Kristina T. Phillips
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Kristina T. Phillips has a total of 91 co-authors. Here are the top co-authors by number of shared publications:
Michael D. Stein
6 shared publications- Michael M. Phillips
School of Psychological Sciences, Campus Box 94, University of Northern Colorado, Greeley, CO 80639, United States
6 shared publications - Peter Washington
Division of Clinical Informatics and Digital Transformation (DoC-IT), Department of Medicine, University of California, San Francisco, USA
6 shared publications - Aditi Jaiswal
Department of Information and Computer Sciences, University of Hawaii at Manoa, Honolulu, HI 96822, USA
5 shared publications - Ali Kargarandehkordi
Department of Information and Computer Sciences, University of Hawaiʻi at Mānoa, Honolulu, USA
5 shared publications - Christopher Slade
Department of Information and Computer Science, University of Hawaii, Honolulu, HI 96822, United States
5 shared publications
Latest Publications
- 14 July 2026
- Published by Elsevier in Addictive Behaviors
- Vol. 183, 108797
Associations of motivation to quit smoking or vaping, quitting self-efficacy, and quit attempts with oral nicotine pouch use
- 14 July 2026
- Published by Elsevier in Addictive Behaviors
- Vol. 183, 108797
Associations of motivation to quit smoking or vaping, quitting self-efficacy, and quit attempts with oral nicotine pouch use
E‐cigarette and cannabis use and co‐use patterns and dependence trajectories among young adults in the United States
Background and aims: Co‐use of cannabis and e‐cigarettes for nicotine is increasing among younger adults in the United States (US). This study aimed to determine (1) the patterns of cannabis and e‐cigarette use and co‐use over time and (2) the trajectories of e‐cigarette and cannabis dependence symptoms across latent classes representing different patterns of use and co‐use. Design: Prospective cohort study, with data collected every six months at six time points. Setting: Hawaii, USA. Participants: N = 1305 individuals [mean age = 25.0 (standard deviation = 2.2); 52% women], who reported past‐30‐day cannabis and/or e‐cigarette use at any of the six data collection time points between March 2021 and February 2024. Measurements: For the repeated measures latent class analysis (RMLCA): past‐30‐day cannabis use, e‐cigarette use and co‐use (any past‐30‐day use of both cannabis and e‐cigarettes for nicotine). RMLCA covariates: demographic variables and cigarette smoking. For the latent growth curve analyses: e‐cigarette dependence [Penn State E‐Cigarette Dependence Index (PSEDI), range: 0–20] and cannabis dependence symptoms [Cannabis Abuse Screening Test (CAST), range: 0–24]. Findings: Four latent classes were identified: individuals who, across time points, consistently showed (1) low probabilities (≤ 0.17) of using e‐cigarettes or cannabis, including co‐use (44% of the sample); and (2) high probabilities (> 0.50) of primarily e‐cigarettes‐only use (18%); (3) primarily cannabis‐only use (20%); and (4) primarily co‐use (18%). For co‐use class, the average trajectory for cannabis dependence remained flat [mean linear slope (LS) = 0.01 (standard error [SE] = 0.03), P > 0.05] whereas that for e‐cigarette dependence increased over time [LS = 0.15 (0.03), P < 0.0001] before showing a slight decline [mean quadratic slope (QS) = −0.02 (0.007), P < 0.01]. For e‐cigarette‐only use class, e‐cigarette dependence increased over time [LS = 0.91 (0.05), P < 0.0001], before showing a slight decline [QS = −0.02 (0.006), P < 0.0001]. The levels of e‐cigarette dependence were higher for e‐cigarette‐only class than the co‐use class across all time points. For cannabis‐only class, the average trajectory of cannabis dependence symptoms did not change over time [LS = 0.03 (0.03), P > 0.05]. The levels of cannabis dependence symptoms were higher for the co‐use class than the cannabis‐only class across all time points. Conclusions: Among young adults who reported past‐30‐day cannabis and/or e‐cigarette use, on average, e‐cigarette dependence increased over time for both e‐cigarette‐only and co‐use classes before showing slight declines. Compared with the co‐use class, the e‐cigarette‐only class tended to show higher levels of e‐cigarette dependence across all time points. Cannabis use dependence did not change statistically significantly for either co‐use or cannabis‐only classes.
E‐cigarette and cannabis use and co‐use patterns and dependence trajectories among young adults in the United States
Background and aims: Co‐use of cannabis and e‐cigarettes for nicotine is increasing among younger adults in the United States (US). This study aimed to determine (1) the patterns of cannabis and e‐cigarette use and co‐use over time and (2) the trajectories of e‐cigarette and cannabis dependence symptoms across latent classes representing different patterns of use and co‐use. Design: Prospective cohort study, with data collected every six months at six time points. Setting: Hawaii, USA. Participants: N = 1305 individuals [mean age = 25.0 (standard deviation = 2.2); 52% women], who reported past‐30‐day cannabis and/or e‐cigarette use at any of the six data collection time points between March 2021 and February 2024. Measurements: For the repeated measures latent class analysis (RMLCA): past‐30‐day cannabis use, e‐cigarette use and co‐use (any past‐30‐day use of both cannabis and e‐cigarettes for nicotine). RMLCA covariates: demographic variables and cigarette smoking. For the latent growth curve analyses: e‐cigarette dependence [Penn State E‐Cigarette Dependence Index (PSEDI), range: 0–20] and cannabis dependence symptoms [Cannabis Abuse Screening Test (CAST), range: 0–24]. Findings: Four latent classes were identified: individuals who, across time points, consistently showed (1) low probabilities (≤ 0.17) of using e‐cigarettes or cannabis, including co‐use (44% of the sample); and (2) high probabilities (> 0.50) of primarily e‐cigarettes‐only use (18%); (3) primarily cannabis‐only use (20%); and (4) primarily co‐use (18%). For co‐use class, the average trajectory for cannabis dependence remained flat [mean linear slope (LS) = 0.01 (standard error [SE] = 0.03), P > 0.05] whereas that for e‐cigarette dependence increased over time [LS = 0.15 (0.03), P < 0.0001] before showing a slight decline [mean quadratic slope (QS) = −0.02 (0.007), P < 0.01]. For e‐cigarette‐only use class, e‐cigarette dependence increased over time [LS = 0.91 (0.05), P < 0.0001], before showing a slight decline [QS = −0.02 (0.006), P < 0.0001]. The levels of e‐cigarette dependence were higher for e‐cigarette‐only class than the co‐use class across all time points. For cannabis‐only class, the average trajectory of cannabis dependence symptoms did not change over time [LS = 0.03 (0.03), P > 0.05]. The levels of cannabis dependence symptoms were higher for the co‐use class than the cannabis‐only class across all time points. Conclusions: Among young adults who reported past‐30‐day cannabis and/or e‐cigarette use, on average, e‐cigarette dependence increased over time for both e‐cigarette‐only and co‐use classes before showing slight declines. Compared with the co‐use class, the e‐cigarette‐only class tended to show higher levels of e‐cigarette dependence across all time points. Cannabis use dependence did not change statistically significantly for either co‐use or cannabis‐only classes.
- 17 May 2026
- Published by Elsevier in Drug and Alcohol Dependence
- Vol. 285, 113199
Oral nicotine pouch use motives and dependence among individuals who use both cigarettes and e-cigarettes
- 17 May 2026
- Published by Elsevier in Drug and Alcohol Dependence
- Vol. 285, 113199
Oral nicotine pouch use motives and dependence among individuals who use both cigarettes and e-cigarettes
- 14 December 2025
- Published by World Scientific Pub Co Pte Ltd in Biocomputing 2020
- Vol. 31, 566-579
Barriers to Designing Inclusive Ecological Momentary Assessment and Wearable Data Collection Protocols for AI-Driven Substance Use Monitoring in Hawai‘i
Ecological momentary assessment (EMA) and wearable sensors offer unprecedented opportunities to capture the dynamics of substance use through real-time, high-resolution behavioral and physiological data. These data streams are increasingly used to train AI/ML models for digital phenotyping and predictive intervention, raising critical questions about fairness, bias, and inclusivity in model development. However, the adoption of these technologies, or the lack thereof, among diverse and historically marginalized groups raises questions and challenges of equity, cultural relevance, and participant trust. In this study, we conducted a four-week observational study with adults in Hawai.i where we combined continuous Fitbit monitoring and daily EMA surveys to document substance use patterns and cravings. Through semi-structured interviews and grounded theory analysis, we identified six primary barriers to study participation and adherence: (1) disruptions to daily routines, (2) physical and psychosocial discomfort associated with wearing the Fitbit device, (3) concerns about aesthetic compatibility and professional appearance, (4) phonerelated issues, (5) challenges related to substance use and cravings, and (6) socially sensitive contexts. We also highlight participant-identified facilitators, such as the value of participant-driven scheduling, motivational feedback, and contextually adaptive protocols. Drawing on these collective findings, we propose a set of design guidelines aimed at advancing the inclusivity, engagement, and fairness of wearable-based EMA research.
- 14 December 2025
- Published by World Scientific Pub Co Pte Ltd in Biocomputing 2020
- Vol. 31, 566-579
Barriers to Designing Inclusive Ecological Momentary Assessment and Wearable Data Collection Protocols for AI-Driven Substance Use Monitoring in Hawai‘i
Ecological momentary assessment (EMA) and wearable sensors offer unprecedented opportunities to capture the dynamics of substance use through real-time, high-resolution behavioral and physiological data. These data streams are increasingly used to train AI/ML models for digital phenotyping and predictive intervention, raising critical questions about fairness, bias, and inclusivity in model development. However, the adoption of these technologies, or the lack thereof, among diverse and historically marginalized groups raises questions and challenges of equity, cultural relevance, and participant trust. In this study, we conducted a four-week observational study with adults in Hawai.i where we combined continuous Fitbit monitoring and daily EMA surveys to document substance use patterns and cravings. Through semi-structured interviews and grounded theory analysis, we identified six primary barriers to study participation and adherence: (1) disruptions to daily routines, (2) physical and psychosocial discomfort associated with wearing the Fitbit device, (3) concerns about aesthetic compatibility and professional appearance, (4) phonerelated issues, (5) challenges related to substance use and cravings, and (6) socially sensitive contexts. We also highlight participant-identified facilitators, such as the value of participant-driven scheduling, motivational feedback, and contextually adaptive protocols. Drawing on these collective findings, we propose a set of design guidelines aimed at advancing the inclusivity, engagement, and fairness of wearable-based EMA research.
- Open Access
- 23 September 2025
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 27, e69831
Associations Between Social Determinants of Health and Adherence in Mobile-Based Ecological Momentary Assessment: Scoping Review
Background: Ecological momentary assessment (EMA) involves repeated prompts to capture real-time self-reported health outcomes and behaviors via mobile devices. With the rise of mobile health (mHealth) technologies, EMA has been applied across diverse populations and health domains. However, the extent to which EMA engagement and data quality vary across social determinants of health (SDoH) remains underexplored. Emerging evidence suggests that EMA adherence and data completeness may be sometimes associated with participant characteristics such as socioeconomic status, race/ethnicity, and education level. These associations may sometimes influence who engages with EMA protocols and the types of contextual data captured. Despite growing interest in these patterns, no review to date has synthesized evidence on how SDoH relate to EMA compliance and engagement. Objective: We conducted a scoping review to study two research questions: (R1) how EMA compliance rates in health studies can differ across SDoH and (R2) what types of SDoH have been identified through EMA health studies. Methods: Following PRISMA-ScR guidelines, we searched PubMed, Web of Science, and EBSCOhost using two sets of queries targeting EMA and its relationship to SDoH. Eligible studies were peer reviewed, were published in English between 2013 and 2024, and used mobile-based EMA methods. Studies were included if they (1) reported on differences in EMA compliance by SDoH or (2) reported at least one SDoH observed or uncovered during an EMA study. We used the social ecological model (SEM) as a guiding framework to categorize and interpret SDoH across individual, interpersonal, community, and societal levels. A qualitative thematic synthesis was conducted to iteratively and collaboratively extract, categorize, and review determinants. Results: We analyzed 48 eligible studies, of which 35 addressed R1 by examining compliance patterns across various SDoH. Using the SEM, we identified 13 determinants categorized across 4 levels: individual (eg, daily routine, biological sex, age, socioeconomic status, language, education, and race or ethnicity), interpersonal (eg, social support), community and organizational (eg, social context, social acceptance, stigmatization, and youth culture), and policy or societal (eg, systemic and structural barriers). These studies described differences in EMA response rates, compliance, and dropout associated with these determinants, often among vulnerable populations. The remaining 13 studies addressed R2, demonstrating examples of the types of SDoH that EMA research can uncover, including family culture, social support, social contexts, stigmatization, gender norms, heroic narratives, LGBTQ+ culture, racial discrimination, and systematic and structural barriers. Conclusions: This scoping review illustrates how EMA compliance rates can differ across SDoH and highlights the potential of EMA to uncover social and cultural factors linked to health behaviors and outcomes. Our findings underscore the importance of integrating SDoH considerations into EMA study designs to capture context-specific sociocultural dynamics.
- 23 September 2025
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 27, e69831
Open AccessAssociations Between Social Determinants of Health and Adherence in Mobile-Based Ecological Momentary Assessment: Scoping Review
Background: Ecological momentary assessment (EMA) involves repeated prompts to capture real-time self-reported health outcomes and behaviors via mobile devices. With the rise of mobile health (mHealth) technologies, EMA has been applied across diverse populations and health domains. However, the extent to which EMA engagement and data quality vary across social determinants of health (SDoH) remains underexplored. Emerging evidence suggests that EMA adherence and data completeness may be sometimes associated with participant characteristics such as socioeconomic status, race/ethnicity, and education level. These associations may sometimes influence who engages with EMA protocols and the types of contextual data captured. Despite growing interest in these patterns, no review to date has synthesized evidence on how SDoH relate to EMA compliance and engagement. Objective: We conducted a scoping review to study two research questions: (R1) how EMA compliance rates in health studies can differ across SDoH and (R2) what types of SDoH have been identified through EMA health studies. Methods: Following PRISMA-ScR guidelines, we searched PubMed, Web of Science, and EBSCOhost using two sets of queries targeting EMA and its relationship to SDoH. Eligible studies were peer reviewed, were published in English between 2013 and 2024, and used mobile-based EMA methods. Studies were included if they (1) reported on differences in EMA compliance by SDoH or (2) reported at least one SDoH observed or uncovered during an EMA study. We used the social ecological model (SEM) as a guiding framework to categorize and interpret SDoH across individual, interpersonal, community, and societal levels. A qualitative thematic synthesis was conducted to iteratively and collaboratively extract, categorize, and review determinants. Results: We analyzed 48 eligible studies, of which 35 addressed R1 by examining compliance patterns across various SDoH. Using the SEM, we identified 13 determinants categorized across 4 levels: individual (eg, daily routine, biological sex, age, socioeconomic status, language, education, and race or ethnicity), interpersonal (eg, social support), community and organizational (eg, social context, social acceptance, stigmatization, and youth culture), and policy or societal (eg, systemic and structural barriers). These studies described differences in EMA response rates, compliance, and dropout associated with these determinants, often among vulnerable populations. The remaining 13 studies addressed R2, demonstrating examples of the types of SDoH that EMA research can uncover, including family culture, social support, social contexts, stigmatization, gender norms, heroic narratives, LGBTQ+ culture, racial discrimination, and systematic and structural barriers. Conclusions: This scoping review illustrates how EMA compliance rates can differ across SDoH and highlights the potential of EMA to uncover social and cultural factors linked to health behaviors and outcomes. Our findings underscore the importance of integrating SDoH considerations into EMA study designs to capture context-specific sociocultural dynamics.
- 15 August 2025
- Published by Elsevier in Osteoarthritis and Cartilage
- Vol. 33 (12) , 1522-1533
Interpretable machine learning integrates multi-source biomarkers for osteoarthritis diagnosis and mechanistic insights: A temporomandibular joint model
- 15 August 2025
- Published by Elsevier in Osteoarthritis and Cartilage
- Vol. 33 (12) , 1522-1533
Interpretable machine learning integrates multi-source biomarkers for osteoarthritis diagnosis and mechanistic insights: A temporomandibular joint model
- Open Access
- 10 June 2025
- Published by Research Society on Marijuana in Cannabis
- Vol. 8 (2) , 112-128
Cannabis Use Patterns and Co-Use of Alcohol and Nicotine in Adults Over 50 by Demographic Factors and Medical Cannabis Use
Objective: Cannabis use among adults over age 50 is increasing, but data on specific products, co-use, and cannabis-related problems in this age group are lacking. The current study assessed differences in cannabis use patterns and alcohol and nicotine co-use by select demographic factors and medical cannabis status, as well as associations with problem cannabis use, among adults over 50. Method: Adults over age 50 who used cannabis use in the past 30 days were recruited from a healthcare system and invited to complete an online survey. Results: Participants (N = 367) were 43% female, with an average age of 65.9 (SD = 8.6), and predominantly White (67.9%), Native Hawaiian or Pacific Islander (NHPI; 12.3%), or Asian (9.3%). Mean frequency of cannabis use within the past 30 days was 18.7 days (SD = 11.5), and 27% reported use of non-medical cannabis-only. Co-use of cannabis with other substances was common, especially alcohol. Although group comparisons showed several differences by age, sex, and medical cannabis status, greater differences were found by race/ethnicity. White participants were more likely than others to have a state medical cannabis card, source their cannabis solely from medical dispensaries, and report alcohol co-use. NHPI participants were most likely to smoke cannabis and use tobacco. Smoking, as compared to consuming edibles and “other” methods, and greater cannabis frequency, were associated with problem cannabis use. Conclusions: Findings illustrate patterns of cannabis and other substance use, with important demographic differences. Future research among older adults should include development of targeted interventions to address cannabis use problems and polysubstance use.
- 10 June 2025
- Published by Research Society on Marijuana in Cannabis
- Vol. 8 (2) , 112-128
Open AccessCannabis Use Patterns and Co-Use of Alcohol and Nicotine in Adults Over 50 by Demographic Factors and Medical Cannabis Use
Objective: Cannabis use among adults over age 50 is increasing, but data on specific products, co-use, and cannabis-related problems in this age group are lacking. The current study assessed differences in cannabis use patterns and alcohol and nicotine co-use by select demographic factors and medical cannabis status, as well as associations with problem cannabis use, among adults over 50. Method: Adults over age 50 who used cannabis use in the past 30 days were recruited from a healthcare system and invited to complete an online survey. Results: Participants (N = 367) were 43% female, with an average age of 65.9 (SD = 8.6), and predominantly White (67.9%), Native Hawaiian or Pacific Islander (NHPI; 12.3%), or Asian (9.3%). Mean frequency of cannabis use within the past 30 days was 18.7 days (SD = 11.5), and 27% reported use of non-medical cannabis-only. Co-use of cannabis with other substances was common, especially alcohol. Although group comparisons showed several differences by age, sex, and medical cannabis status, greater differences were found by race/ethnicity. White participants were more likely than others to have a state medical cannabis card, source their cannabis solely from medical dispensaries, and report alcohol co-use. NHPI participants were most likely to smoke cannabis and use tobacco. Smoking, as compared to consuming edibles and “other” methods, and greater cannabis frequency, were associated with problem cannabis use. Conclusions: Findings illustrate patterns of cannabis and other substance use, with important demographic differences. Future research among older adults should include development of targeted interventions to address cannabis use problems and polysubstance use.
- Top Cited PapersOpen Access
- 21 March 2025
- Published by MDPI AG in Biosensors
- Vol. 15 (4) , 202
Fusing Wearable Biosensors with Artificial Intelligence for Mental Health Monitoring: A Systematic Review
The development of digital instruments for mental health monitoring using biosensor data from wearable devices can enable remote, longitudinal, and objective quantitative benchmarks. To survey developments and trends in this field, we conducted a systematic review of artificial intelligence (AI) models using data from wearable biosensors to predict mental health conditions and symptoms. Following PRISMA guidelines, we identified 48 studies using a variety of wearable and smartphone biosensors including heart rate, heart rate variability (HRV), electrodermal activity/galvanic skin response (EDA/GSR), and digital proxies for biosignals such as accelerometry, location, audio, and usage metadata. We observed several technical and methodological challenges across studies in this field, including lack of ecological validity, data heterogeneity, small sample sizes, and battery drainage issues. We outline several corresponding opportunities for advancement in the field of AI-driven biosensing for mental health.
- 21 March 2025
- Published by MDPI AG in Biosensors
- Vol. 15 (4) , 202
Top Cited PapersOpen AccessFusing Wearable Biosensors with Artificial Intelligence for Mental Health Monitoring: A Systematic Review
The development of digital instruments for mental health monitoring using biosensor data from wearable devices can enable remote, longitudinal, and objective quantitative benchmarks. To survey developments and trends in this field, we conducted a systematic review of artificial intelligence (AI) models using data from wearable biosensors to predict mental health conditions and symptoms. Following PRISMA guidelines, we identified 48 studies using a variety of wearable and smartphone biosensors including heart rate, heart rate variability (HRV), electrodermal activity/galvanic skin response (EDA/GSR), and digital proxies for biosignals such as accelerometry, location, audio, and usage metadata. We observed several technical and methodological challenges across studies in this field, including lack of ecological validity, data heterogeneity, small sample sizes, and battery drainage issues. We outline several corresponding opportunities for advancement in the field of AI-driven biosensing for mental health.
- Open Access
Monitoring Substance Use with Fitbit Biosignals: A Case Study on Training Deep Learning Models Using Ecological Momentary Assessments and Passive Sensing
Substance use disorders affect 17.3% of Americans. Digital health solutions that use machine learning to detect substance use from wearable biosignal data can eventually pave the way for real-time digital interventions. However, difficulties in addressing severe between-subject data heterogeneity have hampered the adaptation of machine learning approaches for substance use detection, necessitating more robust technological solutions. We tested the utility of personalized machine learning using participant-specific convolutional neural networks (CNNs) enhanced with self-supervised learning (SSL) to detect drug use. In a pilot feasibility study, we collected data from 9 participants using Fitbit Charge 5 devices, supplemented by ecological momentary assessments to collect real-time labels of substance use. We implemented a baseline 1D-CNN model with traditional supervised learning and an experimental SSL-enhanced model to improve individualized feature extraction under limited label conditions. Results: Among the 9 participants, we achieved an average area under the receiver operating characteristic curve score across participants of 0.695 for the supervised CNNs and 0.729 for the SSL models. Strategic selection of an optimal threshold enabled us to optimize either sensitivity or specificity while maintaining reasonable performance for the other metric. Conclusion: These findings suggest that Fitbit data have the potential to enhance substance use monitoring systems. However, the small sample size in this study limits its generalizability to diverse populations, so we call for future research that explores SSL-powered personalization at a larger scale.
Open AccessMonitoring Substance Use with Fitbit Biosignals: A Case Study on Training Deep Learning Models Using Ecological Momentary Assessments and Passive Sensing
Substance use disorders affect 17.3% of Americans. Digital health solutions that use machine learning to detect substance use from wearable biosignal data can eventually pave the way for real-time digital interventions. However, difficulties in addressing severe between-subject data heterogeneity have hampered the adaptation of machine learning approaches for substance use detection, necessitating more robust technological solutions. We tested the utility of personalized machine learning using participant-specific convolutional neural networks (CNNs) enhanced with self-supervised learning (SSL) to detect drug use. In a pilot feasibility study, we collected data from 9 participants using Fitbit Charge 5 devices, supplemented by ecological momentary assessments to collect real-time labels of substance use. We implemented a baseline 1D-CNN model with traditional supervised learning and an experimental SSL-enhanced model to improve individualized feature extraction under limited label conditions. Results: Among the 9 participants, we achieved an average area under the receiver operating characteristic curve score across participants of 0.695 for the supervised CNNs and 0.729 for the SSL models. Strategic selection of an optimal threshold enabled us to optimize either sensitivity or specificity while maintaining reasonable performance for the other metric. Conclusion: These findings suggest that Fitbit data have the potential to enhance substance use monitoring systems. However, the small sample size in this study limits its generalizability to diverse populations, so we call for future research that explores SSL-powered personalization at a larger scale.