Effects of Counseling by Peer Human Advisors vs Computers to Increase Walking in Underserved Populations: The COMPASS Randomized Clinical Trial

“Technology-enabled eHealth programs represent potentially cost-efficient and practical means for customized PA [physical activity] guidance to diverse groups. Most people targeted by eHealth, however, are well educated, younger than 50 years, and of non-Hispanic White ancestry, potentially intensifying health disparities.

[..] This investigation tested whether a virtual advisor could increase 12-month walking to an extent similar to a comparably structured human advisor program among Latino adults. The human advisor program was delivered by trained peer advisors—a resource-efficient approach that is well accepted by Latino and other diverse groups but may be less convenient and scalable than computer-based programs. [..] Both interventions were considered “light-touch” given that they were delivered using primarily non–health professional staff and resources.

[..] COMPASS (Computerized Physical Activity Support for Seniors) was a single-blind, cluster-randomized noninferiority parallel trial conducted by Stanford University and Northeastern University. [..] They received a $10 gift card/assessment. Such modest remunerations have been found to have minimal influence on PA change.

Participants were recruited from community centers in Santa Clara and San Mateo counties, California. The centers were randomized in pairs (1:1 allocation) based on locale to either virtual or human advisors based on a computerized randomization sequence. [..] Both arms received a similar 12-month behavioral PA instruction/support program at their designated center based on Active Choices—an individually tailored program with demonstrated effectiveness and translatability that has merited formal recognition by the Administration on Aging, National Council on Aging, and the Centers for Disease Control and Prevention.

[..] The following eligibility criteria were used: (1) age 50 years or older; (2) insufficiently active (ie, engaged in <100 min/wk of moderate-intensity PA over the past month, based on the Community Healthy Activities Model Program for Seniors [CHAMPS] instrument); (3) able to safely engage in moderate-intensity PA, such as walking, based on the PA Readiness Questionnaire; (4) live less than 5 miles from a study-designated community center; (5) be able to read and understand English or Spanish sufficiently to provide informed consent and participate in study procedures, including computer use; and (6) plan to live in the area for a year.

[After an introductory session,] Participants then received up to 28 brief (10-15 minutes) advising sessions across 12 months. Sessions followed a standard, evidence-based protocol and allowed for customization based on participant preferences and availability. Each session consisted of an introductory dialogue, brief check for health changes, review of pedometer steps and minutes walked since last session, problem-solving around personal PA barriers, goal-setting, and next-session scheduling. Cultural tailoring occurred in a variety of ways.

The virtual advisor, described elsewhere, was an interactive, animated computer agent simulating face-to-face counseling using simple speech (synthetic English or Spanish) and nonverbal behaviors (eg, facial cues and hand gestures). The virtual agent, named Carmen, was successful previously in increasing 4-month walking levels among Latino aging adults relative to controls. Individuals interacted with Carmen by touching simple conversation boxes on the computer screen offered in English and Spanish and targeted at a first- to third-grade reading level. Participants received a private log-in and were encouraged to wear headphones for privacy. They also were encouraged to download their pedometer data on the virtual computer at each session via a USB port. One difference between this system and some other automated health information systems is that participants could interact directly with this system at their convenience, receiving real-time customized advice through the multi-modality visual plus auditory interface.

[..] Change in walking minutes per week was assessed using the 4 walking items from the validated CHAMPS survey (interview format) for older adults, which is available in English and Spanish. Such standardized self-report instruments represent the most direct, reliable, and cost-efficient means for assessing specific PA types typically targeted in community interventions, given that device-based assessment tools (accelerometers) capture more general movement beyond such purposeful PA behavior and currently lack sufficient normative data, especially among older adults, to allow direct linkages with the PA guidelines and evidence base.

[..] A total of 241 participants (98.4%) were Latino, 193 participants were women (78.8%), and 52 participants were men (21.2%). Mean (SD) age was 62.3 (8.4) years (range, 50-87 years). One hundred seven participants (43.7%) had an education level of high school or lower. Mean (SD) years of residence in the US was 47.4 (17.0) years. Mean (SD) BMI was 32.8 (6.8), with 156 participants (63.7%) in the obese range. One hundred nineteen participants (48.6%) were taking antihypertensive medications. Reported walking time was low (mean [SD] total, 70 [98] min/wk; approximately 10-minute/d). Baseline accelerometry-derived MVPA [moderate to vigorous physical activity] per week also was low (mean [SD] virtual advisor, 40.7 [25.8]; human advisor, 43.6 [31.2]). One hundred six participants (43.3%) chose to receive their intervention in Spanish (virtual advisor, 48 [39.0%]; human advisor, 58 [47.5%]) (χ2 = 1.81, P = .18).

The mean (SD) number of total advising sessions completed for virtual vs human advisors was 18.8 (11.8) (64.8%) vs 18.4 (8.9) (63.4%) (t test, 0.30; P = .76). Mean length of postintroductory advising sessions was significantly shorter in virtual (8.4 [4.8] min/session) vs human (21.0 [7.1] min/session) (t test, 16.0, P < .001). This finding translates into a mean total intervention volume for virtual vs human of 3.2 hours (56.6) vs 6.9 hours (63.0) (t test, 27.8; P < .001). Although both programs encouraged pedometer use and reporting at each session, pedometer reporting was sporadic in both programs across 12 months (eg, month 12 reporting for virtual advisor, 35.8% vs human advisor, 32.8%; χ2 = 0.24; P = .62).

Twelve-month change in walking minutes per week [..] supported noninferiority (ie, lower limit of the 95% CI [−20.6] lay to the right of the noninferiority margin [−30]). Mean walking increases indicated a somewhat larger increase in the virtual advisor cohort (153.9 min/wk; 95% CI, 126.3 min/wk to infinity vs 131.9 min/wk; 95% CI, 101.4 min/wk to infinity; difference, 22.0, with lower limit of 1-sided 95% CI, −20.6 to infinity). Similar results were obtained using per-protocol analyses (mean [SD] increases in walking time for the virtual advisor cohort [n = 117]: 158.6 [217.1] min/wk vs human advisor cohort [n = 114]: 134.8 [192.1] min/wk; P = .02). At baseline, no participant was at the nationally recommended target range of 150 min/wk or more of MVPA; at 12 months, 29.3% of the virtual advisor and 31.1% of the human advisor cohorts achieved that target (χ2 = 0.10, P = .75).

[..] Both arms reported improvements in overall well-being, with the magnitude of the within-arm improvement somewhat larger in the human advisor (mean [SE] change, 3.5 [0.6]; t = 4.1; P < .001) compared with the virtual advisor (mean [SE] change, 1.1 [0.6]; t = 1.8; P = .06) cohorts. The human advisor cohort reported significant pre-post improvements in all 10 domains (eg, good appetite and few aches or pain), while the virtual advisor cohort reported significant improvements in 3 domains (sleep well, feel rested, and full of pep and energy).

[..] The results of the present trial expand the small PA eHealth evidence base for older adults and Latino populations across longer time frames. Both programs also produced reported decreases in prevalent sedentary behaviors, including television/video viewing, that have been linked independently with detrimental health outcomes.”

Full article, King AC, Campero MI, Sheats JL et al. JAMA Internal Medicine 2020.9.28