Sleep Chatbot Intervention for Emerging Black/African American Adults

Artificial Intelligence Sleep Chatbot in Emerging Black/African American Adults With Cardiometabolic Risk Factors: a Feasibility Study

Unhealthy sleep and cardiometabolic risk are two major public health concerns in emerging Black/African American (BAA) adults. Evidence-based sleep interventions such as cognitive-behavioral therapy for insomnia (CBT-I) are available but not aligned with the needs of this at-risk group. Innovative work on the development of an artificial intelligence sleep chatbot using CBT-I guidelines will provide scalable and efficient sleep interventions for emerging BAA adults.

Abnormal metabolic syndrome (MetS) components affect up to 40% of emerging adults (18-25 years), particularly Black/African Americans (BAA). MetS risk in early life tracks into adulthood and predicts cardiovascular diseases and type 2 diabetes mellitus later in life. Unhealthy sleep is a known modifiable factor for MetS components. However, the prevalence of unhealthy sleep (up to 60%) in emerging adults is alarming, potentially exacerbating downstream future cardiometabolic health. Cognitive-behavioral therapy for insomnia (CBT-I) is an evidence-based intervention for unhealthy sleep that improves both sleep quantity and quality. Compared with traditional in-person intervention paradigms, digital CBT-I has comparable efficacy with enhanced accessibility and affordability. However, current digital CBT-I based programs are unable to deliver tailored content and interactive services in a humanlike way, thus are unable to meet the needs of emerging BAA adults at risk for MetS. Building on prior work by the team, the investigators will leverage artificial intelligence (AI) technologies and refine an AI sleep chatbot using CBT-I guidelines and examine its feasibility and efficacy in a 4-week clinical trial in short-or-poor sleeping, emerging BAA adults with at least one MetS factor.

Using the pretest-posttest design, the investigators will test the efficacy of the 4-week sleep chatbot intervention on improving sleep health (primary) and metabolic syndrome factors (exploratory).

Using CBT-I principles, participants will receive a four-week intervention delivered through a chatbot. The self-administered intervention is comprised of personalized behavioral prescriptions based on stimulus control principles and sleep schedule modification goals using sleep efficiency (SE) criteria. Participants are allowed to self-adjust expectations and make realistic decisions on sleep schedules. Other CBT-I components will be used as on-demand content. The chatbot will facilitate sleep goal setting with the participant, communicate weekly behavioral prescription and CBT-I educational modules, collect sleep diary and provide adaptive feedback and reactive services (e.g. Q&A conversations) 24/7.

Personalized intervention algorithms will be developed based on CBT-I guidelines, focus group data, individual sleep baseline information and self-reported prioritized sleep goals. The CBT-I intervention will focus on principles of sleep restriction and stimulus control, with other CBT-I components used as on-demand content. The sleep chatbot system will facilitate sleep goal-setting with the participant and communicate weekly behavioral prescriptions and educational modules. After baseline data collection, the research coordinator will provide intervention orientation and set up the first-week sleep modification goal during the in-person/Zoom meeting. Sleep modification goals in the remaining weeks will be developed through the participant-chatbot interaction. The Chatbot system will send sleep-related information and behavioral reminders/feedback based on the interactive conversation with participants. Participants will also complete a sleep diary prompted by a chatbot.

The total amount of sleep time (hours) will be estimated each night for seven consecutive days using a wrist-worn ActiGraph GT9X Link. The average sleep time over a week will be used in data analysis.

Sleep efficiency (percentage of time spent asleep while in bed) will be estimated each night for seven consecutive days using a wrist-worn ActiGraph GT9X Link. The average sleep efficiency over a week will be used in data analysis. This variable indicates sleep quality.

Sleep time and awakening time will be estimated for seven consecutive days using a wrist-worn ActiGraph GT9X Link. Mid-sleep time each night refers to the mid-point between sleep time and awakening time. Intra-individual variability in midsleep times will be calculated as the standard deviation of the mid-sleep time over a week for each participant. This variable reflects the regularity of sleep, with higher values showing greater irregularity.

Change from baseline data of intra-individual variability in midsleep times in the end of intervention and 4-week follow-up.

The Insomnia Severity Index is composed of 7 items measuring insomnia-related sleep disturbance. and daytime dysfunction. The seven answers are added up to get a total score (0-28), with higher scores indicating severer insomnia.

Change from baseline score of Insomnia Severity Index in the end of intervention and 4-week follow-up.

The total number of metabolic syndrome components, including high waist circumference, high blood pressure, high fasting triglycerides and glucose, and low HDL, will be calculated to indicate metabolic health (higher value, worse metabolic health). A point-of-care test will provide the fasting glucose and cholesterol panel.

Change from baseline number of metabolic syndrome components in the end of intervention and 4-week follow-up.

A self-assessment questionnaire, Horne and Ostberg Morningness/Eveningness Questionnaire, will be used to determine morningness-eveningness in circadian rhythms---the degree to which respondents are active and alert at certain times of the day. The scale requires between 10 and 15 min for completion. The sum gives a score ranging from 16 to 86; scores of 41 and below indicate "evening types", scores of 59 and above indicate "morning types", and scores between 42-58 indicate "intermediate types".

Change from baseline score of Horne and Ostberg Morningness/Eveningness Questionnaire in the end of intervention and 4-week follow-up.

The Epworth Sleepiness Scale will be used to assess daytime sleepiness. The total score (the sum of 8 item scores, 0-3) can range from 0 to 24. The higher score suggests the higher that person's average sleep propensity in daily life, or 'daytime sleepiness'.

Change from baseline score of Epworth Sleepiness Scale in the end of intervention and 4-week follow-up.

The Dysfunctional Beliefs and Attitudes about Sleep Scare (DBAS-16) is a 16-item self-report measure designed to evaluate a subset of those sleep-related cognition/beliefs (e.g., beliefs, attitudes, expectations, appraisals, attributions). For each item, a higher score suggests a greater dysfunctional belief about sleep. Items with scores > 5 are concerning.

Change from baseline scores of Dysfunctional Beliefs and Attitudes about Sleep Scare in the end of intervention and 4-week follow-up.

Inclusion Criteria: male or female ages 18-25 years old elf-identified as Black/African Americans (BAA), self-report short sleep (<7 hours) or poor sleep [Insomnia severity index (ISI) >10] MetS factors: at least one of the MetS factors confirmed by fasting blood testing during the first lab visit (fasting blood glucose ≥110mg/dL, high-density lipoprotein ≤ 40 mg/dL for males and ≤ 50 mg/dL for females, triglycerides ≥150mg/dL, blood pressure ≥130/85mmHg, waist circumference≥40 inches for males, ≥35 inches for females) own a smartphone (iPhone or Android). Exclusion Criteria: self-report medical conditions [i.e., major depressive disorder [Patient Health Questionnaire-9 (PHQ-9) ≥10) diagnosed obstructive apnea] that may affect sleep regular use of medications with substantial impact on sleep and cardio-metabolic markers shift worker smoker alcohol abuse (Alcohol Use Disorders Identification Test--short form score ≥7 for males and ≥5 for females) self-report pregnancy/lactation.

Nolan PB, Carrick-Ranson G, Stinear JW, Reading SA, Dalleck LC. Prevalence of metabolic syndrome and metabolic syndrome components in young adults: A pooled analysis. Prev Med Rep. 2017 Jul 19;7:211-215. doi: 10.1016/j.pmedr.2017.07.004. eCollection 2017 Sep.

Raynor LA, Schreiner PJ, Loria CM, Carr JJ, Pletcher MJ, Shikany JM. Associations of retrospective and concurrent lipid levels with subclinical atherosclerosis prediction after 20 years of follow-up: the Coronary Artery Risk Development in Young Adults (CARDIA) study. Ann Epidemiol. 2013 Aug;23(8):492-7. doi: 10.1016/j.annepidem.2013.06.003.

Kocevska D, Lysen TS, Dotinga A, Koopman-Verhoeff ME, Luijk MPCM, Antypa N, Biermasz NR, Blokstra A, Brug J, Burk WJ, Comijs HC, Corpeleijn E, Dashti HS, de Bruin EJ, de Graaf R, Derks IPM, Dewald-Kaufmann JF, Elders PJM, Gemke RJBJ, Grievink L, Hale L, Hartman CA, Heijnen CJ, Huisman M, Huss A, Ikram MA, Jones SE, Velderman MK, Koning M, Meijer AM, Meijer K, Noordam R, Oldehinkel AJ, Groeniger JO, Penninx BWJH, Picavet HSJ, Pieters S, Reijneveld SA, Reitz E, Renders CM, Rodenburg G, Rutters F, Smith MC, Singh AS, Snijder MB, Stronks K, Ten Have M, Twisk JWR, Van de Mheen D, van der Ende J, van der Heijden KB, van der Velden PG, van Lenthe FJ, van Litsenburg RRL, van Oostrom SH, van Schalkwijk FJ, Sheehan CM, Verheij RA, Verhulst FC, Vermeulen MCM, Vermeulen RCH, Verschuren WMM, Vrijkotte TGM, Wijga AH, Willemen AM, Ter Wolbeek M, Wood AR, Xerxa Y, Bramer WM, Franco OH, Luik AI, Van Someren EJW, Tiemeier H. Sleep characteristics across the lifespan in 1.1 million people from the Netherlands, United Kingdom and United States: a systematic review and meta-analysis. Nat Hum Behav. 2021 Jan;5(1):113-122. doi: 10.1038/s41562-020-00965-x. Epub 2020 Nov 16.

Matricciani L, Paquet C, Fraysse F, Grobler A, Wang Y, Baur L, Juonala M, Nguyen MT, Ranganathan S, Burgner D, Wake M, Olds T. Sleep and cardiometabolic risk: a cluster analysis of actigraphy-derived sleep profiles in adults and children. Sleep. 2021 Jul 9;44(7):zsab014. doi: 10.1093/sleep/zsab014.

Griggs S, Conley S, Batten J, Grey M. A systematic review and meta-analysis of behavioral sleep interventions for adolescents and emerging adults. Sleep Med Rev. 2020 Dec;54:101356. doi: 10.1016/j.smrv.2020.101356. Epub 2020 Jul 8.

Stock AA, Lee S, Nahmod NG, Chang AM. Effects of sleep extension on sleep duration, sleepiness, and blood pressure in college students. Sleep Health. 2020 Feb;6(1):32-39. doi: 10.1016/j.sleh.2019.10.003. Epub 2019 Nov 19.

Nicol G, Wang R, Graham S, Dodd S, Garbutt J. Chatbot-Delivered Cognitive Behavioral Therapy in Adolescents With Depression and Anxiety During the COVID-19 Pandemic: Feasibility and Acceptability Study. JMIR Form Res. 2022 Nov 22;6(11):e40242. doi: 10.2196/40242.

Stephens TN, Joerin A, Rauws M, Werk LN. Feasibility of pediatric obesity and prediabetes treatment support through Tess, the AI behavioral coaching chatbot. Transl Behav Med. 2019 May 16;9(3):440-447. doi: 10.1093/tbm/ibz043.

Edinger JD, Arnedt JT, Bertisch SM, Carney CE, Harrington JJ, Lichstein KL, Sateia MJ, Troxel WM, Zhou ES, Kazmi U, Heald JL, Martin JL. Behavioral and psychological treatments for chronic insomnia disorder in adults: an American Academy of Sleep Medicine clinical practice guideline. J Clin Sleep Med. 2021 Feb 1;17(2):255-262. doi: 10.5664/jcsm.8986.