Device based monitoring in digital care and its impact on hospital service use
To our knowledge, this is the first systematic review that provides an overview of the impact of components and various design aspects of DRM on hospital service use. Overall, 72% of the included studies showed a decrease in at least some aspects of hospital service use. By analyzing the components of technology design, patient monitoring, and support, integration of DRM into clinical care, and patient engagement, we were able to further define the crucial aspects within those themes and their reported effect on hospital service use in detail and to provide valuable insights into optimizing DRM. The technology design components of DRM were most frequently reported. Non-implantable devices (e.g., wearables and scales) were the most common device type used to measure data, but were less likely to show a decrease in hospital service use compared to implanted or mobile devices (69% vs 89% and 76%). Providing 24/7 support for patients using DRM was the most common, with 81% showing a decrease in hospital service use. Regarding the integration of DRM into clinical care, the impact of differences in design and healthcare provider involvement was small. Differences in DRM in addition to or replacing usual care and involved healthcare providers mostly vary from 3% to 10%. Active patient engagement in data collection and transmission more effectively reduces hospital service use compared to automated processes (75% vs 72%). Various patient engagement components, such as the patient’s role in data collection or education type, did not differentiate in outcomes.
Non-implantable devices were the most common device type used to measure data, possibly reflecting the state of technology development, but were less likely to show a decrease in hospital service use than implanted or mobile devices. Implanted devices used for DRM are commonly medical-grade devices, while this is not always the case for remote monitoring using non-implantable and mobile devices, e.g., smartwatches, apps with questionnaires, or mobile phone use. Based on the provided information in the studies, we could not make the distinction between medical and non-medical devices. Implanted devices, such as those used to assess arrhythmias, can often be read remotely and are not primarily designed for high-frequency monitoring but to regulate the heart’s rhythm or to decide on medication at fixed consultation times. Non-implantable devices have other purposes, such as detecting deterioration and providing data to manage the condition. In DRM with implantable medical devices, the patient is often not involved in the collection and transmission of data, as this is automated. Previous research stated that data transmission must be reliable, and, if possible, automatic9. While patient engagement increases the impact of DRM and self-management, an active role of the patient in data collection and transmission is more susceptible to errors. Furthermore, implanted medical devices are often used in more severe diseases with a higher risk of hospitalization. This has to be taken into account when listing these studies in an overview. While mobile devices for monitoring are more likely to report a decrease in hospital service use than non-implantable devices, it is interesting to note that mobile devices such as activity trackers, are slightly less effective than wearables (a type of ”non-implantable device”) in predicting patients at high risk of rehospitalization18. However, mobile devices or application-based surveys are sometimes used in combination with non-implantable devices. What is the most simple-, least burdensome, and most appropriate technology to be used in DRM, is an extremely relevant question. In addition, prediction- and intervention models especially for care processes using non-implantable devices warrant further development.
The need for an internet connection did not show any differences in effect on hospital service use compared to DRM where, for example, a telephone line was used instead of the internet, possibly reflecting the state of technology development. It may become more relevant over time and affect other factors such as patients suitable for remote monitoring, timely response, accessibility, and communication with the patient.
Optimally integrating DRM into clinical care presents technical workflow challenges, such as incorporating the data into the EMR8. Most studies (66%) do not have monitoring data available (either integrated or manually) in the EMR, but still tend to show a decrease in hospital service use. However, studies with EMR integration perform better in terms of hospital service use reduction than those requiring manual data entry. These findings support the literature stating that interoperability is necessary for sustainability and for seamless interaction between patients and the healthcare system9,19.
In our review, we found that daily measurement and data transmission of monitoring data are most common but show the lowest percentage of studies measuring a decrease in hospital service use. Half of the studies assess the data continuously or daily, while studies that assess the data less frequently, weekly, and less than weekly, all show a decrease in hospital service use. This contradicts expectations, as higher frequency assessments should theoretically improve outcomes by identifying early warning signs and enabling timely interventions20. Continuous monitoring may involve automated processes with lower error sensitivity, and high-intensity monitoring is often associated with high-risk diseases, anticipated to have a higher reduction of hospital service use when measured in terms of hospitalization9. Of the included studies that assessed data continuously, only 15% focused on high-risk (acute) diseases, possibly explaining why we did not find a large reduction in hospital services. However, those focusing on high-risk diseases did not show remarkably different outcomes.
Alerts to initiate contact between the healthcare provider and the patient (in 63% of studies) did not appear to affect hospital service use. Personalized DRM is given in 41% of studies with no clear impact on hospital service use. This seems counterintuitive and does not support a previous review stating that personalized alert thresholds should be carefully determined to successfully impact the effectiveness of DRM on acute care use9. Optimal and appropriate timing of patients’ contact with a healthcare provider and its frequency thus remain important research issues.
The impact of differences in design and healthcare provider involvement on hospital service use was small, and the influence was reported in very few studies, is probably not well understood and its effect may be underestimated. Until now, it remains unknown whether the effects are caused by the technology of the DRM itself or by other aspects of the implementation21,22. The effect could be linked to broader health systems issues. In cases where the care process was redesigned, in terms of DRM replacing usual care and involving other healthcare providers, the percentage decrease in hospital service use seemed to be higher than in cases where the care process was almost similar to usual care. This suggests that an overall redesign of the care process is a crucial component of the implementation of DRM and its effect, not just the technology itself.
Implementation of DRM has initiated a transformation in traditional healthcare roles, involving patients managing their own conditions and interdisciplinary collaboration in the interpretation of monitoring data8. However, providers lack the required tools or understanding to handle extensive information on clinical parameters. Centralized approaches, such as a virtual care center, have the opportunity to train professionals for these tasks. They offer scalability advantages and streamline healthcare professionals’ focus on core tasks and complex care23. Centers with dedicated professionals to monitor data are a factor in intervention success9. It also explains why they are increasingly implemented among hospitals. However, this review does not differentiate between the clinical expertise levels of virtual care center employees, who may range from specialized nurses in targeted diseases, to employees of a company distributing the technology serving as conduits of information. Such variability in clinical involvement level and expertise may affect the impact on hospital service use.
Moreover, there is a growing trend toward automated data assessment, wherein alerts are generated by either automated scoring tools24 or artificial intelligence systems25, relieving healthcare professionals of certain tasks. The methodology for sorting and analyzing DRM data is evolving, partly due to the use of artificial intelligence systems, and the optimal presentation format that works most efficiently for DRM data is being investigated8.
Face-to-face consultations as a standard communication type in DRM have a smaller impact on hospital service use than other communication types, such as phone or video calls. This might be explained by the delays associated with face-to-face consultations, where scheduling is a time-consuming factor. Another possible explanation is that face-to-face contact indicates a higher urgency of the reason for contact, leading to the need for a hospital visit and hospitalization. However, interactive human-human contact has been shown to facilitate effective DRM26,27.
In only 25% of the studies monitoring data was accessible to patients. Of these studies, 70% show a decrease in hospital service use. The few studies reporting that patients did not have access to their data (n = 3) all show a decrease in hospital service use, which is not consistent with the existing literature. Most research indicates that access to data provides patients with a better understanding of their health, helps them to be proactive in meeting health goals, and increases the benefit from DRM8. This aligns with the majority of papers found in our review. Patients also need data literacy through education, trust in the systems that manage it, and shared decision-making to fully benefit from DRM8,9. While self-management is an important aspect of DRM, it is not a standalone solution for improving quality of life, self-efficacy, and self-care. Previous research demonstrated that when DRM is integrated into the self-management process, significant improvements occur26.
Education is essential for improving digital literacy among patients, which will ultimately contribute to narrowing the digital divide. This review revealed that education for patients is available in at least 77% of the studies included. Another review performed in 2020 showed that 18% of the studies used customized education, with the majority of education occurring in DRM using non-implantable devices17. Thus, there seems to be more awareness of appropriate education.
According to a previous review, automated data transmission is preferred for a positive effect on the use of hospital services9, which is the case in 42% of the articles we analyzed. However, in our review, the decrease in hospital service use is higher when patients transmit the data compared to instances where the transmission is automated. This suggests that an active patient role may positively influence hospital service use.
This systematic review is, to our knowledge, the first to assess the effect of components of DRM on reported hospital service use, whereas previous research focused on general telehealth10, or mentioned components in discussions of studies involving remote monitoring without reporting specific effects of each component on hospital service use9. The large number of RCTs included in this current review and the variety of components analyzed give a comprehensive characterization of the DRM literature.
It is important to acknowledge that some effects may be the result of the interplay between multiple components, such as the inevitable combination of continuous monitoring, automated measurements, and implantable devices. In this study, we have primarily investigated components in isolation. Understanding the combined effects of these components is challenging, as it is often unclear whether and to what extent the intervention considered the individual components, given that these are frequently not reported in detail. Consequently, considerable ambiguity persists regarding the impact of individual components on hospital service use.
Although we found reductions in hospital service use in most studies for at least one of the outcome measures related to hospital service use, the impact of reductions in practice may differ considerably. This is also reflected in some recent reviews on health economic aspects in DRM28,29. Reductions do not always lead to savings in FTE personnel. For instance, the volume per department needs to be considerable in order to actually save one 24/7 shift or dedicate the staff to other patients5. It is however clear that generally speaking reducing admissions and in-patient length of stay reduction commonly brings the highest impact in either percentages of efficiency gain, actual savings, or capacity for other- or waiting patients. Due to the variation in outcome measures across studies (as illustrated by the example of hospitalizations and length of stay in Supplementary Figs. 1–8), this is not straightforward and does not lead to substantially different conclusions about the impact of DRM components. The extent of change in hospital service use, and hospitalizations and length of stay in particular, should be investigated more rigorously. The effect, i.e., decrease, neutral, increase, on hospital service use is conflicting for the different outcome measures reported in 19% of the studies (Supplementary Table 4). Four RCTs analyzed showed an increase in outpatient clinic visits while hospitalizations decreased30,31,32,33. This could indicate a shift from more severe to less severe care when DRM is used. These four studies examined heart failure, COPD, diabetes, and pregnancy-related hypertension. The increase in outpatient visits observed in these studies may be due to alerts from DRM prompting patients to visit their healthcare provider, driving up healthcare use, but potentially preventing the progression of chronic diseases8. This is not what one usually wants to achieve as for chronic diseases, the primary aim of using DRM is often to reduce outpatient clinic visits.
The extent of change in hospital service use may provide valuable information to detect those components that may influence service use the most. Since in this review hospital service use is based on multiple outcome measures, and because of the different nature of reported outcome measures (e.g., mean, median, and percentage), it is not possible to draw any conclusions on the statistical significance. In addition, the effect on hospital service use is measured at different time points and follow-up periods in the various studies, which could potentially impact the results. The variations may reflect different stages of a patient’s condition, capture short-term vs long-term effects, and account for differences in patient compliance. In addition, some of the studies measure hospital service use after the monitoring period, and some during the monitoring period. However, an RCT using a different monitoring and observation period showed no significant changes in hospital service use during or after telemonitoring34.
This review has some unexplored components we recommend to evaluate in future studies. We focused on the design of DRM and the impact on hospital service use, without differentiating between diseases. Different diseases may benefit from the use of DRM differently. It is unclear what the consequences are of making a comparison without distinguishing the disease or condition. A more significant impact on hospital service use is to be expected in high-risk populations. This is corroborated by a previous review: ‘patients who are more likely to present to the hospital multiple times have a greater chance of reducing admissions due to more timely interventions’9. However, DRM is mostly used in patients with a more favorable prognosis, where the effect on hospital service use may be less evident short term. In addition, DRM can be short-term, such as for telerehabilitation or early discharge patients, or long-term, such as for chronic diseases, which can affect the desired outcome of hospital service use. DRM could also change patients’ perception of required support, or impact clinical decision-making on referring or admitting patients35. Future studies should evaluate the nature and severity of diseases in relation to hospital service use.
Furthermore, evaluating condition-specific (digitally accessible or biochemical) biomarkers, such as heart rate variability for heart failure or blood glucose levels in diabetes, remotely could enhance clinical value, as the relevance and application of these biomarkers may vary between different health conditions36. So far our findings show a majority of studies reporting on combinations of vital sign-based DRM and survey use. Only a minority dealt with conditions that used biochemical biomarkers in DRM, such as glucose for diabetes37,38,39,40 and bilirubin values for neonatal hyperbilirubinemia25. These biomarkers can be expected to play an increasingly important role as technology progresses.
The main findings in our review on the integration of DRM in clinical care relate to replacing usual care and involving healthcare providers. However, evidence on the optimal design of virtual hospital care services with regard to integration is scarce and a recent policy paper pointed out the important aspects41. The main aspects were the characteristics of the technology, the nature and intensity of provider involvement and supervision, the degree of self-management by the patient and his environment, the relation and cooperation mechanisms with other providers such as home care, general practitioners, and other specialist care, the matter of scale and the uniformity of processes over geographic regions and providers.
In order to improve the adoption and dissemination of DRM in clinical practice, some aspects should be considered that were shown to be related to higher degrees of effect and reduction of capacity use. Recommendations based on the findings of this current review for the design of DRM include active patient involvement with sufficient support, automated processes for the healthcare provider, accurately redesigning the care process by replacing usual care or integrating DRM, and the use of designated healthcare providers for DRM. While hospital service use is important, a broader perspective is essential in future research. This includes examining user impact, implications for healthcare providers, and disease-specific DRM design considerations. Additionally, healthcare economics should be examined through comprehensive studies on the budget impact and cost-effectiveness of DRM.
User characteristics and practical barriers, such as socioeconomic level, education, (digital) health literacy, and self-efficacy, which are described in the literature as components that affect hospital service use9,15, were not considered in this review. Digital health literacy is crucial for the adoption of digital health technology, by patients and healthcare providers42,43. In our review, only two studies actually mentioned the digital health literacy of the patients44,45. We therefore recommend that future impact studies should also include digital health literacy, especially if patients need to play an active role in DRM.
Additionally, patient and provider perspectives (e.g., satisfaction, safety, and motivation), adherence, usability, and duration of the intervention were also not taken into account in our review. Ethical and legal issues are also outside the scope of this review, but were seldom reported upon. Previous research showed that technological barriers can be addressed through training, change-management techniques, and alternating between DRM and face-to-face consultation. However, resistance to change, costs, and reimbursement might need additional policy evaluation46. These aspects, along with the previously mentioned user characteristics and practical barriers, should be evaluated in future studies.
Only a few of the papers in our review report on economic aspects or cost-effectiveness and commonly without a firm methodological approach. Specific reviews on this aspect are still scarce and commonly of a very recent nature28,29. These present a mixed picture regarding methods, use of guideline-based analysis and of results, with a tendency towards being cost-effective and sometimes cost-saving, especially on the case level. It is seldom explained whether this leads to actual savings on an institutional or system level. Most papers suggest efficiency gains in general terms. Further research on the economic aspects of DRM on institutional and system levels is thus certainly needed.
Generating thorough evidence is essential for wider implementation. We suggest developing more uniform reporting formats in DRM research, including a wider perspective of quality and safety-related aspects. Moreover, previous research recommends considering evaluation designs tailored to the context, which do not always have to be RCTs, but can adopt more pragmatic approaches19.
In addition, the standardization of processes and the use of artificial intelligence, which appeared in several RCTs, were not assessed. The influence of automated and standardized processes should also be considered in future research.
This systematic review shows that DRM has the potential to reduce hospital service use when patients are actively involved with sufficient support, processes are automated for the healthcare provider, the care process is accurately redesigned by replacing usual care, and healthcare providers designated for DRM are involved. If these recommendations are incorporated into the design of DRM, it may be more likely to reduce hospital service use. Future studies on DRM should also consider clear reporting of the components mentioned in this review. In particular, we suggest improving reporting on the integration of DRM into clinical care and patient engagement. This review also highlights the need for future studies to consider the impact of the nature and severity of the targeted condition, appropriate timing of alerts and contacts as well as practical barriers such as implementation of DRM into clinical care, adherence, digital health literacy, and socioeconomic status.
link
