Telemonitoring Data Research Articles

Hypertension management and outcomes during the covid-19 pandemic: diagnosis, blood pressure telemonitoring use and mortality

Abstract Introduction The effects of COVID-19 on people with hypertension are poorly understood, but potentially severe both due to increased risk of cardiovascular events post covid (1) and by altering the way people accessed healthcare (2). This study aimed to explore the immediate impact of COVID-19 on the hypertensive population in Scotland. Methods The study used linked routine clinical data. Hypertensive patients from five NHS Scotland Health Boards were identified through the Prescribing Information System (PIS) between March 2019 and February 2021. Linked data included Blood Pressure (BP) telemonitoring data from the Connect Me BP service, outcome data (cardiovascular events and mortality) from Scottish Morbidity Records, and National Records of Scotland COVID-19 testing and vaccination records from Public Health Scotland (PHS). Follow up was until February 2022. EDRIS, the research arm of PHS, linked and pseudonymised the data and transferred it to the secure national safe haven analysis platform for access by the research team. The study was approved by the Public Benefit and Privacy Panel for Health and Social Care. Results Almost 450,000 people with hypertension were identified across the whole adult age range, 52% male. Of those, 89% had been diagnosed prior to the study. New diagnoses dropped by over 50% during the first year of the COVID-19 pandemic. The use of long term BP telemonitoring increased rapidly during the pandemic from 5000 to over 10,000. Overall mortality was over 8%. Mortality rates showed a similar pattern to the general population during the pandemic and remained above pre-pandemic levels until the end of the follow up period in February 2022. Discussion BP telemonitoring supported healthcare delivery at a time when normal services were severely disrupted due to COVID-19. However the data suggests that there may be an increase in the number of people with undiagnosed hypertension with associated raised cardiovascular risk. Mortality in this population was increased by the COVID-19 pandemic and remained above pre-pandemic levels.

Establishment of a digital electrocardiogram telemonitoring system and remote consultation service.

The use of telemonitoring electrocardiogram (ECG) improves the detection rate of various arrhythmias as it offers an extended recording time and does not impose activity restrictions. However, many community hospitals lack the resources to conduct this test due to a lack of cardiac function specialists and antiquated screening equipment. To establish a digital ECG telemonitoring system and remote consultation service between tertiary hospitals and community medical service stations to improve diagnosis and treatment effectiveness. We used the PI ECG telemonitoring data acquisition system, with a personal digital assistant (PDA) or computer serving as the platform for ECG acquisition, storage, display, printing, and transmission. We introduced this system to standardize the storage and transmission of ECG telemonitoring data, and ensure accurate transmission, reproducibility, and preservation of data. The implementation of the PI ECG telemonitoring data acquisition system enabled the sharing of remote ambulatory electrocardiography data between Beijing Shijitan Hospital and its affiliated community hospitals. This has resulted in reduced waiting times for patients to receive reports (from 0.3 to 1 hour, to 0.1 or less), shortened consultation times (from 2 hours to 0.1 hour), and improved patient satisfaction with consultations (> 20% of the patients reported they were highly satisfied). The system successfully combined the goals of intelligence, real-time synchronization, and efficiency, revolutionizing ECG telemonitoring diagnosis in the network era. Sharing digital telematics information between tertiary and community hospitals enables a tiered approach to diagnosis and treatment and allows for more efficient, accurate, and convenient remote diagnostics.

Deep recurrent neural networks distributed on a Hadoop/Spark cluster for fall detection

Falls detection approaches struggle with both Big Data scalability and upholding individual privacy, this research work proposed a novel approach for posture recognition followed by fall detection, taking advantage of the synergy between Random Forests and Uniform Local Binary Patterns (uLBP) histograms for an accurate and fast posture identification while respecting privacy. Additionally, it referred to deep recurrent neural networks distributed on a Hadoop and Spark platform for time series analysis in fall detection. This combination of methods allowed us to achieve acceptable real-time monitoring precision. This study, therefore addressed two objectives simultaneously: efficiency and scalability in posture recognition using Random Forests and uLBP, and fall detection relying on the recurrent neural network (RNN) for time series processing. The suggested solution is designed for home telemonitoring, where scalability and effective data management are supported through Hadoop/Spark. The integration of these technologies promotes reliable detection without any privacy violation, paving the way for a wider adoption of home monitoring systems for an increasing population of dependent individuals.

Advancing Rheumatology Care Through Machine Learning.

Rheumatologic diseases are marked by their complexity, involving immune-, metabolic- and mechanically mediated processes which can affect different organ systems. Despite a growing arsenal of targeted medications, many rheumatology patients fail to achieve full remission. Assessing disease activity remains challenging, as patients prioritize different symptoms and disease phenotypes vary. This is also reflected in clinical trials where the efficacy of drugs is not necessarily measured in an optimal way with the traditional outcome assessment. The recent COVID-19 pandemic has catalyzed a digital transformation in healthcare, embracing telemonitoring and patient-reported data via apps and wearables. As a further driver of digital medicine, electronic medical record (EMR) providers are actively engaged in developing algorithms for clinical decision support, heralding a shift towards patient-centered, decentralized care. Machine learning algorithms have emerged as valuable tools for handling the increasing volume of patient data, promising to enhance treatment quality and patient well-being. Convolutional neural networks (CNN) are particularly promising for radiological image analysis, aiding in the detection of specific lesions such as erosions, sacroiliitis, or osteoarthritis, with several FDA-approved applications. Clinical predictions, including numerical disease activity forecasts and medication choices, offer the potential to optimize treatment strategies. Numeric predictions can be integrated into clinical workflows, allowing for shared decision making with patients. Clustering patients based on disease characteristics provides a personalized care approach. Digital biomarkers, such as patient-reported outcomes and wearables data, offer insights into disease progression and therapy response more flexibly and outside patient consultations. In association with patient-reported outcomes, disease-specific digital biomarkers via image recognition or single-camera motion capture enables more efficient remote patient monitoring.Digital biomarkers may also play a major role in clinical trials in the future as continuous, disease-specific outcome measurement facilitating decentralized studies. Prediction models can help with patient selection in clinical trials, such as by predicting high disease activity.Efforts are underway to integrate these advancements into clinical workflows using digital pathways and remote patient monitoring platforms.In summary, machine learning, digital biomarkers, and advanced imaging technologies hold immense promise for enhancing clinical decision support and clinical trials in rheumatology. Effective integration will require a multidisciplinary approach and continued validation through prospective studies.

Seasonal changes in positive airway pressure adherence.

Through their effects on sleep duration, bedroom environments, and pollen allergies, seasonal variations may impact positive airway pressure (PAP) adherence. We analyzed daily PAP telemonitoring data from 25,846 adults (median age 64 years, 67.8% male) treated with PAP for at least 4 months [mean (standard deviation, SD) duration of PAP: 5.5 years (SD 4.1)] to examine seasonal changes in PAP adherence, leaks, and residual apnea-hypopnea index. We demonstrate a significant decrease in PAP adherence in June compared to January (mean (SD): 0.37 (1.54) h/night) that achieved the minimal clinically important difference (MCID) of 30 min in 13.9% of adults. Furthermore, we provide novel data supporting the association of rising temperatures with seasonal changes in PAP use. Indeed, the most pronounced decline in PAP adherence was observed during the hottest days, while PAP adherence was only slightly reduced during the coolest days of June. Clinicians should be aware of seasonal changes in PAP adherence that are likely to be exacerbated by climate change.

Positive airway pressure telehealth models and long-term therapy termination: a healthcare database analysis.

Telemonitoring-guided interventions can improve short-term positive airway pressure (PAP) therapy adherence, but long-term effects are unknown. This study investigated long-term PAP therapy termination in patients with sleep apnoea managed with standard care, telemonitoring-guided proactive care or telemonitoring-guided proactive care + patient engagement tool. German healthcare provider data were analysed retrospectively. Individuals aged 18-100 years who started PAP from 2014 to 2019 and had device type/interface data were included. Time-to-termination periods were analysed using Kaplan-Meier plots and Cox proportional hazards regression, adjusted for age, sex, insurance type, and device and mask type. The per-protocol population (valid telemonitoring data) included 104 612 individuals (71% male; 95% aged >40 years). Mean follow-up was 3.3±2.0 years. The overall therapy termination rate was significantly lower in the telemonitoring-guided proactive care group versus standard care (20% versus 27%; p<0.001), and even lower in the telemonitoring-guided care + patient engagement tool group (11%; p<0.001 versus other treatment groups). Adjusted risk of therapy termination was lower versus standard care (hazard ratio 0.76, 95% confidence interval 0.74-0.78; and 0.41 (0.38-0.44) for telemonitoring-guided proactive care alone + patient engagement). Age <50 or >59 years and use of a nasal pillows or full-face mask were significant predictors of therapy termination; male sex, use of telemonitoring-guided proactive care (± patient engagement) and private insurance were significantly associated with lower therapy termination rates. Use of telemonitoring-guided proactive care and a patient engagement tool was associated with lower rates of PAP therapy termination.

A Novel Semi-Supervised Learning Model for Smartphone-Based Health Telemonitoring

Telemonitoring is the use of electronic devices such as smartphones to remotely monitor patients. It provides great convenience and enables timely medical decisions. To facilitate the decision making for each patient, a model is needed to translate the data collected by the patient’s smartphone into a predicted score for his/her disease severity. To train a robust predictive model, semi-supervised learning (SSL) provides a viable approach by integrating both labeled and unlabeled samples to leverage all the available data from each patient. There are two challenging issues that need to be simultaneously addressed in using SSL for this problem: (1) feature selection from high-dimensional noisy telemonitoring data; and (2) instance selection from many, possibly redundant unlabeled samples. We propose a novel SSL model allowing for simultaneous feature and instance selection, namely the S2SSL model. We present a real-data application of telemonitoring for patients with Parkinson’s Disease using their smartphone-collected activity data such as tapping and speaking. A total of 382 features were extracted from the activity data of each patient. 74 labeled and 563 unlabeled instances from 37 patients were used to train S2SSL. The trained model achieved a high accuracy of 0.828 correlation between the true and predicted disease severity scores on a validation dataset. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Telemonitoring is an emerging health care platform enabled by smartphones and wearables. Because it allows for health data to be collected anytime and anywhere, patients can be frequently monitored and medical decisions can be made more timely and effectively. This paper addresses the data science challenges in leveraging the telemonitoring platform to benefit patient care. Specifically, we propose a new model, S2SSL, to tackle these challenges and provide better robustness, accuracy, and efficiency. This paper may be interesting to health care practitioners seeking advanced analytics capabilities to model and integrate the data collected through telemonitoring devices, with ultimate purposes of improving the decision in treating each patient and increasing patient access to specialized care.

Predicting Parkinson's Disease Severity using Telemonitoring Data and Machine Learning Models: A Principal Component Analysis-based Approach for Remote Healthcare Services during COVID-19 Pandemic

Parkinson's disease (PD) is a progressive and chronic neurological condition that affects about 1% of the world's over-60 population. The COVID-19 pandemic has emphasized the significance of remote healthcare services, such as telemedicine, in managing chronic diseases such as PD. This research intends to construct machine learning (ML) models to predict PD severity utilizing vocal data derived from the UCI database for motor and total Unified Parkinson's disease rating scale (UPDRS) ratings. ML was used to study the association between voice vibration and PD, and PCA and ML models were utilized to minimize model complexity and compare the predictive performance of various statistical models for PD regression. The dataset included 5,875 medical voice records from 42 patients with early-stage PD who participated in a six-month clinical trial. The proposed PCA model simplified the model and achieved a root-mean-square error of 1.78 with an R-squared value of 0.95 for predicting the motor UPDRS and 1.78 with an R-squared value of 0.97 for predicting the total UPDRS. This work can give a framework for developing remote healthcare services for Parkinson's disease and other chronic conditions, which can be helpful during pandemics and other situations where access to in-person care is limited.

Features of Cheyne-Stokes respiration automatically extracted from CPAP airflow signal raw data: Identification of discriminating features to detect heart failure

BackgroundContinuous positive airway pressure (CPAP) telemonitoring data is collected daily from millions of sleep apnea patients. This huge amount of data can be used for the early detection of treatment failures and could reveal incident acute or chronic cardiovascular events. However, the available automatically computed metrics do not fully characterize sleep-disordered breathing. ObjectiveTo find methods to process metrics characterizing Cheyne-Stokes respiration (CSR) that are able to discriminate between CSR related to heart failure (HF) and CSR related to other underlying conditions. MethodsThe raw airflow signals recorded by CPAP devices were analysed. Change point detection methods were used to isolate each respiratory cycle and CSR cycle. Simple algorithms were implemented to extract key features from the signal. Binary logistic regression was performed to identify the characteristics of the CSR airflow signal that is associated with the presence of underlying HF. ResultsLonger CSR cycles, a longer CSR episode, a greater variation in the amplitude of inspiration, a smaller increase in big breaths, a lower inter-cycle variability and a shorter breath duration were features associated with the occurrence of CSR in the context of HF. ConclusionThe proposed automated computation of CSR characteristics presents a novel tool to include in the CPAP software or remote monitoring platforms for monitoring patients’ treatment and comorbidities; and a step towards cross-disciplinary patient management.

Patterns in the Use of Heart Failure Telemonitoring: Post Hoc Analysis of the e-Vita Heart Failure Trial.

Research on the use of home telemonitoring data and adherence to it can provide new insights into telemonitoring for the daily management of patients with heart failure (HF). We described the use of a telemonitoring platform-including remote patient monitoring of blood pressure, pulse, and weight-and the use of the electronic personal health record. Patient characteristics were assessed in both adherent and nonadherent patients to weight transmissions. We used the data of the e-Vita HF study, a 3-arm parallel randomized trial performed in stable patients with HF managed in outpatient clinics in the Netherlands. In this study, data were analyzed from the participants in the intervention arm (ie, e-Vita HF platform). Adherence to weight transmissions was defined as transmitting weight ≥3 times per week for at least 42 weeks during a year. Data from 150 patients (mean age 67, SD 11 years; n=37, 25% female; n=123, 82% self-assessed New York Heart Association class I-II) were analyzed. One-year adherence to weight transmissions was 74% (n=111). Patients adherent to weight transmissions were less often hospitalized for HF in the 6 months before enrollment in the study compared to those who were nonadherent (n=9, 8% vs n=9, 23%; P=.02). The percentage of patients visiting the personal health record dropped steadily over time (n=140, 93% vs n=59, 39% at one year). With univariable analyses, there was no significant correlation between patient characteristics and adherence to weight transmissions. Adherence to remote patient monitoring was high among stable patients with HF and best for weighing; however, adherence decreased over time. Clinical and demographic variables seem not related to adherence to transmitting weight. ClinicalTrials.gov NCT01755988; https://clinicaltrials.gov/ct2/show/NCT01755988.

Factors Associated With Residual Apnea-Hypopnea Index Variability During CPAP Treatment

CPAP is the first-line therapy for OSA. A high or variable residual apnea-hypopnea index (rAHI) reflects treatment failure and potentially is triggered by exacerbation of cardiovascular comorbidities. Previous studies showed that high rAHI and large rAHI variability are associated with underlying comorbidities, OSA characteristics at diagnosis, and CPAP equipment, including mask type and settings. What factors are associated with predefined groups with low to high rAHI variability? This registry-based study included patients with a diagnosis of OSA who were receiving CPAP treatment with at least 90days of CPAP remote monitoring. We applied the hidden Markov model to analyze the day-to-day trajectories of rAHI variability using telemonitoring data. An ordinal logistic regression analysis identified factors associated with a risk of having a higher and more variable rAHI with CPAP treatment. The 1,126 included patients were middle-aged (median age, 66 years; interquartile range [IQR], 57-73 years), predominantly male (n= 791 [70.3%]), and obese (median BMI, 30.6kg/m2 (IQR, 26.8-35.2kg/m2). Three distinct groups of rAHI trajectories were identified using hidden Markov modeling: low rAHI variability (n= 393 [35%]), moderate rAHI variability group (n= 420 [37%]), and high rAHI variability group (n= 313 [28%]). In multivariate analysis, factors associated with high rAHI variability were age, OSA severity at diagnosis, heart failure, opioids and alcohol consumption, mental and behavioral disorders, transient ischemic attack and stroke, an oronasal mask, and level of leaks when using CPAP. Identifying phenotypic traits and factors associated with high rAHI variability will allow early intervention and the development of personalized follow-up pathways for CPAP treatment.

Use of a remote telemonitoring platform significantly improves medication optimisation in heart failure patients

Abstract Background Guideline directed medical therapy (GDMT) has been shown to reduce morbidity and mortality in patients with heart failure with reduced ejection fraction (HFrEF). Despite this, a large number of eligible patients do not receive these treatments or have prolonged delays in achieving optimal doses. Purpose To determine whether a telemonitoring-supported, remote medication optimisation programme could increase the proportion of HFrEF patients reaching maximum tolerated GDMT, in a shorter period of time compared to usual care. Methods A prospective, randomised controlled trial recruited 108 patients with a diagnosis of HFrEF from the ambulatory heart function clinic of a North American cardiac centre. All patients were enrolled onto a non-invasive remote monitoring platform which allowed daily nurse coordinator-led assessment of patient-reported symptoms and trends in heart rate, blood pressure and weight. In the remote titration intervention group, telemonitoring data were used by treating physicians to make decisions on optimisation of GDMT every two weeks, which was enacted by the patient's nurse coordinator, with no physician visit required. Patients in the control group were reviewed in clinic by their treating physician, where medication doses were optimised as per standard of care. The proportion of patients achieving maximum tolerated GDMT, and the time taken for this were compared between groups. Continuous data are presented as mean±standard deviation and compared with Student's t-test, while categorical data are shown as number (%) and compared using the Chi-squared test. Results 108 patients (69.4% male, mean age 54.1±15.4 years) were recruited with a median follow-up of 740 days. Baseline characteristics and medication prescription were similar between groups (56 randomised to remote titration, RT, 52 to usual care, UC, see Table). There were three withdrawals from the RT group and two from the UC group. Significantly more patients in the RT group 52/53 (98.1%) achieved the primary outcome, reaching maximum tolerated GDMT, compared with 42/50 (84.0%) in the UC group (p=0.01). The RT group achieved GDMT earlier (123±70 vs. 183±136 days, p=0.01) with a 40% reduction in clinic visits (p&amp;lt;0.01). In a time-to-event analysis, time to optimisation was significantly shorter in the intervention group (median 105 vs. 165 days, p[log rank] &amp;lt;0.01, see Figure). There was a similar increase in prescription of GDMT in both groups and no differences in hospitalisation or urgent clinic review suggesting that there was no excess hazard of remote titration. Conclusion Remote titration of GDMT in HFrEF patients resulted in more patients achieving maximum tolerated doses, on average two months earlier, with a reduction in clinic visits and no excess adverse outcomes. Telemonitoring-supported remote GDMT titration is effective, safe and could reduce healthcare costs associated with the management of HFrEF. Funding Acknowledgement Type of funding sources: Foundation. Main funding source(s): DHB is supported by a post-doctoral fellowship award from TRANSFORM-HF (Ontario, Canada).

Bayesian Structural Time Series With Synthetic Controls for Evaluating the Impact of Mask Changes in Residual Apnea-Hypopnea Index Telemonitoring Data.

In obstructive sleep apnea patients on continuous positive airway pressure (CPAP) treatment there is growing evidence for a significant impact of the type of mask on the residual apnea-hypopnea index (rAHI). Here, we propose a method for automatically classifying the impact of mask changes on rAHI. From a CPAP telemonitoring database of 3,581 patients, an interrupted time series design was applied to rAHI time series at a patient level to compare the observed rAHI after a mask-change with what would have occurred without the mask-change. rAHI time series before mask changes were modelled using different approaches. Mask changes were classified as: no effect, harmful, beneficial. The best model was chosen based on goodness-of-fit metrics and comparison with blinded classification by an experienced respiratory physician. Bayesian structural time series with synthetic controls was the best approach in terms of agreement with the physician.s classification, with an accuracy of 0.79. Changes from nasal to facial mask were more often harmful than beneficial: 13.4% vs 7.6% (p-value < 0.05), with a clinically relevant increase in average rAHI greater than 8 events/hour in 4.6% of cases. Changes from facial to nasal mask were less often harmful: 6.0% vs 11.4% (p-value < 0.05). We propose an end-to-end method to automatically classify the impact of mask changes over fourteen days after a switchover. The proposed automated analysis of the impact of changes in health device settings or accessories presents a novel tool to include in remote monitoring platforms for raising alerts after harmful interventions.

Hidden Markov model segmentation to demarcate trajectories of residual apnoea-hypopnoea index in CPAP-treated sleep apnoea patients to personalize follow-up and prevent treatment failure.

Continuous positive airway pressure (CPAP), the reference treatment for obstructive sleep apnoea (OSA), is used by millions of individuals worldwide with remote telemonitoring providing daily information on CPAP usage and efficacy, a currently underused resource. Here, we aimed to implement data science methods to provide tools for personalizing follow-up and preventing treatment failure. We analysed telemonitoring data from adults prescribed CPAP treatment. Our primary objective was to use Hidden Markov models (HMMs) to identify the underlying state of treatment efficacy and enable early detection of deterioration. Secondary goals were to identify clusters of rAHI trajectories which need distinct therapeutic strategies. From telemonitoring records of 2860 CPAP-treated patients (age: 66.31 ± 12.92 years, 69.9% male), HMM estimated three states differing in variability within a given state and probability of shifting from one state to another. The daily inferred state informs on the need for a personalized action, while the sequence of states is a predictive indicator of treatment failure. Six clusters of rAHI trajectories were identified ranging from well-controlled patients (cluster 0: 669 (23%); mean rAHI 0.58 ± 0.59 events/h) to the most unstable (cluster 5: 470 (16%); mean rAHI 9.62 ± 5.62 events/h). CPAP adherence was 30 min higher in cluster 0 compared to clusters 4 and 5 (P value < 0.01). This new approach based on HMM might constitute the backbone for deployment of patient-centred CPAP management improving the personalized interpretation of telemonitoring data, identifying individuals for targeted therapy and preventing treatment failure or abandonment. The online version contains supplementary material available at 10.1007/s13167-021-00264-z.

Is 24/7 remote patient management in heart failure necessary? Results of the telemedical emergency service used in the TIM-HF and in the TIM-HF2 trials.

AimsTelemedical emergency services for heart failure (HF) patients are usually provided during business hours. However, many emergencies occur outside of business hours. This study evaluates if a 24/7 telemedical emergency service is needed for the remote management of high‐risk HF patients.Methods and resultsThe study included 1119 patients merged from the TIM‐HF and TIM‐HF2 trials [age 69 ± 11, 73% male, left ventricular ejection fraction 37% ± 13, 557 New York Heart Association (NYHA) II/562 NYHA III]. Patients received a 24/7 physician‐guided emergency service provided by the telemedical centre (TMC) in addition to remote management within business hours. During emergency calls, patient status, symptoms, electronic patient record, and instant telemonitoring data were evaluated by the TMC physician. Following diagnosis, patients were referred for hospital admission or instructed to stay at home. Apart from the TMC, patients could place a call to the public emergency service at any time.Seven hundred sixty‐eight emergency calls were placed over 1383 patient years (0.56 calls/patient year). Five hundred twenty‐six calls (69%) occurred outside business hours. There were 146 (19%) emergency calls for worsening HF, 297 (39%) other cardiovascular, and 325 (42%) non‐cardiac causes, with a similar pattern inside and outside business hours. Of the 1119 patients, 417 (37%) placed at least one emergency call. Patients with NYHA Class III, higher N‐terminal prohormone of brain natriuretic peptide (>1.400 pg/mL) levels, ischaemic aetiology of HF, implanted defibrillator, and impaired renal function had a higher probability of placing emergency calls. During study follow‐up, patients who made an emergency call had a higher all‐cause mortality (22% vs. 11%, P = 0.007 in TIM‐HF; 16% vs. 4%, P < 0.001 in TIM‐HF2) and more unplanned hospitalizations (324 vs. 162, P < 0.001 in TIM‐HF; 545 vs. 180, P < 0.001 in TIM‐HF2). Of the total 1,211 unplanned hospital admissions, 492 (41%) were initiated by a patient emergency call.Three hundred seventy‐nine calls (49%) were placed to the TMC, whereas 389 calls (51%) were made to the public emergency service. Three hundred twenty‐six (84%) of the calls to the public emergency service resulted in acute hospitalizations.The TMC initiated 202 (53%) hospital admissions; 177 (47%) patients were advised to stay at home. All patients that remained at home were alive during a prespecified safety period of 7 days post‐call. Diagnoses made by the TMC physician were confirmed in 83% of cases by the hospital.ConclusionA telemedical emergency service for high‐risk HF patients is safe and should operate 24/7 to reduce unplanned hospitalizations. Emergency calls could be considered as a marker for higher morbidity and mortality.

A Constructive Fuzzy Representation Model for Heart Data Classification.

The early detection of Heart Disease (HD) and the prediction of Heart Failure (HF) via telemonitoring and can contribute to the reduction of patients' mortality and morbidity as well as to the reduction of respective treatment costs. In this study we propose a novel classification model based on fuzzy logic applied in the context of HD detection and HF prediction. The proposed model considers that data can be represented by fuzzy phrases constructed from fuzzy words, which are fuzzy sets derived from data. Advantages of this approach include the robustness of data classification, as well as an intuitive way for feature selection. The accuracy of the proposed model is investigated on real home telemonitoring data and a publicly available dataset from UCI.

463 The Effect of Intervention of Medical Staff on the Adherence of CPAP and Heart Rate in Patients with Obstructive Sleep Apnea

Abstract Introduction Recent studies have indicated that obstructive sleep apnea syndrome (OSAS) is associated with hypertension. OSA is a common cause of sympathetic nervous activity. Increase of sympathetic nervous activity causes hypertension. Continuous positive airway pressure (CPAP) is the most useful treatment for OSAS. Good CPAP adherence treatment improve the risk of hypertension. This study examined the effect of intervention of medical staff on the adherence of CPAP, heart rate and sleep stages in patients with OSA. Methods All patients diagnosed with OSA and undergoing subsequent CPAP were clinically followed for 12 months to examine CPAP adherence, as well as longitudinal changes in blood pressure, average heartrate of 24 hours and sleep stages. They were divided into 2 groups, Group A: patients who had individual consulted in person by sleep physician and technicians before start using CPAP and Group B: patents who did not have individual consulted. Patients in both groups were consulted by sleep physician and technicians after start CPAP with utilizing tele-monitoring. If the adherence were poor, the patients were recommended to stop CPAP. We provided 3D accelerometer and an optical pulse photoplethysmography to all the patients and analyzed the data of heart rate and sleep stages. Results A total of 30 OSA patients underwent CPAP, were enrolled in the study and assessed for changes in mean heart rate and body weight during the study period. We found a significant reduction in mean heart rate in both group A and B compared with baseline (p&amp;lt;-0.05). The patients aged under 50 years old and whose AHI&amp;lt;20 times/hour have higher ratio of dropout CPAP therapy. There was no significant difference between Group A and Group B on the persistency rate of CPAP therapy. Also, no significant association was found between group A and B on the adherence of CPAP. Conclusion We showed the importance of the effect of intervention of medical staff on the adherence of CPAP and heart rate in patients with OSA the consultation after starting CPAP for a while with utilizing tele-monitoring data would be more effective compared with that in person before start using CPAP. Support (if any):

Hidden Markov Model Segmentation to Demarcate Trajectories of Residual Apnoea-Hypopnoea Index in CPAP-Treated Sleep Apnoea Patients: The New Concept of a Telemonitogram

Background: Continuous positive airway pressure (CPAP), the reference treatment for obstructive sleep apnoea (OSA), is used by millions of individuals worldwide with remote telemonitoring providing daily information on CPAP usage and efficacy, a currently underused resource. Here, we aimed to implement state-of-the-art data science methods to describe heterogeneity and diversity of time-series of residual apnoea-hypopnoea indexes (rAHI) from CPAP telemonitoring. Methods: We analysed a CPAP telemonitoring database to model and cluster rAHI trajectories. Our primary objective was to use Hidden Markov models (HMMs) as a probabilistic model-based approach to extract features from rAHI time-series. Secondary goals were to identify clusters of rAHI trajectories and their relation to CPAP treatment outcomes, adherence and leaks. Findings: From telemonitoring records of 2,860 CPAP-treated patients (age: 66·31 ± 12·92 years, 69·9% male), HMM modelling revealed three states differing in variability within a given state and probability of shifting from one state to another. Six clusters of rAHI trajectories were identified ranging from well controlled CPAP-treated patients (Cluster 0: 669 (23%); mean rAHI 0·58 ± 0·59 events/hour) to the most unstable (Cluster 5: 470 (16%); mean rAHI 9·62 ± 5·62 events/hour). CPAP adherence was 30 minutes higher in cluster 0 compared to clusters 4 and 5 (p-value <0·01). Leaks were significantly more frequent in cluster 5. Interpretation: This new approach based on HMM might constitute the backbone for deployment of digital health solutions improving the interpretation of telemonitoring data from CPAP-treated patients. Funding Information: JLP, SB, and RT are supported by the French National Research Agency in the framework of the Investissements d’avenir” program (ANR-15-IDEX-02) and the “e-health and integrated care and trajectories medicine and MIAI artificial intelligence” Chairs of excellence from the Grenoble Alpes University Foundation (ANR-19-P3IA-0003). AM is supported by Probayes and MIAI in the framework of a “Convention Industrielle de Formation par la Recherche” (CIFRE) PhD. The PhD is also supported by the French National Research Agency (grant 2020/0007). Declaration of Interests: JLP reports grants from Air Liquide Foundation, grants and personal fees from Agiradom, grants and personal fees from AstraZeneca, grants from Fisher and Paykel, grants from Mutualia, grants and personal fees from Philips, grants and personal fees from Resmed, grants from Vitalaire, grants from Boehringer Ingelheim, grants from Jazz Pharmaceuticals, grants from Night Balance, grants from Sefam, outside the submitted work; JCB is employee of AGiR-a-dom, a French home-care provider; RT reports other from AGiR-a-dom, grants from Resmed, outside the submitted work; AM, RLH and MCS report other from AGiR-a-dom, and SB have no conflict of interest directly related to this work. Ethics Approval Statement: We analysed a CPAP telemonitoring database [registered and ethically approved by the French C.C.T.I.R.S: N°15.925bis and ethics regulations MR003 N° 1996650v0] to model and cluster patients’ rAHI trajectories.

Positive Airway Pressure Therapy Adherence with Mask Resupply: A Propensity-Matched Analysis.

There are currently few data on the impact of mask resupply on longer-term adherence to positive airway pressure (PAP) therapy. This retrospective analysis investigated the effects of mask/mask cushion resupply on the adherence to PAP versus no resupply. Deidentified patient billing data for PAP supply items were merged with telemonitoring data from Cloud-connected AirSense 10/AirCurve 10 devices via AirViewTM (ResMed). Eligible patients started PAP between 1 July 2014 and 17 June 2016, had ≥360 days of PAP device data, and achieved initial U.S. Medicare adherence criteria. Patients who received a resupply of mask systems/cushions (resupply group) were propensity-score-matched with those not receiving any mask/cushion resupply (control group). A total of 100,370 patients were included. From days 91 to 360, the mean device usage was 5.6 and 4.5 h/night in the resupply and control groups, respectively (p < 0.0001). The proportion of patients with a mean device usage ≥4 h/night was significantly higher in the resupply group versus the control group (77% vs. 59%; p < 0.0001). The therapy termination rate was significantly lower in the resupply group versus the control group (14.7% vs. 31.9%; p < 0.0001); there was a trend toward lower therapy termination rates as the number of resupplies increased. The replacement of mask interface components was associated with better longer-term adherence to PAP therapy versus no resupply.

Detecting COVID-19 and other respiratory infections in obstructive sleep apnoea patients through CPAP device telemonitoring

ObjectiveThe earliest possible detection of individuals with COVID-19 has been essential to curb the spread of infection. Existing digital tools have been scaled up to address this issue. Every night telemonitoring data on continuous positive airway pressure (CPAP) device use, the first-line therapy for obstructive sleep apnoea (OSA), is collected worldwide. We asked whether the changes in CPAP adherence patterns of might constitute an alert for COVID-19.MethodsWe analysed preliminary results of telemonitoring data, recorded between February 1 and April 30, 2020, on OSA patients followed by our sleep clinics and diagnosed with COVID-19.ResultsCPAP telemonitoring data from the first 19 patients diagnosed with COVID-19 showed a clear decrease or halt in adherence in the 20 days immediately preceding COVID-19 diagnosis compared to an earlier period (p < 0.01).ConclusionPatterns of continuous positive airway pressure device use by obstructive sleep apnoea patients collected through telemonitoring can indicate the onset of COVID-19 symptoms. Existing telemonitoring platforms could be immediately used to screen for COVID-19, and for other respiratory infections, in this large at-risk population.