The Digital Therapeutics Real-World Evidence Framework: An Approach for Guiding Evidence-Based Digital Therapeutics Design, Development, Testing, and Monitoring.

Digital therapeutics (DTx) are a promising way to provide safe, effective, accessible, sustainable, scalable, and equitable approaches to advance individual and population health. However, developing and deploying DTx is inherently complex in that DTx includes multiple interacting components, such as tools to support activities like medication adherence, health behavior goal-setting or self-monitoring, and algorithms that adapt the provision of these according to individual needs that may change over time. While myriad frameworks exist for different phases of DTx development, no single framework exists to guide evidence production for DTx across its full life cycle, from initial DTx development to long-term use. To fill this gap, we propose the DTx real-world evidence (RWE) framework as a pragmatic, iterative, milestone-driven approach for developing DTx. The DTx RWE framework is derived from the 4-phase development model used for behavioral interventions, but it includes key adaptations that are specific to the unique characteristics of DTx. To ensure the highest level of fidelity to the needs of users, the framework also incorporates real-world data (RWD) across the entire life cycle of DTx development and use. The DTx RWE framework is intended for any group interested in developing and deploying DTx in real-world contexts, including those in industry, health care, public health, and academia. Moreover, entities that fund research that supports the development of DTx and agencies that regulate DTx might find the DTx RWE framework useful as they endeavor to improve how DTxcan advance individual and population health.

Virtual AppLication-supported Environment To INcrease Exercise (VALENTINE) during cardiac rehabilitation study: Rationale and design.

BACKGROUND: In-person, exercise-based cardiac rehabilitation improves physical activity and reduces morbidity and mortality for patients with cardiovascular disease. However, activity levels may not be optimized and decline over time after patients graduate from cardiac rehabilitation. Scalable interventions through mobile health (mHealth) technologies have the potential to augment activity levels and extend the benefits of cardiac rehabilitation. METHODS: The VALENTINE Study is a prospective, randomized-controlled, remotely-administered trial designed to evaluate an mHealth intervention to supplement cardiac rehabilitation for low- and moderate-risk patients (ClinicalTrials.gov NCT04587882). Participants are randomized to the control or intervention arms of the study. Both groups receive a compatible smartwatch (Fitbit Versa 2 or Apple Watch 4) and usual care. Participants in the intervention arm of the study additionally receive a just-in-time adaptive intervention (JITAI) delivered as contextually tailored notifications promoting low-level physical activity and exercise throughout the day. In addition, they have access to activity tracking and goal setting through the mobile study application and receive weekly activity summaries via email. The primary outcome is change in 6-minute walk distance at 6-months and, secondarily, change in average daily step count. Exploratory analyses will examine the impact of notifications on immediate short-term smartwatch-measured step counts and exercise minutes. CONCLUSIONS: The VALENTINE study leverages innovative techniques in behavioral and cardiovascular disease research and will make a significant contribution to our understanding of how to support patients using mHealth technologies to promote and sustain physical activity.

Clinic, Home, and Kiosk Blood Pressure Measurements for Diagnosing Hypertension: a Randomized Diagnostic Study.

BACKGROUND: The US Preventive Services Task Force recommends blood pressure (BP) measurements using 24-h ambulatory monitoring (ABPM) or home BP monitoring before making a new hypertension diagnosis. OBJECTIVE: Compare clinic-, home-, and kiosk-based BP measurement to ABPM for diagnosing hypertension. DESIGN, SETTING, AND PARTICIPANTS: Diagnostic study in 12 Washington State primary care centers, with participants aged 18-85 years without diagnosed hypertension or prescribed antihypertensive medications, with elevated BP in clinic. INTERVENTIONS: Randomization into one of three diagnostic regimens: (1) clinic (usual care follow-up BPs); (2) home (duplicate BPs twice daily for 5 days); or (3) kiosk (triplicate BPs on 3 days). All participants completed ABPM at 3 weeks. MAIN MEASURES: Primary outcome was difference between ABPM daytime and clinic, home, and kiosk mean systolic BP. Differences in diastolic BP, sensitivity, and specificity were secondary outcomes. KEY RESULTS: Five hundred ten participants (mean age 58.7 years, 80.2% white) with 434 (85.1%) included in primary analyses. Compared to daytime ABPM, adjusted mean differences in systolic BP were clinic (-4.7mmHg [95% confidence interval -7.3, -2.2]; P<.001); home (-0.1mmHg [-1.6, 1.5];P=.92); and kiosk (9.5mmHg [7.5, 11.6];P<.001). Differences for diastolic BP were clinic (-7.2mmHg [-8.8, -5.5]; P<.001); home (-0.4mmHg [-1.4, 0.7];P=.52); and kiosk (5.0mmHg [3.8, 6.2]; P<.001). Sensitivities for clinic, home, and kiosk compared to ABPM were 31.1% (95% confidence interval, 22.9, 40.6), 82.2% (73.8, 88.4), and 96.0% (90.0, 98.5), and specificities 79.5% (64.0, 89.4), 53.3% (38.9, 67.2), and 28.2% (16.4, 44.1), respectively. LIMITATIONS: Single health care organization and limited race/ethnicity representation. CONCLUSIONS: Compared to ABPM, mean BP was significantly lower for clinic, significantly higher for kiosk, and without significant differences for home. Clinic BP measurements had low sensitivity for detecting hypertension. Findings support utility of home BP monitoring for making a new diagnosis of hypertension. TRIAL REGISTRATION: ClinicalTrials.gov NCT03130257 https://clinicaltrials.gov/ct2/show/NCT03130257.

BATCH POLICY LEARNING IN AVERAGE REWARD MARKOV DECISION PROCESSES.

We consider the batch (off-line) policy learning problem in the infinite horizon Markov Decision Process. Motivated by mobile health applications, we focus on learning a policy that maximizes the long-term average reward. We propose a doubly robust estimator for the average reward and show that it achieves semiparametric efficiency. Further we develop an optimization algorithm to compute the optimal policy in a parameterized stochastic policy class. The performance of the estimated policy is measured by the difference between the optimal average reward in the policy class and the average reward of the estimated policy and we establish a finite-sample regret guarantee. The performance of the method is illustrated by simulation studies and an analysis of a mobile health study promoting physical activity.

Use of a mobile app to capture supplemental health information during pregnancy: Implications for clinical research.

PURPOSE: Mobile applications ("apps") may be efficient tools for improving the quality of clinical research among pregnant women, but evidence is sparse. We assess the feasibility and generalizability of a mobile app for capturing supplemental data during pregnancy. METHODS: In 2017, we conducted a pilot study of the FDA MyStudies mobile app within a pregnant population identified through Kaiser Permanente Washington (KPWA), an integrated healthcare delivery system. We ascertained health conditions, medications, and substance use through app-based questionnaires. In a post-hoc analysis, we utilized electronic health records (EHR) to summarize sociodemographic and health characteristics of pilot participants and, for comparison, a pregnant population identified using similar methods. RESULTS: Six percent (64/1070) of contacted women enrolled in the pilot study. Nearly half (23/53) reported taking medication for headaches and one-fourth for constipation (13/53) and nausea (12/53) each. Few instances (2/92) of over-the-counter medication use were identified in electronic dispensing records. One-quarter to one-third of participants with depression and anxiety/panic, respectively, reported recently discontinuing medications for these conditions. Eighty-eight percent of pilot participants reported White race (95%CI: 81-95%), versus 67% of the comparison population (N = 2065). More pilot participants filled ≥1 prescription for antianxiety medication (22% [95%CI: 13-35%]) and antidepressants (19% [95%CI 10-31%]) pre-pregnancy than the comparison population (10 and 9%, respectively). CONCLUSIONS: Mobile apps may be a feasible tool for capturing health data not routinely available in EHR. Pregnant women willing to use a mobile app for research may differ from the general pregnant population, but confirmation is needed.

Toward a Just-in-Time Adaptive Intervention to Reduce Emerging Adult Alcohol Use: Testing Approaches for Identifying When to Intervene.

ABBREVIATIONS: JITAI: Just-in-time adaptive intervention; ROC: receiver operating characteristic; AUC: area under the curve; MRT: micro-randomized trial.

Linear mixed models with endogenous covariates: modeling sequential treatment effects with application to a mobile health study.

Mobile health is a rapidly developing field in which behavioral treatments are delivered to individuals via wearables or smartphones to facilitate health-related behavior change. Micro-randomized trials (MRT) are an experimental design for developing mobile health interventions. In an MRT the treatments are randomized numerous times for each individual over course of the trial. Along with assessing treatment effects, behavioral scientists aim to understand between-person heterogeneity in the treatment effect. A natural approach is the familiar linear mixed model. However, directly applying linear mixed models is problematic because potential moderators of the treatment effect are frequently endogenous-that is, may depend on prior treatment. We discuss model interpretation and biases that arise in the absence of additional assumptions when endogenous covariates are included in a linear mixed model. In particular, when there are endogenous covariates, the coefficients no longer have the customary marginal interpretation. However, these coefficients still have a conditional-on-the-random-effect interpretation. We provide an additional assumption that, if true, allows scientists to use standard software to fit linear mixed model with endogenous covariates, and person-specific predictions of effects can be provided. As an illustration, we assess the effect of activity suggestion in the HeartSteps MRT and analyze the between-person treatment effect heterogeneity.

Why we need a small data paradigm.

BACKGROUND: There is great interest in and excitement about the concept of personalized or precision medicine and, in particular, advancing this vision via various 'big data' efforts. While these methods are necessary, they are insufficient to achieve the full personalized medicine promise. A rigorous, complementary 'small data' paradigm that can function both autonomously from and in collaboration with big data is also needed. By 'small data' we build on Estrin's formulation and refer to the rigorous use of data by and for a specific N-of-1 unit (i.e., a single person, clinic, hospital, healthcare system, community, city, etc.) to facilitate improved individual-level description, prediction and, ultimately, control for that specific unit. MAIN BODY: The purpose of this piece is to articulate why a small data paradigm is needed and is valuable in itself, and to provide initial directions for future work that can advance study designs and data analytic techniques for a small data approach to precision health. Scientifically, the central value of a small data approach is that it can uniquely manage complex, dynamic, multi-causal, idiosyncratically manifesting phenomena, such as chronic diseases, in comparison to big data. Beyond this, a small data approach better aligns the goals of science and practice, which can result in more rapid agile learning with less data. There is also, feasibly, a unique pathway towards transportable knowledge from a small data approach, which is complementary to a big data approach. Future work should (1) further refine appropriate methods for a small data approach; (2) advance strategies for better integrating a small data approach into real-world practices; and (3) advance ways of actively integrating the strengths and limitations from both small and big data approaches into a unified scientific knowledge base that is linked via a robust science of causality. CONCLUSION: Small data is valuable in its own right. That said, small and big data paradigms can and should be combined via a foundational science of causality. With these approaches combined, the vision of precision health can be achieved.

Efficacy of Contextually Tailored Suggestions for Physical Activity: A Micro-randomized Optimization Trial of HeartSteps.

BACKGROUND: HeartSteps is an mHealth intervention that encourages regular walking via activity suggestions tailored to the individuals' current context. PURPOSE: We conducted a micro-randomized trial (MRT) to evaluate the efficacy of HeartSteps' activity suggestions to optimize the intervention. METHODS: We conducted a 6-week MRT with 44 adults. Contextually tailored suggestions could be delivered up to five times per day at user-selected times. At each of these five times, for each participant on each day of the study, HeartSteps randomized whether to provide an activity suggestion, and, if so, whether to provide a walking or an antisedentary suggestion. We used a centered and weighted least squares method to analyze the effect of suggestions on the 30-min step count following suggestion randomization. RESULTS: Averaging over study days and types of activity suggestions, delivering a suggestion versus no suggestion increased the 30-min step count by 14% (p = .06), 35 additional steps over the 253-step average. The effect was not evenly distributed in time. Providing any type of suggestion versus no suggestion initially increased the step count by 66% (167 steps; p < .01), but this effect diminished over time. Averaging over study days, delivering a walking suggestion versus no suggestion increased the average step count by 24% (59 steps; p = .02). This increase was greater at the start of study (107% or 271 additional steps; p < .01), but decreased over time. Antisedentary suggestions had no detectable effect on the 30-min step count. CONCLUSION: Contextually tailored walking suggestions are a promising way of initiating bouts of walking throughout the day. CLINICAL TRIAL INFORMATION: This study was registered on ClinicalTrials.gov number NCT03225521.

Standardized Effect Sizes for Preventive Mobile Health Interventions in Micro-randomized Trials.

Mobile Health (mHealth) interventions are behavioral interventions that are accessible to individuals in their daily lives via a mobile device. Most mHealth interventions consist of multiple intervention components. Some of the components are "pull" components, which require individuals to access the component on their mobile device at moments when they decide they need help. Other intervention components are "push" components, which are initiated by the intervention, not the individual, and are delivered via notifications or text messages. Micro-randomized trials (MRTs) have been developed to provide data to assess the effects of push intervention components on subsequent emotions and behavior. In this paper, we review the micro-randomized trial design and provide an approach to computing a standardized effect size for these intervention components. This effect size can be used to compare different push intervention components that may be included in an mHealth intervention. In addition, a standardized effect size can be used to inform sample size calculations for future MRTs. Here, the standardized effect size is a function of time because the push notifications can occur repeatedly over time. We illustrate this methodology using data from an MRT involving HeartSteps, an mHealth intervention for physical activity as part of the secondary prevention of heart disease.

Modeling individual differences: A case study of the application of system identification for personalizing a physical activity intervention.

BACKGROUND: Control systems engineering methods, particularly, system identification (system ID), offer an idiographic (i.e., person-specific) approach to develop dynamic models of physical activity (PA) that can be used to personalize interventions in a systematic, scalable way. The purpose of this work is to: (1) apply system ID to develop individual dynamical models of PA (steps/day measured using Fitbit Zip) in the context of a goal setting and positive reinforcement intervention informed by Social Cognitive Theory; and (2) compare insights on potential tailoring variables (i.e., predictors expected to influence steps and thus moderate the suggested step goal and points for goal achievement) selected using the idiographic models to those selected via a nomothetic (i.e., aggregated across individuals) approach. METHOD: A personalized goal setting and positive reinforcement intervention was deployed for 14 weeks. Baseline PA measured in weeks 1-2 was used to inform personalized daily step goals delivered in weeks 3-14. Goals and expected reward points (granted upon goal achievement) were pseudo-randomly assigned using techniques from system ID, with goals ranging from their baseline median steps/day up to 2.5× baseline median steps/day, and points ranging from 100 to 500 (i.e., $0.20-$1.00). Participants completed a series of daily self-report measures. Auto Regressive with eXogenous Input (ARX) modeling and multilevel modeling (MLM) were used as the idiographic and nomothetic approaches, respectively. RESULTS: Participants (N = 20, mean age = 47.25 ±â€¯6.16 years, 90% female) were insufficiently active, overweight (mean BMI = 33.79 ±â€¯6.82 kg/m2) adults. Results from ARX modeling suggest that individuals differ in the factors (e.g., perceived stress, weekday/weekend) that influence their observed steps/day. In contrast, the nomothetic model from MLM suggested that goals and weekday/weekend were the key variables that were predictive of steps. Assuming the ARX models are more personalized, the obtained nomothetic model would have led to the identification of the same predictors for 5 of the 20 participants, suggesting a mismatch of plausible tailoring variables to use for 75% of the sample. CONCLUSION: The idiographic approach revealed person-specific predictors beyond traditional MLM analyses and unpacked the inherent complexity of PA; namely that people are different and context matters. System ID provides a feasible approach to develop personalized dynamical models of PA and inform person-specific tailoring variable selection for use in adaptive behavioral interventions.

Tutorial for Using Control Systems Engineering to Optimize Adaptive Mobile Health Interventions.

BACKGROUND: Adaptive behavioral interventions are individualized interventions that vary support based on a person's evolving needs. Digital technologies enable these adaptive interventions to function at scale. Adaptive interventions show great promise for producing better results compared with static interventions related to health outcomes. Our central thesis is that adaptive interventions are more likely to succeed at helping individuals meet and maintain behavioral targets if its elements can be iteratively improved via data-driven testing (ie, optimization). Control systems engineering is a discipline focused on decision making in systems that change over time and has a wealth of methods that could be useful for optimizing adaptive interventions. OBJECTIVE: The purpose of this paper was to provide an introductory tutorial on when and what to do when using control systems engineering for designing and optimizing adaptive mobile health (mHealth) behavioral interventions. OVERVIEW: We start with a review of the need for optimization, building on the multiphase optimization strategy (MOST). We then provide an overview of control systems engineering, followed by attributes of problems that are well matched to control engineering. Key steps in the development and optimization of an adaptive intervention from a control engineering perspective are then summarized, with a focus on why, what, and when to do subtasks in each step. IMPLICATIONS: Control engineering offers exciting opportunities for optimizing individualization and adaptation elements of adaptive interventions. Arguably, the time is now for control systems engineers and behavioral and health scientists to partner to advance interventions that can be individualized, adaptive, and scalable. This tutorial should aid in creating the bridge between these communities.

Sample size calculations for micro-randomized trials in mHealth.

The use and development of mobile interventions are experiencing rapid growth. In "just-in-time" mobile interventions, treatments are provided via a mobile device, and they are intended to help an individual make healthy decisions 'in the moment,' and thus have a proximal, near future impact. Currently, the development of mobile interventions is proceeding at a much faster pace than that of associated data science methods. A first step toward developing data-based methods is to provide an experimental design for testing the proximal effects of these just-in-time treatments. In this paper, we propose a 'micro-randomized' trial design for this purpose. In a micro-randomized trial, treatments are sequentially randomized throughout the conduct of the study, with the result that each participant may be randomized at the 100s or 1000s of occasions at which a treatment might be provided. Further, we develop a test statistic for assessing the proximal effect of a treatment as well as an associated sample size calculator. We conduct simulation evaluations of the sample size calculator in various settings. Rules of thumb that might be used in designing a micro-randomized trial are discussed. This work is motivated by our collaboration on the HeartSteps mobile application designed to increase physical activity. Copyright © 2015 John Wiley & Sons, Ltd.

Effect-Invariant Mechanisms for Policy Generalization.

Policy learning is an important component of many real-world learning systems. A major challenge in policy learning is how to adapt efficiently to unseen environments or tasks. Recently, it has been suggested to exploit invariant conditional distributions to learn models that generalize better to unseen environments. However, assuming invariance of entire conditional distributions (which we call full invariance) may be too strong of an assumption in practice. In this paper, we introduce a relaxation of full invariance called effect-invariance (e-invariance for short) and prove that it is sufficient, under suitable assumptions, for zero-shot policy generalization. We also discuss an extension that exploits e-invariance when we have a small sample from the test environment, enabling few-shot policy generalization. Our work does not assume an underlying causal graph or that the data are generated by a structural causal model; instead, we develop testing procedures to test e-invariance directly from data. We present empirical results using simulated data and a mobile health intervention dataset to demonstrate the effectiveness of our approach.

Getting cozy with causality: Advances to the causal pathway diagramming method to enhance implementation precision.

Background: Implementation strategies are theorized to work well when carefully matched to implementation determinants and when factors-preconditions, moderators, etc.-that influence strategy effectiveness are prospectively identified and addressed. Existing methods for strategy selection are either imprecise or require significant technical expertise and resources, undermining their utility. This article outlines refinements to causal pathway diagrams (CPDs), a method for articulating the causal process through which implementation strategies work and offers illustrations of their use. Method: CPDs are a visualization tool to represent an implementation strategy, its mechanism(s) (i.e., the processes through which a strategy is thought to operate), determinants it is intended to address, factors that may impede or facilitate its effectiveness, and the series of outcomes that should be expected if the strategy is operating as intended. We offer principles for constructing CPDs and describe their key functions. Results: Applications of the CPD method by study teams from two National Institute of Health-funded Implementation Science Centers and a research grant are presented. These include the use of CPDs to (a) match implementation strategies to determinants, (b) understand the conditions under which an implementation strategy works, and (c) develop causal theories of implementation strategies. Conclusions: CPDs offer a novel method for implementers to select, understand, and improve the effectiveness of implementation strategies. They make explicit theoretical assumptions about strategy operation while supporting practical planning. Early applications have led to method refinements and guidance for the field.

Advances to the Causal Pathway Diagramming Method to Enhance Implementation Precision Plain Language Summary Implementation strategies often fail to produce meaningful improvements in the outcomes we hope to impact. Better tools for choosing, designing, and evaluating implementation strategies may improve their performance. We developed a tool, causal pathway diagrams (CPD), to visualize and describe how implementation strategies are expected to work. In this article, we describe refinements to the CPD tool and accompanying approach. We use real illustrations to show how CPDs can be used to improve how to match strategies to barriers, understand the conditions in which those strategies work best, and develop generalizable theories describing how implementation strategies work. CPDs can serve as both a practical and scientific tool to improve the planning, deployment, and evaluation of implementation strategies. We demonstrate the range of ways that CPDs are being used, from a highly practical tool to improve implementation practice to a scientific approach to advance testing and theorizing about implementation strategies.

Text Messages to Promote Physical Activity in Patients With Cardiovascular Disease: A Micro-Randomized Trial of a Just-In-Time Adaptive Intervention.

BACKGROUND: Text messages may enhance physical activity levels in patients with cardiovascular disease, including those enrolled in cardiac rehabilitation. However, the independent and long-term effects of text messages remain uncertain. METHODS: The VALENTINE study (Virtual Application-supported Environment to Increase Exercise) was a micro-randomized trial that delivered text messages through a smartwatch (Apple Watch or Fitbit Versa) to participants initiating cardiac rehabilitation. Participants were randomized 4× per day over 6-months to receive no text message or a message encouraging low-level physical activity. Text messages were tailored on contextual factors (eg, weather). Our primary outcome was step count 60 minutes following a text message, and we used a centered and weighted least squares mean method to estimate causal effects. Given potential measurement differences between devices determined a priori, data were assessed separately for Apple Watch and Fitbit Versa users over 3 time periods corresponding to the initiation (0-30 days), maintenance (31-120 days), and completion (121-182 days) of cardiac rehabilitation. RESULTS: One hundred eight participants were included with 70 552 randomizations over 6 months; mean age was 59.5 (SD, 10.7) years with 36 (32.4%) female and 68 (63.0%) Apple Watch participants. For Apple Watch participants, text messages led to a trend in increased step count by 10% in the 60-minutes following a message during days 1 to 30 (95% CI, -1% to +20%), with no effect from days 31 to 120 (+1% [95% CI, -4% to +5%]), and a significant 6% increase during days 121 to 182 (95% CI, +0% to +11%). For Fitbit users, text messages significantly increased step count by 17% (95% CI, +7% to +28%) in the 60-minutes following a message in the first 30 days of the study with no effect subsequently. CONCLUSIONS: In patients undergoing cardiac rehabilitation, contextually tailored text messages may increase physical activity, but this effect varies over time and by device. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifier: NCT04587882.

A Text Messaging Intervention for Priming the Affective Rewards of Exercise in Adults: Protocol for a Microrandomized Trial.

BACKGROUND: Physical activity is a critical target for health interventions, but effective interventions remain elusive. A growing body of work suggests that interventions targeting affective attitudes toward physical activity may be more effective for sustaining activity long term than those that rely on cognitive constructs alone, such as goal setting and self-monitoring. Anticipated affective response in particular is a promising target for intervention. OBJECTIVE: We will evaluate the efficacy of an SMS text messaging intervention that manipulates anticipated affective response to exercise to promote physical activity. We hypothesize that reminding users of a positive postexercise affective state before their planned exercise sessions will increase their calories burned during this exercise session. We will deploy 2 forms of affective SMS text messages to explore the design space: low-reflection messages written by participants for themselves and high-reflection prompts that require users to reflect and respond. We will also explore the effect of the intervention on affective attitudes toward exercise. METHODS: A total of 120 individuals will be enrolled in a 9-week microrandomized trial testing affective messages that remind users about feeling good after exercise (40% probability), control reminders (30% probability), or no message (30% probability). Two types of affective SMS text messages will be deployed: one requiring a response and the other in a read-only format. Participants will write the read-only messages themselves to ensure that the messages accurately reflect the participants' anticipated postexercise affective state. Affective attitudes toward exercise and intrinsic motivation for exercise will be measured at the beginning and end of the study. The weighted and centered least squares method will be used to analyze the effect of delivering the intervention versus not on calories burned over 4 hours around the time of the planned activity, measured by the Apple Watch. Secondary analyses will include the effect of the intervention on step count and active minutes, as well as an investigation of the effects of the intervention on affective attitudes toward exercise and intrinsic motivation for exercise. Participants will be interviewed to gain qualitative insights into intervention impact and acceptability. RESULTS: Enrollment began in May 2023, with 57 participants enrolled at the end of July 2023. We anticipate enrolling 120 participants. CONCLUSIONS: This study will provide early evidence about the effect of a repeated manipulation of anticipated affective response to exercise. The use of 2 different types of messages will yield insight into optimal design strategies for improving affective attitudes toward exercise. TRIAL REGISTRATION: ClinicalTrials.gov NCT05582369; https://classic.clinicaltrials.gov/ct2/show/NCT05582369. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/46560.

Development and Validation of Multivariable Prediction Algorithms to Estimate Future Walking Behavior in Adults: Retrospective Cohort Study.

BACKGROUND: Physical inactivity is associated with numerous health risks, including cancer, cardiovascular disease, type 2 diabetes, increased health care expenditure, and preventable, premature deaths. The majority of Americans fall short of clinical guideline goals (ie, 8000-10,000 steps per day). Behavior prediction algorithms could enable efficacious interventions to promote physical activity by facilitating delivery of nudges at appropriate times. OBJECTIVE: The aim of this paper is to develop and validate algorithms that predict walking (ie, >5 min) within the next 3 hours, predicted from the participants' previous 5 weeks' steps-per-minute data. METHODS: We conducted a retrospective, closed cohort, secondary analysis of a 6-week microrandomized trial of the HeartSteps mobile health physical-activity intervention conducted in 2015. The prediction performance of 6 algorithms was evaluated, as follows: logistic regression, radial-basis function support vector machine, eXtreme Gradient Boosting (XGBoost), multilayered perceptron (MLP), decision tree, and random forest. For the MLP, 90 random layer architectures were tested for optimization. Prior 5-week hourly walking data, including missingness, were used for predictors. Whether the participant walked during the next 3 hours was used as the outcome. K-fold cross-validation (K=10) was used for the internal validation. The primary outcome measures are classification accuracy, the Mathew correlation coefficient, sensitivity, and specificity. RESULTS: The total sample size included 6 weeks of data among 44 participants. Of the 44 participants, 31 (71%) were female, 26 (59%) were White, 36 (82%) had a college degree or more, and 15 (34%) were married. The mean age was 35.9 (SD 14.7) years. Participants (n=3, 7%) who did not have enough data (number of days <10) were excluded, resulting in 41 (93%) participants. MLP with optimized layer architecture showed the best performance in accuracy (82.0%, SD 1.1), whereas XGBoost (76.3%, SD 1.5), random forest (69.5%, SD 1.0), support vector machine (69.3%, SD 1.0), and decision tree (63.6%, SD 1.5) algorithms showed lower performance than logistic regression (77.2%, SD 1.2). MLP also showed superior overall performance to all other tried algorithms in Mathew correlation coefficient (0.643, SD 0.021), sensitivity (86.1%, SD 3.0), and specificity (77.8%, SD 3.3). CONCLUSIONS: Walking behavior prediction models were developed and validated. MLP showed the highest overall performance of all attempted algorithms. A random search for optimal layer structure is a promising approach for prediction engine development. Future studies can test the real-world application of this algorithm in a "smart" intervention for promoting physical activity.

Idiographic Dynamic Modeling for Behavioral Interventions with Mixed Data Partitioning and Discrete Simultaneous Perturbation Stochastic Approximation.

This paper presents the use of discrete simultaneous perturbation stochastic approximation (DSPSA) as a routine method to efficiently determine features and parameters of idiographic (i.e. single subject) dynamic models for personalized behavioral interventions using various partitions of estimation and validation data. DSPSA is demonstrated as a valuable method to search over model features and regressor orders of AutoRegressive with eXogenous input estimated models using participant data from Just Walk (a behavioral intervention to promote physical activity in sedentary adults); results of DSPSA are compared to those of exhaustive search. In Just Walk, DSPSA efficiently and quickly estimates models of walking behavior, which can then be used to develop control systems to optimize the impacts of behavioral interventions. The use of DSPSA to evaluate models using various partitions of individual data into estimation and validation data sets also highlights data partitioning as an important feature of idiographic modeling that should be carefully considered.

Assessing the Impact of Context Inference Error and Partial Observability on RL Methods for Just-In-Time Adaptive Interventions.

Just-in-Time Adaptive Interventions (JITAIs) are a class of personalized health interventions developed within the behavioral science community. JITAIs aim to provide the right type and amount of support by iteratively selecting a sequence of intervention options from a pre-defined set of components in response to each individual's time varying state. In this work, we explore the application of reinforcement learning methods to the problem of learning intervention option selection policies. We study the effect of context inference error and partial observability on the ability to learn effective policies. Our results show that the propagation of uncertainty from context inferences is critical to improving intervention efficacy as context uncertainty increases, while policy gradient algorithms can provide remarkable robustness to partially observed behavioral state information.