Mental Distress, Label Avoidance, and Use of a Mental Health Chatbot: Results From a US Survey

BACKGROUND

For almost two decades, researchers and clinicians have argued that certain aspects of mental health treatment can be removed from clinicians' responsibilities and allocated to technology, preserving valuable clinician time and alleviating the burden on the behavioral health care system. The service delivery tasks that could arguably be allocated to technology without negatively impacting patient outcomes include screening, triage, and referral.

OBJECTIVE

We pilot-tested a chatbot for mental health screening and referral to understand the relationship between potential users' demographics and chatbot use; the completion rate of mental health screening when delivered by a chatbot; and the acceptability of a prototype chatbot designed for mental health screening and referral. This chatbot not only screened participants for psychological distress but also referred them to appropriate resources that matched their level of distress and preferences. The goal of this study was to determine whether a mental health screening and referral chatbot would be feasible and acceptable to users.

METHODS

We conducted an internet-based survey among a sample of US-based adults. Our survey collected demographic data along with a battery of measures assessing behavioral health and symptoms, stigma (label avoidance and perceived stigma), attitudes toward treatment-seeking, readiness for change, and technology readiness and acceptance. Participants were then offered to engage with our chatbot. Those who engaged with the chatbot completed a mental health screening, received a distress score based on this screening, were referred to resources appropriate for their current level of distress, and were asked to rate the acceptability of the chatbot.

RESULTS

We found that mental health screening using a chatbot was feasible, with 168 (75.7%) of our 222 participants completing mental health screening within the chatbot sessions. Various demographic characteristics were associated with a willingness to use the chatbot. The participants who used the chatbot found it to be acceptable. Logistic regression produced a significant model with perceived usefulness and symptoms as significant positive predictors of chatbot use for the overall sample, and label avoidance as the only significant predictor of chatbot use for those currently experiencing distress.

CONCLUSIONS

Label avoidance, the desire to avoid mental health services to avoid the stigmatized label of mental illness, is a significant negative predictor of care seeking. Therefore, our finding regarding label avoidance and chatbot use has significant public health implications in terms of facilitating access to mental health resources. Those who are high on label avoidance are not likely to seek care in a community mental health clinic, yet they are likely willing to engage with a mental health chatbot, participate in mental health screening, and receive mental health resources within the chatbot session. Chatbot technology may prove to be a way to engage those in care who have previously avoided treatment due to stigma.

Attitudes Toward Payment for Research Participation: Results from a U.S. Survey of People Living with HIV

Little is known about how payment affects individuals' decisions to participate in HIV research. Using data from a U.S. survey of people living with HIV (N = 292), we examined potential research participants' attitudes toward payment, perceived study risk based on payment amount, and preferred payment forms, and how these factors vary by sociodemographic characteristics. Most respondents agreed people should be paid for HIV research participation (96%) and said payment would shape their research participation decisions (80%). Men, less formally educated individuals, and members of some minoritized racial-ethnic groups were less likely to be willing to participate in research without payment. Higher payment was associated with higher perceived study risks, while preferences for form of payment varied by age, gender, education, race-ethnicity, and census region of residence. Findings suggest payment may influence prospective research participants' risk-benefit calculus and participation, and that a one-size-fits-all approach to payment could differentially influence participation among distinct sociodemographic groups.