TechSectors: Medical Software: The Virtual Nurse


Patient adherence to prescribed medication and self-care regimens is generally below 50% in almost all areas, with the rate even lower for patients with complex regimens, low health literacy, or...

Patient adherence to prescribed medication and self-care regimens is generally below 50% in almost all areas, with the rate even lower for patients with complex regimens, low health literacy, or cognitive impairment. Part of the problem is that clinicians do not always have enough time to spend with patients reviewing treatment plans, and patients do not always feel comfortable asking questions of clinicians. True health behavior change also requires intensive counseling and social support over extended periods of time, with behavioral interventions tailored to each patient’s background and motivational state.

At Northeastern University, we have been developing computer-animated health counselors that work with patients to teach them about their prescribed regimens and motivate them to maintain adherence over time. These counselors communicate with patients using simulated face-to-face conversation, including verbal and nonverbal behavior modeled on best practices from the literature on patient—provider health communication.

In particular, relationship-enhancing behavior, such as displays of caring and empathy, are used by these “relational agents” to establish a therapeutic alliance with patients in order to maximize engagement with the agent and adherence to the regimen. Relational agents have several key components:

1) a user interface, in which the agent speaks using synthetic speech and synchronized animated nonverbal behavior, and the patient “speaks” via multiple-choice input;

2) a dialog engine that governs the content of the patient—agent conversation, modeled as hierarchical transition networks;

3) a non-verbal behavior generator that specifies the animations to be used to accompany the delivery of a given sentence by the agent, based on a linguistic analysis of the sentence and rules from studies of human—human conversation; and

4) a persistent memory that tracks key information from past conversations with a given patient, important both for

therapeutic and relational dialog.

These agents have been tested in two randomized trials for physical activity promotion—one with young adults and one with geriatric patients—both of which demonstrated efficacy over standard-of-care control conditions. These trials were conducted on home desktop computers for one-to-two month interventions, in which the agent talked with patients daily about their physical activity, negotiated shorthand long-term behavior goal-setting, and provided positive reinforcement when goals were met and problem solving to overcome obstacles when they weren’t.

In addition, the agent engaged patients in relationship-building dialog, including social chat, humor, and empathy. An additional system is currently undergoing evaluation for antipsychotic medication adherence in a population of patients with schizophrenia.

Another agent is being developed in collaboration with physicians at Boston Medical Center to perform patient education prior to hospital discharge. The target of this system is the automation of a new, “re-engineered” discharge procedure recently adopted as a “safe practice” by the National Quality Forum. This procedure includes a significant patient education component that currently takes discharge nurses an hour, on average, to perform with each patient. Th e relational agent is being deployed on a touch-screen kiosk that can be used by patients from their hospital beds. It will review the patient’s condition(s), prescribed medication regimens, follow-up appointments, new medical devices, and other information with them, taking as much time as the patient needs to repeat and explain information.

The agent will also test for comprehension on key elements. Patient questions and issues that cannot be addressed by the agent are sent to a nurse for follow up. A recent pilot study demonstrated that patients with low health literacy actually prefer the relational agent to a human for this kind of health counseling. Several additional agents are also in development for other health behaviors and for deployment on alternative devices, such as PDAs with integrated accelerometers for physical activity promotion. The overall objective for these systems is to emulate the “gold standard” of one-on-one, face-to-face health counseling between a patient and provider, and to provide this counseling wherever and whenever patients need it. Such systems, we believe, have the greatest potential for maximizing long-term adherence to prescribed regimens.

Dr. Bickmore is an assistant professor at the College of Computer and Information Science, Northeastern University, Boston, MA.

Recent Videos
1 KOL is featured in this Insights series.
Charles C. Wykoff, MD, PhD: Interim Analysis on Ixo-Vec Gene Therapy for nAMD | Image Credit: Retina Consultants of Texas
Sunir J. Garg, MD: Pegcetacoplan Preserves Visual Function on Microperimetry | Image Credit: Wills Eye Hospital
Edward H. Wood, MD: Pharmacodynamics of Subretinal RGX-314 for Wet AMD | Image Credit: Austin Retina Associates
A panel of 5 experts on iron deficiency anemia
Sorana Segal-Maurer, MD, an expert on HIV
Shawn Kwatra, MD, an expert on atopic dermatitis
Shawn Kwatra, MD, an expert on atopic dermatitis
Shawn Kwatra, MD, an expert on atopic dermatitis
Ruth Ann Vleugels, MD, MPH, MBA, an expert on atopic dermatitis
© 2024 MJH Life Sciences

All rights reserved.