibrillation, and acute respiratory failure being the most common. Several combinations of diagnoses identified patients that exceeded 50% post-interhospital transfer mortality.
CONCLUSIONS : Comorbid burden, in combination with active medical problems, were most predictive for those experiencing the highest rates of mortality. Further improving patient level prognostication can facilitate informed decision making between providers and patients to shift the paradigm from transferring all patients to higher level care to only transferring those who will benefit or desire continued care, and reduce futile transfers.
Reimer Andrew P, Schiltz Nicholas K, Koroukian Siran M
2022-Nov-24
Electronic health records, Emergency helicopter, Helicopter ambulance, Machine learning, Transportation of patients
