Effective health care requires patient-centered and integrated health services coordinated across multidisciplinary teams and delivery settings [1,2].

Efficient communication and health information exchange between patients, providers, and caregivers across settings is currently lacking, jeopardizing patient safety and health system costs [3-5]. To address this gap, health systems are increasingly investing in patient portals that enable timely and comprehensive information access through a single channel by patients [6,7]. Broadly speaking, patient portals offer secure electronic access to clinical information collected by one or more health care institutions [8,9]. They are related to but distinct from personal health records, which store health data compiled, managed, and updated by patients [10].

Comprehensive patient portals that enable access to data from multiple health services (ie, internal and external laboratories, diagnostic images, and discharge reports) have been shown to reduce duplicate testing [11] and enhance communication [12,13]. The primary argument for patient portals is not that they intend to create value for money; rather, they foster patient empowerment and self-management and improve patients’ engagement in their care [14]. The downstream anticipated impact of this includes improved health outcomes [12,13,15-18] and reduced health resource usage [19-22]. However, suboptimal implementation threatens the potential realization of benefits. For instance, approximately 90% of the US health care systems and providers offer patient portals; however, only 15% to 30% of patients report using these resources [23,24]. Australia has noted similar adoption rates, with only 22% of citizens registered on their national portal since its launch in 2012 [25]. Several factors hinder adoption and use, including privacy concerns [18], low motivation to enroll [26], design and usability problems [18,27,28], and issues related to health or technical literacy [28] and equitable access [29]. In addition, the implementation of a comprehensive multi-institutional portal requires system interoperability and interorganizational coordination, which is often difficult to achieve across fragmented health services [7,15].

Although barriers and facilitators to patient portal adoption are well documented [9,14,17,30], there is a dearth of evidence on how the implementation process interacts with individual and organizational contexts [13]. This makes it difficult to discern the specific processes or factors that underpin the success or failure of the implementation. In response to this knowledge gap, we conducted an evaluation of a multi-institutional patient portal deployment to understand the experiences of patients, health care providers, and administrators.

The specific objectives are to explore how the implementation approach impacts the adoption and usage of a patient portal and to identify lessons learned when implementing multi-institutional patient portals that transcend traditional boundaries and disparate systems.

This multimethod study included an analysis of routinely collected usage data that were extracted from the portal by the vendor to understand use patterns during the first 15 months of implementation (between August 2018 and October 2019). In addition, qualitative semistructured interviews were conducted with patients (and their informal caregivers when applicable), health care providers, and administrative stakeholders to understand how the implementation approach influenced patient and provider experiences and overall usage. Ethics approval was received from the Research Ethics Board of the Women’s College Hospital (REB approval no. 2019-0035-E). The reporting of this study was guided by the Consolidated Criteria for Reporting Qualitative Research checklist [31].

Southwestern Ontario (SWO) is a region of Ontario, Canada, that includes rural, suburban, and urban populations. The approximately 3.6 million residents of SWO represent 30% of Ontario’s population [32]. Over 45,000 health care providers in SWO securely access publicly stored patient information through a regional webportal (ClinicalConnect) [32]. ClinicalConnect is a provider-facing viewer that consolidates data from 72 acute care hospitals, 4 home and community care organizations, and 4 regional cancer care programs in SWO [32]. It enables access to 4 provincial data repositories that house diagnostic imaging reports, drug information, laboratory results, and acute care information.

In 2017, Hamilton Health Sciences (HHS), a hospital network consisting of 7 hospitals and a regional cancer center, received funding from Canada Health Infoway, a not-for-profit organization funded by the federal government, to deploy a multi-institutional patient portal for residents in the region [33]. This initiative represents one of the largest deployments of a patient-facing digital health access channel in Ontario. To develop the portal, HHS collaborated with Sunnybrook Health Sciences Centre, a Toronto hospital that adapted its hospital-developed patient portal (MyChart) for the SWO region. To enable access to information across multiple systems, MyChart was integrated with ClinicalConnect rather than being directly tethered to individual institutional electronic patient record systems. Organizations in SWO contributing data to ClinicalConnect were required to sign agreements to enable their data to flow to MyChart for patient access.

The MyChart implementation was launched in August 2018, with a target of 65,000 registered users by December 2019. MyChart was rolled out in stages, as site agreements were finalized with the participating organizations. Patient enrollment was only available in person at each site at the time of this study but was later made available via the internet in March 2020 because of the COVID-19 pandemic. Participating sites were responsible for independently determining what data would be uploaded and with how much delay and for creating implementation processes, such as onboarding of the patient and provider to the portal. Participating sites also had the flexibility to control the availability of portal features (eg, direct messaging) for patients accessing care at their sites (a full list of functionalities is given in Textbox 1). As of January 20, 2020, there were 48 sites contributing data to the regional MyChart and 38 sites actively offering the portal to their patients out of a possible 57 sites that had signed data-sharing agreements in SWO.

Aggregate quantitative data were analyzed using descriptive summary statistics (eg, calculating frequencies and averages) with Excel software (Microsoft Corporation) to understand use patterns, including which features of the portal were most frequently used and how often the portal was accessed.

Qualitative data were analyzed using an inductive thematic analysis approach, which aimed to identify recurring patterns across a data set [35]. Two researchers (JB and JG) developed a preliminary codebook by independently coding the first 3 interview transcripts. The 2 researchers then discussed the codes, clarified and resolved discrepancies, and created a project codebook. The codebook was then discussed with the research team to ensure alignment with the study objectives and applied to additional transcripts as interviews were completed. Throughout this process, JB and JG met periodically to iteratively refine the codebook to reflect new codes, merge related codes, and resolve discrepancies. Once coding was complete, codes were synthesized into preliminary themes that were mapped back to the study objectives. Refinements and specifications of thematic categories and subcategories and relationships between themes were determined through in-depth discussion and negotiated consensus between members of the research team (JB, JG, JKF, and LD). NVivo 12 software (QSR International) was used to assist with coding and analysis.

Between August 2018 and October 2019, a total of 15,271 registration emails were sent, with 67.01% (10,233/15,271) of patients registered for an account across 38 sites. The median number of patients registered per site was 19, with considerable variation (range 1-2114). Of the total number of sites, 55% had less than or equal to 30 registered patients (21/38), whereas only 2 sites had over 1000 registered patients. Among the registered patients, 92.00% (9,414/10,233) logged in at least once during the 15-month evaluation period. At the time of this study (October 2019), only 23.69% (2,424/10,233) had logged in more than once during the preceding month, with less than 3.83% (392/10,233) accessing the portal more than five times. High-traffic information pages included radiology exams and laboratories, discharge summaries, pathology results, and medications (Table 1). Only 2.98% (305/10,233) of registered users appointed a delegate, and less than 1% (73/10,233) actively shared information with their providers. A small number of patients (140/10,233, 1.37%,) independently recorded data in their patient health records via the portal.

A total of 42 individual interviews were conducted with stakeholders from 17 organizations within the SWO region to understand how the implementation unfolded and the factors underlying the adoption rates. The majority of participants (11/42, 26%) were recruited from HHS, followed by the Heron Perth Health Alliance (7/42, 17%) and the London Health Sciences Centre (5/42, 12%). The interview participants included 18 administrative stakeholders involved in the MyChart implementation, 16 patients, 7 health care providers, and 1 informal caregiver (Table 2). Administrative stakeholders included individuals involved in clinical leadership, clerical roles, patient experience teams, and information technology (IT). Among the patients and caregiver interviewed, 76.5% (13/17) had used MyChart, 11.8% (2/17) had registered but had not accessed it because of log-in challenges, and 11.8% (2/17) did not have access because of challenges with registration. The average age of patient or caregiver participants was 54 years, and all indicated that they were managing a chronic illness.

Five key themes emerged. They described how the implementation process influenced adoption rates and lessons learned that can be leveraged to increase the uptake of similar patient portal models.

Multi-institutional patient portals, such as MyChart, that collect and house information across organizations within a given region can enable more effective patient management by streamlining communication, access to clinical information, and service coordination [22]. The downstream benefits of such portals include increased patient engagement [14] and a reduction in emergency care and hospital admissions [20,22]. Although randomized controlled trials assessing patient portals are lacking, some randomized controlled trials have shown that patient portals can reduce hospital readmissions [36], reduce office visit rates, and result in greater adherence to treatment in comparison with control groups [16]. Our findings suggest that data comprehensiveness, organization-wide deployment, provider engagement, and senior leadership endorsement are central to achieving these aims.

Despite the high perceived value of multi-institutional patient portals, MyChart was adopted suboptimally. The population in SWO is approximately 3.6 million [32], implying that less than 0.3% of individuals within this region registered to use the portal at the time of this analysis. In comparison, the adoption of patient portals in other jurisdictions is estimated to reach 5% to 10% of the targeted population per year, with uptake in larger scale (ie, national or regional) implementations lagging behind smaller, more targeted deployments [37]. The slow growth of larger scale patient portals is attributed to challenges in obtaining alignment and system interoperability across fragmented health services [7,37]. In the absence of a cohesive regional implementation strategy, participating organizations of MyChart had to independently establish processes for marketing patient enrollment and training, resulting in variations in adoption and possibly heterogeneous patient experience. Further, a lack of dedicated resources, change management support, senior leadership endorsement, and clinical buy-in impeded success in several organizations. Strategies to improve change management processes and facilitate senior leadership and clinician support include clearly articulating the value of the technology, building consensus on key decisions and operational processes built around a strategic vision, and investing in the required infrastructure and resources, such as interoperable systems and staff training, to foster success [38].

Our findings highlight strategies for the successful implementation of patient portals. First, the implementation strategy needs to align with the core objectives of the technology. In this case, the portal should focus on enabling access to comprehensive clinical information as an initial priority, given that this is the core functionality of interest and would provide the most benefit to patients. Other features (ie, direct messaging and appointment booking) can be explored in consultation with end users once a plan for comprehensive data access is established and successfully operationalized. Second, implementation processes should mirror patient experience. A blanketed rollout approach across an entire organization, rather than limiting deployment to certain departments, was perceived to be more conducive to facilitating information access that transcends traditional health system boundaries. Organizations that implemented the portal uniformly across the organization were able to more effectively advertise and communicate to patients about the portal. Conversely, segmented implementation exacerbates fragmented access to information, limits patient and clinician awareness and shared understanding of the purpose of the technology. This can create a disconnect for patients as they interact with different services within and across organizations. It is important to note that introducing functionality across an organization does not necessarily imply a one-size-fits-all approach as some flexibility is required to adapt to specific population needs [38]. For instance, special considerations regarding sharing clinic notes, proxy consent, and age of consent should be made for pediatric, mental health, and geriatric patients because of concerns regarding their capacity and autonomy [39]. However, interviewees raised the important point that these issues were better governed by universal best practices rather than by individual organizational idiosyncrasies.

Patient portals and other digital technologies can create value for organizations and health systems but only if the surrounding sociocultural factors are considered [40]. Effective leadership and clinical endorsement of technology can reduce behavioral resistance to change [41]. Patient portals may also precipitate changing dynamics between patients and providers, which underpins the cultural shift toward patient-centered care. Similar to other studies [42,43], we found that tensions between paternalistic and patient-centered medicine need to be resolved to facilitate widespread portal use. In tandem, individuals must be equipped with the necessary time, resources, and tools to carry out activities required for adoption, such as onsite training, enrollment, and technical support [43].

At the organizational and system levels, identifying and developing the right infrastructure is an essential component of strategic planning [38]. Multi-institutional patient portals require standard policy and technical infrastructure to enable data sharing that is consistent across sites. This includes identifying guidelines and best practices to establish access policies (ie, proxy access, age of consent, and data delays). The lack of basic integration and interoperability across institutional boundaries impedes the data comprehensiveness required for portals to function effectively in line with their goals [7,15,42]. The exponential growth of digital technologies across health systems implies the need for a degree of interoperability in alignment with more integrated health care [40,42].

Future research should focus on effectively describing and evaluating the implementation strategies that surround multi-institutional patient portals (eg, the use of champions, patient and provider training, addressing beliefs, etc) to identify effective strategies for promoting uptake. In addition, further studies should focus on evaluating the effectiveness of patient portals against their stated aims, including increased patient engagement in care, improved knowledge, and improved patient experience.

The generalizability of our findings is limited to the early stages of implementation. Our sample may have been biased toward early adopters of technology or participants with limited user engagement. Most patient and provider participants had limited interactions with the patient portal, and their perspectives may not reflect the experiences that emerge alongside a more mature patient portal model. Despite this, all participants believed in the value of multi-institutional patient portals that offer patients comprehensive access to their health information in contrast to institutionally tethered offerings that do not centralize health information from across the health system. It is important to note that participants with limited user engagement did not endorse the current operationalization in its entirety; rather, they described the features and functions of a future state patient portal that would provide value to patients. Considerable benefits would be gained from evaluating the factors associated with sustained engagement in such a model. Although most interviews were conducted with individuals who had experience with the portal, the level of knowledge of and exposure to the portal varied. Owing to slower-than-anticipated adoption, our recruitment approach did not seek to discern between high and low adopters, as there were few high adopters at the time of the study. Further evaluations of MyChart should examine if there are characteristic differences (eg, based on patient population, region, or institution) between high and low users. Perspectives from diverse and often hard-to-reach patients were not included (eg, newcomers to Canada, non-English speakers, individuals experiencing housing insecurity); they may experience barriers to accessing technology. Consequently, further work is needed to engage with these populations and determine the impact on access and patient engagement from an equity perspective. Despite this, our study provides useful strategies to inform implementation planning at organizational and system levels.

Although multi-institutional patient portals can enable efficient access to clinical information from across the health system, successful implementation can be affected by several factors. Without proper management and planning, portals can suffer from minimal adoption from patients and poor support from providers. Data comprehensiveness is the foundational component of patient portals and requires aligned policies and a key base of IT infrastructure across all participating sites. It is important to look beyond the category of the technology (ie, patient portal) and consider its functionality (eg, data aggregation, appointment scheduling, messaging) to ensure that it aligns with the underlying strategic priorities of the deployment. Finally, it is critical to establish a clear vision and ensure buy-in from organizational leadership and health care providers to support a culture shift that will enable meaningful and widespread engagement.