This was a prospective, pragmatic, two-arm, parallel implementation trial comparing usual care (UC) with a 3-month mHealth-enabled IC intervention. The study was conducted from July 2018 to August 2019 at the University Hospital of Santa Maria (Lleida, Spain) and its network of primary care centers. This corresponds to a large rural area accounting for more than 236,000 citizens with a life expectancy of 80 and 86 years for men and women, respectively [21].

The eligibility criteria were home-dwelling patients elected for primary THA or TKA at the University Hospital of Santa Maria; aged >65 years; being defined as CCP (Charlson index score ≥3, taking four or more pills per day, and having had contact with the health system at least two times in the last 12 months); being classified according to the American Society of Anesthesiologists (ASA) classification as ASA II (mild systemic disease) or ASA III (severe systemic disease); and successfully passing a basic technological test, aimed to ensure the availability of internet connection at home as well as patients’ or caregivers’ competence with the use of a smartphone, tablet, or computer. The basic technological test can be found in Multimedia Appendix 1 [20,22].

The recruitment was done in several waves to match the pace of the CONNECARE project technological developments. In each wave, consecutive potential participants were contacted by a case manager during preoperative assessment at the anesthesiology outpatient clinic. The case manager explained the study protocol and obtained informed consent. These patients formed the intervention arm. After the recruitment of each patient included in the intervention arm, an active search for a control with similar characteristics from the surgery waitlist of the Orthopedics Department of University Hospital of Santa Maria began. This enhanced the similarity of patients in the intervention and control arms, although it implied a certain lag in the recruitment of controls (from some days to few weeks). All patients and their caregivers, regardless of study arm, received a face-to-face explanation about the study and provided informed consent.

Patients in the intervention arm were attended using an IC model, including (1) preliminary assessment of the patient's health status using several questionnaires, tests, and indices specific to their main chronic diseases and social needs; (2) a self-management app, with status and performance reports, a virtual coach with customizable automated feedback, and full communication with the care team; (3) a Fitbit Flex 2 digital activity tracker [23], fully integrated into the self-management app; (4) a patient’s profile in the SACM web-based platform, accessible to all members of the care team (family physicians, hospital surgical team, and social workers) and used to coordinate professionals in the different settings and enable a communication channel among them and with the patient, when needed; and (5) assignment of a case manager in charge of supervising the whole process and being the main patient contact point. Additional details on the CONNECARE IC model and the supporting eHealth platform can be found in Multimedia Appendix 1. Patients in the control arm were attended using care as usual, being managed from primary care after hospital discharge. After discharge from the initial 90 days of UC or IC management, all patients had 3 months of additional passive follow-up.

Variables characterizing the patients were collected at recruitment using the SACM on tablet and/or desktop computers. Collected variables included main baseline characteristics, such as age, sex, main chronic diseases, Charlson index of comorbidities [24], ASA scores [25], Barthel index for activities of daily living [26], Hospital Anxiety and Depression scale [27], assessment of the dwelling characteristics, main medications, Pfeiffer mental status questionnaire [28], and tobacco and alcohol consumption.

The cost estimation of the IC program and used health care resources is described in Multimedia Appendix 1. Briefly, IC costs included the costs of newly required medical personnel (hospital-based nurse case managers) and the costs of licensing and running the IC platform. In contrast, the one-off costs associated with the development of the IC model and supporting technology or the tuning of the system were not considered. Similarly, the required restructuring of health care professionals’ time to include new tasks related to the IC model was assumed to be covered by the health system at no additional cost, as we assumed that such restructuring would not imply any additional time. Costs associated with the use of health care resources, such as medical visits and hospital admissions, were based on official data from the Catalan Health Department [22]. All costs were originally in Euros (€) and were converted into US $ (conversion used: €1=US $1.21).

The main outcomes were (1) intervention effectiveness, as measured by the changes in the 12-item short-form survey (SF-12) health questionnaire’s physical and mental domains (baseline vs discharge) [29]; (2) use of health care resources after 6 months; and (3) cost-effectiveness, based on the improvement in QoL relative to costs, assessed by means of the incremental cost-effectiveness ratio (ICER). The source of health information was the EMR, and the collected information included hospital admissions, emergency room visits, visits to primary care, and visits to hospital specialists.

Participants’ baseline characteristics were described by n (%), mean (SD), or median (P25-P75), as appropriate. Comparisons between IC and control patients’ baseline characteristics were performed using the chi-square test, t test, or Kruskal-Wallis test, as appropriate. A paired t test comparing baseline with discharge was used to assess changes in the SF-12 domains. Linear regression models were used to assess the differences in the changes experienced by IC and control patients. Negative binomial regression models were used to assess the differences in the number of visits and admissions. Models were adjusted for age, sex, and Charlson score. ICER was calculated in relation to the SF-12 total score. Sensitivity analyses assuming two different scenarios, 150% and 200% estimated cost of the IC program, were performed to explore the cost-effectiveness performance of the IC model under unplanned increases in the implementation costs. Data analyses were conducted using Stata version 12.1 (StataCorp). The threshold for significance was set at a P value of .05. All t tests were two-tailed.

This study was approved by the Ethics Committee of Hospital Arnau de Vilanova (CEIC-1685), and all patients provided written informed consent. All collected data were handled and stored in accordance with the current national and international legislation.

The prospective assessment of the implementation of an mHealth-enabled IC program for TKA and THA management showed a reduction in the number of unplanned contacts with the health system after the surgery; generated substantial savings for the health system, while not having any negative impact on QoL or clinical outcomes; and demonstrated cost-effectiveness.

The implemented IC model had several strengths that must be highlighted. First, there was an effort to involve all the stakeholders from different organizations that would be actors in a large-scale deployment of the mHealth-enabled IC program since the very early stages. This is key, as the lack of cooperation among professionals, teams, and organizations is a recurrent barrier for effective IC implementation [11]. Second, informal caregivers played a role in the IC process, as close relatives of the patients can be their link with the health system. Moreover, informal caregivers can play a key role in facilitating the use of self-management apps in older patients. Third, the self-management app for patients (allowing active monitoring, communication with the care team, and a virtual coach with customizable automated feedback) enhanced doctor-patient relationships [30] and facilitated early detection of any problem in the surgical recovery process [31]. Fourth, the assessment and promotion of patients’ physical activity is an effective way of preventing any mobility impairment, which is found in one-third of people aged >65 years [32]. Finally, the region selected for the deployment of the IC program, a large rural area, was especially appropriate as their citizens can benefit the most from community-based IC initiatives that can avoid unnecessary travel to the hospital.

Regarding this study, several strengths and limitations should be noted. Among the strengths, we note the use of a prospective study design with a comparator arm; the use of objectively measured endpoints, such as visits and admissions, in contrast to patient-reported outcomes; and cost and cost-effectiveness assessments. Concerning the limitations, the technological platform supporting the implemented IC model showed substantial improvements throughout the implementation period. This implied that patients recruited near the end of the implementation study had a richer IC experience than those recruited at the very beginning. Similarly, this had an impact on health care professionals, who had to cope with a platform under development and not fully integrated with existing EMRs. Nevertheless, participating patients and professionals showed great acceptability of the IC model and setting [33]. Moreover, directly participating in a dynamic development and implementation process fostered professionals’ engagement and allowed them to propose changes and new features to be developed, which resulted in not a single professional dropping out of the study. Next, the limited number of patients involved in the study had an impact on statistical power. Nevertheless, the current number of participants sufficed to show a significant reduction in robust endpoints, such as unplanned visits and hospital admissions, and showed the cost-effectiveness of the IC program. Finally, only direct costs were considered, although the inclusion of indirect or societal costs would most likely favor the cost-effectiveness of the IC model.

This study aimed to assess the impact of the implementation of an IC model in three domains: (1) patients’ QoL, (2) patients’ use of health services, and (3) health economics. Regarding the QoL domain, the IC model performed as good as the UC arm in the differences-in-differences analysis. This result is in line with the mixed results found in a 2017 review on the impact of IC interventions on QoL [34]. However, it is worth noting that the great increase in QoL obtained after a successful THA or TKA is likely to mask any minor increase in QoL caused by being managed in an IC model. Next, regarding the use of health services, patients in the IC model required 50% fewer unplanned visits after the surgery. This corresponds to the upper margin of benefits reported in reviews about IC interventions between 2000 and 2015, which reported significant reductions in hospital activity ranging from 15% to 50% [35]. These excellent results could be in line with the notion of system-wide cross-organizational care pathways as a successful way to implement IC, in contrast to smaller and narrow interventions [9]. Finally, regarding health economics, IC generated savings from €109.88 (US $132.96) to €126.99 (US $153.66) per patient and was deemed cost-effective. This is in line with reviews stating the potential cost-effectiveness of IC in the management of chronic diseases [36]. However, these savings are lower than the range of US $584-$1434 obtained when applying the same IC model to CCPs with chronic obstructive pulmonary disease or heart failure [13], who are more prone to experiencing unplanned visits and hospitalizations.

When specifically focusing on THA and TKA, previous studies have suggested the usefulness of the different potential components of an eHealth IC model, including telerehabilitation [15], care pathways [18,37], education [38], patient-centeredness [39], mHealth continuous monitoring [40,41], and cross-setting integration [42]. However, to our knowledge, this study is the first to include all these components in a single mHealth-enabled IC model. Overall, our results are similar to or better than those reported in these previous studies, suggesting a moderate additive effect of combining the reported interventions into a single mHealth-enabled IC model.

The World Health Organization has already stated the need for patient-centered IC models to satisfy the health needs of older populations with chronic diseases while keeping costs sustainable [29,43]. With this premise in mind, the CONNECARE project tackled the task of iteratively co-designing a mHealth-enabled IC model with the participation of all key stakeholders: patients; hospital-based surgeons, anesthesiologists, nurses, physiotherapists, and case managers; primary care physicians and nurses; social caregivers; and managers, technical staff, developers, and researchers. This multidisciplinary team envisioned a system-wide cross-organizational patient-centered care pathway, in line with the principles of the 2015 World Report on Ageing and Health [1]. The experience acquired during the co-design and testing process allows us to highlight some key features that future IC models for the management of older citizens undergoing THA or TKA should consider (1) a common cross-setting web-based platform is key for a successful coordination of care across settings and patient monitoring; (2) the habilitation of functional communication channels for the patients can be a key source of savings, as most savings are achieved through the avoidance of unplanned visits; and (3) involving informal caregivers, such as younger family members, can facilitate the adoption of mHealth tools, such as sensors and self-management apps, and make the overall user experience very satisfactory [44]. Moreover, it is key to ensure access to an internet connection at home, as this was the main criterion halting the participation of potential users.

The implementation of a patient-centered mHealth-enabled IC model for the management of patients undergoing THA or TKA successfully empowered patients, effectively connected the different care settings involved, reduced unplanned contacts with the health system, reduced health costs, and was cost-effective. This supports the use of mHealth tools for the implementation of system-wide cross-organizational IC models.