Xinxin Zhu’s research while affiliated with IBM Research - Thomas J. Watson Research Center and other places

BACKGROUND Technology-enabled ecological momentary assessment (EMA) facilitates the calibration of physiological signals against self-reported data and contexts. However, research using this method rarely considers the impact that user experience (UX) has on the quality of data. OBJECTIVE The purpose of this study is to explore the biases that UX factors induce in self-reported data and physiological signals collected through EMA and the UX factors that have the largest impact on the data. METHODS A retrospective analysis on data from a field feasibility study is conducted. The study uses an application on a smartwatch device to measure heart rate variability (HRV) and collect self-reported stress levels. We collected data on event types, age, sex, personality traits, and engineered 66 UX features (e.g., number of screens viewed, perception of notification frequency). We use a series of random forest models, conditional forest models, linear regression models, and correlation analysis to predict self-reported stress, HRV, and their discrepancies. We then use iterated comparative analysis to confirm the effects of UX factors. RESULTS Analysis on 1240.6 hours of data from 29 participants reveal that self-reported stress is correlated with the HRV signal collected after EMA notification (HRV2) but not with the HRV signal collected before the notification (HRV1) or after user interaction starts (HRV3). UX factors explain 6.6% - 10% (P < .001) of the variation in self-reported stress. UX factors do not significantly predict HRV signals but explain 63.8% (P < .001) of the difference between self-reported stress and the HRV signal collected after the EMA notification. In addition, UX factors have a significant but smaller delayed effect on self-reported stress and HRV signals collected in the next user interaction cycle. In almost all models, UX features rank higher in terms of feature importance than the other confounding factors (i.e., age, sex, personality traits) and in some models rank higher than the main effect (i.e., event types). We discuss specific symptoms of UX-induced biases related to EMA instrument design and study design, mere measurement effect and observer effect, and propose topics of examination for future studies. CONCLUSIONS User experience may induce biases in data collected through technology-enabled EMA method. In some cases, the impact of the biases may be larger than that of the main effect, other confounding factors, and the corresponding data used for calibration.

The advances in mobile and wearable sensing have led to a myriad of approaches for stress detection in both laboratory and free-living settings. Most of these methods, however, rely on the usage of some combination of physiological signals measured by the sensors to detect stress. While these solutions work great in a lab or a controlled environment, the performance in free-living situations leaves much to be desired. In this work, we explore the role of context of the user in free-living conditions, and how that affects users' perceived stress levels. To this end, we conducted an 'in-the-wild' study with 23 participants, where we collected physiological data from the users, along with 'high-level' contextual labels, and perceived stress levels. Our analysis shows that context plays a significant role in the users' perceived stress levels, and when used in conjunction with physiological signals leads to much higher stress detection results, as compared to relying on just physiological data.

In modern life, the nonstop and pervasive stress tends to keep us on long-lasting high alert, which over time, could lead to a broad range of health problems from depression, metabolic disorders to heart diseases. However, there is a stunning lack of practical tools for effective stress management that can help people navigate through their daily stress. This paper presents the feasibility evaluation of StressHacker, a smartwatch-based system designed to continuously and passively monitor one's stress level using bio-signals obtained from the on-board sensors. With the proliferation of smartwatches, StressHacker is highly accessible and suited for daily use. Our preliminary evaluation is based on 300 hours of data collected in a real-life setting (12 subjects, 29 days). The result suggests that StressHacker is capable of reliably capturing daily stress dynamics (precision = 86.1%, recall = 91.2%), thus with great potential to enable seamless and personalized stress management.

Knowing the dynamics of one's daily stress is essential to effective stress management in the context of smart and connected health. However, there lacks a practical and unobtrusive means to obtain real-time and longitudinal stress information. In this paper, we attempt to derive a convenient HRV-based (heart rate variability) biomarker named cHRV, which can be used to reliably reflect stress dynamics. cHRV's key advantage lies in its low maintenance and high practicality. It can be efficiently calculated only using data from photoplethysmography (PPG) sensors, the mainstream heart rate sensor embedded in most of the consumer wearables like Apple Watch. Benefiting from the proliferation of wearables, cHRV is ideal for day-to-day stress monitoring. To evaluate its feasibility and performance, we have conducted 14 in-lab controlled experiments. The result shows that the proposed cHRV has strong correlation with the stress dynamics (r > 0.95), therefore exhibits great potential for continuous stress assessment.

Telehealth is the use of technology for remote patient monitoring and care. Wearables are small electronic devices that can seamlessly collect data about a patient for prolonged periods of time and support the implementation of telemedicine in the patient’s natural environment. In a reality where patients are becoming older and sicker, medicine is becoming more and more a multidisciplinary team work and healthcare resources are limited, telehealth holds promise as a way to improve patient care while cutting on costs. It may improve coordination between care providers, allow for bringing top notch expertise to remote, rural settings, provide a more complete picture of the patient’s condition and support independent living of the elderly and patients with chronic diseases. In this chapter, we review some of the related technology and application and portrait how they may be integrated in the near future in the healthcare delivery system.

As the volume and variety of healthcare related data continues to grow, the analysis and use of this data will increasingly depend on the ability to appropriately collect, curate and integrate disparate data from many different sources. We describe our approach to and highlight our experiences with the development of a robust data collection, curation and integration infrastructure that supports healthcare analytics. This system has been successfully applied to the processing of a variety of data types including clinical data from electronic health records and observational studies, genomic data, microbiomic data, self-reported data from surveys and self-tracked data from wearable devices from over 600 subjects. The curated data is currently being used to support healthcare analytic applications such as data visualization, patient stratification and predictive modeling.

Patient engagement can be enhanced through continuous monitoring of patient' health knowledge and self-efficacy levels across different co-morbid conditions and key aspects in health improvement and recovery. While selfreported test batteries and computerized instruments have been designed to perform evaluation of patient literacy and self-efficacy, they are not practical to be used in scenarios where concurrent tests are needed to understand the change over time. In this paper we propose an adaptive system that can progressively compose the most pertinent test for each user based on the provisional estimates of a patient's latent trait being measured. This requires modeling not only the latent health literacy and self-efficacy levels of patients and the difficulty and discriminating factors of test items, but also the temporal dependency among concurrent tests. The goal is to account for changes over the course of patient engagement history as the basis for devising personalized patient plans.

Patient engagement is important to help patients become more informed and active in managing their health. Effective patient engagement demands short, yet valid instruments for measuring self-efficacy in various care dimensions. However, the static instruments are often too lengthy to be effective for assessment purposes. Furthermore, these tests could neither account for the dynamicity of measurements over time, nor differentiate care dimensions that are more critical to certain sub-populations. To remedy these disadvantages, we devise a dynamic instrument composition approach that can model the measurement of patient self-efficacy over time and iteratively select critical care dimensions and appropriate assessment questions based on dynamic user categorization. The dynamically composed instruments are expected to guide patients through self-management reinforcement cycles within or across care dimensions, while tightly integrated into clinical workflow and standard care processes.

Each year thousands of patients die of avoidable medication errors. When a patient is admitted to, transferred within, or discharged from a clinical facility, clinicians should review previous medication orders, current orders and future plans for care, and reconcile differences if there are any. If medication reconciliation is not accurate and systematic, medication errors such as omissions, duplications, dosing errors, or drug interactions may occur and cause harm. Computer-assisted medication applications showed promise as an intervention to reduce medication summarization inaccuracies and thus avoidable medication errors. In this study, a computer-assisted medication summarization application, designed to abstract and represent multi-source time-oriented medication data, was introduced to assist clinicians with their medication reconciliation processes. An evaluation study was carried out to assess clinical usefulness and analyze potential impact of such application. Both quantitative and qualitative methods were applied to measure clinicians' performance efficiency and inaccuracy in medication summarization process with and without the intervention of computer-assisted medication application. Clinicians' feedback indicated the feasibility of integrating such a medication summarization tool into clinical practice workflow as a complementary addition to existing electronic health record systems. The result of the study showed potential to improve efficiency and reduce inaccuracy in clinician performance of medication summarization, which could in turn improve care efficiency, quality of care, and patient safety.