Douglas Spangler: On the Quantitative Evaluation and Enhancement of Prehospital Decisional Capacity

Abstract

Prehospital care involves increasingly complex decision-making processes, necessitating commensurate advances in the methods used to assess and improve the quality of patient triage processes. This doctoral project aimed to advance the measurement of patient outcomes in the evaluation of prehospital decision-making and develop interventions to improve those outcomes. In study I, a set of outcome definitions for evaluating referrals to non-emergency care by dispatch nurses was validated, confirming the ability of systematic data abstraction processes to identify patient harms missed by traditional incident reporting systems. In study II, an intervention delivering feedback on process and outcome metrics to dispatch nurses was evaluated, identifying improvements in some process metrics, while impacts on outcomes remained elusive. In study III a machine learning-based approach to estimating composite risk scores was validated internally for use in prehospital contexts. In study IV, similar models for use in Ambulance care were externally validated in a dataset collected from six Swedish regions, finding that model performance remained superior to traditional rule-based risk assessment instruments even when the models were applied in novel settings. Study V is a randomized controlled trial whereby a clinical decision support tool based on these models was found to enhance the ability of dispatchers to identify and prioritize high-risk patients in resource constrained situations. Future directions for study include the incorporation of additional structured and unstructured data in the prediction models, and efforts to evaluate and enhance their fairness and alignment with human assessments of care need. Open-source software packages implementing these tools are available to enhance the transparency of the work and stimulate further development.