Despite a growing prevalence of patients with complex and multiple chronic conditions (MCCs) [1,2], siloed care models focusing on single conditions have been a barrier to the appropriate management of these patients [3]. In Canada, 12.9% of individuals across all age groups report having ≥2 chronic conditions, and 3.9% report having ≥3 conditions [1]. They are among the highest cost users of health care systems because of a higher frequency of hospitalizations, many of which are thought to be preventable [4,5]. Research suggests that effective patient self-management can reduce the need for urgent care while promoting self-efficacy, improving quality of life, and reducing the risk of adverse psychological effects [6]. However, complex decision-making and often conflicting clinical advice from multiple siloed health care providers (HCPs) make self-management challenging for patients with MCCs [7,8].

Telemonitoring has the potential to empower patients and HCPs by facilitating patient self-management and clinical decision support to manage MCCs. Telemonitoring systems enable patients to track vital signs and symptoms and can enable the automatic generation of self-management instructions [9]. In addition, by delivering these data to the clinical team, HCPs can identify patients showing early signs of exacerbation, which offers an opportunity for reinforcing the principles of self-management at teachable moments [10]. Importantly, telemonitoring allows HCPs to provide remote guidance or make changes to a care plan, thereby stabilizing symptoms before they escalate to the point of hospitalization.

Although research on telemonitoring is rapidly growing [11], these systems typically target a singular condition such as diabetes mellitus (DM), hypertension (HT), or heart failure (HF) [12-14]. Systematic reviews indicate that telemonitoring for single conditions leads to improved health outcomes and quality of life and reductions in health service use and costs [15-21]. Studies that do not report improvements do not often include a self-care component, are difficult to use, or do not target patients who are most ill and frequently hospitalized [22-25]. To date, few studies have focused on the use of telemonitoring among patients with complex conditions, although these patients may benefit the most from such interventions [15,26-32].

The objective of this study is to evaluate the impact of a mobile phone–based telemonitoring program for the management of patients with complex conditions in specialty care settings. Patients with complex conditions are defined as those who are at high risk for hospitalization, exacerbations of their chronic conditions, and disease progression and those with MCCs, including HF, uncontrolled HT, and insulin-requiring DM. The primary research question was the following: what is the impact of a telemonitoring program for patients with complex conditions on health status, self-management, and health service use? The secondary research question was the following: what were the experiences of patients and HCPs related to the telemonitoring program and the way in which it was implemented?

This was a pragmatic, unblinded, 1:1 randomized controlled trial (RCT) comparing the 6-month impact of telemonitoring to support the management of patients with complex conditions with that of the standard of care. An embedded qualitative component was included to understand the results of the trial. Patients with HF were recruited from the Heart Function Clinic at the University Health Network (UHN), a large academic hospital in Toronto, Ontario, Canada, between August 2016 and February 2018. Patients with HT were recruited from an HT clinic at Mount Sinai Hospital between July 2019 and December 2019, and patients with DM were recruited from the UHN Endocrinology Clinic between August 2019 and December 2019. The study received approval from the UHN (15-9995-BE) and Mount Sinai Hospital (16-0093-E) research ethics boards, which approved the procedures to ensure patient privacy, including anonymization of data during storage and analysis.

To be eligible, patients had to be aged ≥18 years; able to speak and read English (or have a caregiver who does); and diagnosed with HF with reduced ejection fraction (<40%), uncontrolled HT (≥140/90 mm Hg auscultatory), or insulin-requiring DM and performing self-capillary glucose monitoring. Exclusion criteria included being on mechanical circulatory support, dialysis, or a transplant list. In addition, patients with a life expectancy <1 year, dementia, uncontrolled psychiatric illness, or residents of a long-term care facility were excluded. Refer to the published protocol for the full criteria [33].

A total of 146 patients with varying chronic conditions were targeted for enrollment (see sample size justification [33]). HCPs familiar with the study’s inclusion criteria identified patients during scheduled outpatient appointments and introduced them to an on-site research coordinator. After confirming eligibility, the coordinator explained the study and obtained informed consent. The patients were then block randomized using blocks of 4 into the intervention and standard of care treatment groups in a 1:1 ratio as per the published stratification protocol [33]. Patients allocated to the intervention arm were provided with the telemonitoring equipment and user manual before receiving face-to-face training on how to use the equipment.

The primary outcome was health status as measured by the 36-Item Short Form Health Survey questionnaire. Secondary outcomes included anxiety and depression, as measured using the Hospital Anxiety and Depression Scale [40], and patients’ self-efficacy to manage their condition, as measured using the Self-Efficacy for Managing Chronic Disease 6-Item scale [41-43]; the number of self-reported interactions with the health system in the previous 6 months was also collected, including hospitalizations and visits to ED, specialty care clinics, and family physicians. Self-care, as measured by the Self-Care of HF Index [44], and HF-specific quality of life, as measured by the Minnesota Living with HF Questionnaire, [45] were also collected for patients with HF.

Questionnaires containing the patient-reported outcome measures were administered at baseline and at 6 months [33]. Demographic questions were included in the baseline questionnaire to characterize the study participants. Patient adherence was calculated from the Medly server log data as a percentage of the completed recommended readings over the course of the 6-month trial.

Posttrial data and change scores were compared between the treatment arms using independent Student t tests and Mann–Whitney tests (for normally and not normally distributed data, respectively). Paired Student t tests and Wilcoxon signed-rank tests were performed to compare baseline and poststudy data within the control and telemonitoring groups. Analyses were performed using SPSS (version 27; IBM Corp) under the intention-to-treat principle and using a significance of P<.05.

All HCPs who used the system and a sample of patients in the intervention group were invited to participate in poststudy semistructured interviews. The interviews began with open-ended questions about the participants’ experiences with the intervention and had probing questions inspired by the constructs of performance expectancy, effort expectancy, social influence, and facilitating conditions from the unified theory of acceptance and use of technology [46]. Patients were identified through a convenience sample with efforts made to include patients from the 3 monitored conditions. Interviews took place in a private clinic room or over the telephone and were audio recorded for later transcription.

Qualitative data were analyzed using a conventional content analysis approach [47] by 2 researchers (AS and PW). An initial round of independent open coding was conducted with the primary objective of organizing quotes into themes that explained the quantitative results. Then, AS and PW met to discuss themes and agree upon a coding framework that was applied in the second round of deductive coding. A final discussion was held to review the results and reach a consensus. NVivo (QSR International; version 11) was used to help organize the source documents and themes.

Recruitment ended after the enrollment of 96 patients, of which 66 (69%) were followed for HF, 22 (23%) were followed for uncontrolled HT, 5 (5%) were followed for insulin-requiring DM, and 3 (3%) were followed for both HT and DM (Figure 2). The decision to stop recruitment before reaching the target sample size was made for 3 key reasons. First, the HF telemonitoring program became standard of care at the UHN Heart Function clinic, meaning there would no longer be a difference between the treatment arms. Second, recruitment challenges were observed for patients with DM because of the rapid emergence and growing use of continuous and flash glucose monitors, which meant that few patients met the criteria of self-capillary glucose monitoring. Finally, and most importantly, the COVID-19 pandemic led to a pause of nonessential research activities and a significant shift toward virtual care, which fundamentally altered the control group after the research pause was lifted.

In addition to ending recruitment, shifting priorities at the onset of the COVID-19 pandemic also affected the collection of poststudy data, as patients followed for HT and DM ended their enrollment during the first wave, which led to a higher rate of incomplete questionnaires. A death was reported in each treatment arm; however, these were not considered as adverse events of the study. Therefore, although 45 and 50 patients were allocated to the telemonitoring and control arms, only 29% (28/96) and 27% (26/96) of complete data sets were available for analysis in the telemonitoring and control arms, respectively. Study participants were predominantly male (54/96, 56%) and had an average age of 59 (SD 12.6) years. These, along with the other demographics presented in Table 1, are representative of the UHN Heart Function Clinic, from which most patient participants were recruited.

Table 2 shows the results from the statistical analyses. The within-group pre–post comparisons, poststudy between-group comparisons, and between-group change score comparisons revealed no significant differences in the primary outcome of health status as measured by the 36-Item Short Form Health Survey questionnaire physical and emotional subscales. Similarly, no statistically significant differences were observed for any of the secondary outcomes relevant to the entire sample, including anxiety and depression (as measured by the Hospital Anxiety and Depression Scale), self-efficacy (as measured by the Self-Efficacy for Managing Chronic Disease 6-Item scale), and self-reported use metric. An exception was a reduction in self-reported hospitalizations across all patients, which was significant in the control group.

Self-care maintenance and physical quality of life improved significantly for patients with HF in the telemonitoring group (P=.04 and P=.046, respectively). However, none of the between-group comparisons were statistically significant, likely because of the general improvement in self-care and quality of life scores of the control group and insufficient sample size for detecting changes in condition-specific metrics.

Of the 41 patients who completed the study in the intervention arm, 29 (71%) used the HF module, 10 (24%) used the HT module, and 2 (5%) used the DM module. On average, patients with HF completed all 4 readings on 78.1% (SD 5.8%) of the days that they were enrolled in the 6-month trial. This use rate increased to 86.2% when looking at the percentage of days the patients with HF reported at least one reading. Average adherence was 64.2% (SD 18.6%) for the patients with HT and 58.3% (SD 24.6%) for the patients with DM based on a minimum of biweekly and weekly readings for HT and DM, respectively. The combined average adherence across all participants was 73.1% (SD 10%) over the study period. Figure 3 depicts the monthly adherence rates for the 3 disease modules. It shows that although adherence was initially high across all conditions and remained relatively stable for patients with HF, it dropped markedly after the second month for patients with HT and DM.

In the absence of high-quality evidence on telemonitoring for managing MCCs [48], this study sought to evaluate the impact of a mobile phone–based telemonitoring program to manage patients with complex conditions in specialty care settings. Through a pragmatic RCT, halted before reaching the intended sample size, we observed no effects of the telemonitoring program on health status, anxiety and depression, self-efficacy, and most health service use metrics. Subgroup analyses of patients with HF in the intervention arm found a significant improvement in physical quality of life and self-care maintenance; however, no differences in these groups were observed between the treatment arms. The reduction in self-reported hospitalization found in the control group could also be seen as a trend in the intervention group but did not reach statistical significance, likely because of the small sample size. This reduction in both arms may be partly attributed to the impact of the HF clinic in stabilizing patients.

Several challenges were encountered during the trial, necessitating deviations from the published protocol, which likely contributed to the null results of the study. Foremost, this trial did not reach its intended sample size of 146. Although the onset of the COVID-19 pandemic and the rapid shift to virtual care for both arms did bring about the decision to stop recruitment permanently, other important challenges contributed to slow recruitment. Importantly, funding and development challenges contributed to the inability to include the planned chronic obstructive pulmonary disease and chronic kidney disease modules. Similarly, although patients who participated in our study had MCCs, we were unable to offer monitoring for all the conditions simultaneously as the HF, HT, and DM modules were not initially available at the same time. Therefore, the MCC model of care was never fully tested as there was never a need to explore communication workflows for multidisciplinary care that are inherent to the care of MCCs. Finally, recruitment of patients from the diabetes clinic was challenging because of the emergence of newer continuous glucose monitoring technologies at the time of recruitment.

Although patient perceptions of the telemonitoring program were largely positive, the perceived benefits and, consequently, use of the system varied across conditions. Indeed, although all patients in this study started with high adherence, these higher levels were only maintained in patients who were followed for HF. These differences may reflect the more actionable nature of the self-care instructions and more frequent clinician alerts from the HF module relative to those provided by the HT and DM modules. HF tends to be more dynamic and requires more parameters to be taken at a higher frequency, which provides the data inputs necessary for a more sophisticated telemonitoring algorithm. Conversely, HT and DM require fewer inputs at a lower frequency, which typically means that less urgency is required when alerts are generated. This does not suggest that HT and DM are not suitable conditions for systems such as Medly, as patients followed up for these conditions did express benefiting from self-management support. Rather, the opinions expressed by the HT and DM HCPs suggest that the way telemonitoring data are presented to HCPs needs to consider the greater importance of long-term trends versus the acute symptom-based alerting needed for optimal HF management. Considering these differences between conditions may increase the perceived relative advantage [46] of telemonitoring to promote comparable use across conditions.

Various factors related to the implementation of the telemonitoring program may also explain the lack of impact observed. We posit that an initially narrow conceptualization of telemonitoring, focused primarily on the technology rather than the model of care, may have contributed to these implementation issues. For instance, relative to HF [49], minimal consideration was given to the model of care associated with the DM and HT modules. Consistent with previous studies, we recommend that future research conceive telemonitoring as a program comprising both the system and its associated model of care. Critically, implementation study designs should be used to assess the feasibility of the program before conducting a trial so that challenges may be identified and addressed [50].

A notable characteristic thought to enhance the feasibility of telemonitoring for patients with complex conditions and MCCs was having a nurse in a central role as it relates to alert management and care coordination. In contrast to the fee-for-service model for physicians, all HCPs in this study noted that the salaried remuneration of nurses offered greater flexibility needed to monitor patients with MCCs through telemonitoring. Indeed, when the same MCC telemonitoring system was applied in a different study within a multidisciplinary nurse-led model of care [51], telemonitoring was not only perceived positively by patients with MCCs but was also normalized within their daily life and routines [52], which was a finding that was only observed among the patients with HF and HCPs in this trial, who also benefited from a nurse-led model. What may explain the stark contrast in the results between these studies is the differences in their underlying program theory. Per Kastner et al [53], effective interventions for older adults with MCCs are often founded upon principles of care coordination, disease prioritization, and patient self-management. From this, we conclude that telemonitoring interventions for patients with complex conditions, especially those with MCCs, should (1) be integrated with usual care workflows and complementary clinical services, (2) be supported by a multidisciplinary model, and (3) include ≥1 element of care coordination (eg, referral pathways, case management, and interoperable information systems). Critically, these 3 elements may be best operationalized by having a nurse play a central role in the intervention.

The outcomes of this trial are consistent with recent reviews [48,54], indicating a dearth of evidence on the effectiveness of telemedicine interventions for the population with MCC. In 2020, Kraef et al [48] identified only 1 study reporting no impact of telemedicine on self-reported health status and only moderate improvements in disease control measures such as hemoglobin A_1c and systolic blood pressure across 7 meta-analyzed studies. Similarly, Pisa et al [54] found no impact of telemedicine interventions in primary care on quality of life and contradictory results on ED visits and hospital admissions across 5 studies [54].

Importantly, trials reporting significant effects on condition-specific outcomes and no implementation challenges often included a role of a nonphysician HCP in the intervention, most commonly a nurse [30,55-57]. Moreover, several authors have attributed frequent and immediate feedback to patient data as important facilitators in promoting the intended use of telemonitoring systems and the success of their trials [30,57,58]. Thus, the results of this trial and the existing literature converge in identifying care coordination as an important component in telemonitoring interventions for MCCs. We suggest that this be considered as a central component of further studies exploring the feasibility and effectiveness of telemonitoring for MCCs.

In addition to the recruitment and implementation challenges previously discussed, this study had limitations attributable to the trial’s pragmatic nature. First, although the study period was 6 months, the study duration lasted approximately 4 years with no overlap in the periods in which patients with HF and patients with HT or DM were active. This meant that in addition to inhibiting the exploration of multidisciplinary care coordination, ending the recruitment because of the COVID-19 pandemic contributed to a difference in sample size across conditions. Second, self-reported health service use introduces potential challenges because of recall. Third, unintended selection bias is possible as the recruitment relied on busy clinicians identifying eligible patients during clinic hours. We do not have data on patients who might have been eligible but were never identified by their care team or on patients who were approached but did not agree to be introduced to the research coordinator. Finally, the interviews were conducted with a convenience sample of patients, and although all HCPs who were involved in the study in any significant manner were interviewed, the number of 5 HCPs was relatively small.

Decades of research suggest that telemonitoring for self-management and clinical decision support may assist in managing single chronic conditions; however, its effectiveness for the growing number of patients with complex conditions, including those with MCCs, remains unclear. This 6-month pragmatic RCT sought to evaluate the impact on patients and HCPs and their experiences with a telemonitoring program to manage patients with complex conditions compared with those in standard care. No significant within- and between-group differences were found in the primary outcome of health status or the secondary outcomes of self-efficacy, anxiety, depression, and health service use. However, improvements in physical quality of life and self-care maintenance were observed in patients with HF who were enrolled in the telemonitoring program. Qualitative data suggest that these null findings may be because of the fact that not all disease modules were available at the same time, implementation challenges within the siloed specialty care, varying degrees of acuity and urgency across chronic conditions, and an insufficient sample size attributable to the onset of the COVID-19 pandemic. On the basis of the study findings, we caution that self-management and clinical decision support are necessary but not sufficient components of telemonitoring for patients with complex conditions. We conclude that telemonitoring for patients with complex conditions or those within multidisciplinary care settings may be best operationalized through nurse-led models of care and that the complex nature of these interventions makes it such that they are best studied through feasibility or pragmatic study designs before conducting RCTs.