Assessing inclusion and representativeness on digital platforms for health education: Evidence from YouTube.

BACKGROUND: Studies confirm that significant biases exist in online recommendation platforms, exacerbating pre-existing disparities and leading to less-than-optimal outcomes for underrepresented demographics. We study issues of bias in inclusion and representativeness in the context of healthcare information disseminated via videos on the YouTube social media platform, a widely used online channel for multi-media rich information. With one in three US adults using the Internet to learn about a health concern, it is critical to assess inclusivity and representativeness regarding how health information is disseminated by digital platforms such as YouTube. METHODS: Leveraging methods from fair machine learning (ML), natural language processing and voice and facial recognition methods, we examine inclusivity and representativeness of video content presenters using a large corpus of videos and their metadata on a chronic condition (diabetes) extracted from the YouTube platform. Regression models are used to determine whether presenter demographics impact video popularity, measured by the video's average daily view count. A video that generates a higher view count is considered to be more popular. RESULTS: The voice and facial recognition methods predicted the gender and race of the presenter with reasonable success. Gender is predicted through voice recognition (accuracy = 78 %, AUC = 76 %), while the gender and race predictions use facial recognition (accuracy = 93 %, AUC = 92 % and accuracy = 82 %, AUC = 80 %, respectively). The gender of the presenter is more significant for video views only when the face of the presenter is not visible while videos with male presenters with no face visibility have a positive relationship with view counts. Furthermore, videos with white and male presenters have a positive influence on view counts while videos with female and non - white group have high view counts. CONCLUSION: Presenters' demographics do have an influence on average daily view count of videos viewed on social media platforms as shown by advanced voice and facial recognition algorithms used for assessing inclusion and representativeness of the video content. Future research can explore short videos and those at the channel level because popularity of the channel name and the number of videos associated with that channel do have an influence on view counts.

Multi-resource allocation and care sequence assignment in patient management: a stochastic programming approach.

To mitigate outpatient care delivery inefficiencies induced by resource shortages and demand heterogeneity, this paper focuses on the problem of allocating and sequencing multiple medical resources so that patients scheduled for clinical care can experience efficient and coordinated care with minimum total waiting time. We leverage highly granular location data on people and medical resources collected via Real-Time Location System technologies to identify dominant patient care pathways. A novel two-stage Stochastic Mixed Integer Linear Programming model is proposed to determine the optimal patient sequence based on the available resources according to the care pathways that minimize patients' expected total waiting time. The model incorporates the uncertainty in care activity duration via sample average approximation.We employ a Monte Carlo Optimization procedure to determine the appropriate sample size to obtain solutions that provide a good trade-off between approximation accuracy and computational time. Compared to the conventional deterministic model, our proposed model would significantly reduce waiting time for patients in the clinic by 60%, on average, with acceptable computational resource requirements and time complexity. In summary, this paper proposes a computationally efficient formulation for the multi-resource allocation and care sequence assignment optimization problem under uncertainty. It uses continuous assignment decision variables without timestamp and position indices, enabling the data-driven solution of problems with real-time allocation adjustment in a dynamic outpatient environment with complex clinical coordination constraints.

A Machine Learning Approach to Support Urgent Stroke Triage Using Administrative Data and Social Determinants of Health at Hospital Presentation: Retrospective Study.

BACKGROUND: The key to effective stroke management is timely diagnosis and triage. Machine learning (ML) methods developed to assist in detecting stroke have focused on interpreting detailed clinical data such as clinical notes and diagnostic imaging results. However, such information may not be readily available when patients are initially triaged, particularly in rural and underserved communities. OBJECTIVE: This study aimed to develop an ML stroke prediction algorithm based on data widely available at the time of patients' hospital presentations and assess the added value of social determinants of health (SDoH) in stroke prediction. METHODS: We conducted a retrospective study of the emergency department and hospitalization records from 2012 to 2014 from all the acute care hospitals in the state of Florida, merged with the SDoH data from the American Community Survey. A case-control design was adopted to construct stroke and stroke mimic cohorts. We compared the algorithm performance and feature importance measures of the ML models (ie, gradient boosting machine and random forest) with those of the logistic regression model based on 3 sets of predictors. To provide insights into the prediction and ultimately assist care providers in decision-making, we used TreeSHAP for tree-based ML models to explain the stroke prediction. RESULTS: Our analysis included 143,203 hospital visits of unique patients, and it was confirmed based on the principal diagnosis at discharge that 73% (n=104,662) of these patients had a stroke. The approach proposed in this study has high sensitivity and is particularly effective at reducing the misdiagnosis of dangerous stroke chameleons (false-negative rate <4%). ML classifiers consistently outperformed the benchmark logistic regression in all 3 input combinations. We found significant consistency across the models in the features that explain their performance. The most important features are age, the number of chronic conditions on admission, and primary payer (eg, Medicare or private insurance). Although both the individual- and community-level SDoH features helped improve the predictive performance of the models, the inclusion of the individual-level SDoH features led to a much larger improvement (area under the receiver operating characteristic curve increased from 0.694 to 0.823) than the inclusion of the community-level SDoH features (area under the receiver operating characteristic curve increased from 0.823 to 0.829). CONCLUSIONS: Using data widely available at the time of patients' hospital presentations, we developed a stroke prediction model with high sensitivity and reasonable specificity. The prediction algorithm uses variables that are routinely collected by providers and payers and might be useful in underresourced hospitals with limited availability of sensitive diagnostic tools or incomplete data-gathering capabilities.

Cognitive workload reduction in hospital information systems : Decision support for order set optimization.

Order sets are a critical component in hospital information systems that are expected to substantially reduce physicians' physical and cognitive workload and improve patient safety. Order sets represent time interval-clustered order items, such as medications prescribed at hospital admission, that are administered to patients during their hospital stay. In this paper, we develop a mathematical programming model and an exact and a heuristic solution procedure with the objective of minimizing physicians' cognitive workload associated with prescribing order sets. Furthermore, we provide structural insights into the problem which lead us to a valid lower bound on the order set size. In a case study using order data on Asthma patients with moderate complexity from a major pediatric hospital, we compare the hospital's current solution with the exact and heuristic solutions on a variety of performance metrics. Our computational results confirm our lower bound and reveal that using a time interval decomposition approach substantially reduces computation times for the mathematical program, as does a K -means clustering based decomposition approach which, however, does not guarantee optimality because it violates the lower bound. The results of comparing the mathematical program with the current order set configuration in the hospital indicates that cognitive workload can be reduced by about 20.2% by allowing 1 to 5 order sets, respectively. The comparison of the K -means based decomposition with the hospital's current configuration reveals a cognitive workload reduction of about 19.5%, also by allowing 1 to 5 order sets, respectively. We finally provide a decision support system to help practitioners analyze the current order set configuration, the results of the mathematical program and the heuristic approach.

Paving the COWpath: Learning and visualizing clinical pathways from electronic health record data.

OBJECTIVE: Clinical pathways translate best available evidence into practice, indicating the most widely applicable order of treatment interventions for particular treatment goals. We propose a practice-based clinical pathway development process and a data-driven methodology for extracting common clinical pathways from electronic health record (EHR) data that is patient-centered, consistent with clinical workflow, and facilitates evidence-based care. MATERIALS AND METHODS: Visit data of 1,576 chronic kidney disease (CKD) patients who developed acute kidney injury (AKI) from 2009 to 2013 are extracted from the EHR. We model each patient's multi-dimensional clinical records into one-dimensional sequences using novel constructs designed to capture information on each visit's purpose, procedures, medications and diagnoses. Analysis and clustering on visit sequences identify distinct types of patient subgroups. Characterizing visit sequences as Markov chains, significant transitions are extracted and visualized into clinical pathways across subgroups. RESULTS: We identified 31 patient subgroups whose extracted clinical pathways provide insights on how patients' conditions and medication prescriptions may progress over time. We identify pathways that show typical disease progression, practices that are consistent with guidelines, and sustainable improvements in patients' health conditions. Visualization of pathways depicts the likelihood and direction of disease progression under varied contexts. DISCUSSION AND CONCLUSIONS: Accuracy of EHR data and diversity in patients' conditions and practice patterns are critical challenges in learning insightful practice-based clinical pathways. Learning and visualizing clinical pathways from actual practice data captured in the EHR may facilitate efficient practice review by healthcare providers and support patient engagement in shared decision making.

YouTube Videos for Public Health Literacy? A Machine Learning Pipeline to Curate Covid-19 Videos.

The COVID-19 pandemic has highlighted the dire necessity to improve public health literacy for societal resilience. YouTube provides a vast repository of user-generated health information in a multi-media-rich format which may be easier for the public to understand and use if major concerns about content quality and accuracy are addressed. This study develops an automated solution to identify, retrieve and shortlist medically relevant and understandable YouTube videos that domain experts can subsequently review and recommend for disseminating and educating the public on the COVID-19 pandemic and similar public health outbreaks. Our approach leverages domain knowledge from human experts and machine learning and natural language processing methods to provide a scalable, replicable, and generalizable approach that can also be applied to enhance the management of many health conditions.

Continuous Remote Patient Monitoring for Post-Discharge Heart Failure Management: Workflow Modeling Using Discrete Event Simulation.

The Cascade-HF protocol is a Continuous Remote Patient Monitoring (CRPM) study at a major health system in the United States to reduce Heart Failure (HF)-related hospitalizations and readmissions using wearable biosensors to collect physiological data over a 30-day period to determine decompensation risk among HF patients. The alerts produced, coupled with electronic patient-reported outcomes, are utilized daily by the home health team, and escalated to the heart failure team as needed, for proactive actions. Limited research has examined anticipating the implementation and workflow challenges of such complex CRPM studies such as resource planning and staffing decisions that leverage the recorded data to drive clinical preparedness and operational efficiency. This preliminary analysis applies discrete event simulation modeling to the Cascade-HF protocol using pilot data from a soft launch to assess workload of the clinical team, evaluate escalation patterns and provide decision support recommendations to enable scale-up for all post-discharge patients.

YouTube Video Analytics for Patient Education: An Exploratory Clustering of Obstructive Sleep Apnea Videos.

With the growing popularity of content-sharing platforms, patients are increasingly using the Internet as a critical source of health information. As one of the most popular video-sharing sites, YouTube provides easy access to health information seekers, but it is difficult and time-consuming to identify and retrieve high-quality videos that may serve as engaging patient education materials. This paper reports on an exploratory analysis of 317 YouTube videos on Obstructive Sleep Apnea (OSA) to better understand some key features of the videos and the relationships between them to facilitate subsequent video classification and recommendation. Features intrinsic to a video, such as video duration, and extrinsic, such as the number of views, are analyzed using unsupervised clustering methods and the Sankey diagram to discover the relationship between the clusters and their significance across different clusters, providing promising insights for the assessment of video quality.

Limitations of receiver operating characteristic curve on imbalanced data: Assist device mortality risk scores.

OBJECTIVE: In the left ventricular assist device domain, the receiver operating characteristic is a commonly applied metric of performance of classifiers. However, the receiver operating characteristic can provide a distorted view of classifiers' ability to predict short-term mortality due to the overwhelmingly greater proportion of patients who survive, that is, imbalanced data. This study illustrates the ambiguity of the receiver operating characteristic in evaluating 2 classifiers of 90-day left ventricular assist device mortality and introduces the precision recall curve as a supplemental metric that is more representative of left ventricular assist device classifiers in predicting the minority class. METHODS: This study compared the receiver operating characteristic and precision recall curve for 2 classifiers for 90-day left ventricular assist device mortality, HeartMate Risk Score and Random Forest for 800 patients (test group) recorded in the Interagency Registry for Mechanically Assisted Circulatory Support who received a continuous-flow left ventricular assist device between 2006 and 2016 (mean age, 59 years; 146 female vs 654 male patients), in whom 90-day mortality rate is only 8%. RESULTS: The receiver operating characteristic indicates similar performance of Random Forest and HeartMate Risk Score classifiers with respect to area under the curve of 0.77 and Random Forest 0.63, respectively. This is in contrast to their precision recall curve with area under the curve of 0.43 versus 0.16 for Random Forest and HeartMate Risk Score, respectively. The precision recall curve for HeartMate Risk Score showed the precision rapidly decreased to only 10% with slightly increasing sensitivity. CONCLUSIONS: The receiver operating characteristic can portray an overly optimistic performance of a classifier or risk score when applied to imbalanced data. The precision recall curve provides better insight about the performance of a classifier by focusing on the minority class.

Developing a common data model approach for DISCOVER CKD: A retrospective, global cohort of real-world patients with chronic kidney disease.

OBJECTIVES: To describe a flexible common data model (CDM) approach that can be efficiently tailored to study-specific needs to facilitate pooled patient-level analysis and aggregated/meta-analysis of routinely collected retrospective patient data from disparate data sources; and to detail the application of this CDM approach to the DISCOVER CKD retrospective cohort, a longitudinal database of routinely collected (secondary) patient data of individuals with chronic kidney disease (CKD). METHODS: The flexible CDM approach incorporated three independent, exchangeable components that preceded data mapping and data model implementation: (1) standardized code lists (unifying medical events from different coding systems); (2) laboratory unit harmonization tables; and (3) base cohort definitions. Events between different coding vocabularies were not mapped code-to-code; for each data source, code lists of labels were curated at the entity/event level. A study team of epidemiologists, clinicians, informaticists, and data scientists were included within the validation of each component. RESULTS: Applying the CDM to the DISCOVER CKD retrospective cohort, secondary data from 1,857,593 patients with CKD were harmonized from five data sources, across three countries, into a discrete database for rapid real-world evidence generation. CONCLUSIONS: This flexible CDM approach facilitates evidence generation from real-world data within the DISCOVER CKD retrospective cohort, providing novel insights into the epidemiology of CKD that may expedite improvements in diagnosis, prognosis, early intervention, and disease management. The adaptable architecture of this CDM approach ensures scalable, fast, and efficient application within other therapy areas to facilitate the combined analysis of different types of secondary data from multiple, heterogeneous sources.

Association of Health Insurance Literacy With Financial Hardship in Patients With Cancer.

Importance: Patient-reported financial hardship is an increasing challenge in cancer care delivery. Health insurance literacy and its association with financial hardship in patients with cancer, especially after controlling for financial literacy, have not been well examined. Objective: To examine the prevalence of and factors in the association between health insurance literacy and financial literacy as well as the overall and individual domains of financial hardship and their association with health insurance literacy, both independently and when adjusted for financial literacy, in patients with cancer. Design, Setting, and Participants: This cross-sectional survey study recruited and enrolled patients from 2 separate ambulatory infusion centers at Mayo Clinic Arizona in Phoenix, Arizona, and the University of Mississippi Medical Center in Jackson, Mississippi. Adult patients aged 18 years or older were enrolled from December 2019 to February 2020 and from August to October 2020 at Mayo Clinic Arizona (n = 299) and from September 2020 through January 2021 at the University of Mississippi Medical Center (n = 105). Survey respondents received a $5 gift card. Exposures: Surveys included questions about sociodemographic characteristics, health insurance literacy and financial literacy, financial knowledge, and financial hardship and its domains (material hardship, psychological hardship, and behavioral hardship). Main Outcomes and Measures: Financial hardship was assessed using the COST-FACIT (Comprehensive Score for Financial Toxicity-Functional Assessment of Chronic Illness Therapy) measure and National Health Interview Survey questions to capture information about the domains of financial hardship. The Health Insurance Literacy Measure is a validated 21-item measure of a consumer's ability to select and use health insurance. Five questions from the National Financial Capability Study assessed financial literacy. Results: A total of 404 participants were enrolled in the study. Median (IQR) age of the respondents was 63 (54-71) years, and 219 were women (54%), 307 were non-Hispanic White individuals (76%), 153 (38%) had private insurance, and 289 (72%) had solid tumors. Overall financial hardship (denoted by median COST-FACIT score <27 points) was reported by 49% (95% CI, 44%-53%) of the cohort. Prevalence of financial hardship was higher using the National Health Interview Survey questions, with 68% (95% CI, 63%-72%) of respondents reporting at least 1 hardship domain (n = 276). Sixty-six percent (95% CI, 60%-69%) of respondents (n = 265) had a high level of financial literacy. The mean (SD) Health Insurance Literacy Measure score was 64.9 (13.3) points. In multivariable analyses, each 10-point increase in the Health Insurance Literacy Measure score was associated with lower odds of financial hardship (odds ratio, 0.82; 95% CI, 0.68-0.99; P = .04). However, this association was no longer significant after adjusting for financial literacy. Conclusions and Relevance: Results of this study showed that, despite a high level of health insurance literacy and financial literacy, the prevalence of financial hardship was high. Although there were lower odds of financial hardship with increased health insurance literacy, the association was no longer significant when financial literacy was added to the model, suggesting that a high level of financial literacy may help mitigate the adverse outcome of lower health insurance literacy levels in patients with cancer.

Continuous Remote Patient Monitoring in Patients With Heart Failure (Cascade Study): Protocol for a Mixed Methods Feasibility Study.

BACKGROUND: Heart failure (HF) is a prevalent chronic disease and is associated with increases in mortality and morbidity. HF is a leading cause of hospitalizations and readmissions in the United States. A potentially promising area for preventing HF readmissions is continuous remote patient monitoring (CRPM). OBJECTIVE: The primary aim of this study is to determine the feasibility and preliminary efficacy of a CRPM solution in patients with HF at NorthShore University HealthSystem. METHODS: This study is a feasibility study and uses a wearable biosensor to continuously remotely monitor patients with HF for 30 days after discharge. Eligible patients admitted with an HF exacerbation at NorthShore University HealthSystem are being recruited, and the wearable biosensor is placed before discharge. The biosensor collects physiological ambulatory data, which are analyzed for signs of patient deterioration. Participants are also completing a daily survey through a dedicated study smartphone. If prespecified criteria from the physiological data and survey results are met, a notification is triggered, and a predetermined electronic health record-based pathway of telephonic management is completed. In phase 1, which has already been completed, 5 patients were enrolled and monitored for 30 days after discharge. The results of phase 1 were analyzed, and modifications to the program were made to optimize it. After analysis of the phase 1 results, 15 patients are being enrolled for phase 2, which is a calibration and testing period to enable further adjustments to be made. After phase 2, we will enroll 45 patients for phase 3. The combined results of phases 1, 2, and 3 will be analyzed to determine the feasibility of a CRPM program in patients with HF. Semistructured interviews are being conducted with key stakeholders, including patients, and these results will be analyzed using the affective adaptation of the technology acceptance model. RESULTS: During phase 1, of the 5 patients, 2 (40%) were readmitted during the study period. The study completion rate for phase 1 was 80% (4/5), and the study attrition rate was 20% (1/5). There were 57 protocol deviations out of 150 patient days in phase 1 of the study. The results of phase 1 were analyzed, and the study protocol was adjusted to optimize it for phases 2 and 3. Phase 2 and phase 3 results will be available by the end of 2022. CONCLUSIONS: A CRPM program may offer a low-risk solution to improve care of patients with HF after hospital discharge and may help to decrease readmission of patients with HF to the hospital. This protocol may also lay the groundwork for the use of CRPM solutions in other groups of patients considered to be at high risk. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/36741.

Using machine learning to improve risk prediction in durable left ventricular assist devices.

Risk models have historically displayed only moderate predictive performance in estimating mortality risk in left ventricular assist device therapy. This study evaluated whether machine learning can improve risk prediction for left ventricular assist devices. Primary durable left ventricular assist devices reported in the Interagency Registry for Mechanically Assisted Circulatory Support between March 1, 2006 and December 31, 2016 were included. The study cohort was randomly divided 3:1 into training and testing sets. Logistic regression and machine learning models (extreme gradient boosting) were created in the training set for 90-day and 1-year mortality and their performance was evaluated after bootstrapping with 1000 replications in the testing set. Differences in model performance were also evaluated in cases of concordance versus discordance in predicted risk between logistic regression and extreme gradient boosting as defined by equal size patient tertiles. A total of 16,120 patients were included. Calibration metrics were comparable between logistic regression and extreme gradient boosting. C-index was improved with extreme gradient boosting (90-day: 0.707 [0.683-0.730] versus 0.740 [0.717-0.762] and 1-year: 0.691 [0.673-0.710] versus 0.714 [0.695-0.734]; each p<0.001). Net reclassification index analysis similarly demonstrated an improvement of 48.8% and 36.9% for 90-day and 1-year mortality, respectively, with extreme gradient boosting (each p<0.001). Concordance in predicted risk between logistic regression and extreme gradient boosting resulted in substantially improved c-index for both logistic regression and extreme gradient boosting (90-day logistic regression 0.536 versus 0.752, 1-year logistic regression 0.555 versus 0.726, 90-day extreme gradient boosting 0.623 versus 0.772, 1-year extreme gradient boosting 0.613 versus 0.742, each p<0.001). These results demonstrate that machine learning can improve risk model performance for durable left ventricular assist devices, both independently and as an adjunct to logistic regression.

Machine Learning Approaches to Analyzing Adverse Events Following Durable LVAD Implantation.

BACKGROUND: This study employed machine learning approaches to analyze sequences of adverse events (AEs) after left ventricular assist device (LVAD) implantation. METHODS: Data on patients implanted with the HeartWare HVAD durable LVAD were extracted from the ENDURANCE and ENDURANCE Supplemental clinical trials, with follow-up through 5 years. Major AEs included device malfunction, major bleeding, major infection, neurological dysfunction, renal dysfunction, respiratory dysfunction, and right heart failure (RHF). Time interval and transition probability analyses were performed. We created a Sankey diagram to visualize transitions between AEs. Hierarchical clustering was applied to dissimilarity matrices based on the longest common subsequence to identify clusters of patients with similar AE profiles. RESULTS: A total of 568 patients underwent HVAD implantation with 3590 AEs. Bleeding and RHF comprised the highest proportion of early AEs after surgery whereas infection and bleeding accounted for most AEs occurring after 3 months. The highest transition probabilities were observed with infection to infection (0.34), bleeding to bleeding (0.31), RHF to bleeding (0.31), RHF to infection (0.28), and bleeding to infection (0.26). Five distinct clusters of patients were generated, each with different patterns of time intervals between AEs, transition rates between AEs, and clinical outcomes. CONCLUSIONS: Machine learning approaches allow for improved visualization and understanding of AE burden after LVAD implantation. Distinct patterns and relationships provide insights that may be important for quality improvement efforts.

Delineating Pathways to Death by Multisystem Organ Failure in Patients With a Left Ventricular Assist Device.

BACKGROUND: This study delineates the sequences of adverse events (AEs) preceding mortality attributed to multisystem organ failure (MSOF) in patients with a left ventricular assist device (LVAD). METHODS: We analyzed 3765 AEs after 536 LVAD implants recorded in The Society of Thoracic Surgeons Intermacs data registry between 2006 and 2015 that resulted in MSOF death. Hierarchical clustering identified and visualized quantitatively unique clusters of patients with similar AE profiles. Markov modeling was used to illustrate the AE sequences that led to MSOF death within the clusters. Cox proportional hazard models determined the risk-adjusted, preimplant predictors of MSOF. RESULTS: We identified 2 distinct MSOF clusters based on their proportion of AE types and survival time. The early-death cluster (418 patients, 2304 AEs) had a median survival of 1 month (interquartile range, 3-6 months), whereas the late-death cluster (118 patients, 1,461 AEs) had a median survival of 11 months (interquartile range, 6-22 months). The predominant AE sequences in the early-death and late-death clusters were renal failure, to respiratory failure, to death (62%) and bleeding, to infection, to respiratory failure, to death (45%), respectively. Significant risk-adjusted preimplant predictors of MSOF included line sepsis (hazard ratio [HR] 3.0; 95% confidence interval [CI], 1.1-8.2), extracorporeal membrane oxygenation (HR, 2.2; 95% CI, 1.2-3.9), and dialysis or ultrafiltration (HR, 2.1; 95% CI, 1.5-3.0). CONCLUSIONS: This analysis identified 2 AE clusters and the predominant sequences that result in MSOF-associated mortality. MSOF develops in 1 cluster of patients after chronic bleeding and repeated infections but has prolonged survival, while another group dies early after renal and respiratory complications.

Continuous Remote Patient Monitoring: Evaluation of the Heart Failure Cascade Soft Launch.

OBJECTIVE: We report on our experience of deploying a continuous remote patient monitoring (CRPM) study soft launch with structured cascading and escalation pathways on heart failure (HF) patients post-discharge. The lessons learned from the soft launch are used to modify and fine-tune the workflow process and study protocol. METHODS: This soft launch was conducted at NorthShore University HealthSystem's Evanston Hospital from December 2020 to March 2021. Patients were provided with non-invasive wearable biosensors that continuously collect ambulatory physiological data, and a study phone that collects patient-reported outcomes. The physiological data are analyzed by machine learning algorithms, potentially identifying physiological perturbation in HF patients. Alerts from this algorithm may be cascaded with other patient status data to inform home health nurses' (HHNs') management via a structured protocol. HHNs review the monitoring platform daily. If the patient's status meets specific criteria, HHNs perform assessments and escalate patient cases to the HF team for further guidance on early intervention. RESULTS: We enrolled five patients into the soft launch. Four participants adhered to study activities. Two out of five patients were readmitted, one due to HF, one due to infection. Observed miscommunication and protocol gaps were noted for protocol amendment. The study team adopted an organizational development method from change management theory to reconfigure the study protocol. CONCLUSION: We sought to automate the monitoring aspects of post-discharge care by aligning a new technology that generates streaming data from a wearable device with a complex, multi-provider workflow into a novel protocol using iterative design, implementation, and evaluation methods to monitor post-discharge HF patients. CRPM with structured escalation and telemonitoring protocol shows potential to maintain patients in their home environment and reduce HF-related readmissions. Our results suggest that further education to engage and empower frontline workers using advanced technology is essential to scale up the approach.

eVisit: a pilot study of a new kind of healthcare delivery.

Patient online eVisits are gaining momentum due to increasing consumer demand for improved access to clinical services, availability of new technologies to deploy such services and development of reimbursement initiatives by major payers. The eVisit service provides patients with an online consultation through a series of structured, secure message exchanges with a physician, providing an alternative for onsite office visits and non-reimbursed phone-based care. In this study, we evaluate a pilot deployment of eVisits in a primary care clinic providing online consultation service for 7 simple health conditions at its three locations. We examine usage data over 3 months and survey and interview results for trends in adoption, demographic and temporal patterns of usage, clinician and patient expectations and experiences, and challenges to sustainability of the service. Based on our analysis, we conclude that the eVisit pilot was a success. Patients valued the new service being offered as demonstrated by a rapid increase in usage. The quality of service was good with fast turnaround times and few exchanges to resolve a request. These positive outcomes combined with a reimbursement model are promising indications of sustainability but several challenges remain.

Social determinants of health in electronic health records and their impact on analysis and risk prediction: A systematic review.

OBJECTIVE: This integrative review identifies and analyzes the extant literature to examine the integration of social determinants of health (SDoH) domains into electronic health records (EHRs), their impact on risk prediction, and the specific outcomes and SDoH domains that have been tracked. MATERIALS AND METHODS: In accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we conducted a literature search in the PubMed, CINAHL, Cochrane, EMBASE, and PsycINFO databases for English language studies published until March 2020 that examined SDoH domains in the context of EHRs. RESULTS: Our search strategy identified 71 unique studies that are directly related to the research questions. 75% of the included studies were published since 2017, and 68% were U.S.-based. 79% of the reviewed articles integrated SDoH information from external data sources into EHRs, and the rest of them extracted SDoH information from unstructured clinical notes in the EHRs. We found that all but 1 study using external area-level SDoH data reported minimum contribution to performance improvement in the predictive models. In contrast, studies that incorporated individual-level SDoH data reported improved predictive performance of various outcomes such as service referrals, medication adherence, and risk of 30-day readmission. We also found little consensus on the SDoH measures used in the literature and current screening tools. CONCLUSIONS: The literature provides early and rapidly growing evidence that integrating individual-level SDoH into EHRs can assist in risk assessment and predicting healthcare utilization and health outcomes, which further motivates efforts to collect and standardize patient-level SDoH information.

Exploring trajectories of emergency department visits using a laboratory-based indicator of serious illness.

Repeated emergency department visits have become a serious challenge worldwide. Despite prior research indicating that laboratory results may provide early alerts about such patients on their upcoming adverse events, few studies have examined their role as a critical indicator of the stability of a patient's medical condition over time. We model and analyze the developmental trajectories of patients' creatinine levels, a key laboratory marker of serious illness, as a potential risk stratification mechanism across many emergency department visits. We apply group-based statistical methodology to electronic health record data of 70,385 patients, with 3-15 emergency department visits, to identify and profile these trajectories for the entire population, for males and for females. Results reveal three distinct creatinine-based trajectory groups over time with significantly differing characteristics that may enable targeted interventions for each group. Future research will incorporate additional disease markers to identify longitudinal factors leading to repeated emergency department visits.

Predicting Volume Responsiveness Among Sepsis Patients Using Clinical Data and Continuous Physiological Waveforms.

The efficacy of early fluid treatment in patients with sepsis is unclear and may contribute to serious adverse events due to fluid non-responsiveness. The current method of deciding if patients are responsive to fluid administration is often subjective and requires manual intervention. This study utilizes MIMIC III and associated matched waveform datasets across the entire ICU stay duration of each patient to develop prediction models for assessing fluid responsiveness in sepsis patients. We developed a pipeline to extract high frequency continuous waveform data and included waveform features in the prediction models. Comparing across five machine learning models, random forest performed the best when no waveform information is added (AUC = 0.84), with mean arterial blood pressure and age identified as key factors. After incorporation of features from physiologic waveforms, logistic regression with L1 penalty provided consistent performance and high interpretability, achieving an accuracy of 0.89 and F1 score of 0.90.