Conducting Community-Based Research with Older Adults: Adapting to Unexpected Demands (Preprint)

Abstract

A wide range of challenges may emerge in recruiting and enrolling older adults in community-based research that often relies on partnership with agencies, services, and providers who facilitate access to the target population. This baseline of usual concerns was deeply affected by the Covid-19 pandemic that presented diverse barriers to accessing older adults to participate in our in-person study with older adults in senior living settings.

This paper discusses barriers and solutions that occurred while conducting a study that evaluated two aims related to predicting late-life depression using vocal patterns among community-dwelling older adults. We highlight unanticipated events that emerged at the onset of funding through three years of pandemic-related problems that were overcome through teamwork. Lessons learned are offered to help guide future community-based researchers as both usual and highly unexpected challenges are encountered.

The observational, repeated measures design included adults aged 65 years and older using in-person interviews to collect background information, depression levels using the 9-item Patient Health Questionnaire and vocal data. We describe the impact of the pandemic on access to older adults, transitions from the initial in-person approach to “no contact” virtual data collection, and factors that extended the timeline.

Our step-wise adaptations to barriers resulted in outcomes that were both below and beyond expectations. Two main adaptations to in-person data collection included a Hybrid approach that had some contact with study team members and a no-contact Virtual approach. A total of 129 participants included 17 in-person/iPad, 18 Hybrid and 87 Virtual group members who were an average age of 68.8 to 81 7 years old, largely white (86.2 – 100%) and female (60.9 – 77.2%). Average depression severity scores ranged from 4.5 to 12.9, and average minutes of vocal recording time ranged from 3 to 17 minutes. Wave2Ved machine learning analytics successfully predicted depression from vocal patterns 98.5% to 100% across groups.

In spite of facing a variety of problems in accessing and interviewing older adults for this study, strong outcomes were achieved in predicting depression from vocal patterns data that was collected using three different approaches. Further, the advancement of new technologies that were devised to overcome challenges hold promise for advancing future vocal research.

Not applicable