Dr. Meghan Dierks

I completed clinical training as a general surgeon at Washington University, St. Louis, MO (Preliminary) and the Lahey Clinic, Burlington, MA (Chief Residency) and research training under NIH training fellowships in Medical Informatics and Molecular Genetics. While a graduate student at MIT, I studied informatics, risk analysis, human factors and systems analysis. I remain clinically active in a part-time capacity, focusing on trauma and critical care, but my major professional commitment is in basic and clinical research.

Currently, my focus is in the areas of Clinical Systems Analysis and Human Factors Engineering, Decision and Risk Analysis, and other aspects of Informatics. This work is highly cross-disciplinary, applying analytical and theoretical principals from the fields of engineering, computer science and decision theory to solve complex clinical problems. In recent years, the theoretical work and clinical field work focuses on the following:

• Systems Engineering/Human Factors Engineering applied to high-risk, complex clinical settings. In recent years, clinical or healthcare systems have begun to take on many of the features of industrial engineered systems in terms of their complexity, technology-dependency and automation. Not surprisingly, the most complex clinical systems also mimic their industrial counterparts in terms of their exposure to unexpected failures or near-failure behavior when they too are pushed to the limits of operating conditions. This work adapts modeling and simulation techniques originally developed for use in industrial process industries and nuclear power plants to explore how the probability of specific outcomes (both favorable and adverse) vary as a function of complex, time-dependent sequences of events and event interdependencies. I specifically focus on : information access and utilization for planning and decision making; influence of physical space configurations on situation awareness and response to changing acuity; cyclic variations in staffing levels, patient acuity and its influence on adverse events; improving human-machine interfaces to reduce risk and improve performance. The output of this research is new predictive models of risk based on properties of the clinical environment (e.g., staffing ratios, access to information, time of day, physical properties, resource availability, scheduling).
• Impact of technology on human performance in high-risk settings (ICU and OR). Theoretical and empiric human performance and reliability models that were developed in industrial domains are being adapted and validated in the healthcare domain, particularly in clinical environments in which there is heavy reliance on technology and semi-automated processes. In particular, I am focusing on the negative and positive impact of technology - how well new technologies integrate into (or disrupt) existing clinical environments - and how the new technologies influences safety and performance.
• Information flow in high-risk, dynamic clinical settings. I study volume, accessibility, utilization, formatting and disambiguation/filtering of noisy data by clinical providers. High information loads, significant noise:signal ratios and procedural and patient-based uncertainty add considerable cognitive load to the clinician. Long-term goals include the improving clinical performance in these complex settings through the use of intelligent monitoring, consolidated information displays and improved filtering of data/information.
• Knowledge representation and conceptual modeling. Using Description Logic as a modeling formalism, I am developing a formal representation for medical adverse events that enables machine-based automated reasoning over semi-structured clinical data. The long-term goals of this area of research are to algorithmically perform causal reasoning and outcome classification on large sets of clinical data. In a separate research activity being carried out for the FDA, I am using Description Logic to develop a robust and extensible ontology of medical devices that will enable more flexible classification of such devices for pre-market approval, post-market surveillance and other regulatory activities.
• Statistical pattern recognition in complex data sets for predictive modeling and outcomes analysis. I am using artificial intelligence and machine-learning techniques to automatically identify and classify adverse event using large clinical and administrative data sets.