An Explainable Artificial Intelligence Software Tool for Weight Management Experts (PRIMO): Mixed Methods Study.

BACKGROUND: Predicting the likelihood of success of weight loss interventions using machine learning (ML) models may enhance intervention effectiveness by enabling timely and dynamic modification of intervention components for nonresponders to treatment. However, a lack of understanding and trust in these ML models impacts adoption among weight management experts. Recent advances in the field of explainable artificial intelligence enable the interpretation of ML models, yet it is unknown whether they enhance model understanding, trust, and adoption among weight management experts. OBJECTIVE: This study aimed to build and evaluate an ML model that can predict 6-month weight loss success (ie, ≥7% weight loss) from 5 engagement and diet-related features collected over the initial 2 weeks of an intervention, to assess whether providing ML-based explanations increases weight management experts' agreement with ML model predictions, and to inform factors that influence the understanding and trust of ML models to advance explainability in early prediction of weight loss among weight management experts. METHODS: We trained an ML model using the random forest (RF) algorithm and data from a 6-month weight loss intervention (N=419). We leveraged findings from existing explainability metrics to develop Prime Implicant Maintenance of Outcome (PRIMO), an interactive tool to understand predictions made by the RF model. We asked 14 weight management experts to predict hypothetical participants' weight loss success before and after using PRIMO. We compared PRIMO with 2 other explainability methods, one based on feature ranking and the other based on conditional probability. We used generalized linear mixed-effects models to evaluate participants' agreement with ML predictions and conducted likelihood ratio tests to examine the relationship between explainability methods and outcomes for nested models. We conducted guided interviews and thematic analysis to study the impact of our tool on experts' understanding and trust in the model. RESULTS: Our RF model had 81% accuracy in the early prediction of weight loss success. Weight management experts were significantly more likely to agree with the model when using PRIMO (χ2=7.9; P=.02) compared with the other 2 methods with odds ratios of 2.52 (95% CI 0.91-7.69) and 3.95 (95% CI 1.50-11.76). From our study, we inferred that our software not only influenced experts' understanding and trust but also impacted decision-making. Several themes were identified through interviews: preference for multiple explanation types, need to visualize uncertainty in explanations provided by PRIMO, and need for model performance metrics on similar participant test instances. CONCLUSIONS: Our results show the potential for weight management experts to agree with the ML-based early prediction of success in weight loss treatment programs, enabling timely and dynamic modification of intervention components to enhance intervention effectiveness. Our findings provide methods for advancing the understandability and trust of ML models among weight management experts.

Study protocol of the Health4Life initiative: a cluster randomised controlled trial of an eHealth school-based program targeting multiple lifestyle risk behaviours among young Australians.

INTRODUCTION: Lifestyle risk behaviours, including alcohol use, smoking, poor diet, physical inactivity, poor sleep (duration and/or quality) and sedentary recreational screen time ('the Big 6'), are strong determinants of chronic disease. These behaviours often emerge during adolescence and co-occur. School-based interventions have the potential to address risk factors prior to the onset of disease, yet few eHealth school-based interventions target multiple behaviours concurrently. This paper describes the protocol of the Health4Life Initiative, an eHealth school-based intervention that concurrently addresses the Big 6 risk behaviours among secondary school students. METHODS AND ANALYSIS: A multisite cluster randomised controlled trial will be conducted among year 7 students (11-13 years old) from 72 Australian schools. Stratified block randomisation will be used to assign schools to either the Health4Life intervention or an active control (health education as usual). Health4Life consists of (1) six web-based cartoon modules and accompanying activities delivered during health education (once per week for 6 weeks), and a smartphone application (universal prevention), and (2) additional app content, for students engaging in two or more risk behaviours when they are in years 8 and 9 (selective prevention). Students will complete online self-report questionnaires at baseline, post intervention, and 12, 24 and 36 months after baseline. Primary outcomes are consumption of sugar-sweetened beverages, moderate-to-vigorous physical activity, sleep duration, sedentary recreational screen time and uptake of alcohol and tobacco use. ETHICS AND DISSEMINATION: This study has been approved by the University of Sydney (2018/882), NSW Department of Education (SERAP no. 2019006), University of Queensland (2019000037), Curtin University (HRE2019-0083) and relevant Catholic school committees. Results will be presented to schools and findings disseminated via peer-reviewed journals and scientific conferences. This will be the first evaluation of an eHealth intervention, spanning both universal and selective prevention, to simultaneously target six key lifestyle risk factors among adolescents. TRIAL REGISTRATION NUMBER: Australian New Zealand Clinical Trials Registry (ACTRN12619000431123), 18 March 2019.

Impact of Baseline Fatigue on a Physical Activity Intervention to Prevent Mobility Disability.

OBJECTIVES: Our aim was to examine the impacts of baseline fatigue on the effectiveness of a physical activity (PA) intervention to prevent major mobility disability (MMD) and persistent major mobility disability (PMMD) in participants from the Lifestyle Interventions and Independence for Elders (LIFE) study. DESIGN: Prospective cohort of individuals aged 65 years or older undergoing structured PA intervention or health education (HE) for a mean of 2.6 years. SETTING: LIFE was a multicenter eight-site randomized trial that compared the efficacy of a structured PA intervention with an HE program in reducing the incidence of MMD. PARTICIPANTS: Study participants (N = 1591) at baseline were 78.9 ± 5.2 years of age, with low PA and at risk for mobility impairment. MEASUREMENTS: Self-reported fatigue was assessed using the modified trait version of the Exercise-Induced Feelings Inventory, a six-question scale rating energy levels in the past week. Responses ranged from 0 (none of the time) to 5 (all of the time). Total score was calculated by averaging across questions; baseline fatigue was based on the median split: 2 or higher = more fatigue (N = 856) and lower than 2 = less fatigue (N = 735). Participants performed a usual-paced 400-m walk every 6 months. We defined incident MMD as the inability to walk 400-m at follow-up visits; PMMD was defined as two consecutive walk failures. Cox proportional hazard models quantified the risk of MMD and PMMD in PA vs HE stratified by baseline fatigue adjusted for covariates. RESULTS: Among those with higher baseline fatigue, PA participants had a 29% and 40% lower risk of MMD and PMMD, respectively, over the trial compared with HE (hazard ratio [HR] for MMD = .71; 95% confidence interval [CI] = .57-.90; P = .004) and PMMD (HR = .60; 95% CI = .44-.82; P = .001). For those with lower baseline fatigue, no group differences in MMD (P = .36) or PMMD (P = .82) were found. Results of baseline fatigue by intervention interaction was MMD (P = .18) and PMMD (P = .05). CONCLUSION: A long-term moderate intensity PA intervention was particularly effective at preserving mobility in older adults with higher levels of baseline fatigue. J Am Geriatr Soc 68:619-624, 2020.

Genetic Predisposition Impacts Clinical Changes in a Lifestyle Coaching Program.

Both genetic and lifestyle factors contribute to an individual's disease risk, suggesting a multi-omic approach is essential for personalized prevention. Studies have examined the effectiveness of lifestyle coaching on clinical outcomes, however, little is known about the impact of genetic predisposition on the response to lifestyle coaching. Here we report on the results of a real-world observational study in 2531 participants enrolled in a commercial "Scientific Wellness" program, which combines multi-omic data with personalized, telephonic lifestyle coaching. Specifically, we examined: 1) the impact of this program on 55 clinical markers and 2) the effect of genetic predisposition on these clinical changes. We identified sustained improvements in clinical markers related to cardiometabolic risk, inflammation, nutrition, and anthropometrics. Notably, improvements in HbA1c were akin to those observed in landmark trials. Furthermore, genetic markers were associated with longitudinal changes in clinical markers. For example, individuals with genetic predisposition for higher LDL-C had a lesser decrease in LDL-C on average than those with genetic predisposition for average LDL-C. Overall, these results suggest that a program combining multi-omic data with lifestyle coaching produces clinically meaningful improvements, and that genetic predisposition impacts clinical responses to lifestyle change.

Just-in-Time Adaptive Interventions (JITAIs) in Mobile Health: Key Components and Design Principles for Ongoing Health Behavior Support.

Background: The just-in-time adaptive intervention (JITAI) is an intervention design aiming to provide the right type/amount of support, at the right time, by adapting to an individual's changing internal and contextual state. The availability of increasingly powerful mobile and sensing technologies underpins the use of JITAIs to support health behavior, as in such a setting an individual's state can change rapidly, unexpectedly, and in his/her natural environment. Purpose: Despite the increasing use and appeal of JITAIs, a major gap exists between the growing technological capabilities for delivering JITAIs and research on the development and evaluation of these interventions. Many JITAIs have been developed with minimal use of empirical evidence, theory, or accepted treatment guidelines. Here, we take an essential first step towards bridging this gap. Methods: Building on health behavior theories and the extant literature on JITAIs, we clarify the scientific motivation for JITAIs, define their fundamental components, and highlight design principles related to these components. Examples of JITAIs from various domains of health behavior research are used for illustration. Conclusions: As we enter a new era of technological capacity for delivering JITAIs, it is critical that researchers develop sophisticated and nuanced health behavior theories capable of guiding the construction of such interventions. Particular attention has to be given to better understanding the implications of providing timely and ecologically sound support for intervention adherence and retention.

Effect of Metabolic Syndrome on the Mobility Benefit of a Structured Physical Activity Intervention-The Lifestyle Interventions and Independence for Elders Randomized Clinical Trial.

OBJECTIVES: To test whether structured physical activity (PA) is associated with a greater reduction in major mobility disability (MMD) in older persons with metabolic syndrome (MetS) than in those without. DESIGN: Data from the Lifestyle Interventions and Independence for Elders (LIFE) Study, a multicenter randomized trial of 1,635 persons with assessments every 6 months (average 2.7 years). SETTING: Eight U.S. centers. PARTICIPANTS: Sedentary men and women aged 70 to 89 with functional limitations (N = 1,535); 100 participants were excluded because of missing MetS data. INTERVENTION: Participants were randomized to a moderate-intensity PA program (n = 766) or a health education program (n = 769). MEASUREMENTS: MetS was defined according to the 2009 multiagency harmonized criteria. Outcomes included incident MMD (loss of ability to walk 400 m) and persistent MMD (two consecutive MMD diagnoses or one MMD diagnosis followed by death). RESULTS: Seven hundred sixty-three (49.7%) participants met criteria for MetS. PA reduced incident MMD more than health education did in participants with MetS (hazard ratio (HR) = 0.72, 95% confidence interval (CI) = 0.57-0.91, P = .007) but not in those without MetS (HR = 0.96, 95% CI = 0.73-1.25, P = .75); the test for statistical interaction was not significant (P = .13). PA reduced the risk of persistent MMD in participants with MetS (HR = 0.57, 95% CI = 0.41-0.79, P < .001) but not in those without MetS (HR = 0.97, 95% CI = 0.67-1.41, P = .87). The test for statistical interaction was significant (P = .04). CONCLUSION: Moderate-intensity PA substantially reduces the risk of persistent MMD in older persons with functional limitations with MetS but not in those without MetS. Comparable results were observed for incident MMD. The LIFE PA program may be an effective strategy for reducing mobility disability in vulnerable older persons with MetS.

Understanding and Promoting Effective Engagement With Digital Behavior Change Interventions.

This paper is one in a series developed through a process of expert consensus to provide an overview of questions of current importance in research into engagement with digital behavior change interventions, identifying guidance based on research to date and priority topics for future research. The first part of this paper critically reflects on current approaches to conceptualizing and measuring engagement. Next, issues relevant to promoting effective engagement are discussed, including how best to tailor to individual needs and combine digital and human support. A key conclusion with regard to conceptualizing engagement is that it is important to understand the relationship between engagement with the digital intervention and the desired behavior change. This paper argues that it may be more valuable to establish and promote "effective engagement," rather than simply more engagement, with "effective engagement" defined empirically as sufficient engagement with the intervention to achieve intended outcomes. Appraisal of the value and limitations of methods of assessing different aspects of engagement highlights the need to identify valid and efficient combinations of measures to develop and test multidimensional models of engagement. The final section of the paper reflects on how interventions can be designed to fit the user and their specific needs and context. Despite many unresolved questions posed by novel and rapidly changing technologies, there is widespread consensus that successful intervention design demands a user-centered and iterative approach to development, using mixed methods and in-depth qualitative research to progressively refine the intervention to meet user requirements.

Breast cancer survivors' beliefs and preferences regarding technology-supported sedentary behavior reduction interventions.

PURPOSE: Less time spent in sedentary behaviors is associated with improved health and disease outcomes in breast cancer survivors. However, little is known about survivors' interest in sedentary behavior reduction interventions and how to effectively reduce this risk behavior. The purpose of this study was to explore breast cancer survivors' interest in and preferences for technology-supported sedentary behavior reduction interventions. METHODS: Breast cancer survivors [n=279; Mage =60.7 (SD=9.7)] completed a battery of online questionnaires. Descriptive statistics were calculated for all data. To examine potential relationships between demographic, disease and behavioral factors, and survivors' interest in a technology-supported sedentary behavior reduction intervention, we conducted logistic regression analyses. These same factors were examined in relation to the perceptions of the effectiveness of such intervention using multiple regression analyses. RESULTS: On average, survivors spent 10.1 (SD=4.3) hours/day in sedentary activity. They believed prolonged periods of sedentary behavior were harmful to their health (87.0%) and that reducing sedentary behavior could improve their health (88.4%). Survivors believed they should move around after 30-60 (56.7%) or ≥60 (29.9%) minutes of sedentary behavior and indicated they were most likely to replace sedentary behaviors with walking around (97.1%) or walking in place (73.4%). The majority of survivors (79.9%) was interested in participating in a technology-supported sedentary behavior reduction intervention and indicated they would use a smartphone application (61.3%) 2-3 times/day (48.0%), 6 to 7 days/week (52.0%). Most survivors (73.5%) believed reminders would help them decrease sedentary behavior and preferred they be delivered after sitting for 60 minutes (60.5%) via vibrations on a wrist worn activity tracker (77.3%) or text messages (54.4%). CONCLUSIONS: Technology-supported sedentary behavior reduction interventions may be feasible and acceptable to breast cancer survivors. Data regarding user preferences for content, features, delivery mode and design will aid researchers in developing sedentary interventions that are potentially more relevant and effective from the outset.

Effect of Structured Physical Activity on Sleep-Wake Behaviors in Sedentary Elderly Adults with Mobility Limitations.

OBJECTIVES: To evaluate the effect of structured physical activity on sleep-wake behaviors in sedentary community-dwelling elderly adults with mobility limitations. DESIGN: Multicenter, randomized trial of moderate-intensity physical activity versus health education, with sleep-wake behaviors prespecified as a tertiary outcome over a planned intervention period ranging from 24 to 30 months. SETTING: Lifestyle Interventions and Independence for Elders Study. PARTICIPANTS: Community-dwelling persons aged 70 to 89 who were initially sedentary and had a Short Physical Performance Battery score less than 10 (N = 1,635). MEASUREMENTS: Sleep-wake behaviors were evaluated using the Insomnia Severity Index (ISI) (≥8 defined insomnia), Epworth Sleepiness Scale (ESS) (≥10 defined daytime drowsiness), and Pittsburgh Sleep Quality Index (PSQI) (>5 defined poor sleep quality) administered at baseline and 6, 18, and 30 months. RESULTS: The randomized groups were similar in terms of baseline demographic variables, including mean age (79) and sex (67% female). Structured physical activity resulted in a significantly lower likelihood of having poor sleep quality (adjusted odds ratios (aOR) for PSQI >5 = 0.80, 95% confidence interval (CI) = 0.68-0.94), including fewer new cases (aOR for PSQI >5 = 0.70, 95% CI = 0.54-0.89), than health education but not in resolution of prevalent cases (aOR for PSQI ≤5 = 1.13, 95% CI = 0.90-1.43). No significant intervention effects were observed for the ISI or ESS. CONCLUSION: Structured physical activity resulted in a lower likelihood of developing poor sleep quality (PSQI >5) over the intervention period than health education but had no effect on prevalent cases of poor sleep quality or on sleep-wake behaviors evaluated using the ISI or ESS. These results suggest that the benefit of physical activity in this sample was preventive and limited to sleep-wake behaviors evaluated using the PSQI.

Comparison of breast cancer screening rates between Korean women in America versus Korea.

BACKGROUND: Breast cancer is the most commonly diagnosed cancer among Korean American women and among Korean women in Korea. However, many Asian women are not aware of the importance of screening mammography. This study compares breast cancer screening in Korean women in California and in Korea. METHODS: The 2005 California Health Interview Survey (CHIS) and 2005 Korean National Cancer Screening Survey (KNCSS) were used to examine breast cancer screening and related factors. The sample included 252 Korean American women in California and 889 Korean women aged > or = 40 years. Logistic regression was used to predict the odds of having had a screening mammography in the past 2 years. For CHIS data, SUDAAN software (Research Triangle Institute, Research Triangle Park, NC) was used to take into account the design of the complex and multistage samples. RESULTS: Korean women reported a lower rate than Korean American women of ever having had a screening mammography (50.8% vs. 77.8%). Korean women also had lower rates of having had a screening mammography in the past 2 years than Korean American women (39.5% vs. 57.2%). Korean Americans were less likely to undergo screening mammography if they were older. Korean women were less likely to undergo screening mammography if they were older, unemployed, lacked private cancer insurance, lacked health checkups, did not exercise, had poor health status, or currently smoked. CONCLUSIONS: We found significant differences in cancer screening behavior in Korean women depending on where they lived. These differences may indicate an important influence of social factors on preventive health behavior.

HPA axis response to stress predicts short-term snack intake in obese women.

Prior research has linked heightened cortisol reactivity to stress with increased food consumption. This pilot study tested corollaries of the hypothesis that cortisol stress reactivity promotes obesity. Thirty-four lean and obese women completed an acute stress task and a non-stressful control task in counterbalanced order. Contrary to expectations, higher post-stress cortisol was associated with decreased post-stress snack intake in obese women but was unrelated to snack intake in lean women. Stress also blunted an expected rise in hunger only among obese women. Findings suggest that some obese women may be more sensitive to short-term anorectic effects of HPA axis activation.

Predicting the stages of adoption of cervical cancer screening among Korean women.

OBJECTIVE: We examined the Transtheoretical Model (TTM) of behavioral change based on cervical cancer screening among Korean women. We assessed the stages of Pap smear adoption, positive and negative attitudes toward Pap smear, and contributory factors. METHOD: This study was based on the 2007 Korean National Cancer Screening Survey, which included a total of 1224 women > or =30 years. Multiple logistic analyses were conducted to determine the factors that influenced the women's stages of adoption. RESULTS: Fifty-seven percent of the women reported on-schedule screening while 43% did not. Those subjects who were 50-64 years old, had higher incomes, married, lived in a rural area, exercised regularly, and had regular checkups were more likely to be in maintenance than in precontemplation. However, the con scores were negatively associated with a transition from precontemplation to maintenance. Further, a high con score was associated with being in relapse risk or relapse rather than in maintenance while the pro scores were not significant across the stages of adoption. CONCLUSION: Our results confirm the applicability of the TTM to Pap smear screening in an Asian context. Age, income, marital status, residency, regular exercise, health checkups, and con scores were significantly associated with the stage of adoption.

Carbohydrate craving: a double-blind, placebo-controlled test of the self-medication hypothesis.

Carbohydrate craving, the overwhelming desire to consume carbohydrate-rich foods in an attempt to improve mood, remains a scientifically controversial construct. We tested whether carbohydrate preference and mood enhancement could be demonstrated in a double-blind, placebo-controlled self-administration trial. Overweight females who met strict operational criteria for carbohydrate craving participated in two 3-day discrete choice trials over a 2-week period. Participants reported their mood before and at several time points after undergoing a dysphoric mood induction and ingesting, either a carbohydrate beverage or a taste and calorie-matched protein-rich balanced nutrient beverage. Every third testing day, participants were asked to self-administer the beverage they preferred based on its previous mood effect. Results showed that, when rendered mildly dysphoric, carbohydrate cravers chose the carbohydrate beverage significantly more often than the protein-rich beverage and reported that carbohydrate produced greater mood improvement. The carbohydrate beverage was perceived as being more palatable by the carbohydrate cravers, although not by independent taste testers who performed the pre-trial taste matching. This study, performed under rigorous study conditions, supports the existence of a carbohydrate craving syndrome in which carbohydrate self-administration improves mildly dysphoric mood.

Ethnic similarities and differences in reasons for smoking.

We examined reasons for smoking among Black and White smokers. We hypothesized that Blacks would be more likely than Whites to cite tension reduction and less likely to cite weight control as a reason for smoking. Black (n = 100) and White (n = 100) female smokers completed the Reasons for Smoking Scale and the Smoking Situations Questionnaire. Regardless of treatment status, Blacks and Whites differed in their reasons for smoking [F(6,191) = 2.24; p < .05]. Blacks, as compared to Whites, less strongly endorsed weight concern as a reason for smoking [F(1,198) = 6.10; p < .05]. Results failed to reveal differences in tension reduction, stimulation, handling, relaxation, addiction, and habit as a function of ethnicity. Results suggest that Blacks and Whites exhibit more similarity than difference in their reasons for smoking.

Weight concerns and smoking in Black and White female smokers.

We examined whether the weight concerns of Blacks and Whites who enroll in smoking-cessation treatment differed from women who declined treatment. Black (n=100) and White (n=100) female smokers completed four measures of weight concern. Whites reported more general weight concern and smoking-specific weight concern than Blacks did. Treatment enrollers reported more general and smoking-specific weight concerns than decliners did. After controlling for BMI, SES, and number of cigarettes, ethnicity accounted for significant variance in general and smoking-specific weight concerns. Overall, Blacks reported less weight concerns than the Whites did, but when Blacks enrolled in treatment, these differences were less apparent.

Eating pathology among Black and White smokers.

Among White smokers, many females use smoking as a weight control strategy. Little is known about the relationship between eating pathology and smoking among Black females, and whether smokers who enroll in treatment differ in eating pathology from smokers who decline treatment. We examined eating pathology among Black and White smokers who enrolled in a smoking cessation treatment and those who declined treatment. Participants were 100 Black and 100 White female smokers (ages 18-65) who completed three measures of eating pathology. After controlling for BMI, Whites reported greater levels of overall eating pathology than Blacks [F(1,195)=4.1; p<0.05]. Treatment enrollers reported greater levels of disinhibited eating [F(1,197)=15.0; p<0.001) and hunger awareness [F(1,197)=5.6, p<0.05] than treatment decliners, but did not differ in bulimic symptoms. Status of treatment enrollment accounted for the most variance in disinhibited eating (7%; p<.001) and hunger awareness (2.9%; p<0.05). Results indicate that overall pathological eating behaviors are greater among White than Black smokers. However, once females seek smoking cessation treatment, these ethnic differences are not apparent.