Adapting Diabetes Treatment Expert Systems to Patient's Expectations and Psychobehavioral
Characteristics in Type 1 Diabetes.
Glucose variability (GV) in type 1 diabetes (T1DM) is commonly viewed as a primary marker of glycemic control,
potentially responsible, along with chronic hyperglycemia, for diabetes complications. This proposed project
continues 20 years of research, which identified physiological and behavioral correlates of GV and successfully
tested feedback control policies to reduce GV via simultaneous protection against hypoglycemia and systematic
hyperglycemia in T1DM.
Our primary hypothesis is that: Reducing glucose variability in T1DM can be optimally achieved by technology
that is informed of, and adapted to, the individual psychobehavioral and metabolic profiles of patients/users. This
can be achieved through personalization and automated adaptation of treatment policies, and through treatment
intervention that corresponds to each patient's level of technology acceptance and is designed to maximize
successful system use by tracking and reinforcing trust in the intervention.
Therefore, in this project we plan to (i) confirm the efficacy of two previously designed technological interventions
- Informative Decision Support System (iDSS) and Prescriptive Decision Support System (pDSS) - in reducing
GV in T1DM patients during a 6-month long randomized cross-over clinical trial; (ii) show that subjects
participating in this study will have technology intervention preferences (e.g. iDSS vs pDSS) that can be predicted
by key parameters of their psychobehavioral profile and are prognostic of the level of GV control achievable by
the intervention; and finally, we propose to define and validate a novel, measureable, index of technology
acceptance and trust, by automatically observing user/system interactions.
In summary, this project will demonstrate that CGM-based decision support systems can significantly reduce
GV in T1DM, and that performance is predicted by psychobehavioral characteristics and expectations. We
further introduce a novel index tracking technology acceptance and trust, predictive of system performance.
Such index would ultimately enable future optimal self-adaptation of automated treatment strategies.