An Algorithm for Activity Recognition and Assessment of Adults Poststroke.

IMPORTANCE: Stroke is the leading cause of long-term disability in the United States. Providers have no robust tools to objectively and accurately measure the activity of people with stroke living at home. OBJECTIVE: To explore the integration of validated upper extremity assessments poststroke within an activity recognition system. DESIGN: Exploratory descriptive study using data previously collected over 3 mo to report on algorithm testing and assessment integration. SETTING: Data were collected in the homes of community-dwelling participants. PARTICIPANTS: Participants were at least 6 mo poststroke, were able to ambulate with or without an assistive device, and self-reported some difficulty using their arm in everyday activities. OUTCOMES AND MEASURES: The activity detection algorithm's accuracy was determined by comparing its activity labels with manual labels. The algorithm integrated assessment by describing the quality of upper extremity movement, which was determined by reporting extent of reach, mean and maximum speed during movement, and smoothness of movement. RESULTS: Sixteen participants (9 women, 7 men) took part in this study, with an average age of 63.38 yr (SD = 12.84). The algorithm was highly accurate in correctly identifying activities, with 87% to 95% accuracy depending on the movement. The algorithm was also able to detect the quality of movement for upper extremity movements. CONCLUSIONS AND RELEVANCE: The algorithm was able to accurately identify in-kitchen activities performed by adults poststroke. Information about the quality of these movements was also successfully calculated. This algorithm has the potential to supplement clinical assessments in treatment planning and outcomes reporting. Plain-Language Summary: This study shows that clinical algorithms have the potential to inform occupational therapy practice by providing clinically relevant data about the in-home activities of adults poststroke. The algorithm accurately identified activities that were performed in the kitchen by adults poststroke. The algorithm also identified the quality of upper extremity movements of people poststroke who were living at home.

A semi-supervised approach to unobtrusively predict abnormality in breathing patterns using hydraulic bed sensor data in older adults aging in place.

Shortness of breath is often considered a repercussion of aging in older adults, as respiratory illnesses like COPD1 or respiratory illnesses due to heart-related issues are often misdiagnosed, under-diagnosed or ignored at early stages. Continuous health monitoring using ambient sensors has the potential to ameliorate this problem for older adults at aging-in-place facilities. In this paper, we leverage continuous respiratory health data collected by using ambient hydraulic bed sensors installed in the apartments of older adults in aging-in-place Americare facilities to find data-adaptive indicators related to shortness of breath. We used unlabeled data collected unobtrusively over the span of three years from a COPD-diagnosed individual and used data mining to label the data. These labeled data are then used to train a predictive model to make future predictions in older adults related to shortness of breath abnormality. To pick the continuous changes in respiratory health we make predictions for shorter time windows (60-s). Hence, to summarize each day's predictions we propose an abnormal breathing index (ABI) in this paper. To showcase the trajectory of the shortness of breath abnormality over time (in terms of days), we also propose trend analysis on the ABI quarterly and incrementally. We have evaluated six individual cases retrospectively to highlight the potential and use cases of our approach.

Development and Testing of a Daily Activity Recognition System for Post-Stroke Rehabilitation.

Those who survive the initial incidence of a stroke experience impacts on daily function. As a part of the rehabilitation process, it is essential for clinicians to monitor patients' health status and recovery progress accurately and consistently; however, little is known about how patients function in their own homes. Therefore, the goal of this study was to develop, train, and test an algorithm within an ambient, in-home depth sensor system that can classify and quantify home activities of individuals post-stroke. We developed the Daily Activity Recognition and Assessment System (DARAS). A daily action logger was implemented with a Foresite Healthcare depth sensor. Daily activity data were collected from seventeen post-stroke participants' homes over three months. Given the extensive amount of data, only a portion of the participants' data was used for this specific analysis. An ensemble network for activity recognition and temporal localization was developed to detect and segment the clinically relevant actions from the recorded data. The ensemble network, which learns rich spatial-temporal features from both depth and skeletal joint data, fuses the prediction outputs from a customized 3D convolutional-de-convolutional network, customized region convolutional 3D network, and a proposed region hierarchical co-occurrence network. The per-frame precision and per-action precision were 0.819 and 0.838, respectively, on the test set. The outcomes from the DARAS can help clinicians to provide more personalized rehabilitation plans that benefit patients.

Tracking personalized functional health in older adults using geriatric assessments.

BACKGROUND: Higher levels of functional health in older adults leads to higher quality of life and improves the ability to age-in-place. Tracking functional health objectively could help clinicians to make decisions for interventions in case of health deterioration. Even though several geriatric assessments capture several aspects of functional health, there is limited research in longitudinally tracking personalized functional health of older adults using a combination of these assessments. METHODS: We used geriatric assessment data collected from 150 older adults to develop and validate a functional health prediction model based on risks associated with falls, hospitalizations, emergency visits, and death. We used mixed effects logistic regression to construct the model. The geriatric assessments included were Activities of Daily Living (ADL), Instrumental Activities of Daily Living (IADL), Mini-Mental State Examination (MMSE), Geriatric Depression Scale (GDS), and Short Form 12 (SF12). Construct validators such as fall risks associated with model predictions, and case studies with functional health trajectories were used to validate the model. RESULTS: The model is shown to separate samples with and without adverse health event outcomes with an area under the receiver operating characteristic curve (AUC) of > 0.85. The model could predict emergency visit or hospitalization with an AUC of 0.72 (95% CI 0.65-0.79), fall with an AUC of 0.86 (95% CI 0.83-0.89), fall with hospitalization with an AUC of 0.89 (95% CI 0.85-0.92), and mortality with an AUC of 0.93 (95% CI 0.88-0.97). Multiple comparisons of means using Turkey HSD test show that model prediction means for samples with no adverse health events versus samples with fall, hospitalization, and death were statistically significant (p < 0.001). Case studies for individual residents using predicted functional health trajectories show that changes in model predictions over time correspond to critical health changes in older adults. CONCLUSIONS: The personalized functional health tracking may provide clinicians with a longitudinal view of overall functional health in older adults to help address the early detection of deterioration trends and decide appropriate interventions. It can also help older adults and family members take proactive steps to improve functional health.

Human-centered approaches that integrate sensor technology across the lifespan: Opportunities and challenges.

Children, parents, older adults, and caregivers routinely use sensor technology as a source of health information and health monitoring. The purpose of this paper is to describe three exemplars of research that used a human-centered approach to engage participants in the development, design, and usability of interventions that integrate technology to promote health. The exemplars are based on current research studies that integrate sensor technology into pediatric, adult, and older adult populations living with a chronic health condition. Lessons learned and considerations for future studies are discussed. Nurses have successfully implemented interventions that use technology to improve health and detect, prevent, and manage diseases in children, families, individuals and communities. Nurses are key stakeholders to inform clinically relevant health monitoring that can support timely and personalized intervention and recommendations.

Technology for Healthy Independent Living: Creating a Tailored In-Home Sensor System for Older Adults and Family Caregivers.

Sensing technologies hold enormous potential for early detection of health changes that can dramatically affect the aging experience. In previous work, we developed a health alert system that captures and analyzes in-home sensor data. The purpose of this research was to collect input from older adults and family members on how the health information generated can best be adapted, such that older adults and family members can better self-manage their health. Five 90-minute focus groups were conducted with 23 older adults (mean age = 80 years; 87% female) and five family members (mean age = 64; 100% female). Participants were asked open-ended questions about the sensor technology and methods for interacting with their health information. Participants provided feedback regarding tailoring the technology, such as delegating access to family and health care providers, receiving health messages and alerts, interpreting health messages, and graphic display options. Participants also noted concerns and future likelihood of technology adoption. [Journal of Gerontological Nursing, 46(7), 35-40.].

Using Sensor Signals in the Early Detection of Heart Failure: A Case Study.

Early detection of heart failure in older adults will be a significant issue for the foreseeable future. The current article presents a case study to describe how monitoring ballistocardiogram (BCG) waveforms captured non-invasively using sensors placed under a bed mattress can detect early heart failure changes. Heart and respiratory rates obtained from the bed sensor of a female older adult who was hospitalized with acute mixed congestive heart failure, clinic notes, and data from computer simulations reflecting increasing diastolic dysfunction were analyzed. Mean heart and respiratory rate trends obtained from her bed sensor in the prior 2 months did not indicate heart failure. BCG waveforms resulting from the simulations demonstrated changes associated with decreasing cardiac output as diastolic function worsened. Developing new methods for clinically interpreting BCG waveforms presents a significant opportunity for improving early heart failure detection. [Journal of Gerontological Nursing, 46(7), 41-46.].

Relationship Between 2-Dimensional Frontal Plane Measures and the Knee Abduction Angle During the Drop Vertical Jump.

Context: Knee abduction angle (KAA), as measured by 3-dimensional marker-based motion capture systems during jump-landing tasks, has been correlated with an elevated risk of anterior cruciate ligament injury in females. Due to the high cost and inefficiency of KAA measurement with marker-based motion capture, surrogate 2-dimensional frontal plane measures have gained attention for injury risk screening. The knee-to-ankle separation ratio (KASR) and medial knee position (MKP) have been suggested as potential frontal plane surrogate measures to the KAA, but investigations into their relationship to the KAA during a bilateral drop vertical jump task are limited. Objective: To investigate the relationship between KASR and MKP to the KAA during initial contact of the bilateral drop vertical jump. Design: Descriptive. Setting: Biomechanics laboratory. Participants: A total of 18 healthy female participants (mean age: 24.1 [3.88] y, mass: 65.18 [10.34] kg, and height: 1.63 [0.06] m). Intervention: Participants completed 5 successful drop vertical jump trials measured by a Vicon marker-based motion capture system and 2 AMTI force plates. Main Outcome Measure: For each jump, KAA of the tibia relative to the femur was measured at initial contact along with the KASR and MKP calculated from planar joint center data. The coefficient of determination (r2) was used to examine the relationship between the KASR and MKP to KAA. Results: A strong linear relationship was observed between MKP and KAA (r2 = .71), as well as between KASR and KAA (r2 = .72). Conclusions: Two-dimensional frontal plane measures show strong relationships to the KAA during the bilateral drop vertical jump.

Older Adults' Perceptions of and Preferences for a Fall Risk Assessment System: Exploring Stages of Acceptance Model.

Aging in place is a preferred and cost-effective living option for older adults. Research indicates that technology can assist with this goal. Information on consumer preferences will help in technology development to assist older adults to age in place. The study aim was to explore the perceptions and preferences of older adults and their family members about a fall risk assessment system. Using a qualitative approach, this study examined the perceptions, attitudes, and preferences of 13 older adults and five family members about their experience living with the fall risk assessment system during five points in time. Themes emerged in relation to preferences and expectations about the technology and how it fits into daily routines. We were able to capture changes that occurred over time for older adult participants. Results indicated that there was acceptance of the technology as participants adapted to it. Two themes were present across the five points in time-safety and usefulness. Five stages of acceptance emerged from the data from preinstallation to 2 years postinstallation. Identified themes, stages of acceptance, and design and development considerations are discussed.

Comparison of 3D Joint Angles Measured With the Kinect 2.0 Skeletal Tracker Versus a Marker-Based Motion Capture System.

The Microsoft Kinect is becoming a widely used tool for inexpensive, portable measurement of human motion, with the potential to support clinical assessments of performance and function. In this study, the relative osteokinematic Cardan joint angles of the hip and knee were calculated using the Kinect 2.0 skeletal tracker. The pelvis segments of the default skeletal model were reoriented and 3-dimensional joint angles were compared with a marker-based system during a drop vertical jump and a hip abduction motion. Good agreement between the Kinect and marker-based system were found for knee (correlation coefficient = 0.96, cycle RMS error = 11°, peak flexion difference = 3°) and hip (correlation coefficient = 0.97, cycle RMS = 12°, peak flexion difference = 12°) flexion during the landing phase of the drop vertical jump and for hip abduction/adduction (correlation coefficient = 0.99, cycle RMS error = 7°, peak flexion difference = 8°) during isolated hip motion. Nonsagittal hip and knee angles did not correlate well for the drop vertical jump. When limited to activities in the optimal capture volume and with simple modifications to the skeletal model, the Kinect 2.0 skeletal tracker can provide limited 3-dimensional kinematic information of the lower limbs that may be useful for functional movement assessment.

A New Paradigm of Technology-Enabled 'Vital Signs' for Early Detection of Health Change for Older Adults.

Environmentally embedded (nonwearable) sensor technology is in continuous use in elder housing to monitor a new set of 'vital signs' that continuously measure the functional status of older adults, detect potential changes in health or functional status, and alert healthcare providers for early recognition and treatment of those changes. Older adult participants' respiration, pulse, and restlessness are monitored as they sleep. Gait speed, stride length, and stride time are calculated daily, and automatically assess for increasing fall risk. Activity levels are summarized and graphically displayed for easy interpretation. Falls are detected when they occur and alerts are sent immediately to healthcare providers, so time to rescue may be reduced. Automated health alerts are sent to healthcare staff, based on continuously running algorithms applied to the sensor data, days and weeks before typical signs or symptoms are detected by the person, family members, or healthcare providers. Discovering these new functional status 'vital signs', developing automated methods for interpreting them, and alerting others when changes occur have the potential to transform chronic illness management and facilitate aging in place through the end of life. Key findings of research in progress at the University of Missouri are discussed in this viewpoint article, as well as obstacles to widespread adoption.

Enhanced registered nurse care coordination with sensor technology: Impact on length of stay and cost in aging in place housing.

BACKGROUND: When planning the Aging in Place Initiative at TigerPlace, it was envisioned that advances in technology research had the potential to enable early intervention in health changes that could assist in proactive management of health for older adults and potentially reduce costs. PURPOSE: The purpose of this study was to compare length of stay (LOS) of residents living with environmentally embedded sensor systems since the development and implementation of automated health alerts at TigerPlace to LOS of those who are not living with sensor systems. Estimate potential savings of living with sensor systems. METHODS: LOS for residents living with and without sensors was measured over a span of 4.8 years since the implementation of sensor-generated health alerts. The group living with sensors (n = 52) had an average LOS of 1,557 days (4.3 years); the comparison group without sensors (n = 81) was 936 days (2.6 years); p = .0006. Groups were comparable based on admission age, gender, number of chronic illnesses, SF12 physical health, SF12 mental health, Geriatric Depression Scale (GDS), activities of daily living, independent activities of daily living, and mini-mental status examination scores. Both groups, all residents living at TigerPlace since the implementation of health alerts, receive registered nurse (RN) care coordination as the standard of care. DISCUSSION: Results indicate that residents living with sensors were able to reside at TigerPlace 1.7 years longer than residents living without sensors, suggesting that proactive use of health alerts facilitates successful aging in place. Health alerts, generated by automated algorithms interpreting environmentally embedded sensor data, may enable care coordinators to assess and intervene on health status changes earlier than is possible in the absence of sensor-generated alerts. Comparison of LOS without sensors TigerPlace (2.6 years) with the national median in residential senior housing (1.8 years) may be attributable to the RN care coordination model at TigerPlace. Cost estimates comparing cost of living at TigerPlace with the sensor technology vs. nursing home reveal potential saving of about $30,000 per person. Potential cost savings to Medicaid funded nursing home (assuming the technology and care coordination were reimbursed) are estimated to be about $87,000 per person. CONCLUSIONS: Early alerts for potential health problems appear to enhance the current RN care coordination care delivery model at TigerPlace, increasing LOS for those living with sensors to nearly twice that of those who did not. Sensor technology with care coordination has cost saving potential for consumers and Medicaid.

Automated fall detection with quality improvement "rewind" to reduce falls in hospital rooms.

The purpose of this study was to test the implementation of a fall detection and "rewind" privacy-protecting technique using the Microsoft® Kinect™ to not only detect but prevent falls from occurring in hospitalized patients. Kinect sensors were placed in six hospital rooms in a step-down unit and data were continuously logged. Prior to implementation with patients, three researchers performed a total of 18 falls (walking and then falling down or falling from the bed) and 17 non-fall events (crouching down, stooping down to tie shoe laces, and lying on the floor). All falls and non-falls were correctly identified using automated algorithms to process Kinect sensor data. During the first 8 months of data collection, processing methods were perfected to manage data and provide a "rewind" method to view events that led to falls for post-fall quality improvement process analyses. Preliminary data from this feasibility study show that using the Microsoft Kinect sensors provides detection of falls, fall risks, and facilitates quality improvement after falls in real hospital environments unobtrusively, while taking into account patient privacy.

Smart Technology for Aging and Reducing Disability (STAReD): Study protocol for a randomized pragmatic clinical trial.

BACKGROUND: There is a critical need to improve quality of life for community-dwelling older adults with disabilities. Prior research has demonstrated that a smart, in-home sensor system can facilitate aging in place for older adults living in independent living apartments with care coordination support by identifying early illness and injury detection. Self-management approaches have shown positive outcomes for many client populations. Pairing the smart, in-home sensor system with a self-management intervention for community-dwelling older adults with disabilities may lead to positive outcomes. METHODS: This study is a prospective, two-arm, randomized, pragmatic clinical trial to compare the effect of a technology-supported self-management intervention on disability and health-related quality of life to that of a health education control, for rural, community-dwelling older adults. Individuals randomized to the self-management study arm will receive a multidisciplinary (nursing, occupational therapist, and social work) self-management approach coupled with the smart-home sensor system. Individuals randomized to the health education study arm will receive standard health education coupled with the smart-home sensor system. The primary outcomes of disability and health-related quality of life will be assessed at baseline and post-intervention. Generalizable guidance to scale the technology-supported self-management intervention will be developed from qualitatively developed exemplar cases. CONCLUSION: This study has the potential to impact the health and well-being of rural, community-dwelling older adults with disabilities. We have overcome barriers including recruitment in a rural population and supply chain issues for the sensor system. Our team remains on track to meet our study aims.

Evaluation of health alerts from an early illness warning system in independent living.

Passive sensor networks were deployed in independent living apartments to monitor older adults in their home environments to detect signs of impending illness and alert clinicians so they can intervene and prevent or delay significant changes in health or functional status. A retrospective qualitative deductive content analysis was undertaken to refine health alerts to improve clinical relevance to clinicians as they use alerts in their normal workflow of routine care delivery to older adults. Clinicians completed written free-text boxes to describe actions taken (or not) as a result of each alert; they also rated the clinical significance (relevance) of each health alert on a scale of 1 to 5. Two samples of the clinician's written responses to the health alerts were analyzed after alert algorithms had been adjusted based on results of a pilot study using health alerts to enhance clinical decision-making. In the first sample, a total of 663 comments were generated by seven clinicians in response to 385 unique alerts; there are more comments than alerts because more than one clinician rated the same alert. The second sample had a total of 142 comments produced by three clinicians in response to 88 distinct alerts. The overall clinical relevance of the alerts, as judged by the content of the qualitative comments by clinicians for each alert, improved from 33.3% of the alerts in the first sample classified as clinically relevant to 43.2% in the second. The goal is to produce clinically relevant alerts that clinicians find useful in daily practice. The evaluation methods used are described to assist others as they consider building and iteratively refining health alerts to enhance clinical decision making.

In-home fall risk assessment and detection sensor system.

Falls are a major problem in older adults. A continuous, unobtrusive, environmentally mounted (i.e., embedded into the environment and not worn by the individual), in-home monitoring system that automatically detects when falls have occurred or when the risk of falling is increasing could alert health care providers and family members to intervene to improve physical function or manage illnesses that may precipitate falls. Researchers at the University of Missouri Center for Eldercare and Rehabilitation Technology are testing such sensor systems for fall risk assessment (FRA) and detection in older adults' apartments in a senior living community. Initial results comparing ground truth (validated measures) of FRA data and GAITRite System parameters with data captured from Microsoft(®) Kinect and pulse-Doppler radar are reported.

Novel clinically-relevant assessment of upper extremity movement using depth sensors.

BACKGROUND: For individuals post-stroke, home-based programs are necessary to deliver additional hours of therapy outside of the limited time in the clinic. Virtual reality (VR)-based approaches show modest outcomes in improving client function when delivered in the home. The movement sensors used in these VR-based approaches, such as the Microsoft Kinect® have been validated against gold standards tools but have not been used as an assessment of upper extremity movement quality in the stroke population. OBJECTIVES: The purpose of this study was to explore the use of a movement sensor paired with a VR-based intervention to assess upper extremity movement for individuals post-stroke. METHODS: Movement data captured with the Microsoft Kinect® from four separate studies were aggregated for analysis (n = 8 individuals post-stroke, n = 30 individuals without disabilities). For all participants, the skeletal data (x, y, z coordinates for 15 tracked joints) for each game play session were processed in MatLab and movement variables (normalized jerk, movement path ratio, average path sway) were calculated using an OPTICS density-based cluster algorithm. RESULTS: Data from the 30 healthy individuals created a normative baseline for the three kinematic variables. Individuals post-stroke were less efficient and had more jerky movements in both upper extremities as compared to healthy individuals. CONCLUSION: It is feasible to use a movement sensor paired with a VR-based intervention to quantify and qualify upper extremity movement for individuals post-stroke. Further research with a larger cohort is necessary to establish clinical sensitivity and specificity.

Stepping Beyond Assessment: Fall Risk Prediction Models Among Older Adults from Cumulative Change in Gait Parameter Estimates.

Falls significantly affect the health of older adults. Injuries sustained through falls have long-term consequences on the ability to live independently and age in place, and are the leading cause of injury death in the United States for seniors. Early fall risk detection provides an important opportunity for prospective intervention by healthcare providers and home caregivers. In-home depth sensor technologies have been developed for real-time fall detection and gait parameter estimation including walking speed, the sixth vital sign, which has been shown to correlate with the risk of falling. This study evaluates the use of supervised classification for estimating fall risk from cumulative changes in gait parameter estimates as captured by 3D depth sensors placed within the homes of older adult participants. Using recall as the primary metric for model success rate due to the severity of fall injuries sustained by false negatives, we demonstrate an enhancement of assessing fall risk with univariate logistic regression using multivariate logistic regression, support vector, and hierarchical tree-based modeling techniques by an improvement of 18.80%, 31.78%, and 33.94%, respectively, in the 14 days preceding a fall event. Random forest and XGBoost models resulted in recall and precision scores of 0.805 compared to the best univariate regression model of Y-Entropy with a recall of 0.639 and precision of 0.527 for the 14-day window leading to a predicted fall event.

Fall Risk Prediction in Older Adults Using Free-Text Nursing Notes and Medications in Electronic Health Records.

Nursing notes in Electronic Health Records (EHR) contain critical health information, including fall risk factors. However, an exploration of fall risk prediction using nursing notes is not well examined. In this study, we explored deep learning architectures to predict fall risk in older adults using text in nursing notes and medications in the EHR. EHR predictor data and fall events outcome data were obtained from 162 older adults living at TigerPlace, a senior living facility located in Columbia, MO. We used pre-trained BioWordVec embeddings to represent the words in the clinical notes and medications and trained multiple recurrent neural network-based natural language processing models to predict future fall events. Our final model predicted falls with an accuracy of 0.81, a sensitivity of 0.75, a specificity of 0.83, and an F1 score of 0.82. This preliminary exploratory analysis provides supporting evidence that fall risk can be predicted from clinical notes and medications. Future studies will utilize additional data modalities available in the EHR to potentially improve fall risk prediction from EHR data.

Cardiovascular sex-differences: insights via physiology-based modeling and potential for noninvasive sensing via ballistocardiography.

In this study, anatomical and functional differences between men and women in their cardiovascular systems and how these differences manifest in blood circulation are theoretically and experimentally investigated. A validated mathematical model of the cardiovascular system is used as a virtual laboratory to simulate and compare multiple scenarios where parameters associated with sex differences are varied. Cardiovascular model parameters related with women's faster heart rate, stronger ventricular contractility, and smaller blood vessels are used as inputs to quantify the impact (i) on the distribution of blood volume through the cardiovascular system, (ii) on the cardiovascular indexes describing the coupling between ventricles and arteries, and (iii) on the ballistocardiogram (BCG) signal. The model-predicted outputs are found to be consistent with published clinical data. Model simulations suggest that the balance between the contractile function of the left ventricle and the load opposed by the arterial circulation attains similar levels in females and males, but is achieved through different combinations of factors. Additionally, we examine the potential of using the BCG waveform, which is directly related to cardiovascular volumes, as a noninvasive method for monitoring cardiovascular function. Our findings provide valuable insights into the underlying mechanisms of cardiovascular sex differences and may help facilitate the development of effective noninvasive cardiovascular monitoring methods for early diagnosis and prevention of cardiovascular disease in both women and men.