Telemonitoring System for Early Diagnosis of COPD Exacerbations.

A tailored management of COPD patients would obviously allow to reduce costs for hospitalizations and improve quality of life. This management could benefit of the Information and Comunication Technology support, which can offer the possibility of telemonitoring patients without the need of repeated hospital visits and improving the efficacy of healthcare services. Moreover, the high frequency of exacerbations and their often atypical clinical presentation in the aged patient make particularly desirable the availability of a telemonitoring system which could guarantee continuous control and early intervention in case of necessity. The aim of the present study is to test an innovative telemonitoring system in patients with COPD.

Chronic obstructive pulmonary disease (COPD) is a global health problem throughout the world. It's not only the fourth cause of death with 600 million deaths every year, but it's also the most rapidly increasing pathology in terms of mortality in industrialized countries and in 2020 it will become the second cause of death (WHO data). Moreover it's impact is largely underestimated. Only one patient on four is diagnosed, often with delay, and this obviously reduces therapeutic chance. Consequently, COPD is often identified and cured only in late stages, while it constitutes an important health problem also in younger people (from 45-50 years). Recent studies estimated an average cost of 1.300 euro per patient every year, which can increase to 7.000 euro in more severe stages. A tailored management of COPD patients would obviously allow to reduce costs for hospitalizations and improve quality of life. This management could benefit of the Information and Comunication Technology support, which can offer the possibility of telemonitoring patients without the need of repeated hospital visits and improving the efficacy of healthcare services. Moreover, the high frequency of exacerbations and their often atypical clinical presentation in the aged patient make particularly desirable the availability of a telemonitoring system which could guarantee continuous control and early intervention in case of necessity. The use of telemonitoring systems in COPD patients has already demonstrated a moderate efficacy in reducing hospidalizations and other related outcomes (Pedone C et al. BMC Health Serv Res 2003), but evidence is not homogeneous (Pedone C e Lelli D Pneumonol Alergol Pol 2015). A further improvement of the impact of telemonitoring systems on health status of COPD patients could derive from Decision Support System (DSS), able to predict clinical events (i.e. exacerbations) and to assist healthcare providers in monitoring or predicting events on the basis of variation of clinical parameters, not directly collected with conventional strategies. Our research group (L.I.A., Università Campus Biomedico di Roma, Asl n. 4 Lanusei) has already developed a predictive algorithm for exacerbations or clinical worrisome events before the onset of symptoms. The project has been financed by Provincia Autonoma di Bolzano (Italy) and presented at the IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI) in Las Vegas on February 24-27th 2016 \["On the Remote Detection of COPD-Related Worrisome Events" - BHI 2016\]. The full paper has been published on the IEEE Journal of Biomedical and Health Informatics \["A decision support system for tele-monitoring COPD-related worrisome events"\] on March 2017. Exploiting our results, an innovative medical device for health status monitoring has been developed to detect exacerbations or clinical worrisome events through the consideration of early signs the patient is not still aware of and, accordingly, suggesting him to refer to doctors before the onset of symptoms. The device is made up of a Bluetooth digital pulse oximeter and of an APP. The APP connects with the pulse oximeter that measure heart rate and blood oxygen saturation. The APP, at first, adapts to the physiological characteristics of each patient, thus personalizing its responses, setting its own coefficients for the analysis of the parameters used on the basis of the continuous flow of data received from the pulse oximeter. The initial period of "adaptation" for the basic setting of the machine has been estimated in 5 days, after which the device will continue to model itself on the patient both through self-learning using the input data, and through a possible, but not obligatory, intervention from medical doctors modifying the configuration parameters for a better response from the machine. Our results suggest the algorithm is able to detect potentially critical situations with a sensitivity index of 98% and a specificity of 100%. 3\. Study objectives To test the algorithm on a larger population to verify the performance of the current system. Based on the number of descriptors used by the system, considering a drop-off rate of 15%, error margin of 0.02, p = 0.05, the number of patients enrolled is 120. To train the algorithm for the recognition of exacerbations, defined as worsening of dyspnoea or sputum of entities superior to the normal circadian variability and requiring treatment modifications, ascertained through clinical evaluation; To calibrate and validate the algorithm for recognition of exacerbation episodes. To evaluate the effectiveness of the system in terms of costs and quality of life of the patient: Efficacy: it will be based on the comparison of the frequency of ER and hospital admissions in the three years before the intervention and during the intervention period in order to verify the expected decline in the need for extraordinary care. To this end, after authorization, the File A and File L of the Scheda di Dimissione Ospedaliera (SDO) will be used. In this way, all the data concerning ER and hospital admissions, causes and days of hospitalization will be retrieved for each patients. A significant reduction in the cumulative frequency of the outcome will be considered proof of efficacy. The comparison between the year of intervention and the previous year will be supplemented by the analysis of the time series including the previous 3 years and the intervention to avoid that an apparently significant difference between the year of intervention and the previous year reflects only the evolution of a trend. Moreover, since detected parameters can be modified also by incident not only respiratory, but also of other origin (e.g. cardiac) pathology, the causes of ER or hospital admissions will be detected and catalogued for the purpose of an analytical evaluation of the effects of the intervention. the analysis of objective outcomes, that are expression of the care needs (efficacy check 1), will be complemented by that of the health effects expressed by a baseline and final oligodimensional evaluation including a measure of dyspnea (MRC), a physical performance index ( 6'WT), a personal autonomy scale (ADL, IADL) and a health status questionnaire correlated with respiratory disease (CAT). Since there is no comparison for this outcome, the same will be considered achieved in the case of stability of performance compared to baseline. Cost/effectiveness: assuming a weighted average cost of hospitalization for exacerbated COPD and estimating the cost of the proposed intervention (see the attached analytical scheme), the intervention will be considered cost / effective if the savings achieved will exceed at least 25% the expenditure incurred. This threshold, albeit arbitrary, is convenient and reasonable because lower values could be penalized by variously localized defects in the procedure for calculating the variables involved. Patients with COPD with frequent exacerbations will be enrolled regardless from bronchial obstruction degree or GOLD stage. According to the 2018 GOLD guidelines, two or more exacerbations every year or at least one leading to hospitalization are considered frequent. Exacerbation means any change in respiratory health that requires a change in drug therapy, regardless of the need for ER or hospital admission, which conversely characterize the proportion of exacerbations constituting the first efficacy outcome. Patients in long-term oxygen therapy, with cognitive impairment hampering, in the opinion of the investigator, the possibility of adhering to the protocol and with a life expectancy limited to one year, will be excluded. No other exclusion criteria will be foreseen, in particular co-morbidity, since the study is designed for a real life dimension and the algorithm can be used to convey information of interest for worsening of non-respiratory pathology. Each patient will be provided with the BPCOmedia Kit for measuring the status of patient's health COPD. The kit is composed of a Bluetooth pulse oximeter and an APP for Android system downloadable from Google Play and installed on the Android smartphone of the patient from version 4.1 on. During sperimentation, a dedicated case manager nurse will care, for the professional team leader (GP) and for the entire team where the specialist pneumologist is present, an impartial and transversal approach for the development of a personalized treatment plan, facilitating the coordination and appropriate use of the various services. Measurements of oxyhemoglobinic saturation and heart rate will be performed 3 times a day, according to a predefined monitoring plan, and whenever the patient experiences symptoms. The patient will firstly refer to the case manager nurse or to the GP, if the first is unavailable, whenever the algorithm will detect a potentially dangerous situation or when its perceived health status will deviate from measurement performed by the telemonitoring system. On the other hand, the case manager nurse and the GP will daily monitor the parameters and will evaluate if the received values are compatible with an ongoing or imminent deterioration of the state of health ("worrisome event"). In this case, the patient will be contacted by telephone to investigate the symptoms or the presence of signs of exacerbation (dyspnoea, asthenia, increased secretions). On the basis of the obtained data, the worrisome event will be reclassified as follows: False alarm; Minor event (clinical worsening not requiring therapeutical intervention or other interventions) Possible exacerbation. In case of possible exacerbation, the patient will be visited by the GP who will establish the actual presence of an exacerbation (with consequent modification of therapy, indication of further diagnostic investigations, or sending to higher levels of treatment, depending on the indication), or reclassify the event as "Minor Event". The pneumologist will be always available for the interpretation of situations involving specialist knowledge and will be contacted by both the case manager nurse and the GP depending on availability. In case of particular situations, for example important desaturations, preferential routes will be activated so that the specialist pneumologist can intervene promptly. In order to obtain a sufficient number of events for the objectives of the project, it is planned to enroll at least 120 patients, who will be followed for a period of 12 months. The main outcome will be bronchial exacerbations, which are assumed to have a Poisson distribution with an average of 2/patient/year. Considering as clinically significant a reduction of 0.25 exacerbations/patient/year, using Lehr's equation with Type I and Type II error of 5% and 20%, respectively, we obtain an estimate of the sample size of 480 people followed for a year; it is considered appropriate to increase to 500 to offset any losses during follow-up. This sample size is also sufficient for the development of the event prediction algorithm: in fact, based on the number of descriptors used by the system and considering a drop-off rate of 15%, error margin of 0.02, p = 0.05 , the number of patients required is 120. The instrumentation will be delivered within a training session in which the patient will be instructed on the functioning of the telemonitoring system and on the behavior to be taken in case of changes in health status. Appropriate understanding will be verified at the end of the session, and the study staff will offer the patient additional sessions even after some time in case of need. The patient will also be specified that the investigating system is not substitutive of the usual assistance and that in case of any variation of the symptoms it is necessary to follow the usual procedures by contacting the primary care physician or the specialist as needed. All patients will sign an informed consent. The telemonitoring system used is not intended as a substitute for normal assistance and the study procedures do not provide emergency interventions, therefore participants will be advised to contact their doctor for the evaluation of any health problems and to activate the usual emergency procedures in case of critical situations. Data analysis Data will be analyzed according to descriptive statistics methods to evaluate the performance, calculating accuracy, precision, specificity and sensitivity. Initially, the training phase will use data from a reference population. Subsequently, using the leave-one person out, the system training will be evaluated again using the data acquired during the present experimentation, then proceeding to the estimate of the performances. 5\. Telemonitoring system Basic characteristics are summarized below: The system consisting of the app and of a pulse oximeter and is certified as a medical device (next June 2017); The patient receives a kit consisting of a pulse oximeter and the Android app that can be downloaded from Google Play and can be installed on your Android smartphone from version 4.1 and up; The patient uses the app to perform measurements at home, without the need for external support; The app works in three modes: configuration, training / calibration, monitoring; During configuration, the user creates an account associated with the purchased pulse oximeter and the app, in order to reduce the abuse of the device by people who have not purchased it; During the training phase, the app acquires the necessary information to calibrate the algorithm and to make it able to identify the occurrence of one of the aforementioned dangerous events; During the monitoring phase the patient uses the app to monitor the onset of one of the aforementioned dangerous events; The user, using the credentials provided during the configuration, can access a web portal called the patient portal where he can check the history of the performed measurements; The patient portal allows the end user to view their data in a graphical and intuitive form and to share measurements with their doctor; Doctors access a portal dedicated to them and called the medical portal; In the medical portal, the doctor can consult the measurements performed by his patients provided that the patient has enabled the doctor to consult his data; Possibility for the doctor to obtain reports of the measures performed by the patients; The app and both the patient portal and the medical portal support the main European languages starting from Italian and English. Installation and configuration phase The user downloads the Android app from the Google Play store. At the time of installation, the app guides the user to register a new account, if not already in possession of one, and to register the pulse oximeter purchased, by entering the alphanumeric commercial code available inside the package. A single account can be linked to the commercial code. During the account creation and association with the pulse oximetry wizard, the app requires the minimum information necessary for registration and user profiling: Name and surname Email address Telephone number Username and password In order to proceed with registration, the patient must view the privacy statement and consent to the processing of personal health data as required by current legislation. At the end of the installation / configuration phase the app starts in training / calibration mode. Training and calibration phase For the correct functioning, the app needs an initial training phase of the predefined duration of N = 5 days. The training phase is based on a monitoring plan consisting of three measurements / day divided by predefined time period: morning, afternoon, evening. A measurement is required for each time slot. The monitoring plan includes a series of reminders that, through the push notification associated with a sound signal, remind the patient of the measurements to be performed. The monitoring plan can be consulted in read only on the app in a special section. At each measurement indicated by M the app acquires the value of the percentage of hemoglobin saturation SpO2, the value of the HR heart rate measured by the pulse oximeter and the date and time when the measurement takes place. The input data are formally indicated with the triad (SPO2, HR, t), where the value t indicates both the date and the time of measurement M and therefore serves to classify the band of measurement: morning, afternoon, evening. At the end of the 5 days, ie of the training phase, using the measurements and a dataset of patient profiles pre-stored in the APP, it is automatically passed during the monitoring phase. The APP displays in a special band always clearly visible the phase in which it finds: training or monitoring. During the training phase, the app stores the data of the individual measurements and the result of the calibration on the backend at the end of the training phase. The app also works in an independent way from the backend so the data are synchronized in the background in a transparent way to the user even in the absence of connectivity. Monitoring phase Once the training and calibration phase is complete, the app goes into the monitoring phase in which the patient performs the measures to keep his health checked and monitored. The information to be memorized in this phase are: HR heart rate; SpO2 hemoglobin saturation; Date and Time of detection According to the monitoring plan, the app invites the user to perform the measurement by wearing and turning on the pulse oximeter. Upon detection, the app alerts the user by inviting him to remove the pulse oximeter. The single measurement, completely transparent to the user, is implemented as a series of measures more appropriately filtered (indicatively the average of the measurements performed in 5 - 10 seconds). At each measurement, the app, received the parameters from the pulse oximeter, queries the algorithm by sending SpO2, heart rate, date and time and displays the result of the processing by the algorithm. This result is presented to the patient through easily comprehensible text and with appropriate icons that provide a clear and immediate understanding of the patient's state of health. The app is able to perform the measurement even in the absence of connectivity, so all configuration parameters (see monitoring plan) are stored locally to the app and from time to time synchronized with what is present remotely to the backend. After the measurement, the parameters of the measurement itself and the result of the processing are stored locally and, if possible, on the cloud. In case it is not possible to immediately store them on the cloud due to lack of connection, the alignment of the data will happen in a transparent way to the patient. The app also keeps a measurement history locally so as to allow the patient to consult the measurements directly from the app. These measures are presented in tabular and graphic mode. During the monitoring phase the patient must have the possibility to contact the relative doctor in case of a worsening of the perceived conditions and if he finds a disagreement between his / her perceived condition and the result of the app. It is therefore possible directly from the app to start a new training / training phase to recalibrate the algorithm. Settings The app provides a special section dedicated to personal settings. In this section the patient can consult his personal data, the monitoring plan and perform a reset to restart the training and calibration phase. Algorithm library The algorithm is encapsulated in a library developed by the Biomedical Campus of the University of Rome and owned by the L.I.A of Giuseppe Capasso. Patients' interface Using the credentials chosen during the initial configuration phase, the patient can access a web portal dedicated to patients. After access to this portal the patient has the possibility of: Manage your personal data and change the password for accessing the portal and the app. Consult the standard monitoring plan available as read-only. Consult the list of measurements made and display them in graphical mode, with the possibility to filter and view them for time periods. Set up your own doctor with whom you can then share your monitoring data. The patient invites the doctor to access his data in a manner similar to that used by the most common social networks to share information. Doctors' interface This portal allows treating physicians to access their patients' data via the web. By following the appropriate link 'Register as a doctor on the login page of the medical portal, the doctor can register at the portal. The activation must take place after adequate verification by the administrators, who may request further documentation to ascertain the identity of the doctor requesting registration. Once registered, the doctor can use his credentials to access the portal where he can manage his personal data entered during registration and access the measurements of patients who have shared access. The doctor can in turn search for patients and send a request for access to data. The patient can then accept the request or not. The doctor can not make any changes to the patient's data, he has them read-only. There are known fields where you can enter other information deriving from a medical history or a visit. Doctors' interface - specific features In order to support the achievement of the objectives of the study / experimentation, a series of functionalities are foreseen in order to collect a set of "certain" data, that is validated by the doctor near the detected measure so that the same doctor has the possibility to confront the patient and confirm or redirect the result generated by the algorithm. From a technical point of view it is about collecting the data sent by the app in the backend but allows the doctor to confirm the data or reclassify the result that, according to his opinion, should have been generated. Therefore, as already foreseen, the doctor will access the back end with the possibility to select the patient in question, to examine the measures interested in the judgment confirming the outcome or modifying it according to a series of labels that will be indicated by the scientific committee. The information can then be extracted ad hoc and provided for analysis at the Biomedical Campus. Working scheme The following diagram represents the logical architecture of the whole system. The diagram also shows which parts reside in the smartphone (Smarpthone area) and which are instead ancillary components that reside in the web (Web area). It is important to keep in mind that the medical device consists of the app and the algorithm included in the area marked as Smartphone. The users of the system are the patient and the doctor. The patient uses the smartphone for monitoring and consultation, while the web part is used for simple consultation both by the patient and by the doctor in the appropriate web portals. Principal components are described below. App This represents the app installed in the patient's smartphone. The app receives measurements from the pulse oximetry during the monitoring phase and interacts with the predictive algorithm to determine the patient's health status. The data useful for processing health status and for consultation are then saved in a local database so that you can work independently from the web-side connectivity. The app interacts with the backend for the sole purpose of saving measurements and health data remotely. Predictive algorithm Although it is physically integrated into the app, the algorithm is locally a block in itself, given its criticality. The app takes care of managing the time slots of the measures based on the monitoring plan and to call accordingly the predictive algorithm passing the information collected by the pulse oximeter and other information processed from the historical data stored in the local database. Starting from this information, the algorithm determines the patient's health status. Local database The local database is used by the app to maintain locally the necessary measures in the first instance to the algorithm to perform the processing and to determine the health status of the patient, and secondly to allow the patient to consult the previous measures directly on the smartphone . For the correct processing the algorithm requires the measurements of the last 90 days. These are the measures stored in the local database. Pulse oximeter. It is the medical device used by the patient to perform the measurement of SpO2 and heart rate. Interacts via bluetooth with the app to share the measurement performed. Backend This is the remote hardware / software component on which the measurements made by the various patients and the patient profiles are saved. Here the data are stored in a remote database and are always available to be downloaded on smartphones in case of need, to be consulted by patients and doctors through the appropriate portals, or to be used by researchers of the Rome Biomedical Campus to refine and further develop the algorithm. The following section dedicated to the detailed architecture will also describe the various protocols used for the interaction with the other components of the system. Remote Database Remote database on which patient measurements and patient and medical profile data are saved. Web interface These are the components of the system that allow patients and physicians to consult the data stored in the backend. Structure The following diagram represents the physical architecture of the whole system and how the components of the whole system interact / communicate with each other.