Remote Monitoring in Telehealth: Advancements, Feasibility and Implications

Abstract

Over the past several decades, telehealth has evolved within various medical fields, gaining momentum with sequential technological advancements. The development of remote monitoring specifically expands the function of telehealth by facilitating the ongoing review of patients remotely. Through leveraging of technologies such as wearable sensors, mobile apps, and implantable devices, remote patient monitoring (RPM) enables the collection of biometric data for clinical decision-making. The utilisation of decision-making algorithms in addition to this can flag patient deterioration prompting for a clinician review. This narrative review summarises disease-specific applications, patient and clinician perspectives, and potential future acute care applications, highlighting RPM as a promising tool that, when combined with telehealth, could revolutionise healthcare delivery in the near future.

1. Introduction

Telehealth has been used for several decades within various fields of medical practice. However, progressive technological advancements, particularly in the development of secure videoconferencing, the Internet of Things and cloud computing now provide us with the option of managing appropriately selected patients outside the traditional hospital or clinic setting. This, in combination with demand pressures exerted by the COVID-19 pandemic, has facilitated a greater acceptance of remote delivery of care by both clinicians and patients and has paved the way for more sophisticated telehealth care models.

Leveraging these technologies presents not only the opportunity to enhance patient experience by facilitating hospital level care at a patient’s residence but also potentially leads to improved patient outcomes, because of better access to a wide range of expertise. Telehealth also potentially reduces the negative consequences of hospital admissions (e.g. nosocomial infections, delirium in the patient with cognitive impairments). Further, by mitigating the barriers of distance on a health service level, the integration of these tools allows for operational efficiency and greater scalability opportunities.

Previous experience suggests that the use of these technologies is a safe means of delivering care. This narrative review aims to provide some insight into the potential applications of remote patient monitoring in conjunction with telehealth, based on current published literature.

2. Remote patient monitoring interventions

Remote patient monitoring (RPM) refers to the collection of biometric data without an in-person clinical team present, to inform clinical decision-making. Utilising technology, health-related data from an individual patient can be transferred from their location to a healthcare provider situated elsewhere. This process employs various devices, such as wearable sensors, mobile apps, or medical instruments to track vital signs and other health metrics. RPM facilitates early detection of deterioration, thus enabling the initiation of timely care.

2.1 Self-monitoring

Self-monitoring refers to technologies that enable patients to manually record clinical parameters. A patient may receive a monitoring device and is then prompted to record data. Objective variables such as heart rate and weight have been utilised as clinical parameters [1]. Additionally, teleconsultation may be used in conjunction with self-monitoring, by requesting patients to report data as part of the consultation.

2.2 Non-invasive monitoring

Non-invasive continuous monitoring encompasses technologies or devices that monitor one or more clinical parameters and transmit this data to treating teams. These teams may either review the data regularly or be alerted when a variable sits outside a specified range.

Wearable technologies capable of such monitoring include smartwatches, handheld devices and skin patches. A diverse range of parameters can be measured using remote non-invasive monitoring, with the most common being vital signs such as oxygen saturations, heart rate, blood pressure.

2.3 Invasive monitoring

Invasive monitoring involves devices which have been implanted within the patient. A notable example is implantable haemodynamic monitors. Elevations in left ventricular filling pressures and pulmonary artery pressures are closely correlated with clinical congestion, functional limitation and prognosis in patients with heart failure [2]. Invasive continuous monitoring devices can be integrated into other devices already used for managing heart failure, such as cardiac resynchronisation therapy devices and pacemakers. Additionally, these monitors can also be employed for heart rhythm monitoring.

3. Decision making algorithms

Remote monitoring frequently employs algorithms to support data analysis. These algorithms can be divided into two categories. Firstly, pre-processing algorithms extract meaningful information from physiological signals (such as removing erratic spikes in heart rate data caused my movement artefacts or identifying patterns indicating irregular heartbeats). Secondly, inference algorithms convert clinical variables into alarms and determine risk probabilities [3].

Inference algorithms are further categorised into three types:

• Heuristic threshold-based algorithms: these use predefined thresholds or rules to detect anomalies or trigger alerts based on specific criteria (e.g. a range of acceptable heart rates). While simple and easy to implement, they lack adaptability to individual patients and may miss complex patterns or gradual changes.

• Non-sequential models: employing machine learning, these algorithms analyse data without considering the sequential order of events. They predict outcomes or infer patterns without taking time dependencies into account. Capable of handling complex relationships between various patient parameters, they offer predictions on outcomes but may not capture changing trends associated with clinical deterioration.

• Sequential models: utilising machine learning in a more sophisticated manner, these models specifically analyse time-dependent patterns and relationships. While they can be more complex to implement, they offer a more nuanced analysis of sequential data.

Inference algorithms can be employed to trigger alerts about patient decompensation by analysing the physiological data obtained from remote patient monitoring devices.

4. Disease specific applications

4.1 Cardiac applications

The cardiac diseases most extensively evaluated to date in the context of remote monitoring are arrhythmias and heart failure [4]. The capability to measure respiratory rate, heart rate, electrocardiogaphy (ECG) and oxygen saturations through remote monitoring broadens the application to a variety of cardiac conditions.

4.1.1 Arrhythmias

Remote monitoring for arrhythmias has existed for some time, initially with Holter monitors and then implantable loop recorders. Recently, developments include mobile phone-based photoplethysmography (PPG) and hand-held ECG recorders, which are used to monitor heart rate and heart rate variability.

PPG sensors detect variations in light intensity through transmission or reflection from tissue, with these variations correlating to changes in blood perfusion. Hand-held ECG recorders require an additional external sensor unit and enable recording of a standard lead I ECG.

The most studied devices for arrythmia monitoring are handheld ECG recorders and mobile phone-based PPG which yield sensitivities and specificities of over 90% in detecting atrial fibrillation [5, 6]. Some devices have yielded particularly impressive results. For instance, a study by Lau et al. involving the AliveCor ECG recorder reported a sensitivity of 87% and a specificity of 97% in detecting atrial fibrillation, compared with a cardiologist’s interpretation. Modifying the algorithm to emphasise the absence of P waves improved sensitivity to 97%, maintaining the same level of specificity [7].

The Apple Heart Study, involving 419,297 self-enrolled participants, utilised the Apple Watch and Heart Study iOS app. Notifications were based on irregular pulse detection via PPG. Subjects were notified to contact the study physician via telehealth to determine the necessity of further arrhythmia work up. Those with urgent symptoms were directed to attend in person emergency care, however those that were felt to not require urgent treatment were sent an ECG patch to wear for 7 days which was subsequently reviewed by trained technicians. 0.52% of participants received a notification and of the 450 participants who returned ECG patches, 34% of participants were found to have atrial fibrillation. Notably, in those who had been notified of an irregular pulse, the positive predictive value was 0.84 for observing atrial fibrillation on the ECG simultaneously when a subsequent irregular pulse was picked up by PPG [8]. Only half of those who had been notified of an irregular rhythm failed to contact the study clinician. This study highlighted the need for accessible follow-up care to enhance the utility of remote monitoring in screening.

Many cardiac implantable electronic devices (CIED) like permanent pacemakers, implantable cardioverter-defibrillators (ICDs) and implantable cardiac monitors are now able to provide remote monitoring, enabling the detection of life-threatening arrhythmias such as ventricular tachyarrhythmias, along with enabling a timely reaction to shocks delivered by ICDs (be they appropriate or inappropriate). Pacemaker and ICD rhythms can be monitored through sensing electrical activity by leads attached to cardiac tissue. With insertable cardiac monitors (ICM), far field subcutaneous electrocardiograms are obtained by sensing from a subcutaneous electrode.

Dual chamber devices can facilitate the early recognition of atrial fibrillation (AF) or atrial flutter in those at risk, such as patients with heart failure. Atrial arrhythmias pose a risk in heart failure patients with CIEDs due to risks of systemic thromboembolism, inappropriate shocks, or decompensation [9]. Episodes picked up as potentially being AF or Atrial Flutter are termed Atrial High-Rate episodes (AHREs). Among 2718 individuals with heart failure and CIEDs, AHREs were picked up in 34.8% of patients, with 91% of these being confirmed as AF [10]. Similarly, in a observational study involving 304 heart failure patients with CRT, 57.9% of patients had AHREs detected within 2.5 years post implantation, with 89.2% of these being confirmed as AF on follow up [11]. However, there is the potential to pick up short “sub-clinical” episodes of AF and the risk of thromboembolic events with such episodes has yet to be comprehensively evaluated. Consequently, a consensus on the appropriate timing for anticoagulation in such cases has yet to be established [12].

The traditional management of patients with CIEDs involves scheduled in-person device interrogation combined with the event-triggered alerts from the device. Advanced adaptations in CIEDs now allow remote interrogation through wand-based radiofrequency data transfer via home transceivers through phone lines or cellular connections to a remote server. Most recently low-energy Bluetooth allows coupling the CIED with the patient’s smartphone, allowing the smartphone to act as the transceiver [13]. This not only facilitates data transmission but also provides information on device functionality, such as battery status or electrode function [13].

4.1.2 Heart failure

Early iterations of remote monitoring in heart failure primarily focused on non-invasive methods, especially monitoring vital parameters such as weight, blood pressure and heart rate [14]. However, CIEDs now have the capability to monitor parameters like intracardiac filling pressures or their surrogates, detecting haemodynamic congestion before clinical decompensation occurs.

Weight gain is a well-established clinical parameter for monitoring exacerbations of heart failure, with guidelines [15, 16] recommending daily body weight monitoring for these patients. Educating patients on the importance of daily weight recordings and encouraging regular compliance has long been considered crucial.

External blood pressure monitoring alone has not shown accuracy in predicting heart failure exacerbation [17, 18, 19, 20]. However, some studies [21, 22], have demonstrated that combining blood pressure data with weight measurements can yield more accurate predictive models.

Heart rate and respiratory rate monitoring are standard practices for heart failure patients. A higher heart rate has been associated with increased hospitalizations due to heart failure [23] and Goetze et al. [24] reported significant increases in daily minimum, maximum, and median respiratory rates in the 30 days preceding a heart failure decompensation event.

Studies using multiple non-invasive parameters to predict decompensation have shown varied performances. Differences in the parameters used and associated algorithms contribute to some heterogeneity. The effectiveness of these monitoring systems is significantly influenced by patient compliance with self-monitoring [25]. Mobile technologies could aid compliance through programmed reminders and electronic charting of measurements via Bluetooth-capable devices.

Expert consensus recommends remote follow up with CIEDs for all patients with devices implanted for heart failure [26]. A recent meta-analysis compared the outcomes of remote haemodynamic assessment versus standard care in patients with heart failure [27]. It found that in the remote monitoring group containing 7733 patients compared to the control group of 7567 there was a 32% lower risk of CHF-related hospitalisation (RR 0.68, 95% CI 0.87–1.07, p < 0.001), however there was no significant difference between the group in terms of all cause mortality (RR 0.97, 95% CI 0.87–1.07, p = 0.53). These results did not specify the outcomes for those receiving both non-invasive and haemodynamic monitoring simultaneously. Further assessment of trend detection algorithms utilising a combination of invasive and non-invasive parameters might offer greater predictive accuracy for decompensation episodes.

4.1.3 Myocardial infarction

The ALERTS trial [28] employed an implanted device connected to a right ventricle apical pacemaker lead. The system was designed to store and capture electrocardiogram data and detect rapid progressive ST-segment shifts relative to a subjected baseline. Upon detecting changes, the device would alert the patient to seek medical attention through a vibratory alarm, along with a visual and auditory alarm transmitted wirelessly to a pager. When comparing outcomes for patients who had active alerts turned on compared to controls who had alerts disabled, there was no significant difference in the occurrence of pre-set primary endpoints (cardiac or unexplained death, new Q-wave myocardial infarction or detection to presentations > 2 hours). This result may be partly attributed to the low number of cardiac events in the study. However, 96.7% of the participants experienced no system related complications, suggesting that early detection of myocardial infarction could be another potential safe application for implantable devices.

4.2 Respiratory applications

4.2.1 Chronic Obstructive Pulmonary Disease (COPD)

Early recognition of exacerbations in Chronic Obstructive Pulmonary Disease (COPD) can potentially reduce hospitalisations through prompt initiation of treatment. A change in respiratory effort has been shown to be a strong indicator of exacerbations [29] and can be detected through impedance pneumography. This method involves injecting current into the chest tissue using two drive electrodes and measuring the potential difference between them. The difference in potential varies with the impendence of the tissue, which changes with respiration. Alternatively, respiratory rate can be assessed using stretch sensors that detect chest wall movements, or flow thermography, which identifies temperature differences between inhaled and exhaled air.

Pulse oximetry, which provides continuous measurement of peripheral oxygen saturation (SpO₂), is a well-established practice in medicine. The market offers numerous inexpensive and user-friendly devices.

Diagnostic spirometry measures the volumes of air inspired and expired by the lungs. The forced expiratory volume in one second (FEV1 can be a useful indicator in assessing COPD exacerbations. Portable handheld spirometers, now available for remote monitoring, are considered less useful during acute exacerbations [30].

Peak expiratory flow metres are simpler and cheaper than spirometers and determine the maximum speed of forced expiration, rather than the volume of expired air. Low readings from these metres could indicate an acute COPD exacerbation.

A recent systematic review highlighted that most telemonitoring intervention studies for COPD utilised online platforms connected to devices that collect and transmit various parameters such as clinical symptoms and vital signs [31]. Compared to usual care, telemonitoring did not reduce hospital admissions for COPD but did decrease emergency department presentations.

A scoping review focused on telemedicine applications in managing patients post-hospitalisation for COPD exacerbations. The reviewed studies employed varied strategies, combining phone or video consultations with a nurse and ongoing symptom and vital sign monitoring. Out of 27 studies examining hospital readmissions, 18 indicated a reduction, but only 13 of these were statistically significant [32].

4.2.2 Asthma

In asthma management, wearable home monitors can longitudinally measure symptoms, physiological parameters, and assess risk factors. Focusing on environmental triggers such as allergens (e.g., pollen, dust), air pollution, and weather (known exacerbators of asthma) wearable devices equipped with sensors have been developed [33, 34]. These devices detect levels of specific triggers and integrate additional data to alert patients about the risk of an exacerbation. However, there is currently a lack of intervention studies involving such devices.

The use of peak flow meters in asthma management is a topic of ongoing discussion. Peak flow monitoring can potentially enhance self-management, improve trigger identification, and decrease emergency department presentations and hospital admissions [35]. While some guidelines [36] consider peak flow measurements appropriate for confirming asthma exacerbations, reliance on this method alone is generally cautioned against [37]. Remote spirometry systems that record and transmit peak expiratory flow (PEF) readings have been developed and may play a role in remote monitoring [38].

Patient engagement is crucial for effective asthma control. Mobile health applications (apps) are considered useful tools for helping patients monitor symptoms and medication use, as well as for providing education. There are numerous apps designed for self-monitoring, but more sophisticated versions can communicate with inhaler sensors via Bluetooth. These apps collect data on medication usage, which can then be transmitted to healthcare providers [39]. Additionally, they can remind patients to take preventive inhalers [40]. The use of such technology has been shown to improve medication adherence, increase symptom-free days, and has received positive patient feedback [41].

Asthma exacerbations are complex and multidimensional phenomena. Predicting an individual patient’s risk depends on factors like history of exacerbations, lung function, smoking status, and environmental exposures. Machine learning, using data from patient cohorts with previous exacerbations, can develop algorithms to predict exacerbation risks in the near future. A recent meta-analysis [42] looked at 11 studies including 23 prediction models designed to determine if individual patients were at risk of asthma exacerbations. The pooled area under the curve of receiver operating characteristics (AUROC) of 11 studies was 0.80 (95% CI 0.77–0.83). The most common important predictors were systemic steroid use, short acting beta2-agonist use, emergency department visits, age and exacerbation history. Boosting, Logistic regression (LR) and Random Forest (RF) were the top three popular machine learning methods for asthma exacerbation prediction. Boosting was the best performing algorithm with the overall pooled AUROC of boosting being 0.84 (95% CI 0.81–0.87). These results suggest that machine learning models for asthma exacerbation prediction can achieve good discrimination. Applying this to data from remote monitoring could help identify patients requiring closer follow-up and more regular reviews.

4.3 Early post-operative period

In the post-operative phase, patients experience a potent stress response and, during convalescence, often have decreased mobility and are subject to the side effects of medications, such as opioids. Consequently, post-operative patients are at an increased risk of complications including falls, myocardial infarction, and pneumonia. Traditionally, post-operative monitoring has included regular reviews of pain, mood, wound healing, oral intake, fluid balance, and bowel movements. With an increasing focus on facilitating shorter hospital stays, telemedicine combined with remote monitoring offers an opportunity to extend this approach to a broader patient population.

A scoping review by Amin et al. [43] explored published research on the use of wearable devices for remote patient monitoring across a range of surgical disciplines, including general, cardiac, orthopaedic and obstetric procedures. In most of the studies reviewed (23 out of 24), devices employing either accelerometry or pedometer functions were utilised to track physical activity thereby serving as markers of postoperative function. The potential development of remote ambulatory monitoring in this context could enable earlier hospital discharge and improve continuity of care at home, thereby enhancing support during the rehabilitation period.

4.4 Other disease applications for remote patient monitoring

In the existing literature, a significant focus of remote patient monitoring (RPM) lies in cardio-pulmonary diseases. However, RPM has also shown promise in managing other conditions. Beyond monitoring for deterioration, RPM can assist in enhancing control of chronic conditions like diabetes and hypertension. This is achieved by facilitating data collection (e.g., blood sugar or blood pressure readings), which can then be reported back to clinicians or patients themselves [44]. More regular intervention theoretically enables better disease control and improved long-term outcomes [45].

In geriatric care, RPM can play a crucial role in fall detection. Falls among the elderly are a major cause of morbidity and mortality, and early detection may mitigate harm by flagging patients for welfare checks, potentially initially through teleconferencing. These systems often use accelerometers to detect movement patterns associated with falls, such as sudden increases in acceleration followed by rapid deceleration upon impact [46]. Machine learning could further enhance the accuracy of fall detection by analysing accelerometer data.

In obstetric care, especially during the later stages of pregnancy, a high degree of self-monitoring is necessary to identify maternal and fetal complications, such as pre-eclampsia or fetal distress. RPM may improve both safety and convenience in this context. Proposed external devices for monitoring pregnant patients include blood pressure monitors, electronic scales for weight tracking, and electrodes for monitoring uterine contractions and fetal electrocardiography [47].

This review has outlined just a few applications of RPM. The technology holds the potential to assist in managing a variety of conditions beyond the traditional healthcare setting. By tracking vital signs, physical activity, symptom monitoring, and medication adherence, RPM can facilitate care in a patient’s usual residence. This not only enhances convenience from the patient’s perspective but also provides superior care across a broad spectrum of diseases and patient populations. The utility of RPM is further amplified when combined with teleconferencing. As technology advances, additional applications of RPM with more sophisticated models of care are likely to emerge in the near future.

5. Patient and clinician perspectives

Qualitative studies, involving structured interviews with clinicians and patients engaged in telemonitoring for chronic conditions [48, 49] and those in “virtual wards” [50], have provided valuable insights into the patient journey. These insights are crucial for developing more patient-centered interventions. The studies have identified themes highlighting both the benefits and potential drawbacks of remote care via telemonitoring from a patient perspective.

In managing chronic conditions, patients have reported an increased sense of security with telemonitoring. They feel more visible to the healthcare system and safer, knowing that their condition is being monitored and help is available if needed. Despite the initial burden of self-monitoring and reporting, many patients find that these activities become routine and part of their daily schedule over time. In some patient groups there is a potential for apprehension regarding the use of technology if they are unfamiliar, however some are able to overcome this with experience. Patients have reported that whilst they felt that telemonitoring provided better access to care, they have some concerns that over reliance on technology might reduce direct clinician contact and that the desire to interact with a professional remains.

Both clinicians and patients report that the process of self-monitoring helps patients understand their conditions better, empowering them to manage their own health more effectively. In particular it enabled them to identify and mitigate deteriorations in their conditions. Nonetheless, some patients have experienced anxiety due to alerts, in particular when there were notifications of deterioration.

Regarding the “virtual ward” model, patients have expressed greater satisfaction due to proximity to friends and family. This model is especially beneficial for those with access to informal social support. Conversely, the virtual ward model was perceived to be more challenging to patients lacking such support. Caution must be taken when admitting patients to virtual wards, particularly if they might find using the required technology difficult.

6. Potential future acute care applications

While the majority of existing literature on remote patient monitoring (RPM) focuses on chronic condition management, the COVID-19 pandemic has highlighted its potential in acute care. The strain on healthcare resources, including a lack of inpatient bed capacity and heightened risk of disease transmission, led to an expansion in the use of “virtual wards,” where patients remain in their usual residence and are monitored through online platforms, telephone calls and wearable sensors. COVID-19 has been the acute condition to which virtual wards have been applied most commonly [51]. These wards enable monitoring for patient deterioration, triggering unscheduled tele-conferencing reviews and decisions about inpatient admission or alternative care escalation (e.g., community outreach teams, medication dispensing). Although this model of care has yet to be extensively applied to other conditions, it does provide a potential means to address some of the issues facing healthcare systems worldwide.

Applying this model to other conditions could address some global healthcare system challenges. Emergency Department overcrowding is a widespread problem and the associated effects on morbidity and mortality as well as staff burnout are well documented. Boarding is largely driven by “access block” or the inability to access an appropriate hospital bed for an admitted patient. Telehealth in combination with RPM would allow more patients to be managed at home, without the need to take up an inpatient bed, thus potentially reducing the strain on the healthcare system as a whole.

RPM could also enhance care in lower acuity settings, such as rural hospitals or residential aged care facilities. By remotely monitoring patients, the burden of observations on local staff is reduced, thereby easing their workload and enabling the care of higher acuity patients within these environments. Telehealth also allows such centres to access expertise that might otherwise only be available in urban centres, this prevents the need for costly transfers for some patients.

In urban centers, RPM can be beneficial not only in lower acuity areas but also in high-demand situations where patient safety is at risk. Emergency Department waiting rooms might be made safer through being able to continuously monitor patient’s vital signs with wearable devices.

7. Future research priorities

New remote patient monitoring devices continue to be developed however the utility of such technology in a real-world setting is dependent on multiple factors, including the reliability of network coverage, user acceptability and cost-effectiveness. Wearable devices, particularly in the form of smart watches may be appropriate devices for the purposes of remote patient monitoring. The latest commercially available smart watches are able to measure oxygen saturations, heart rate, ECGs, blood pressure and alert for falls using built in accelerometers and gyroscopes. If reliable, these provide the means to assess patients and detect deterioration in the same what that conventionally obtained patient vital signs might do in the hospital setting.

Whilst there have been some studies looking into the accuracy of such devices [52], there has been little done to look into their reliability when being used in a clinical setting and little research into the ability of these devices to improve patient outcomes [53].

Future research should be focused on translational research looking into the application of these technologies into models of care which supplement monitoring with ongoing remote consultations. Implementation of such technology should be accompanied by robust, well rounded evaluations which include both qualitative and quantitative components assessing patient and clinician experience as well as determining patient outcomes. Validation of the reliability of data obtained from wearable devices in the clinical setting would be the first step prior to utilisation within a pilot. Graded expansion of telehealth systems making use of remote patient monitoring can then progress if evaluations are shown to be of benefit, again accompanied by robust evaluation with each stage of expansion. In this way, mature telehealth systems can be developed which can gradually increase the range of conditions covered in both the acute and outpatient setting, whilst also providing the infrastructure to which more sophisticated remote monitoring technologies might be introduced, making use of more complex decision-making algorithms and additional sensors.

8. Conclusions

There has been a significant increase in literature surrounding the use of remote patient monitoring (RPM). These studies have focused on chronic disease management and a limited range of medical conditions. RPM has demonstrated a facility to enhance tighter control of chronic diseases by allowing more regular review of relevant parameters. It also has been shown to be effective in early detection of deterioration allowing for timely intervention. To date a majority of studies related to RPM are proof of concept or pilot studies with relatively few studies operating within mature telehealth models of care.

In conjunction with telehealth, developments in remote monitoring devices hold the potential to cause a paradigm shift in health care. Many conditions that would have traditionally required inpatient care can be managed safely within a patient’s own residence. However, aside from management of COVID-19, the literature has not concentrated on the application of RPM to acute conditions. Given the current strain on many health systems, further research into the utility of virtual wards is warranted. Future analyses should not only consider patient outcomes but also encompass patient satisfaction and cost analysis to justify the implementation of such care models.

RPM represents an exciting and evolving tool in healthcare delivery. While further studies looking into patient outcomes as well as the broader application of RPM are required, existing literature shows significant promise and the technology has the potential to revolutionise the delivery of healthcare going forward.

Conflict of interest

The authors declare no conflict of interest.