Sunnybrook Health Sciences Centre is a major academic hospital in Toronto, Canada. Toronto is a city of 6 million individuals and the fourth largest urban area in North America. In a typical year during the study, Sunnybrook had over 600 acute care beds, 35,000 admissions, 60,000 emergency cases, 500,000 outpatient visits, 16,000 surgical operations, and 4000 newborn deliveries (total exceeding 600,000 patients annually) [15,16]. Sunnybrook is one of five general hospitals affiliated with the University of Toronto, funded under the universal health insurance of the Ontario government, staffed by more than 10,000 health care professionals including volunteers, and run with a budget of Can $1 billion annually [17]. For perspective, the Ontario region had a population of 14 million individuals, 155 hospitals distributed over 238 sites, and 213 acute care emergency departments [18,19].

The patient portal (named MyChart, not to be mistaken with Epic MyChart) was developed by Sunnybrook in 2005, had an early launch in 2006, became widely available in 2007, and later expanded to 10 other institutions [20,21]. The core functions provided patients read-only access to blood test results, imaging reports, and physician notes. Additional features later developed included access to clinic schedules, pathology reports, microbiology data, electrocardiogram results, and patient diaries. The MyChart system did not support secure messaging, patient annotation, prescription renewals, or targeted education (presumably, patients obtain such services through other channels). A MyChart account is available at no cost to patients, does not require a clinician’s consent, can be accessed by the Internet, and has a standard interface (Multimedia Appendix 1).

The Sunnybrook MyChart system has been subjected to early analyses and small studies. The median user was a middle-aged woman who lived within 100 km of the hospital [22]. Individual patient demographic profiles varied substantially and spanned a wide range of diagnoses. Most patients were ambulatory, had at least one chronic medical disease, and expressed a particular interest in accessing their laboratory test results [23]. Patients had originally learned of Sunnybrook MyChart from word-of-mouth, brochures, or posters in high-traffic areas of the hospital and subsequently taught themselves how to use it without a formal Sunnybrook training program [24]. About half shared their personal access identifier and password with at least one family member. Some Sunnybrook physicians did not immediately embrace the initiative for this patient portal.

We identified the most recent 4 consecutive calendar years of utilization data for the Sunnybrook MyChart system encompassing all complete years of system information (January 1, 2012, to December 31, 2015). This selection interval excluded the initial years of system development, early testing, and major upgrades that preceded widespread launch (April 1, 2007, to December 31, 2011). The analyses focused on aggregate traffic details to respect medical privacy and removed identifiable information to preserve patient confidentiality. This analytic strategy also included patterns where an individual patient might have shared their account details with family or other trusted delegates. The study protocol was approved by the Research Ethics Board of the Sunnybrook Research Institute including a waiver for patient consent on the condition that the study examined only aggregated anonymous utilization data.

Individual patients obtained a MyChart account by submitting a simplified application with personal identification for authentication (denoted as a MyChart registration). The onsite contact office for handling submissions was located adjacent to the hospital cafeteria and delivered as an on-demand service with no appointment. MyChart registration provided patients a temporary password, instructions for website navigation, and access to an unlimited number of website visits (denoted as a MyChart login). In turn, each MyChart login allowed an unlimited number of connections to individual features (denoted as a MyChart pageview). A MyChart pageview could contain differing amounts of information; for example, a single pageview of laboratory results could encompass dozens of individual test results.

The MyChart system routinely tracked activity as part of system monitoring, quality assurance, and ongoing maintenance. These aggregated data were automatically indexed to time and available in an aggregated manner grouped as consecutive days with no exclusions and no missing data. By design, the aggregated data were stripped of individual patient identifiers and expunged of actual detailed demographic or medical information. The purpose was to maintain full patient privacy and data security yet still provide health service research data necessary for evaluating comprehensive patterns of utilization (denoted as MyChart traffic information). MyChart traffic information is not traceable to individual patients, does not reveal individual medical characteristics, and does not breach doctor-patient confidentiality.

The primary analysis used time-trend analysis to describe MyChart traffic information over the study interval. Separate analyses were conducted for each of the primary measures of MyChart activity, specifically, registrations, logins, and pageviews. Secondary analyses examined specific priority pageviews of blood test results, medical imaging reports, and medical consultation notes. We conducted comparative analyses to also examine daily trends (contrasting weekends of Saturdays and Sundays), seasonal trends (focusing on January and July), annual trends (highlighting the first and final year of observation), and resiliency to unforeseen shocks (evaluating the 7-day Toronto ice storm of December 20-26, 2013) [25,26]. The unit-of-analysis throughout was the individual day to avoid small cell sizes and allow adequate statistical power.

We studied the online patient portal at one large hospital to assess aggregated utilization over 4 years. We found that new registrations were uneven and amounted to about 1% of total patients treated at the hospital (8081/600,000). Despite the low uptake, total utilization grew substantially and exceeded 1000 logins daily in the final year. Utilization declined on weekends yet remained reliable despite an unforeseen extreme ice storm. Patients averaged 5 pageviews per login, showed relatively high interest in viewing clinical notes, and had relatively little interest in maintaining a diary [27,28]. Together, these data suggest that online patient portals can provide a popular and reliable system for distributing medical information to a self-selected subgroup of patients.

An important limitation of our study is the absence of detailed data linking patient characteristics, diagnoses, treatments, and medical outcomes: this limitation was necessary for the ethics safeguards to analyze patient activity without patient consent. Our study also did not distinguish how much utilization occurred from one patient with multiple logins versus multiple patients with one login each. The digital data, furthermore, do not track a patient’s experience to test how a recent medical encounter changes subsequent online portal utilization or further nuances that examine which pageviews tend to be viewed first and for how long. The ultimate positive benefits of online patient portals for increasing patient satisfaction, knowledge, motivation, equity, and health outcomes are all topics that remain for future research.

A second limitation relates to generalizability because the Sunnybrook system is not the epitome of patient portals. The user interface can be awkward for some patients with cognitive impairment or those unfamiliar with medical data [29]. The system lacks some functions relevant to ongoing patient care such as secure messaging and prescription medication renewals. The system also has few of the connections necessary for networking between hospitals or physicians outside the region. The background software provides no customization based on an individual’s past search history or tailoring to a patient’s unique profile. All patient portals involve design compromises between functionality, simplicity, reliability, and security [30-34] and are likely to evolve further in future years.

Another limitation of our study is the absence of data on possible adverse effects of communicating large amounts of personal health information to individual patients. One worry is that patients lack formal medical training, misinterpret innocuous items, and can be upset about inconsequential anomalies [35,36]. A related concern is that some patients may obsess about minutiae, repeatedly refresh for updated data, or attempt to embarrass health care providers in a deliberate strategy of one-upmanship [37]. The direct viewing of physician notes also allows patients to detect undiplomatic language, spelling mistakes, and other awkward professional lapses [38]. Ultimately, these medical portals empower patients and thereby shift the balance of power around a clinical encounter.

A related limitation is the lack of data on physician behavior since health care providers are aware of patient portals and may adjust their documentation accordingly. One reaction is to stifle candid collegial dialogue due to concerns about later misinterpretation by the patient [39,40]. Another reaction involves unnecessary clutter or auto-text inserted as a medico-legal defensive strategy [41,42]. A further reaction is for health care providers or families to focus excess attention on online data and neglect the patient’s symptoms, physical findings, or other changes found at the bedside, not online [43,44]. A comprehensive online patient portal, moreover, could be counterproductive for a patient diagnosed with a personality disorder who can immediately access a physician’s notes.

The Sunnybrook online patient portal has strengths that justify future research. The system provides a flexible platform for patients to help integrate clinical services, enhance self-management, and potentially improve health outcomes. Patients can easily share their medical information with family members or community physicians for broader system integration [45]. Patients can also use the system to incorporate tracking tools for personal health status (eg, ambulatory blood pressure monitors) [46]. Patients have frequent opportunities to comment on their experiences through user surveys and, in turn, developers are integrated with the larger health care team to facilitate system advances [47]. All these benefits might increase further with ongoing system expansion. The sustained high rates of utilization found in this study therefore suggest potentially sufficient sample size and statistical power necessary for planning future research.

The utilization of online patient portals observed in this study indicates the technology is popular for a small subgroup of patients who register [48,49]. If each pageview were assumed equal in value to a 50¢ payphone call, for example, the net utilization in this study would be worth Can $1 million annually. If uptake could increase to 20% of total patients and patterns persisted, furthermore, the net utilization might be worth Can $25 million annually (Can $10 billion if extrapolated to all of Canada). Regardless of assumed uptake, this study also implies such value can accrue for years without a major security breach or software failure. As the Internet continues to transform society and individuals grow increasingly comfortable online, more patients may demand information technology solutions to help access personal health data [50,51]. Other hospitals, therefore, might consider an online patient portal for future years.