Background: The personalization of conversational agents with natural language user interfaces is seeing increasing use in health care applications, shaping the content, structure, or purpose of the dialogue between humans and conversational agents. Objective: The goal of this systematic review was to understand the ways in which personalization has been used with conversational agents in health care and characterize the methods of its implementation. Methods: We searched on PubMed, Embase, CINAHL, PsycInfo, and ACM Digital Library using a predefined search strategy. The studies were included if they: (1) were primary research studies that focused on consumers, caregivers, or health care professionals; (2) involved a conversational agent with an unconstrained natural language interface; (3) tested the system with human subjects; and (4) implemented personalization features. Results: The search found 1958 publications. After abstract and full-text screening, 13 studies were included in the review. Common examples of personalized content included feedback, daily health reports, alerts, warnings, and recommendations. The personalization features were implemented without a theoretical framework of customization and with limited evaluation of its impact. While conversational agents with personalization features were reported to improve user satisfaction, user engagement and dialogue quality, the role of personalization in improving health outcomes was not assessed directly. Conclusions: Most of the studies in our review implemented the personalization features without theoretical or evidence-based support for them and did not leverage the recent developments in other domains of personalization. Future research could incorporate personalization as a distinct design factor with a more careful consideration of its impact on health outcomes and its implications on patient safety, privacy, and decision-making.

Clinicians spend a large amount of time on clinical documentation of patient encounters, often impacting quality of care and clinician satisfaction, and causing physician burnout. Advances in artificial intelligence (AI) and machine learning (ML) open the possibility of automating clinical documentation with digital scribes, using speech recognition to eliminate manual documentation by clinicians or medical scribes. However, developing a digital scribe is fraught with problems due to the complex nature of clinical environments and clinical conversations. This paper identifies and discusses major challenges associated with developing automated speech-based documentation in clinical settings: recording high-quality audio, converting audio to transcripts using speech recognition, inducing topic structure from conversation data, extracting medical concepts, generating clinically meaningful summaries of conversations, and obtaining clinical data for AI and ML algorithms.

Objective: The study sought to understand the potential roles of a future artificial intelligence (AI) documentation assistant in primary care consultations and to identify implications for doctors, patients, healthcare system, and technology design from the perspective of general practitioners. Materials and Methods: Co-design workshops with general practitioners were conducted. The workshops focused on (1) understanding the current consultation context and identifying existing problems, (2) ideating future solutions to these problems, and (3) discussing future roles for AI in primary care. The workshop activities included affinity diagramming, brainwriting, and video prototyping methods. The workshops were audio-recorded and transcribed verbatim. Inductive thematic analysis of the transcripts of conversations was performed. Results: Two researchers facilitated 3 co-design workshops with 16 general practitioners. Three main themes emerged: professional autonomy, human-AI collaboration, and new models of care. Major implications identified within these themes included (1) concerns with medico-legal aspects arising from constant recording and accessibility of full consultation records, (2) future consultations taking place out of the exam rooms in a distributed system involving empowered patients, (3) human conversation and empathy remaining the core tasks of doctors in any future AI-enabled consultations, and (4) questioning the current focus of AI initiatives on improved efficiency as opposed to patient care. Conclusions: AI documentation assistants will likely to be integral to the future primary care consultations. However, these technologies will still need to be supervised by a human until strong evidence for reliable autonomous performance is available. Therefore, different human-AI collaboration models will need to be designed and evaluated to ensure patient safety, quality of care, doctor safety, and doctor autonomy.

Background: Research in psychology has shown that the way a person walks reflects that person’s current mood (or emotional state). Recent studies have used mobile phones to detect emotional states from movement data. Objective: The objective of our study was to investigate the use of movement sensor data from a smart watch to infer an individual’s emotional state. We present our findings of a user study with 50 participants. Methods: The experimental design is a mixed-design study: within-subjects (emotions: happy, sad, and neutral) and between-subjects (stimulus type: audiovisual “movie clips” and audio “music clips”). Each participant experienced both emotions in a single stimulus type. All participants walked 250 m while wearing a smart watch on one wrist and a heart rate monitor strap on the chest. They also had to answer a short questionnaire (20 items; Positive Affect and Negative Affect Schedule, PANAS) before and after experiencing each emotion. The data obtained from the heart rate monitor served as supplementary information to our data. We performed time series analysis on data from the smart watch and a t test on questionnaire items to measure the change in emotional state. Heart rate data was analyzed using one-way analysis of variance. We extracted features from the time series using sliding windows and used features to train and validate classifiers that determined an individual’s emotion. Results: Overall, 50 young adults participated in our study; of them, 49 were included for the affective PANAS questionnaire and 44 for the feature extraction and building of personal models. Participants reported feeling less negative affect after watching sad videos or after listening to sad music, P<.006. For the task of emotion recognition using classifiers, our results showed that personal models outperformed personal baselines and achieved median accuracies higher than 78% for all conditions of the design study for binary classification of happiness versus sadness. Conclusions: Our findings show that we are able to detect changes in the emotional state as well as in behavioral responses with data obtained from the smartwatch. Together with high accuracies achieved across all users for classification of happy versus sad emotional states, this is further evidence for the hypothesis that movement sensor data can be used for emotion recognition.

Cardiovascular disease (CVD) risk prediction is important for guiding the intensity of therapy in CVD prevention. Whilst current risk prediction algorithms use traditional statistical approaches, machine learning (ML) presents an alternative method that may improve risk prediction accuracy. This systematic review and meta-analysis aimed to investigate whether ML algorithms demonstrate greater performance compared with traditional risk scores in CVD risk prognostication. MEDLINE, EMBASE, CENTRAL, and SCOPUS Web of Science Core collections were searched for studies comparing ML models to traditional risk scores for CVD risk prediction between the years 2000 and 2021. We included studies that assessed both ML and traditional risk scores in adult (≥18 year old) primary prevention populations. We assessed the risk of bias using the Prediction Model Risk of Bias Assessment Tool (PROBAST) tool. Only studies that provided a measure of discrimination [i.e. C-statistics with 95% confidence intervals (CIs)] were included in the meta-analysis. A total of 16 studies were included in the review and meta-analysis (3302 515 individuals). All study designs were retrospective cohort studies. Out of 16 studies, 3 externally validated their models, and 11 reported calibration metrics. A total of 11 studies demonstrated a high risk of bias. The summary C-statistics (95% CI) of the top-performing ML models and traditional risk scores were 0.773 (95% CI: 0.740–0.806) and 0.759 (95% CI: 0.726–0.792), respectively. The difference in C-statistic was 0.0139 (95% CI: 0.0139–0.140), P < 0.0001. ML models outperformed traditional risk scores in the discrimination of CVD risk prognostication. Integration of ML algorithms into electronic healthcare systems in primary care could improve identification of patients at high risk of subsequent CVD events and hence increase opportunities for CVD prevention. It is uncertain whether they can be implemented in clinical settings. Future implementation research is needed to examine how ML models may be utilized for primary prevention. This review was registered with PROSPERO (CRD42020220811).