Conversational assessment using artificial intelligence is as clinically useful as depression scales and preferred by users

BACKGROUND

Depression is prevalent, chronic, and burdensome. Due to limited screening access, depression often remains undiagnosed. Artificial intelligence (AI) models based on spoken responses to interview questions may offer an effective, efficient alternative to other screening methods.

OBJECTIVE

The primary aim was to use a demographically diverse sample to validate an AI model, previously trained on human-administered interviews, on novel bot-administered interviews, and to check for algorithmic biases related to age, sex, race, and ethnicity.

METHODS

Using the Aiberry app, adults recruited via social media (N = 393) completed a brief bot-administered interview and a depression self-report form. An AI model was used to predict form scores based on interview responses alone. For all meaningful discrepancies between model inference and form score, clinicians performed a masked review to determine which one they preferred.

RESULTS

There was strong concurrent validity between the model predictions and raw self-report scores (r = 0.73, MAE = 3.3). 90 % of AI predictions either agreed with self-report or with clinical expert opinion when AI contradicted self-report. There was no differential model performance across age, sex, race, or ethnicity.

LIMITATIONS

Limitations include access restrictions (English-speaking ability and access to smartphone or computer with broadband internet) and potential self-selection of participants more favorably predisposed toward AI technology.

CONCLUSION

The Aiberry model made accurate predictions of depression severity based on remotely collected spoken responses to a bot-administered interview. This study shows promising results for the use of AI as a mental health screening tool on par with self-report measures.

US-Mexico Border Stress and Daily Substance Use among Latino Youth: a Cross-Sectional Study

Background

Historically marginalized youth are at risk for daily substance use. Daily use may be associated with social and environmental factors.

Methods

In March 2018, we surveyed primarily Latino adolescents ages 14-18 who lived on the US-Mexico border and assessed associations between daily substance use, neighborhood stress, border community and immigration stress, and family support.

Results

Of 443 surveyed adolescents, 41 (9%) reported daily use. Those who used daily were more likely to be older, identify as male, and reported lower social support and higher neighborhood and border community stress compared to those who did not use daily. Perceived neighborhood stress (OR = 1.95, 95% CI 1.37-2.80) and border community and immigration stress (OR = 1.55, 95% CI 1.12-2.02) were associated with increased odds of daily substance use.

Discussion

Latino adolescents who live near the US-Mexico border experience unique socioenvironmental stress which is associated with daily substance use.

Opportunities to expand postpartum support for those in recovery from opioid use disorder: Results from a qualitative study

Introduction

Over the past two decades the national prevalence of opioid use disorder (OUD) during pregnancy has increased more than 600%. Managing recovery from OUD during the postpartum period can be particularly challenging. Thus, we sought to identify ways to expand perinatal OUD treatment to ultimately reduce risk of postpartum return to opioid misuse.

Methods

We conducted in-depth semi-structured interviews with pregnant or postpartum (i.e., gave birth within the past year) mothers who have OUD, as well as with professionals who work with this population. Interviews were audio-recorded, transcribed, and coded for themes using Dedoose software using an eco-social framework.

Results

Participants included 7 mothers (median age 32 years old; 100% receiving treatment for OUD) and 11 professionals (average of 12.5 years in the field; n=7 healthcare providers, n=4 child safety caseworkers). A total of 10 major themes emerged in three levels. First, at an individual level themes included mental health, personal responsibility, and individual agency. Second, at the inter-individual level themes included support from friends and family, and other sources of support. Next, at the systems/institutional level themes included culture of healthcare systems, an ill-equipped healthcare system, social determinates of health, and continuum of care. Finally, a theme identified across all three levels included keeping mother and baby together.

Conclusions

Several opportunities to enhance the support and clinical care of OUD during the perinatal period were identified. Additional work is needed to explore how these themes may be incorporated into existing programs and/or the development of new interventions.

Perceived Normalization of Drug Trafficking and Adolescent Substance Use on the US-Mexico Border

BACKGROUND

Historically marginalized youth are at risk for daily substance use. Daily use may be associated with social and environmental factors.

METHODS

In March 2018, we surveyed primarily Latino adolescents ages 14-18 who lived on the US-Mexico border and assessed associations between daily substance use, neighborhood stress, border community and immigration stress, and family support.

RESULTS

Of 443 surveyed adolescents, 41 (9%) reported daily use. Those who used daily were more likely to be older, identify as male, and reported lower social support and higher neighborhood and border community stress compared to those who did not use daily. Perceived neighborhood stress (OR = 1.95, 95% CI 1.37-2.80) and border community and immigration stress (OR = 1.55, 95% CI 1.12-2.02) were associated with increased odds of daily substance use.

DISCUSSION

Latino adolescents who live near the US-Mexico border experience unique socioenvironmental stress which is associated with daily substance use.

Abstract P6-11-13: Identification of barriers to breast cancer screening that affect compliance

Background: In 2015, the latest year for which the Center for Disease Control (CDC) has incidence data, 242,476 new cases of Female Breast Cancer were reported, and 41,523 women died of Female Breast Cancer in the United States. For every 100,000 women, 125 new Female Breast Cancer cases were reported and 20 died of cancer. In Arizona, breast was the leading site of new cancers in 2015, with 110.9 new breast cancer diagnoses per 100,000 women, and is second for cancer deaths, reporting 19 deaths per 100,000 women. This study specifically addresses low screening rates among women in Pima County in Southern Arizona through a telephone navigation program with the goal of increasing follow-up mammograms and recommended breast imaging. Our short-term goal is to increase both initial screening and diagnostic resolution rates among women in Southern Arizona. Our long-term plan is to develop a sustainable model to increase breast cancer screening among women that is expected to inform programs state-wide and nationally. Specific aims are: 1) to increase adherence or repeat screening rate by 20% for women lost to follow-up; 2) Establish the framework for a community-academic partnership that can be replicated in other Hispanic-based areas. Methods: Potential study participants were identified using Electronic Health Records and categorized by Breast Imaging-Reporting and Data System (BI-RADS) 0 to 5. Inclusion criteria include non-compliant women of any race/ethnicity between the ages of 40 and 75 who have been referred to, or seen at a breast imaging clinic in Tucson, Arizona (Pima County) from January 1, 2014 to September 30, 2017. Of those non-compliant, 47% had a BI-RADS 0,3,4, or 5. With IRB approval from University of Arizona, we adapted a questionnaire validated for use in colon cancer to measure screening knowledge, motivations and barriers to adherence, and self-efficacy. Two navigators reached out to participants to recruit and re-engage them in the care process. Results: The sample is n=9661 of noncompliant patients from a clinic in Southern Arizona, 26% were Hispanic and 65% non-Hispanic, with 10% with 9% requiring an interpreter. From this sample, 123 study participants have been recruited (300 target accrual) with a 21% recruitment rate, 46% unable to reach rate, and a 47% decline rate. Notably, of those who declined, 66% reported being compliant at another clinic. The participants were 46% non-Hispanic and 52% Hispanic, with 25% requiring an English/Spanish interpreter. Nearly half had less than or a high school degree and similarly, nearly half had a monthly family income of $2000 or less. On the questionnaire section of knowledge, there were no statistically significant results indicating extensive knowledge, except for how often one should have a mammogram, in which 93% answered correctly. On the Barriers to Getting a Mammogram section, when the data was disaggregated by ethnicity, there was statistically significant (P⇒.05) responses. One barrier identified was financial reasons (P=.05), but interestingly no (0) Hispanics reported financial reasons to be a barrier. Conclusion: The high number of declined due to compliant elsewhere points to the need to develop a business case to sustain lay navigators across the breast cancer continuum, from screening to diagnostic resolution, treatment, survivorship, and/or palliative care. Further, breast cancer and breast screening education materials will be developed in the next phase of this study and interventions to address financial reasons. For more information on this study, please contact Drs. Jorge Gomez (Jorgejgomez@email.arizona.edu), Allison Huff MacPherson (allison7@email.arizona.edu).

Citation Format: Allison Huff MacPherson, Jorge Gomez, Elizabeth Calhoun, Marisa Borders, Kimberly Fitzpatrick, Yessenia Prado, Dora Bezies-Lopez, Chiu-Hsieh Hsu. Identification of barriers to breast cancer screening that affect compliance [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P6-11-13.