Using Adaptive Bandit Experiments to Increase and Investigate Engagement in Mental Health.

Digital mental health (DMH) interventions, such as text-message-based lessons and activities, offer immense potential for accessible mental health support. While these interventions can be effective, real-world experimental testing can further enhance their design and impact. Adaptive experimentation, utilizing algorithms like Thompson Sampling for (contextual) multi-armed bandit (MAB) problems, can lead to continuous improvement and personalization. However, it remains unclear when these algorithms can simultaneously increase user experience rewards and facilitate appropriate data collection for social-behavioral scientists to analyze with sufficient statistical confidence. Although a growing body of research addresses the practical and statistical aspects of MAB and other adaptive algorithms, further exploration is needed to assess their impact across diverse real-world contexts. This paper presents a software system developed over two years that allows text-messaging intervention components to be adapted using bandit and other algorithms while collecting data for side-by-side comparison with traditional uniform random non-adaptive experiments. We evaluate the system by deploying a text-message-based DMH intervention to 1100 users, recruited through a large mental health non-profit organization, and share the path forward for deploying this system at scale. This system not only enables applications in mental health but could also serve as a model testbed for adaptive experimentation algorithms in other domains.

Incorporating participants' welfare into sequential multiple assignment randomized trials.

Dynamic treatment regimes (DTRs) are sequences of decision rules that recommend treatments based on patients' time-varying clinical conditions. The sequential, multiple assignment, randomized trial (SMART) is an experimental design that can provide high-quality evidence for constructing optimal DTRs. In a conventional SMART, participants are randomized to available treatments at multiple stages with balanced randomization probabilities. Despite its relative simplicity of implementation and desirable performance in comparing embedded DTRs, the conventional SMART faces inevitable ethical issues, including assigning many participants to the empirically inferior treatment or the treatment they dislike, which might slow down the recruitment procedure and lead to higher attrition rates, ultimately leading to poor internal and external validities of the trial results. In this context, we propose a SMART under the Experiment-as-Market framework (SMART-EXAM), a novel SMART design that holds the potential to improve participants' welfare by incorporating their preferences and predicted treatment effects into the randomization procedure. We describe the steps of conducting a SMART-EXAM and evaluate its performance compared to the conventional SMART. The results indicate that the SMART-EXAM can improve the welfare of the participants enrolled in the trial, while also achieving a desirable ability to construct an optimal DTR when the experimental parameters are suitably specified. We finally illustrate the practical potential of the SMART-EXAM design using data from a SMART for children with attention-deficit/hyperactivity disorder.

Microrandomized Trials: Developing Just-in-Time Adaptive Interventions for Better Public Health.

Just-in-time adaptive interventions (JITAIs) represent an intervention design that adapts the provision and type of support over time to an individual's changing status and contexts, intending to deliver the right support on the right occasion. As a novel strategy for delivering mobile health interventions, JITAIs have the potential to improve access to quality care in underserved communities and, thus, alleviate health disparities, a significant public health concern. Valid experimental designs and analysis methods are required to inform the development of JITAIs. Here, we briefly review the cutting-edge design of microrandomized trials (MRTs), covering both the classical MRT design and its outcome-adaptive counterpart. Associated statistical challenges related to the design and analysis of MRTs are also discussed. Two case studies are provided to illustrate the aforementioned concepts and designs throughout the article. We hope our work leads to better design and application of JITAIs, advancing public health research and practice. (Am J Public Health. 2023;113(1):60-69. https://doi.org/10.2105/AJPH.2022.307150).

Daily Motivational Text Messages to Promote Physical Activity in University Students: Results From a Microrandomized Trial.

BACKGROUND: Low physical activity is an important risk factor for common physical and mental disorders. Physical activity interventions delivered via smartphones can help users maintain and increase physical activity, but outcomes have been mixed. PURPOSE: Here we assessed the effects of sending daily motivational and feedback text messages in a microrandomized clinical trial on changes in physical activity from one day to the next in a student population. METHODS: We included 93 participants who used a physical activity app, "DIAMANTE" for a period of 6 weeks. Every day, their phone pedometer passively tracked participants' steps. They were microrandomized to receive different types of motivational messages, based on a cognitive-behavioral framework, and feedback on their steps. We used generalized estimation equation models to test the effectiveness of feedback and motivational messages on changes in steps from one day to the next. RESULTS: Sending any versus no text message initially resulted in an increase in daily steps (729 steps, p = .012), but this effect decreased over time. A multivariate analysis evaluating each text message category separately showed that the initial positive effect was driven by the motivational messages though the effect was small and trend-wise significant (717 steps; p = .083), but not the feedback messages (-276 steps, p = .4). CONCLUSION: Sending motivational physical activity text messages based on a cognitive-behavioral framework may have a positive effect on increasing steps, but this decreases with time. Further work is needed to examine using personalization and contextualization to improve the efficacy of text-messaging interventions on physical activity outcomes. CLINICALTRIALS.GOV IDENTIFIER: NCT04440553.

Adaptive learning algorithms to optimize mobile applications for behavioral health: guidelines for design decisions.

OBJECTIVE: Providing behavioral health interventions via smartphones allows these interventions to be adapted to the changing behavior, preferences, and needs of individuals. This can be achieved through reinforcement learning (RL), a sub-area of machine learning. However, many challenges could affect the effectiveness of these algorithms in the real world. We provide guidelines for decision-making. MATERIALS AND METHODS: Using thematic analysis, we describe challenges, considerations, and solutions for algorithm design decisions in a collaboration between health services researchers, clinicians, and data scientists. We use the design process of an RL algorithm for a mobile health study "DIAMANTE" for increasing physical activity in underserved patients with diabetes and depression. Over the 1.5-year project, we kept track of the research process using collaborative cloud Google Documents, Whatsapp messenger, and video teleconferencing. We discussed, categorized, and coded critical challenges. We grouped challenges to create thematic topic process domains. RESULTS: Nine challenges emerged, which we divided into 3 major themes: 1. Choosing the model for decision-making, including appropriate contextual and reward variables; 2. Data handling/collection, such as how to deal with missing or incorrect data in real-time; 3. Weighing the algorithm performance vs effectiveness/implementation in real-world settings. CONCLUSION: The creation of effective behavioral health interventions does not depend only on final algorithm performance. Many decisions in the real world are necessary to formulate the design of problem parameters to which an algorithm is applied. Researchers must document and evaulate these considerations and decisions before and during the intervention period, to increase transparency, accountability, and reproducibility. TRIAL REGISTRATION: clinicaltrials.gov, NCT03490253.

Clinician-reported symptomatic adverse events in cancer trials: are they concordant with patient-reported outcomes?

AIM: We investigate the concordance, in terms of favoring the same treatment arm, between clinician-reported symptomatic adverse events (AEs) and information obtained via patient-reported outcomes (PRO) measures in cancer randomized controlled trials (RCTs). METHODS: We conducted a systematic literature search to identify all RCTs conducted in breast, colorectal, lung and prostate cancer, published between 2004 and 2017. RESULTS: We identified 207 RCTs. In the majority of RCTs (n=133, 64.2%) a discordance between PROs and AEs was found. In 104 studies (50.2%), PRO data favored the experimental arm when AEs did not, while the opposite situation was found in 29 trials (14.0%). CONCLUSION: Frequently, information obtained via PRO measures and clinician-reported AEs do not favor the same treatment arm in RCT settings.

Modeling strategies to improve parameter estimates in prognostic factors analyses with patient-reported outcomes in oncology.

PURPOSE: The inclusion of patient-reported outcome (PRO) questionnaires in prognostic factor analyses in oncology has substantially increased in recent years. We performed a simulation study to compare the performances of four different modeling strategies in estimating the prognostic impact of multiple collinear scales from PRO questionnaires. METHODS: We generated multiple scenarios describing survival data with different sample sizes, event rates and degrees of multicollinearity among five PRO scales. We used the Cox proportional hazards (PH) model to estimate the hazard ratios (HR) using automatic selection procedures, which were based on either the likelihood ratio-test (Cox-PV) or the Akaike Information Criterion (Cox-AIC). We also used Cox PH models which included all variables and were either penalized using the Ridge regression (Cox-R) or were estimated as usual (Cox-Full). For each scenario, we simulated 1000 independent datasets and compared the average outcomes of all methods. RESULTS: The Cox-R showed similar or better performances with respect to the other methods, particularly in scenarios with medium-high multicollinearity (ρ = 0.4 to ρ = 0.8) and small sample sizes (n = 100). Overall, the Cox-PV and Cox-AIC performed worse, for example they did not select one or more prognostic collinear PRO scales in some scenarios. Compared with the Cox-Full, the Cox-R provided HR estimates with similar bias patterns but smaller root-mean-squared errors, particularly in higher multicollinearity scenarios. CONCLUSIONS: Our findings suggest that the Cox-R is the best approach when performing prognostic factor analyses with multiple and collinear PRO scales, particularly in situations of high multicollinearity, small sample sizes and low event rates.

Distribution- and anchor-based methods to determine the minimally important difference on patient-reported outcome questionnaires in oncology: a structured review.

BACKGROUND: Interpretation of differences or changes in patient-reported outcome scores should not only consider statistical significance, but also clinical relevance. Accordingly, accurate determination of the minimally important difference (MID) is crucial to assess the effectiveness of health care interventions, as well as for sample size calculation. Several methods have been proposed to determine the MID. Our aim was to review the statistical methods used to determine MID in patient-reported outcome (PRO) questionnaires in cancer patients, focusing on the distribution- and anchor-based approaches and to present the variability of criteria used as well as possible limitations. METHODS: We performed a systematic search using PubMed. We searched for all cancer studies related to MID determination on a PRO questionnaire. Two reviewers independently screened titles and abstracts to identify relevant articles. Data were extracted from eligible articles using a predefined data collection form. Discrepancies were resolved by discussion and the involvement of a third reviewer. RESULTS: Sixty-three articles were identified, of which 46 were retained for final analysis. Both distribution- and anchor-based approaches were used to assess the MID in 37 studies (80.4%). Different time points were used to apply the distribution-based method and the most frequently reported distribution was the 0.5 standard deviation at baseline. A change in a PRO external scale (N = 13, 30.2%) and performance status (N = 15, 34.9%) were the most frequently used anchors. The stability of the MID over time was rarely investigated and only 28.2% of studies used at least 3 assessment timepoints. The robustness of anchor-based MID was questionable in 37.2% of the studies where the minimal number of patients by anchor category was less than 20. CONCLUSION: Efforts are needed to improve the quality of the methodology used for MID determination in PRO questionnaires used in oncology. In particular, increased attention to the sample size should be paid to guarantee reliable results. This could increase the use of these specific thresholds in future studies.

Evaluating methodological quality of Prognostic models Including Patient-reported HeAlth outcomes iN oncologY (EPIPHANY): a systematic review protocol.

INTRODUCTION: While there is mounting evidence of the independent prognostic value of patient-reported outcomes (PROs) for overall survival (OS) in patients with cancer, it is known that the conduct of these studies may hold a number of methodological challenges. The aim of this systematic review is to evaluate the quality of published studies in this research area, in order to identify methodological and statistical issues deserving special attention and to also possibly provide evidence-based recommendations. METHODS AND ANALYSIS: An electronic search strategy will be performed in PubMed to identify studies developing or validating a prognostic model which includes PROs as predictors. Two reviewers will independently be involved in data collection using a predefined and standardised data extraction form including information related to study characteristics, PROs measures used and multivariable prognostic models. Studies selection will be reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, with data extraction form using fields from the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist for multivariable models. Methodological quality assessment will also be performed and will be based on prespecified domains of the CHARMS checklist. As a substantial heterogeneity of included studies is expected, a narrative evidence synthesis will also be provided. ETHICS AND DISSEMINATION: Given that this systematic review will use only published data, ethical permissions will not be required. Findings from this review will be published in peer-reviewed scientific journals and presented at major international conferences. We anticipate that this review will contribute to identify key areas of improvement for conducting and reporting prognostic factor analyses with PROs in oncology and will lay the groundwork for developing future evidence-based recommendations in this area of research. PROSPERO REGISTRATION NUMBER: CRD42018099160.