Ethics in digital phenotyping: considerations regarding Alzheimer’s disease, speech and artificial intelligence

Products and services based on digital wellbeing technologies typically include mobile device apps as well as browser-based apps to a lesser extent, and can include telephony-based services, text-based chatbots and voice activated chatbots. Many of these digital products and services are simultaneously available across many channels in order to maximize availability for users. Digital wellbeing technologies offer useful methods for real time data capture of the interactions of users with the products and services. We can design what data are recorded, how and where it may be stored, and crucially, how it can be analyzed to reveal individual or collective usage patterns. Digital phenotyping is the term given to the capturing and use of user log data from health and wellbeing technologies used in apps and cloud-based services. Digital phenotyping was originally proposed to correlate a person’s mental state by using their metadata and even sensor data on their smartphone. In some cases, the data is physiological, for example pulse or movement-related, and it is collected automatically. In other cases, the data is actually metadata, for example, when a call is made and the call duration rather than the content of the call. Oftentimes, as would be expected from a personal device located on the body of the user, rich data pertaining to geo-location, social media use and interaction is gathered. Health and wellbeing-related, scientifically validated assessment scales may also generate digital phenotype data. Another form of digital phenotype data is Ecological Momentary Assessment (EMA), which originally made use of paper-diary techniques to enable people to record their observations or answers to specific questions and combined the ecological validity with the rigorous measurement techniques of psychometric research. EMA secures data about both behavioural and intrapsychic aspects of individuals' daily activities, and it obtains reports about the experience as it occurs, thereby minimizing the effects of reliance on memory and reconstruction which can often be impaired by hindsight bias or recall bias. The use of digital phenotyping data and its analysis using machine learning and artificial intelligence is important since many national public health organizations are exploring how to use digital technologies such as health apps and cloud-based services for the self-management of diseases and thus logging user interactions allows for greater insight into user needs and provides ideas for improving these digital interventions, for example through enhanced personalization. Public health services benefit since the data can be automatically and hence cost-effectively collected. Such data may facilitate new ways for digital epidemiological analyses and provide data to inform health policies. If the public health organizations promote health apps and digital phenotyping analysis using machine learning and artificial intelligence is taken up by these organizations, then there is clear need for guidelines on the ethical application of these ‘democratized’ algorithms and techniques. My keynote talk begins by reviewing the evolution of the use of technology to support peoples’ health and wellbeing, from telecare and telehealth through to personalised healthcare, the growth of the idea of ‘quantified self’ and ultimately, self-managed care. I then discuss the growing use of commercially available digital devices and software for selfcare, and the explosion in the data arising from their use in society. The opportunities for the application of machine learning to the data, including EMA data are explored and the implications are discussed, across such areas as big data for research study design, ethics, the ‘servitization’ of machine learning, bias, surveillance, and health and wellbeing services. In order to illustrate my work, I will draw upon case studies from digital health and wellbeing, including maternal mental health, crisis helplines and apps for people living with dementia.

BackgroundDigital diabetes prevention programs (dDPPs) are effective “digital prescriptions” but have high attrition rates and program noncompletion. To address this, we developed a personalized automatic messaging system (PAMS) that leverages SMS text messaging and data integration into clinical workflows to increase dDPP engagement via enhanced patient-provider communication. Preliminary data showed positive results. However, further investigation is needed to determine how to optimize the tailoring of support technology such as PAMS based on a user’s preferences to boost their dDPP engagement.ObjectiveThis study evaluates leveraging machine learning (ML) to develop digital engagement phenotypes of dDPP users and assess ML’s accuracy in predicting engagement with dDPP activities. This research will be used in a PAMS optimization process to improve PAMS personalization by incorporating engagement prediction and digital phenotyping. This study aims (1) to prove the feasibility of using dDPP user-collected data to build an ML model that predicts engagement and contributes to identifying digital engagement phenotypes, (2) to describe methods for developing ML models with dDPP data sets and present preliminary results, and (3) to present preliminary data on user profiling based on ML model outputs.MethodsUsing the gradient-boosted forest model, we predicted engagement in 4 dDPP individual activities (physical activity, lessons, social activity, and weigh-ins) and general activity (engagement in any activity) based on previous short- and long-term activity in the app. The area under the receiver operating characteristic curve, the area under the precision-recall curve, and the Brier score metrics determined the performance of the model. Shapley values reflected the feature importance of the models and determined what variables informed user profiling through latent profile analysis.ResultsWe developed 2 models using weekly and daily DPP data sets (328,821 and 704,242 records, respectively), which yielded predictive accuracies above 90%. Although both models were highly accurate, the daily model better fitted our research plan because it predicted daily changes in individual activities, which was crucial for creating the “digital phenotypes.” To better understand the variables contributing to the model predictor, we calculated the Shapley values for both models to identify the features with the highest contribution to model fit; engagement with any activity in the dDPP in the last 7 days had the most predictive power. We profiled users with latent profile analysis after 2 weeks of engagement (Bayesian information criterion=−3222.46) with the dDPP and identified 6 profiles of users, including those with high engagement, minimal engagement, and attrition.ConclusionsPreliminary results demonstrate that applying ML methods with predicting power is an acceptable mechanism to tailor and optimize messaging interventions to support patient engagement and adherence to digital prescriptions. The results enable future optimization of our existing messaging platform and expansion of this methodology to other clinical domains.Trial RegistrationClinicalTrials.gov NCT04773834; https://www.clinicaltrials.gov/ct2/show/NCT04773834International Registered Report Identifier (IRRID)RR2-10.2196/26750

Digital phenotyping is the term given to the capturing and use of user log data from health and wellbeing technologies used in apps and cloud-based services. This paper explores ethical issues in making use of digital phenotype data in the arena of digital health interventions. Products and services based on digital wellbeing technologies typically include mobile device apps as well as browser-based apps to a lesser extent, and can include telephony-based services, text-based chatbots, and voice-activated chatbots. Many of these digital products and services are simultaneously available across many channels in order to maximize availability for users. Digital wellbeing technologies offer useful methods for real-time data capture of the interactions of users with the products and services. It is possible to design what data are recorded, how and where it may be stored, and, crucially, how it can be analyzed to reveal individual or collective usage patterns. The paper also examines digital phenotyping workflows, before enumerating the ethical concerns pertaining to different types of digital phenotype data, highlighting ethical considerations for collection, storage, and use of the data. A case study of a digital health app is used to illustrate the ethical issues. The case study explores the issues from a perspective of data prospecting and subsequent machine learning. The ethical use of machine learning and artificial intelligence on digital phenotype data and the broader issues in democratizing machine learning and artificial intelligence for digital phenotype data are then explored in detail.

BackgroundThe digital health care community has been urged to enhance engagement and clinical outcomes by analyzing multidimensional digital phenotypes.ObjectiveThis study aims to use a machine learning approach to investigate the performance of multivariate phenotypes in predicting the engagement rate and health outcomes of digital cognitive behavioral therapy.MethodsWe leveraged both conventional phenotypes assessed by validated psychological questionnaires and multidimensional digital phenotypes within time-series data from a mobile app of 45 participants undergoing digital cognitive behavioral therapy for 8 weeks. We conducted a machine learning analysis to discriminate the important characteristics.ResultsA higher engagement rate was associated with higher weight loss at 8 weeks (r=−0.59; P<.001) and 24 weeks (r=−0.52; P=.001). Applying the machine learning approach, lower self-esteem on the conventional phenotype and higher in-app motivational measures on digital phenotypes commonly accounted for both engagement and health outcomes. In addition, 16 types of digital phenotypes (ie, lower intake of high-calorie food and evening snacks and higher interaction frequency with mentors) predicted engagement rates (mean R2 0.416, SD 0.006). The prediction of short-term weight change (mean R2 0.382, SD 0.015) was associated with 13 different digital phenotypes (ie, lower intake of high-calorie food and carbohydrate and higher intake of low-calorie food). Finally, 8 measures of digital phenotypes (ie, lower intake of carbohydrate and evening snacks and higher motivation) were associated with a long-term weight change (mean R2 0.590, SD 0.011).ConclusionsOur findings successfully demonstrated how multiple psychological constructs, such as emotional, cognitive, behavioral, and motivational phenotypes, elucidate the mechanisms and clinical efficacy of a digital intervention using the machine learning method. Accordingly, our study designed an interpretable digital phenotype model, including multiple aspects of motivation before and during the intervention, predicting both engagement and clinical efficacy. This line of research may shed light on the development of advanced prevention and personalized digital therapeutics.Trial RegistrationClinicalTrials.gov NCT03465306; https://clinicaltrials.gov/ct2/show/NCT03465306

Digital phenotyping is a rapidly growing research field promising to transform how psychiatry measures, classifies, predicts, and explains human behavior. This article advances the social-scientific examination of digital phenotyping's epistemology and knowledge claims. Drawing on the notion of a “neuromolecular gaze” in psychiatry since the 1960s, it suggests that digital phenotyping concerns a new psychiatric gaze—the “digital gaze.” Rather than privileging neuromolecular explanations, the digital gaze privileges the “deep” physiological, behavioral, and social “truths” afforded by digital technologies and big data. The article interrogates two concepts directing the digital gaze: “digital phenotype” and “digital biomarkers.” Both concepts make explicit an epistemic link between “the digital” and “the biological.” The article examines the soundness and construction of this link to, first, offer a “reality check” of digital phenotyping's claims and, second, more clearly delineate and demarcate the digital gaze. It argues there is evidence of significant mis- and overstatements about digital phenotyping's basis in biology, including in much-hyped psychiatric digital biomarker research. Rather than driving the biologization of digital traces, as some have suggested, digital mental health phenotyping so far seems mainly concerned with physiological, behavioral, and social processes that can be surveilled by means of digital devices.

Digital phenotyping is the real-time collection of individual-level active and passive data from users in naturalistic and free-living settings via personal digital devices, such as mobile phones and wearable devices. Given the novelty of research in this field, there is heterogeneity in the clinical use cases, types of data collected, modes of data collection, data analysis methods, and outcomes measured. The primary aim of this scoping review was to map the published research on digital phenotyping and to outline study characteristics, data collection and analysis methods, machine learning approaches, and future implications. We utilized an a priori approach for the literature search and data extraction and charting process, guided by the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-analyses Extension for Scoping Reviews). We identified relevant studies published in 2020, 2021, and 2022 on PubMed and Google Scholar using search terms related to digital phenotyping. The titles, abstracts, and keywords were screened during the first stage of the screening process, and the second stage involved screening the full texts of the shortlisted articles. We extracted and charted the descriptive characteristics of the final studies, which were countries of origin, study design, clinical areas, active and/or passive data collected, modes of data collection, data analysis approaches, and limitations. A total of 454 articles on PubMed and Google Scholar were identified through search terms associated with digital phenotyping, and 46 articles were deemed eligible for inclusion in this scoping review. Most studies evaluated wearable data and originated from North America. The most dominant study design was observational, followed by randomized trials, and most studies focused on psychiatric disorders, mental health disorders, and neurological diseases. A total of 7 studies used machine learning approaches for data analysis, with random forest, logistic regression, and support vector machines being the most common. Our review provides foundational as well as application-oriented approaches toward digital phenotyping in health. Future work should focus on more prospective, longitudinal studies that include larger data sets from diverse populations, address privacy and ethical concerns around data collection from consumer technologies, and build "digital phenotypes" to personalize digital health interventions and treatment plans.

CORR Synthesis: When Should the Orthopaedic Surgeon Use Artificial Intelligence, Machine Learning, and Deep Learning?

Migraine is a complex neurobiological disorder characterized by diverse phenotypes and unpredictable therapy outcomes. Digital phenotyping (DP), defined as the real-time collection of behavioral and physiological data in natural environments to identify individual phenotypes, represents a promising approach with the potential to enhance clinicians' ability to identify migraine subtypes. Additionally, DP offers new insights into the intricate neurobiological and behavioral patterns, as well as environmental influences, associated with each phase of a migraine attack, including potential triggers, pre-attack symptoms, the characteristics of an attack and response to treatment. Moreover, a DP-based approach has the potential to revolutionize migraine research and clinical trials by enabling more personalized, patient-centred diagnostics and tailored acute and preventive treatments. Despite the limited literature available and the heterogeneity of study designs, migraine DP may lay the groundwork for future digital twin models and the discovery of digital biomarkers for diagnosis, therapy optimization and outcome evaluation. Furthermore, DP could serve as a predictive marker for migraine attacks, empowering patients to monitor their condition and adopt a proactive approach to treatment. Integrating DP into migraine studies could also contribute to the development of an updated international migraine classification that incorporates neurobiological and psychosocial factors alongside clinical symptomatology. To fully realize its potential in migraine research and care, experts should prioritize collaboration with artificial intelligence (AI) specialists, data scientists and medical engineers. Establishing a multidisciplinary ecosystem will be essential to developing robust and clinically meaningful DP tools for migraine research. This review aims to show the current landscape of both active and passive DP methodologies, which leverage smartphones, wearable biosensors and AI-driven analytics to capture real-time physiological, cognitive and environmental data, at the same time as pointing to the future ahead of migraine DP.

Decisional and Dispositional Capacity Determinations: Neuropsychiatric Illness and an Integrated Clinical Paradigm

Advances in high-throughput technologies, digital phenotyping, and increased accessibility of publicly available datasets offer opportunities for big data to be applied in infectious disease surveillance, diagnosis, treatment, and outcome prediction. Artificial intelligence (AI) and machine learning (ML) have emerged as promising tools to analyze complex clinical and molecular data. However, it remains unclear which AI or ML models are most suitable for infectious disease management, as most existing studies use non-scoping literature reviews to recommend AI and ML models for data analysis. This scoping literature review thus examines the ML models and applications that are most relevant for infectious disease management, with a proposed actionable workflow for implementing ML models in clinical practice. We conducted a literature search on PubMed, Google Scholar, and ScienceDirect, including papers published in English between January 2020 and April 2024. Search keywords included AI, ML, public health, surveillance, diagnosis, prognosis, and infectious disease, to identify published studies using AI and ML in infectious disease management. Studies without public datasets or lacking descriptions of the ML models were excluded. This review included a total of 77 studies applied in surveillance, prognosis, and diagnosis. Different types of input data from infectious disease surveillance, clinical diagnosis, and prognosis required different ML and AI models to achieve the maximum performance in infectious disease management. Our findings highlight the potential of Explainable AI and ensemble learning models to be more broadly applicable in different aspects of infectious disease management, which can be integrated in clinical workflows to improve infectious disease surveillance, diagnosis, and prognosis. Explainable AI and ensemble learning models can be suitably used to achieve high accuracy in prediction. However, as most of the studies have not been validated in different cohorts, it remains unclear whether these ML models can be broadly applicable to different populations. Nonetheless, the findings encourage deploying ML and AI to complement clinicians and augment clinical decision-making.

This 8-week study was designed to explore any correlation between a passive data collection approach using a wearable device (i.e., digital phenotyping), active data collection (patient’s questionnaires), and a traditional clinical evaluation [Montgomery-Åsberg Depression Rating Scale (MADRS)] in patients with major depressive disorder (MDD) treated with trazodone once a day (OAD). Overall, 11 out of 30 planned patients were enrolled. Passive parameters measured by the wearable device included number of steps, distance walked, calories burned, and sleep quality. A relationship between the sleep score (derived from passively measured data) and MADRS score was observed, as was a relationship between data collected actively (assessing depression, sleep, anxiety, and warning signs) and MADRS score. Despite the limited sample size, the efficacy and safety results were consistent with those previously reported for trazodone. The small population in this study limits the conclusions that can be drawn about the correlation between the digital phenotyping approach and traditional clinical evaluation; however, the positive trends observed suggest the need to increase synergies among clinicians, patients, and researchers to overcome the cultural barriers toward implementation of digital tools in the clinical setting. This study is a step toward the use of digital data in monitoring symptoms of depression, and the preliminary data obtained encourage further investigations of a larger population of patients monitored over a longer period of time.

Early detection of autism, a neurodevelopmental condition associated with challenges in social communication, ensures timely access to intervention. Autism screening questionnaires have been shown to have lower accuracy when used in real-world settings, such as primary care, as compared to research studies, particularly for children of color and girls. Here we report findings from a multiclinic, prospective study assessing the accuracy of an autism screening digital application (app) administered during a pediatric well-child visit to 475 (17-36 months old) children (269 boys and 206 girls), of which 49 were diagnosed with autism and 98 were diagnosed with developmental delay without autism. The app displayed stimuli that elicited behavioral signs of autism, quantified using computer vision and machine learning. An algorithm combining multiple digital phenotypes showed high diagnostic accuracy with the area under the receiver operating characteristic curve = 0.90, sensitivity = 87.8%, specificity = 80.8%, negative predictive value = 97.8% and positive predictive value = 40.6%. The algorithm had similar sensitivity performance across subgroups as defined by sex, race and ethnicity. These results demonstrate the potential for digital phenotyping to provide an objective, scalable approach to autism screening in real-world settings. Moreover, combining results from digital phenotyping and caregiver questionnaires may increase autism screening accuracy and help reduce disparities in access to diagnosis and intervention.

Minority populations are utilizing mobile health applications more frequently to access health information. One group that may benefit from using mHealth technology is underserved women, specifically those on community supervision. Discuss methodological approaches for navigating digital health strategies to address underserved women's health disparities. Using an intersectional lens, we identified strategies for conducting research using digital health technology and artificial intelligence amongst the underserved, particularly those with community supervision. We explore (1) methodological approaches that combine traditional research methods with precision medicine, digital phenotyping, and ecological momentary assessment; (2) implications for artificial intelligence; and (3) ethical considerations with data collection, storage, and engagement. Researchers must address gendered differences related to health, social, and economic disparities concurrently with an unwavering focus on the protection of human subjects when addressing the unique needs of underserved women while utilizing digital health methodologies. Women on community supervision in South Central Texas helped inform the design of JUN, the mHealth app we reported in the case exemplar. JUN is named after the Junonia shell, a native shell to South Texas, which means strength, power, and self-sufficiency, like the participants in our preliminary studies.

Artificial intelligence (AI)-based computer perception technologies (e.g., digital phenotyping and affective computing) promise to transform clinical approaches to personalized care in psychiatry and beyond by offering more objective measures of emotional states and behavior, enabling precision treatment, diagnosis, and symptom monitoring. At the same time, passive and continuous nature by which they often collect data from patients in non-clinical settings raises ethical issues related to privacy and self-determination. Little is known about how such concerns may be exacerbated by the integration of neural data, as parallel advances in computer perception, AI, and neurotechnology enable new insights into subjective states. Here, we present findings from a multi-site NCATS-funded study of ethical considerations for translating computer perception into clinical care and contextualize them within the neuroethics and neurorights literatures. We conducted qualitative interviews with patients (n = 20), caregivers (n = 20), clinicians (n = 12), developers (n = 12), and clinician developers (n = 2) regarding their perspective toward using PC in clinical care. Transcripts were analyzed in MAXQDA using Thematic Content Analysis. Stakeholder groups voiced concerns related to (1) perceived invasiveness of passive and continuous data collection in private settings; (2) data protection and security and the potential for negative downstream/future impacts on patients of unintended disclosure; and (3) ethical issues related to patients' limited versus hyper awareness of passive and continuous data collection and monitoring. Clinicians and developers highlighted that these concerns may be exacerbated by the integration of neural data with other computer perception data. Our findings suggest that the integration of neurotechnologies with existing computer perception technologies raises novel concerns around dignity-related and other harms (e.g., stigma, discrimination) that stem from data security threats and the growing potential for reidentification of sensitive data. Further, our findings suggest that patients' awareness and preoccupation with feeling monitored via computer sensors ranges from hypo- to hyper-awareness, with either extreme accompanied by ethical concerns (consent vs. anxiety and preoccupation). These results highlight the need for systematic research into how best to implement these technologies into clinical care in ways that reduce disruption, maximize patient benefits, and mitigate long-term risks associated with the passive collection of sensitive emotional, behavioral and neural data.

A few weeks ago, we published an editorial discussion on whether artificial intelligence applications should be authors of academic articles [1] . We were delighted to receive more than one interesting reply letter to this editorial in a short time [2, 3] . We hope that opinions on this