Background

Chronic obstructive pulmonary disease (COPD) is a common, preventable lung disease characterized by long-term breathing problems and poor airflow due to airway and alveolar dysfunction commonly caused by smoking [1, 2]. COPD is one of the leading causes of morbidity and mortality worldwide, with a substantial economic and social burden on individuals and society [2, 3]. Patients with COPD often suffer from comorbid diseases including heart failure, diabetes, and depression, making management of these patients complex and multifactorial [4].

Previous studies have shown that acute exacerbation are common in patients with COPD, and increasing frequency of exacerbations is associated with a decrease in lung function [5, 6], and an increase in the use of health services [7]. Integrating remote home monitoring (RHM) into clinical care may support patient self management, and lead to improvements in symptoms and quality of life, while reducing COPD exacerbations burden and healthcare utilization [1, 2, 8,9,10]. Tomasic et al. have described remote monitoring as encompassing “automatic continuous physiological data transmission and processing decision support, the prediction of deterioration and alarming” [9]. Although, RHM has the capacity to prevent exacerbations, evidence concerning its safety and effectiveness remains unclear. Therefore, the objectives of this study was to determine the effectiveness of RHM programs for patients with COPD. This study was part of a project commissioned by Alberta Health to optimize care of patients with COPD in Alberta, Canada.

Methods

A systematic review of peer-reviewed primary studies was conducted following the Cochrane Handbook and Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRSMA) guidelines [11, 12].

Search strategy

An experienced medical information specialist in consultation with the research team iteratively developed a comprehensive, structured search strategy. It was peer-reviewed by another senior information specialist for quality assurance using the Peer Review of Electronic Search Strategies (PRESS) checklist [13]. The search strategy was applied to the following databases: Ovid MEDLINE, Embase, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, Health Technology Assessment and the NHS Economic Evaluation Database. We also searched CINAHL and EconLit on the Ebsco platform and Web of Science. Details of the search strategy are presented in online supplementary Table S1. The search was conducted from March 1st to March 13th, 2020. The electronic searches were also supplemented by manual searches of reference lists from included studies. Results from the search strategy were compiled into Reference Manager which was used to manage all references.

Eligibility criteria

Two reviewers independently screened the titles and abstracts of all citations to identify studies for a full review. Full papers corresponding to potentially relevant citations were retrieved, divided among, and assessed by three reviewers for inclusion/exclusion according to criteria (Table 1). Although RCTs are considered the gold standard in assessing interventions under specific settings, observational studies may provide evidence on the effectiveness of RHM compared to usual care in the “real world”. As this study was commissioned to inform policy decisions, studies were not exluded based on design and quality. Reviewers met to compare results and agree on the final set of studies to include. At both screening steps, consensus between reviewers was assessed using the Kappa statistics and found to be “substantial”.

Table 1 Remote Home Monitoring PICOS elements of the clinical effectiveness review protocol

Data extraction and synthesis

Extracted data were tabulated to identify trends or patterns across studies and facilitate qualitative and quantitative comparative analyses. Key characteristics of included studies, their quality, potential sources of bias, and findings were synthesized narratively. A narrative synthesis of effectiveness outcomes across the studies was undertaken. Analysis was based on the types of technologies used for home monitoring which were grouped into three groups: (1) smartphones, apps, tablets; (2) dedicated home monitoring devices; (3) phone calls and text messages. Additionally, studies were assessed to determine whether patient populations, designs, and outcomes were similar enough to perform meta-analyses. Results were pooled if outomes were assessed with the same measures and follow-up times. Heretogeneity was assumed to be too substancial to pool data when the I2 statistic was equal to or greater than 50% [11]. Forest plots were used to display individual and pooled results. A p value < 0.05 was considered statistically significant in all analyses. Pooled risk ratios for categorical data and mean differences with 95% confidence intervals (CIs) for continuous outcomes were reported. Publication bias was assessed using funnel plots, where sufficient data (i.e. at least ten studies) were available from the meta-analyses [14]. Multiple studies published with an overlap of outcomes and patients were combined.

Assessment of study quality

RCTs were appraised using the Cochrane Collaboration Risk of Bias tool (ROB) [15]. The methodological quality of the non-RCT interventional and comparative observational studies were evaluated using the Cochrane Risk of Bias Assessment Tool for Non-Randomized Studies (ACROBAT-NRSI [16]. The quality of evidence relating to key outcomes of interest were assessed using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) tool [17]. Prior to conducting the systematic review, a small questionnaire was conducted with members of an Expert Advisory Group (EAG) to rank the outcomes according to their importance. The six outcomes with the highest rank were included in the GRADE assessment [11, 18]. The EAG was arranged to oversee the project and involved clinicians, COPD program coordinators, policy makers and researchers. In this review, GRADE assessment was conducted by two independent reviewers and based on the following outcomes: health-related quality of life (HRQoL), patient experience and number of exacerbations, number of emergency room (ER) visits, COPD-related hospital admissions, and adherence as the proportion of patients who completed the study. Discrepancies between the reviewers were resolved through discussion.

Results

Search results

Four thousand nine hundred ninety-three discrete citations were identified through the literature searches and screened, of which 239 were retrieved for full consideration. Twenty papers met the criteria for inclusion in the review. They comprised 17 RCTs and 2 comparative observational studies. Literature search results described using the PRISMA flow diagram are shown in Fig. 1.

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram of literature search and study selection for efficacy/effectiveness review of remote home monitoring (RHM)

Characteristics of studies

Seventeen [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35] of the 19 studies [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38] compared remote home monitoring (RHM) to ‘no remote home monitoring’ (no RHM), and two [36, 37] compared RHM with provider’s feedback to RHM without feedback. The 19 studies were conducted between 2006 and 2018 and had sample sizes ranging from 34 to 1,238 patients (details of participants' characteristics included in the studies are presented in online supplementary Table S3). Collectively, they included 3,144 patients with COPD. Ten studies [19,20,21,22, 25, 26, 30, 32, 34, 35] recruited patients from multiple centres, and one [21] spanned five European countries (Spain, the United Kingdom, Slovenia, Estonia and Sweden). The remaining eighteen studies were conducted in Australia (2) [24, 27], Canada (1) [30], Denmark (1) [32], Germany (2) [28, 37], Hong Kong (1) [31], Italy (1) [25], Netherlands (3) [20, 22, 23], South Korea (1) [19], Spain (2) [26, 29], the United Kingdom (1) [33], and the United States (3) [34,35,36] (Table 2).

Table 2 Characteristics of included studies

RHM program characteristics

The length of the monitoring period varied from one to 12 months (comparison of what was monitored remotely and when in the included studies is presented in Table 3). At the beginning of the program, nurses taught patients how to use the technology, typically in patients’ homes. However, two studies [19, 20] held group sessions at outpatient clinics and two studies [22, 23] trained patients in their homes and outpatient clinics. In four studies [19, 20, 23, 24], patients also participated in outpatient group education and exercise sessions (a detailed description of the home monitoring program, technology and its components is presented in online supplementary Tables S4, S5, and S6).

Table 3 Comparison of what was monitored remotely and when in the included studies

Comparator interventions in the studies were ‘no RHM’, or RHM without feedback and alerts from healthcare providers. ‘No RHM’ generally comprised usual care based on local practices, in which patients were instructed to contact their healthcare provider if they experienced worsening of symptoms. Five studies [19, 20, 24, 27, 32] reported that patients in the ‘no RHM’ group received education and exercise training sessions or materials similar to those received by the RHM group. RHM without feedback and alerts consisted of patients measuring parameters and transmitting data, but with no healthcare providers' feedback.

The RHM technology and integrated peripheral devices varied across studies; most of the interventions had a dedicated device for home monitoring. Four studies [19, 20, 22, 23] used smartphones and apps as the main device, and in one study [19], education material and exercise training were also available through the app. One study [21] used a tablet to facilitate recording and transmission of data. Two studies [36, 37] used telephone and text messages to monitor patients. Included integrated peripheral devices were: pulse oximeter (in 11 studies) [20, 21, 24,25,26,27, 29, 31,32,33, 35], blood pressure cuff (7) [21, 24, 26, 27, 29, 32, 35], spirometer (6) [20, 24, 26, 28, 29, 32], thermometer (5) [20, 21, 24, 27, 33], heart rate monitor (4) [21, 24, 26, 29], weight scale (4) [24, 27, 32, 35], accelerometer (3) [22, 23, 28], pedometer (2) [19, 32], glucometer (1) [24], peak flow meter (1) [26], and respiration sensor (1) [31]. In the majority of studies, patients were asked to collect data daily. The most common parameters monitored were symptoms (15), oxygen saturation (11) [20, 21, 24,25,26,27, 29, 31,32,33, 35], blood pressure (8) [21, 24, 26, 27, 29, 32, 34, 38], and lung function (7) [20, 21, 24, 26, 28, 29, 32]. All studies reported that transmission of data occurred immediately. In ten studies [19, 24, 27, 28, 30,31,32,33,34, 37] healthcare providers and/or nurses were responsible for monitoring data. Four studies [25, 26, 29, 35] used a central management unit to monitor data and five [20,21,22,23, 36] used algorithms and decision trees to monitor and detect changes in symptoms and clinical parameters. In most studies, healthcare providers or nurses contacted patients if clinical worsening was observed and/or data were not recorded for several days. None of the studies discussed data privacy.

Risk of bias

Results of risk of bias assessment

Quality assessment was conducted for the seventeen [19,20,21,22,23,24,25,26,27,28,29,30, 32, 33, 36,37,38] RCTs and two non-randomized studies [34, 35] (detailed description in online supplementary Table S7).

RCTs

Six studies [19,20,21, 23, 25, 27] assigned patients to treatment groups via a computer-generated sequence and three [31, 32, 40] used drawing of lots. Four trials [23, 29, 33, 36, 38] were likely at high risk of selection bias since three [23, 29, 33, 38] reported an unequal distribution of patients’ characteristics between groups, and one [36] had assigned seventeen patients to the control group without randomization. Only five [23, 26, 30, 32, 33, 38] provided details around allocation concealment. Four [23, 30, 32, 33, 38] adhered to adequate methods for keeping patients and investigators unaware of treatment allocation prior to assignment. One study [26] randomized the clinics, rather than patients themselves. Due to the nature of remote monitoring, neither patient nor staff were blinded to the intervention. Thus, the risk of performance bias was high. Patient reported outcomes were also at high risk of observer bias because patients were the assessors and not blinded to the type of intervention they had received. Eight studies [20,21,22, 24, 25, 32, 36, 37] were at low risk of observer bias. Six RCTs [22, 26, 28,29,30, 33, 38] provided insufficient information to determine the presence of observer bias. Four RCTs [23, 24, 31, 36] were at high risk of attrition bias. Three studies reported differences in the frequency of missing data and reasons for dropouts between groups. The risk of attrition bias was low in eleven of the RCTs [19,20,21,22, 25,26,27,28,29,30, 32], where the extent of missing data was small and similar between groups. Two studies [33, 37, 38] did not provide sufficient information to determine the risk of attrition bias. Ten trials [19, 22,23,24, 27,28,29,30,31,32] did not publish or register their protocols and five [19, 32, 33, 37, 38] were considered to have an incomplete follow-up data on outcome measures described in trial registrations and study methods sections (Figs. 2 and 3).

Fig. 2
figure 2

Cochrane risk of bias summary

Fig. 3
figure 3

Risk of bias graph with each risk of bias presented as a percentage across all included RCTs

Non-randomized studies

Both non-randomized studies were at serious risk of bias due to confounding and patient selection. They used methods to adjust for socioeconomic variables, but did not measure and adjust for clinical confounders (e.g. FEV1, severity of COPD). Further, recruitment into these studies was based on availability of the technology and patients’ preferences [34, 35]. One non-randomized study [34] measured objective outcomes unlikely to be influenced by knowledge of the intervention received. The second non-randomized study [35] did not blind the outcome assessor to intervention type. However, in both of these studies, the intervention status remained the same throughout their duration, minimizing the risk of bias in the measurement of interventions. In the two non-randomized studies [34, 35], data were reasonably complete. None of the non-randomized studies [34, 35] discussed the care received by the comparator group (Figs. 4 and 5).

Fig. 4
figure 4

ACROBAT-NRSI summary

Fig. 5
figure 5

ACROBAT-NRSI graph with each risk of bias item presented as a percentage across all included non-randomized studies

Results from GRADE assessment

GRADE assessment was conducted on the selected outcomes (Tables 4, 5 and 6). The GRADE level or certainty of the evidence for these outcomes was very low to low for all outcomes in studies comparing RHM (smartphones, apps, tablets) to no RHM, RHM (dedicated monitoring devices) to no RHM, and RHM (phone calls, text messages) to no RHM.

Table 4 Studies comparing remote home monitoring (smartphones, apps, and tablets) to no remote home monitoring
Table 5 Studies comparing remote home monitoring (dedicated monitoring devices) to no remote home monitoring
Table 6 Studies comparing remote home monitoring (phone calls, text messages) to no remote home monitoring

Summary results of effectiveness

Health-related quality of life (HRQoL)

RHM (smartphones, apps, tablets) versus no RHM

Two studies [20, 23] that measured changes in HRQoL from baseline using the CCQ reported no statistically significant differences between groups. Across studies [19,20,21, 23] that used other HRQoL measures (SF-36, NCSI, and EQ-5D), there were no statistically significant differences between changes in the two groups.

RHM (dedicated monitoring devices) versus no RHM

None of the studies comparing RHM with no RHM showed a statistically significant difference between groups in the change in HRQoL over time, regardless of the instrument used (CAT, CRQ, SGRQ or EQ5D) [25, 27, 29, 33, 38].

RHM with feedback vs RHM without feedback

After 6 months, CAT scores had statistically significantly improved within both groups in the cross-over RCT suggesting that the feedback component had minimal to no effect on HRQoL [37] (details are presented in online supplementary Tables S8, S9 and S10).

Patient experience

RHM (smartphones, apps, tablets) versus no RHM

Three studies [19, 20, 22] that examined patient experiences and satisfaction with RHM demonstrated comparably high satisfaction levels.

RHM (dedicated monitoring devices) versus no RHM

Seven studies [24, 27, 29,30,31, 33, 35] explored patients’ experiences with the dedicated monitoring devices. Few difficulties with the devices were reported. In general, patients felt the technology was easy to operate and were satisfied with the support received when troubleshooting clinical and technical problems. Six studies [24, 27, 29,30,31, 35] assessed perceived benefits related to RHM. They included: better control over/management of their disease, less anxiety, improved ability to cope with their disease, and reduced burden on family members. In all five studies [24, 30, 31, 33, 35] that measured overall satisfaction, the proportion of patients satisfied was high – at least 80% were reported (details presented in online supplementary Table S11).

Frequency of exacerbations

RHM (smartphones, apps, tablets) versus no RHM

No difference was reported [20, 21].

RHM (dedicated monitoring devices) versus no RHM

One study [28] reported a statistically significantly higher number in the no RHM group, but the larger study [34] found no difference between groups.

Healthcare resource utilization (hospital admissions, ER visits, and physician visits)

RHM (smartphones, apps, tablets) versus no RHM

The impact of RHM on healthcare resource utilization was assessed using numbers of hospital admissions due to COPD [19,20,21], ER visits [19, 22], and physician visits among patients who received or did not receive RHM [19, 20]. These were similar between groups (detailed description in online supplementary Table S12).

RHM (dedicated monitoring devices) versus no RHM

Nine studies [24,25,26,27,28, 30, 31, 33, 34, 38] assessed the extent to which RHM with dedicated monitoring devices affected COPD-related hospitalizations. In seven studies, values were similar between groups [24, 25, 27, 28, 30, 31, 33, 38]. However in two studies [26, 34], there were statistically significantly fewer admissions in the RHM group. Of the five studies [25, 27,28,29, 33, 38] measuring visits to specialists or primary care physicians, four [25, 27, 29, 33, 38] found no statistically significant differences between groups in specialist or primary care physician visits. In one study [28], however, the number of visits to a primary care physician was higher among patients who did not receive RHM.

RHM with feedback vs RHM without feedback

One study compared the total number of COPD-related hospital admissions over 8 months between treatment groups. The group who received continuous feedback on self-reported monitoring data from a healthcare provider had a statistically significantly lower number of admissions (over 8 months) than the group who did not [36].

Adherence to/compliance with treatment

RHM (smartphones, apps, tablets) versus no RHM

In four studies [19,20,21, 23], adherence with treatment appeared to be similar between groups, but in the fifth study [22], it was almost 5 times higher in the RHM group than in the comparator group (no RHM). Risk ratios for the two studies [21, 23] demonstrated conflicting results (Fig. 6).

Fig. 6
figure 6

Forest plot of risk ratios for treatment adherence at 9 months of follow-up

RHM (dedicated monitoring devices) versus no RHM

In the studies, adherence with treatment appeared to be similar between groups. The exception was a small 12-month study [24] of 21 patients who received RHM and 21 patients who had usual care (no RHM) (Fig. 7) [24, 25].

Fig. 7
figure 7

Forest plot of risk ratios for treatment adherence at 12 months of follow-up

RHM with feedback vs RHM without feedback

One study [36] reported a 76% compliance for RHM and 68% for no RHM, but there was no information on the statistical significance of the difference.

Safety

One study [19] (RHM using smartphones, apps or tablets vs. no RHM) reported data on adverse events, and no statistically significant differences between treatment groups were found. Eight studies [20, 21, 24,25,26, 29, 33, 35, 38] reported deaths from all causes and were similar between treatment groups.

Exercise capacity and activity levels

RHM (smartphones, apps, tablets) versus no RHM

Exercise capacity and activity levels improved statistically significantly in the RHM group, but the between groups difference was not statistically significant [19].

RHM (dedicated monitoring devices) versus no RHM

Patients who received RHM statistically significantly increased the 6-min walk distance, but those in the no RHM group did not [28].

RHM with feedback vs RHM without feedback

Total leisure activity at 6 months in patients who received RHM with feedback was statistically significantly higher than in the group without feedback (details presented in online supplementary Table S14) [37].

Mental health

RHM (smartphones, apps, tablets) versus no RHM

Neither study reported statistically significant changes in POMS or PHQ-9 (tension-anxiety and depression) scores within or between groups after 6 months [19, 21].

RHM (dedicated monitoring devices) versus no RHM

No statistically significant differences in HADS values were reported among patients who received RHM compared to those who did not [25, 33] (details presented in online supplementary Table S15).

Self-efficacy

RHM (smartphones, apps, tablets) versus no RHM

Neither of two studies reported statistically significant differences in self-efficacy measures between the RHM and usual care groups at baseline or at the end of the follow-up period [19, 20] (details are presented in online supplementary Table S16).

Cost per patient

RHM (smartphones, apps, tablets) versus no RHM

In the single study [21] reporting per patient costs with and without RHM, no statistically significant differences were seen between groups.

RHM (dedicated monitoring devices) versus no RHM

Two studies [24, 32] compared the cost of hospital admission and one [30] compared all costs (from 12 months prior to and 6 months after the start of RHM). All concluded that there were no statistically significant difference between groups (details presented in online supplementary Table S17).

Provider experience

None of the included studies reported on providers' experiences involved in RHM (smartphones, apps, tablets) versus no RHM and RHM with feedback vs RHM without feedback comparisons.

RHM (dedicated monitoring devices) versus no RHM

Two studies [24, 29] reported findings from surveys designed to obtain feedback from providers. Perceptions around the dedicated home monitoring device's usability and value improved as provider experience increased; however, only six providers participated in the two studies (details presented in online supplementary Table S18).

Lung function and symptoms

RHM (smartphones, apps, tablets) versus no RHM

No statistically significant differences in baseline or follow-up scores in validated measures in 2 studies [19, 20].

RHM (dedicated monitoring devices) versus no RHM

In the two studies [28, 31], no statistically significant differences were found in predicted values for FEV1 at baseline and at the end of follow-up between groups (details presented in online supplementary Table S19).

Discussion

Several aspects distinguish this work from previously published literature reviews [3, 41,42,43,44,45,46]. This review yielded more studies due to the broader inclusion criteria of home monitoring technology and its components, outcomes, and types of included studies. For example, previous systematic reviews included small numbers of studies (between 3 [41] and 10 [42]). Further, previous reviews measured a relatively small number of outcomes [45], had unclearly defined outcomes [3], had a substantive difference between defined and measured outcomes [41], or considered satisfaction from the patient perspective only [42, 46]. In the current review, in addition to defining and measuring outcomes such as adherence (exercise, self-management, diary, and medication use), exacerbation, hospitalizations, and patient satisfaction, the focus was also on including other outcomes relevant to health services and program planning such as safety, cost per patient and provider experience. Finally, this study extends previous reviews [42, 47, 48] by synthesizing findings according to type of technology and feedback provided.

This review concludes that HRQoL was not significantly improved with RHM as compared to usual care, regardless of monitoring technology; this finding is similar to previous reviews [42, 48, 49]. HRQoL is a complex construct, and while programs such as pulmonary rehabilitation consistently show improvement in HRQoL in COPD [50, 51], other disease management interventions (e.g., pharmaceutical care, patient education and action plan) do not consistently improve HRQoL in this disease [52, 53]. Unfortunately, no study was identified that used RHM during pulmonary rehabilitation. Instead, studies that included elements of pulmonary rehabilitation such as patient education, and/or exercise in both usual care and RHM groups, showed no greater benefit in HRQoL within the RHM group [19, 20, 22, 23]. These findings would suggest that regardless of the disease management program used, RHM did not improve HRQoL over and above usual care.

Remote home monitoring has the potential to improve disease self-management by making patients more aware of day-to-day changes in their symptoms and physical function [54, 55], thus improving disease management and reducing the risk of exacerbation. While previous studies have shown a significant reduction in health care utilization in COPD patients using self-management programs [56,57,58], this review found no consistent impact of RHM on patient self-efficacy, physician visits, ER visits or hospitalizations. Behaviour change is required for proper disease self-management, and time is required for patients to adopt and adhere to new behaviours. Most trials were under 12 months, and there may have been insufficient time to develop appropriate behaviour change that would lead to better disease management and reduced health care utilization.

Numerous studies have evaluated the impact of disease management programs in COPD, but due to heterogeneity in content, duration, and frequency of follow-ups, it is challenging to identify the key components of these programs. This review suggests that COPD related hospital admissions improved when RHM was coupled with feedback from healthcare professionals. No RCTs have investigated patient-provider communication in COPD specifically, but within other chronic disease states, more frequent and positive patient-provider communication was associated with improved health outcome and higher levels of self-efficacy [59, 60]. A recent qualitative study aimed to explore the views of patients and professionals on telemonitoring found that patients and health care professionals considered relationship-based care important in COPD telemonitoring services [61]. Therefore, RHM that facilitates regular communication with a healthcare professional appears to be important.

While considered usual care, patients with COPD are often not referred to pulmonary rehabilitation. Several barriers, including lack of available programs and travel/transportation needs, prevent patients from attending conventional centre-based rehabilitation programs [62,63,64]. Home-based alternatives are needed, but these are currently underdeveloped and the complexity of COPD patients raises concerns regarding patient safety. Future studies should aim to evaluate the additional benefits of RHM in patients undergoing (virtual) pulmonary rehabilitation. Further work should also evaluate patient behaviour to determine if RHM is effective at changing key behaviours that are foundational to improved disease management.

Limitations

This systematic review has several limitations. One drawback of this review is the lack of its protocol registration in the PROSPERO database as recommended by guidelines [11, 12]. Any protocol changes were documented and discussed within the research group to minimize bias.Second, there is the possible risk of bias due to missing information in the included studies. Furthermore, included studies provided limited descriptions of the study randomization process, and the studies varied in components of the interventions. Third, the study was restricted to English language studies, which might have led to the exclusion of possibly relevant studies. In addition, it was not possible to perform a meta-analysis on outcomes due to a high level of heterogeneity and limited data. Finally, there is the possibility of an impact on the findings by unpublished negative studies.

Conclusion

By applying objective, high-quality methods for gathering and synthesizing information from primary studies, this systematic review was conducted to review evidence from 19 studies, 17 of them RCTs, of remote home monitoring effectiveness in patients with COPD. Although a considerable amount of evidence to the effectiveness of RHM exists, due to heterogeneity of care settings, RHM components and the low-quality evidence, they should be interpreted with caution.