Feasibility of Deploying Home-Based Digital Technology, Environmental Sensors, and Web-Based Surveys for Assessing Behavioral Symptoms and Identifying Their Precipitants in Older Adults: Longitudinal, Observational Study.

BACKGROUND: Apathy, depression, and anxiety are prevalent neuropsychiatric symptoms experienced by older adults. Early detection, prevention, and intervention may improve outcomes. OBJECTIVE: We aim to demonstrate the feasibility of deploying web-based weekly questionnaires inquiring about the behavioral symptoms of older adults with normal cognition, mild cognitive impairment, or early-stage dementia and to demonstrate the feasibility of deploying an in-home technology platform for measuring participant behaviors and their environment. METHODS: The target population of this study is older adults with normal cognition, mild cognitive impairment, or early-stage dementia. This is an observational, longitudinal study with a study period of up to 9 months. The severity of participant behavioral symptoms (apathy, depression, and anxiety) was self-reported weekly through web-based surveys. Participants' digital biomarkers were continuously collected at their personal residences and through wearables throughout the duration of the study. The indoor physical environment at each residence, such as light level, noise level, temperature, humidity, or air quality, was also measured using indoor environmental sensors. Feasibility was examined, and preliminary correlation analysis between the level of symptoms and the digital biomarkers and between the level of symptoms and the indoor environment was performed. RESULTS: At 13 months after recruitment began, a total of 9 participants had enrolled into this study. The participants showed high adherence rates in completing the weekly questionnaires (response rate: 275/278, 98.9%), and data collection using the digital technology appeared feasible and acceptable to the participants with few exceptions. Participants' severity of behavioral symptoms fluctuated from week to week. Preliminary results show that the duration of sleep onset and noise level are positively correlated with the anxiety level in a subset of our participants. CONCLUSIONS: This study is a step toward more frequent assessment of older adults' behavioral symptoms and holistic in situ monitoring of older adults' behaviors and their living environment. The goal of this study is to facilitate the development of objective digital biomarkers of neuropsychiatric symptoms and to identify in-home environmental factors that contribute to these symptoms.

Differences in Life Space Activity Patterns Between Older Adults With Mild Cognitive Impairment Living Alone or as a Couple: Cohort Study Using Passive Activity Sensing.

BACKGROUND: Measuring function with passive in-home sensors has the advantages of real-world, objective, continuous, and unobtrusive measurement. However, previous studies have focused on 1-person homes only, which limits their generalizability. OBJECTIVE: This study aimed to compare the life space activity patterns of participants living alone with those of participants living as a couple and to compare people with mild cognitive impairment (MCI) with cognitively normal participants in both 1- and 2-person homes. METHODS: Passive infrared motion sensors and door contact sensors were installed in 1- and 2-person homes with cognitively normal residents or residents with MCI. A home was classified as an MCI home if at least 1 person in the home had MCI. Time out of home (TOOH), independent life space activity (ILSA), and use of the living room, kitchen, bathroom, and bedroom were calculated. Data were analyzed using the following methods: (1) daily averages over 4 weeks, (2) hourly averages (time of day) over 4 weeks, or (3) longitudinal day-to-day changes. RESULTS: In total, 129 homes with people living alone (n=27, 20.9%, MCI and n=102, 79.1%, no-MCI homes) and 52 homes with people living as a couple (n=24, 46.2%, MCI and n=28, 53.8%, no-MCI homes) were included with a mean follow-up of 719 (SD 308) days. Using all 3 analysis methods, we found that 2-person homes showed a shorter TOOH, a longer ILSA, and shorter living room and kitchen use. In MCI homes, ILSA was higher in 2-person homes but lower in 1-person homes. The effects of MCI status on other outcomes were only found when using the hourly averages or longitudinal day-to-day changes over time, and they depended on the household type (alone vs residing as a couple). CONCLUSIONS: This study shows that in-home behavior is different when a participant is living alone compared to when they are living as a couple, meaning that the household type should be considered when studying in-home behavior. The effects of MCI status can be detected with in-home sensors, even in 2-person homes, but data should be analyzed on an hour-to-hour basis or longitudinally.

Tunable White Light for Elders (TWLITE): A Protocol Demonstrating Feasibility and Acceptability for Deployment, Remote Data Collection, and Analysis of a Home-Based Lighting Intervention in Older Adults.

Sleep disturbances are common in older adults and may contribute to disease progression in certain populations (e.g., Alzheimer's disease). Light therapy is a simple and cost-effective intervention to improve sleep. Primary barriers to light therapy are: (1) poor acceptability of the use of devices, and (2) inflexibility of current devices to deliver beyond a fixed light spectrum and throughout the entirety of the day. However, dynamic, tunable lighting integrated into the native home lighting system can potentially overcome these limitations. Herein, we describe our protocol to implement a whole-home tunable lighting system installed throughout the homes of healthy older adults already enrolled in an existing study with embedded home assessment platforms (Oregon Center for Aging & Technology-ORCATECH). Within ORCATECH, continuous data on room location, activity, sleep, and general health parameters are collected at a minute-to-minute resolution over years of participation. This single-arm longitudinal protocol collected participants' light usage in addition to ORCATECH outcome measures over a several month period before and after light installation. The protocol was implemented with four subjects living in three ORCATECH homes. Technical/usability challenges and feasibility/acceptability outcomes were explored. The successful implementation of our protocol supports the feasibility of implementing and integrating tunable whole-home lighting systems into an automated home-based assessment platform for continuous data collection of outcome variables, including long-term sleep measures. Challenges and iterative approaches are discussed. This protocol will inform the implementation of future clinical intervention trials using light therapy in patients at risk for developing Alzheimer's disease and related conditions.

Monitoring Behaviors of Patients With Late-Stage Dementia Using Passive Environmental Sensing Approaches: A Case Series.

OBJECTIVE: To show the feasibility of using different unobtrusive activity-sensing technologies to provide objective behavioral markers of persons with dementia (PwD). DESIGN: Monitored the behaviors of two PwD living in memory care unit using the Oregon Center for Aging & Technology (ORCATECH) platform, and the behaviors of two PwD living in assisted living facility using the Emerald device. SETTING: A memory care unit in Portland, Oregon and an assisted living facility in Framingham, Massachusetts. PARTICIPANTS: A 63-year-old male with Alzheimer's disease (AD), and an 80-year-old female with frontotemporal dementia, both lived in a memory care unit in Portland, Oregon. An 89-year-old woman with a diagnosis of AD, and an 85-year-old woman with a diagnosis of major neurocognitive disorder, Alzheimer's type with behavioral symptoms, both resided at an assisted living facility in Framingham, Massachusetts. MEASUREMENTS: These include: sleep quality measured by the bed pressure mat; number of transitions between spaces and dwell times in different spaces measured by the motion sensors; activity levels measured by the wearable actigraphy device; and couch usage and limb movements measured by the Emerald device. RESULTS: Number of transitions between spaces can identify the patient's episodes of agitation; activity levels correlate well with the patient's excessive level of agitation and lack of movement when the patient received potentially inappropriate medication and neared the end of life; couch usage can detect the patient's increased level of apathy; and periodic limb movements can help detect risperidone-induced side effects. This is the first demonstration that the ORCATECH platform and the Emerald device can measure such activities. CONCLUSION: The use of technologies for monitoring behaviors of PwD can provide more objective and intensive measurements of PwD behaviors.

Real-Time Arrhythmia Detection Using Hybrid Convolutional Neural Networks.

Background Accurate detection of arrhythmic events in the intensive care units (ICU) is of paramount significance in providing timely care. However, traditional ICU monitors generate a high rate of false alarms causing alarm fatigue. In this work, we develop an algorithm to improve life threatening arrhythmia detection in the ICUs using a deep learning approach. Methods and Results This study involves a total of 953 independent life-threatening arrhythmia alarms generated from the ICU bedside monitors of 410 patients. Specifically, we used the ECG (4 channels), arterial blood pressure, and photoplethysmograph signals to accurately detect the onset and offset of various arrhythmias, without prior knowledge of the alarm type. We used a hybrid convolutional neural network based classifier that fuses traditional handcrafted features with features automatically learned using convolutional neural networks. Further, the proposed architecture remains flexible to be adapted to various arrhythmic conditions as well as multiple physiological signals. Our hybrid- convolutional neural network approach achieved superior performance compared with methods which only used convolutional neural network. We evaluated our algorithm using 5-fold cross-validation for 5 times and obtained an accuracy of 87.5%±0.5%, and a score of 81%±0.9%. Independent evaluation of our algorithm on the publicly available PhysioNet 2015 Challenge database resulted in overall classification accuracy and score of 93.9% and 84.3%, respectively, indicating its efficacy and generalizability. Conclusions Our method accurately detects multiple arrhythmic conditions. Suitable translation of our algorithm may significantly improve the quality of care in ICUs by reducing the burden of false alarms.

Real-time machine learning-based intensive care unit alarm classification without prior knowledge of the underlying rhythm.

AIMS: This work attempts to develop a standalone heart rhythm alerting system for the intensive care unit (ICU), where life-threatening arrhythmias have to be identified/alerted more precisely and more instantaneously (i.e. with lower latency) than existing bedside monitors. METHODS AND RESULTS: We use the dataset from the PhysioNet 2015 Challenge, which contains records that led to true and false arrhythmic alarms in the ICU. These records have been re-annotated as one of eight classes, namely (i) asystole, (ii) extreme bradycardia, (iii) extreme tachycardia, (iv) ventricular fibrillation (VF), (v) ventricular tachycardia (VT), (vi) normal sinus rhythm, (vii) sinus tachycardia, and (viii) noise/artefacts. Arrhythmia-specific features and features that measure the signal quality were extracted from all the records. To improve VF detection, an improved, over an existing, single-lead R-wave detection was developed that takes into account the R-waves detected in all electrocardiographic (ECG) leads. To avoid false R-wave detection due to pacing spikes, ECG signals were filtered with a low pass filter prior to R-wave detection, while the raw signals were used for feature extraction. Random forest was used as the classifier, and 10-time five-fold cross-validation, resulted in a macro-average sensitivity of 81.54%. CONCLUSIONS: In conclusion, comparing with the bedside monitors used in the PhysioNet 2015 competition, we find that our method achieves higher positive predictive values for asystole, extreme bradycardia, VT, and VF; furthermore, our method is able to alert the presence of arrhythmia instantaneously, i.e. up to 4 s earlier.

Step Count Standardization: Validation of Step Counts from the Withings Activite using PiezoRxD and wGT3X-BT.

Daily step counts from the Withings Activite were validated against those collected concurrently from the PiezoRxD Pedometer and the wGT3X-BT Actigraph worn on the waist and on the wrist in free-living conditions from 10 older adult volunteers. The Withings Activite underestimated step counts but showed good correlations with the other devices (Pearson correlation coefficient: 0.850 - 0.891).Clinical Relevance - Although the Withings Activite underestimated steps, they may be used in studies to estimate relative level of physical activity in free-living conditions since they have good correlations with other well-validated devices. Underestimation of steps may be corrected using linear transformation.

Sensing a problem: Proof of concept for characterizing and predicting agitation.

INTRODUCTION: Agitation, experienced by patients with dementia, is difficult to manage and stressful for caregivers. Currently, agitation is primarily assessed by caregivers or clinicians based on self-report or very brief periods of observation. This limits availability of comprehensive or sensitive enough reporting to detect early signs of agitation or identify its precipitants. The purpose of this article is to provide proof of concept for characterizing and predicting agitation using a system that continuously monitors patients' activities and living environment within memory care facilities. METHODS: Continuous and unobtrusive monitoring of a participant is achieved using behavioral sensors, which include passive infrared motion sensors, door contact sensors, a wearable actigraphy device, and a bed pressure mat sensor installed in the living quarters of the participant. Environmental sensors are also used to continuously assess temperature, light, sound, and humidity. Episodes of agitation are reported by nursing staff. Data collected for 138 days were divided by 8-hour nursing shifts. Features from agitated shifts were compared to those from non-agitated shifts using t-tests. RESULTS: A total of 37 episodes of agitation were reported for a male participant, aged 64 with Alzheimer's disease, living in a memory care unit. Participant activity metrics (eg, transitions within the living room, sleep scores from the bedmat, and total activity counts from the actigraph) significantly correlated with occurrences of agitation at night (P < 0.05). Environmental variables (eg, humidity) also correlated with the occurrences of agitation at night (P < 0.05). Higher activity levels were also observed in the evenings before agitated nights. DISCUSSION: A platform of sensors used for unobtrusive and continuous monitoring of participants with dementia and their living space seems feasible and shows promise for characterization of episodes of agitation and identification of behavioral and environmental precipitants of agitation.

The Collaborative Aging Research Using Technology Initiative: An Open, Sharable, Technology-Agnostic Platform for the Research Community.

INTRODUCTION: Future digital health research hinges on methodologies to conduct remote clinical assessments and in-home monitoring. The Collaborative Aging Research Using Technology (CART) initiative was introduced to establish a digital technology research platform that could widely assess activity in the homes of diverse cohorts of older adults and detect meaningful change longitudinally. This paper reports on the built end-to-end design of the CART platform, its functionality, and the resulting research capabilities. METHODS: CART platform development followed a principled design process aiming for scalability, use case flexibility, longevity, and data privacy protection while allowing sharability. The platform, comprising ambient technology, wearables, and other sensors, was deployed in participants' homes to provide continuous, long-term (months to years), and ecologically valid data. Data gathered from CART homes were sent securely to a research server for analysis and future data sharing. RESULTS: The CART system was created, iteratively tested, and deployed to 232 homes representing four diverse cohorts (African American, Latinx, low-income, and predominantly rural-residing veterans) of older adults (n = 301) across the USA. Multiple measurements of wellness such as cognition (e.g., mean daily computer use time = 160-169 min), physical mobility (e.g., mean daily transitions between rooms = 96-155), sleep (e.g., mean nightly sleep duration = 6.3-7.4 h), and level of social engagement (e.g., reports of overnight visitors = 15-45%) were collected across cohorts. CONCLUSION: The CART initiative resulted in a minimally obtrusive digital health-enabled system that met the design principles while allowing for data capture over extended periods and can be widely used by the research community. The ability to monitor and manage health digitally within the homes of older adults is an important alternative to in-person assessments in many research contexts. Further advances will come with wider, shared use of the CART system in additional settings, within different disease contexts, and by diverse research teams.

Reduction of false alarms in the intensive care unit using an optimized machine learning based approach.

This work attempts to reduce the number of false alarms generated by bedside monitors in the intensive care unit (ICU), as a majority of current alarms are false. In this study, we applied methods that can be categorized into three stages: signal processing, feature extraction, and optimized machine learning. At the stage of signal processing, we ensured that the heartbeats were properly annotated. During feature extraction, besides extracting features that are relevant to the arrhythmic alarms, we also extracted a set of signal quality indices (SQIs), which we used to distinguish noise/artifact from normal physiological signals. When applying a machine learning algorithm (Random Forest), we performed feature selection in order to reduce the complexity of the models and improve the efficiency of the algorithm. The dataset used is from Reducing False Arrhythmia Alarms in the ICU: the PhysioNet/Computing in Cardiology Challenge 2015. Using the performance metric "score" from the Challenge, we achieved a score of 83.08 in the real-time category on the hidden test set, which is the highest in all published work.

Development and validation of warning system of ventricular tachyarrhythmia in patients with heart failure with heart rate variability data.

Implantable-cardioverter defibrillators (ICD) detect and terminate life-threatening ventricular tachyarrhythmia with electric shocks after they occur. This puts patients at risk if they are driving or in a situation where they can fall. ICD's shocks are also very painful and affect a patient's quality of life. It would be ideal if ICDs can accurately predict the occurrence of ventricular tachyarrhythmia and then issue a warning or provide preventive therapy. Our study explores the use of ICD data to automatically predict ventricular arrhythmia using heart rate variability (HRV). A 5 minute and a 10 second warning system are both developed and compared. The participants for this study consist of 788 patients who were enrolled in the ICD arm of the Sudden Cardiac Death-Heart Failure Trial (SCD-HeFT). Two groups of patient rhythms, regular heart rhythms and pre-ventricular-tachyarrhythmic rhythms, are analyzed and different HRV features are extracted. Machine learning algorithms, including random forests (RF) and support vector machines (SVM), are trained on these features to classify the two groups of rhythms in a subset of the data comprising the training set. These algorithms are then used to classify rhythms in a separate test set. This performance is quantified by the area under the curve (AUC) of the ROC curve. Both RF and SVM methods achieve a mean AUC of 0.81 for 5-minute prediction and mean AUC of 0.87-0.88 for 10-second prediction; an AUC over 0.8 typically warrants further clinical investigation. Our work shows that moderate classification accuracy can be achieved to predict ventricular tachyarrhythmia with machine learning algorithms using HRV features from ICD data. These results provide a realistic view of the practical challenges facing implementation of machine learning algorithms to predict ventricular tachyarrhythmia using HRV data, motivating continued research on improved algorithms and additional features with higher predictive power.

SCD-HeFT: Use of R-R interval statistics for long-term risk stratification for arrhythmic sudden cardiac death.

BACKGROUND: In the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT), a significant fraction of the patients with congestive heart failure ultimately did not die suddenly of arrhythmic causes. Patients with CHF will benefit from better tools to identify if implantable cardioverter-defibrillator (ICD) therapy is needed. OBJECTIVES: We aimed to identify predictor variables from baseline SCD-HeFT patients' R-R intervals that correlate to arrhythmic sudden cardiac death (SCD) and mortality and to design an ICD therapy screening test. METHODS: Ten predictor variables were extracted from prerandomization Holter data from 475 patients enrolled in the ICD arm of the SCD-HeFT by using novel and traditional heart rate variability methods. All variables were correlated to SCD using the Mann-Whitney-Wilcoxon test and receiver operating characteristic analysis. ICD therapy screening tests were designed by minimizing the cost of false classifications. Survival analysis, including log-rank test and Cox models, was also performed. RESULTS: A short-term fractal exponent, α1, and a long-term fractal exponent, α2, from detrended fluctuation analysis, the ratio of low- to high-frequency power, the number of premature ventricular contractions per hour, and the heart rate turbulence slope are all statistically significant for predicting the occurrences of SCD (P < .001) and survival (log-rank, P < .01). The most powerful multivariate predictor tool using the Cox proportional hazards regression model was α2 with a hazard ratio of 0.0465 (95% confidence interval 0.00528-0.409; P < .01). CONCLUSION: Predictor variables extracted from R-R intervals correlate to the occurrences of SCD and distinguish survival functions among patients with ICDs in SCD-HeFT. We believe that SCD prediction models should incorporate Holter-based R-R interval analysis to refine ICD patient selection, especially to exclude patients who are unlikely to benefit from ICD therapy.