Communication behaviors, including information seeking, information giving, and responding to emotions, can be measured within in-person interpersonal health communication between Veterans and healthcare providers. Investigators have developed reliable coding schemas to extract communication behaviors from audiotapes of clinical encounters. Using these schemas, including the Roter Interaction Analysis System (RIAS), patterns of communication behaviors have been positively associated with patient satisfaction, trust in providers, and positive changes in Veteran self-management (e.g., medication adherence). Recently, RIAS has been adapted for use with telehealth and asynchronous written communication (like email). With the advent of Secure Messaging, VA has a new opportunity to directly measure communication behaviors written into these messages. Over the past five years, our team has demonstrated that communication behaviors are present in Secure Messages and can reliably be extracted using the same coding schemas validated for in-person interpersonal exchanges. In this project, we propose to advance knowledge and methods related to communication behaviors measurable through asynchronous Secure Messages. We propose the following specific aims:
Specific Aim 1 : Mine communication behaviors. Using a national corpus of Secure Messages, we will develop a sentence classification system incorporating machine learning techniques to detect communication in Secure Message responses from primary care doctors and clinical staff.
Specific Aim 2 : Define communication behavior indicators (CBIs) that represent clinically meaningful measures of Secure Message communication patterns between Veterans and Clinical Teams, then test the association of CBIs with measures of Veteran Experience (2.a) and Patient-reported behavior (2.b), medication adherence. We will identify and survey a sample of Veterans (CASES) with high CBI rates (top tertile) and a matched set of (CONTROLS) with low rates (bottom tertile).
Aim 2. a Veteran experience with Secure Messaging and CBIs: We hypothesize (H1) that CASES (Veterans with high rates of communication behaviors (CBIs)) will rate the experience with physician communication through Secure Messaging more positively than CONTROL Veterans.
Aim 2. b. Veteran-reported medication adherence: In prior studies of in-person communication, patterns of communication behaviors are strongly associated with measures of medication adherence. In our survey, we will measure patient-reported medication adherence and assess the association of adherence reports with secure messaging CBIs. We hypothesize (H2) that CASES will have better self-reported medication adherence, compared with CONTROLS.
Specific Aim 3 : Understand experiences of providers with high rates of CBIs in messages. A high priority for the VA Under Secretary for Health is to collect and disseminate best practices in VA.
In Aim 3, we will collect best practices from physicians (N = 30) with high rates of these positive communication behaviors from Secure Messages, and a comparison sample of 30 with low rates of CBIs. 1
In this project, we propose to advance knowledge and methods related to communication behaviors measurable through asynchronous Secure Messages. Working with the Office of Clinical Analytics and Reporting and the Office of Connected Care, this project is a collaboration between the VA Bedford informatics team (Houston and colleagues) with over ten years of experience evaluating Secure Messaging and an experienced text data mining team (Finch and colleagues at VA Tampa, with consultation) that has assembled a corpus of Secure Messages to use for testing. Additionally, for this resubmission, we have added the expertise of Dr. Byron Wallace, computer scientist who brings complementary expertise outside VA. 1
