Between 2018 and 2024, OMD conducted an annual survey of clinicians as an evaluative mechanism (internal purposes only) to provide insight into the impact of physician-facing initiatives and to inform future priorities. Within this survey, there was a section on challenges, with optional open-text questions that were reviewed and coded for themes. The cumulative findings suggested that a focused interrogation of the impact of technology on physicians was warranted.

We designed and conducted a self-administered electronic questionnaire on HIT-related burnout, including potential factors, guided by (1) a literature review on HIT-related factors contributing to physician burnout, (2) a review of qualitative responses to annual OMD surveys on challenges with HIT, and (3) team discussions of evidence to date. The questionnaire was piloted by the Nova Scotia E-Health Committee, comprising 9 physicians and 2 medical learners, and then refined with support from OMD-affiliated physicians. The final 25-item instrument consisted of 23 mandatory closed-ended and 2 open-ended questions. The first question asked whether respondents were currently using HIT in their primary practice; those who responded “no” were excluded from subsequent questions on HIT. Three potentially sensitive questions in the “About You” section included “prefer not to answer” as an option. Two versions containing the same content specified regional consent and identification details (Multimedia Appendix 1). Microsoft Dynamics 365 Customer Voice was used to develop and distribute both questionnaires separately via links embedded in email and on social media. Respondents could complete the questionnaire at their own pace and edit previous responses, if necessary. No incentives were provided for survey completion. Survey design is reported according to the CHERRIES checklist [19].

The questionnaire was launched on February 14, 2024, with an in-field period until April 5, 2024. In Nova Scotia, distribution used nonprobability voluntary response sampling [20] via Dalhousie Faculty of Medicine (FoM) communications, the website, social media, and department newsletters. The Doctors Nova Scotia e-newsletter and discussion forum and the Nova Scotia Health (NSH) newsletter also distributed the e-questionnaire and reminders. In Ontario, the questionnaire was distributed using OMD’s customer relationship management software to an estimated sampling frame of between 24,000 and 31,000 Ontario physicians and other clinical staff. Reminder emails to complete the questionnaire were sent at regular intervals. Additionally, social media posts were shared throughout the distribution period on behalf of both OMD and the Ontario Medical Association. Unique identifiers in the form of College of Physicians and Surgeons of Ontario registration numbers precluded duplicate responses. Eligibility criteria of “current use of HIT” and “family physician” or “specialist” were applied to ensure a valid sample of respondents according to study objectives.

The Nova Scotia questionnaire was approved by the Research Ethics Board at Dalhousie University (Halifax, Nova Scotia, Canada, file number: 2023‐6993). As OMD operates under both a provincial government and professional association structure without an institutional ethics board, the Ontario survey was deployed with informed consent language (in consultation with the Dalhousie team and OMD’s legal team) in the survey header, asking participants to read the consent text, which also cites OMD’s privacy policy on information sharing [27]. Participants were asked to confirm they had read the text and that they agreed to participate. Emailed invitations to complete the survey also used language regarding the voluntary nature of participation.

Because the objective of the study was to explore physician experiences with HIT, 138 Ontario and 20 Nova Scotia nonphysician clinicians and clinic staff, as well as 24 Ontario and 4 Nova Scotia non-HIT physician users, were excluded, resulting in 1245 Ontario and 136 Nova Scotia physician HIT user respondents. Data on clinician and clinical characteristics are reported in Table 2. Significant differences in self-reported tech-related burnout by characteristic are noted for the Ontario sample only.

Figures 2 and 3 show the order from most to least burdensome administrative tasks, with bars representing the percentages of the top 3 burdens. Over half of the Ontario sample ranked the following administrative tasks as a top 3 burden: managing reports sent through your EMR (929/1245, 74.6%), managing communications related to patient care (810/1245, 65.1%), and inputting patient data into your EMR (675/1245, 54.2%). The top 3 burdens for the Nova Scotia sample were managing communications related to patient care (n=72, 47.1%), logging in and out of health information technology platforms (60/136, 44.1%), and inputting patient data into your EMR (55/136, 40.4%). For Ontario specialists, the top 3 most burdensome tasks were managing communications related to patient care, inputting patient data into your EMR, and managing reports sent through your EMR. For Nova Scotia specialists, the top 3 most burdensome tasks were logging in and out of HIT platforms, managing communications related to patient care, and retrieving patient information. For both Ontario and Nova Scotia family physicians, the top 3 most burdensome tasks were managing reports sent through your EMR, managing communications related to patient care, and inputting patient data into your EMR.

Table 3 provides data on perceptions of HIT, as well as motivations for using HIT. For the Ontario sample, significant differences in burnout levels across the agreement categories for certain perception and motivation items are noted. Most Ontario and Nova Scotia physicians agreed with the following beliefs regarding HIT: HIT facilitates coordination of patient care (978/1245, 78.6% in Ontario; 82/136, 60.3% in Nova Scotia) and HIT improves the quality of patient care (819/1245, 65.8% in Ontario; 76/136, 55.9% in Nova Scotia). Most Ontario physicians agreed that HIT makes it convenient for me to access data (861/1245, 69.2%), while over half of Nova Scotia physicians agreed that HIT will revolutionize the future of medicine (82/136, 60.3%). In terms of motivations for using HIT, the most frequently cited reason for using HIT among Ontario physicians was that they either believed HIT would help them provide better patient care or believed HIT would be of value to them, both internally driven sources of motivation. Among Nova Scotia physicians, the most frequently cited reason for using HIT was that they were required to use it, an externally driven source of motivation.

Table 4 shows descriptive statistics on physicians’ work week. Significant associations with technology-related burnout were noted for the Ontario sample. Ontario and Nova Scotia physicians in our sample reported an average work week of 37.4 (SD 14.0) and 39.1 (SD 14.5) hours, respectively, about half of which was spent on direct patient care (50.2%, SD 16.4) in the Ontario sample. Around 26% and 20% of Ontario and Nova Scotia physicians’ time was spent on administration related to patient care. Ontario and Nova Scotia physicians worked an average of 8.6 (SD 7.1) and 4.8 (SD 5.7) additional hours per week, respectively, offsite on administration related to patient care.

In terms of HIT support, over 40% in Ontario (539/1245, 43.3%) received support from their technology vendor; over half in Nova Scotia (69/136, 50.7%) reported support from their on-site technical support. In both cases, these forms of support were deemed at least acceptable (75.4% for Ontario and 79.7% for Nova Scotia; Table 5). For the Ontario sample, significant differences by tech-related burnout are noted.

In the Ontario and Nova Scotia samples, the median level of self-reported tech-related burnout was 7 and 6, respectively; levels with the highest frequency counts were 7 or 8 in the Ontario sample and 5, 6, or 7 in the Nova Scotia sample. A little over half of the Ontario sample reported a tech-related burnout level of at least 7, while almost two-thirds reported a tech-related burnout level of at least 6. In the Nova Scotia sample, most respondents reported a tech-related burnout level of at least 5. Table 6 shows the findings from an additional question on tech-related burnout.

A series of KW rank sum tests on the Ontario data, given its larger sample size, yielded significant results for support quality (KW χ₄²=90.6; P<.001), proficiency (KW χ₄²=25.0; P<.001), and clinician type (KW χ₁²=21.8; P<.001). In this sample, higher quality support and greater proficiency yielded lower levels of self-reported tech-related burnout, whereas lower-quality support and greater (less) proficiency yielded lower (higher) levels of self-reported tech-related burnout. Specialists also tended to report lower tech-related burnout. More unfavorable or favorable perceptions of HIT yielded higher or lower levels of self-reported tech-related burnout. Specifically, agreeing that HIT is the reason for job stress (KW χ₄²=523.98; P<.001), prevents patient care (KW χ₄²=305.6; P<.001), or negatively impacts their practice (KW χ₄²=360.8; P<.001) yielded significantly higher levels of self-reported tech-related burnout. Conversely, all other items yielded statistically significant differences in median self-reported tech-related burnout, with more favorable views of perception linked to lower levels of self-reported tech-related burnout. Pearson product moment correlation tests also yielded a significant correlation between the overall score on perceptions and tech-related burnout (r=−0.56; P<.001). Hours per week at practice (r=0.19; P<.001), hours per week (r=0.33; P<.001) or percent of time (r=0.21; P<.001) spent on administration related to patient care, and extra hours per week spent on administration related to patient care (r=0.34; P<.001) yielded weak-to-moderate positive correlations with self-reported tech-related burnout.

Systematic stepwise regression tests yielded a robust, well-fitting model explaining about 60% of the variation in self-reported tech-related burnout (R²=0.60; F(7,1237)=266.515; P<.001). Table 7 shows the performance of individual factors in the regression model. Percentage burnout tech (ie, the respondent’s estimated percentage that technology contributes to their burnout) was subsequently removed due to difficulties in interpretation, which resulted in a model accounting for about half of the variation in self-reported tech-related burnout, with 6 stable, meaningful, and potentially actionable factors. All ordinal factors evidenced monotonic, near-linear trends. Follow-up statistical tests were conducted to confirm the regression model. Neither multicollinearity (ie, through examination of variance inflation factors) nor autocorrelation among residuals (ie, through the Durbin-Watson test) was observed. A Bonferroni test for outliers indicated no studentized outliers (P<.05).

Both coding teams began by assigning responses to the main qualitative question, “How does health information technology contribute to or help to alleviate your burnout?” into 2 themes: negative (corresponding to “contributes to”) and positive (corresponding to “alleviates”). With the larger Ontario sample, QSR NVivo software was used to facilitate coding, and automated sentiment analysis was attempted but was found to be inaccurate upon review—with frequent misattributions of sentiment due to expressive ambiguity or sarcasm, for example. As described in the “Methods” section, blank or difficult-to-interpret responses were excluded from coding.

Using this approach, 146 references were coded as negative and 31 as positive in the Nova Scotia sample, with 864 coded as negative and 657 as positive in the Ontario sample. Each coding team then proceeded with open coding for themes that emerged under each of the 2 larger categories. Codebooks were developed for each province’s sample with codenames and definitions. Once this step was completed, the 2 coding teams compared codebooks to identify similarities across themes—with a focus on aligning code definitions rather than codenames.

From this we discovered that the 2 most common positive themes in both groups were aligned, although in reverse order. The first of these in the Ontario sample dealt with HIT’s capacity to facilitate data management and accessibility (coded as “data management”—168 references; in the Nova Scotia sample coded as “accessibility”—7 references, the second most common). The second refers to administrative efficiency (coded as “regular task completion”—112 references; in the Nova Scotia sample coded as “efficiency”—11 references, the most common).

The considerably larger set of references for the negative statements also offered greater complexity and some differences between the samples. The highest frequency of negative statements in the Ontario sample referred to report management. The second most common theme in the Ontario sample and the most common in the Nova Scotia sample were aligned: the challenges of systems that “don’t talk to each other,” that is, do not interact easily (some variation of which we saw frequently), so that accessing all the necessary patient data requires logging in and out of multiple unconnected platforms. These responses often cited the interruptive nature of recalling different login information and completing dual-factor authentication. These types of references were initially coded as “interoperability and integration” in the Ontario sample and “fragmented” in the Nova Scotia sample.

Otherwise, there were additional differences in the remaining negative codes. Table 8 shows the 4 most common negative subthemes coded in each sample, including the codename, definition, number of references for each code, and a sample quote.

Despite notable differences between the provinces in HIT maturity, complexity, adoption rates, practice models, and physician characteristics, common themes emerged surrounding HIT involvement in burnout. While physicians identified many positive aspects of HIT, our findings show that most physicians self-assessed their tech-related burnout level as high: 6 out of 10 in Nova Scotia and 7 out of 10 in Ontario. On average, Ontario respondents reported that 43% of their overall burnout is tech-related, while Nova Scotia respondents reported 37%. Ontario respondents reported greater tech-related burnout and attributed a higher percentage of their overall burnout to technology. Over a third of burnout in our respondents’ self-reports is attributable to HIT.

The top-ranked administrative burdens share commonalities across provinces. Nova Scotia respondents identified managing communications as their primary burden; in Ontario, this ranked second after managing reports due to the effects of nonstandardized implementation of Health Report Manager (HRM) in hospitals (a long-standing issue, which is in the process of being addressed in the province). Both samples ranked inputting patient data into their EMR in third place and logging in and out of technology platforms in their top 5 (second in Nova Scotia and fifth in Ontario). This issue emerged strongly in the qualitative material, which revealed frustrations with logging in and out of non-EMR-integrated platforms for referrals, data portals, etc, which all require unique logins, passwords, and dual-factor authentication. Responses suggest that physicians experience these as both adding to cognitive burden and taking time away from patient care. In both samples, ongoing change initiatives—the provincial rollout of OPOR in Nova Scotia and both HRM improvement initiatives [28] and the Primary Care Action Plan in Ontario [29]—are likely to have an impact on the future experience of administrative burdens and perhaps shift these rankings. Other initiatives at the national level to improve vendor attention to interoperability and standards for HIT solutions brought to market may also have an impact on the physician experience. Bill S-5, the Connected Care for Canadians Act, would require HIT vendors to make the technology they sell to clinics interoperable, with penalties for “data blocking.” [30-32] Further, a national effort to standardize data content nationwide is underway with the Pan-Canadian Health Data Content Framework [33]. In the meantime, expensive HIT tools are outpacing our health system’s capacity to adapt. For example, auto-population of forms, patient data integration, and AI assistance are not uniformly regulated or implemented across Canadian practices and provinces [34].

One of the strengths of our approach is the elicitation of candid physician comments via the open text survey questions and our subsequent analysis. This has allowed us to see more layers of the physician experience, especially the complex “love/hate” relationship physicians have with their HIT. In the qualitative material, we found commonality across the Ontario and Nova Scotia samples in terms of the expressed positives of HIT, particularly for data management and administrative efficiency. These align with the reasons HIT was developed and implemented, particularly where government investment is involved. We note with interest that while both samples showed common issues with interoperability and integration (in line with the quantitative results), there were differences in how respondents expressed their other challenges with HIT. In the Ontario sample, we found high counts of references to specific issues, for example, incoming document management and referrals; this was not the case in the Nova Scotia sample, where “usability” was a common theme. This may be due to differences in the samples (ie, higher counts of specific mentions in the larger Ontario sample). We also suggest that the sample from Ontario’s more mature HIT environment may have progressed beyond general usability concerns to complaints about more specific challenges.

Perceptions about HIT showed some agreement with positive statements about technology. This tended to be stronger in Ontario, perhaps due to that province being a more seasoned HIT landscape than Nova Scotia and some anticipatory concern in Nova Scotia about the province’s upcoming move from a largely paper-based system to OPOR. In both samples, more than half of respondents agreed with positive statements about the impact of HIT on patient care and data access. However, about half of each sample also strongly agreed that HIT prevents spending more time with patients. Indeed, both samples reported spending about half their time on direct patient care, with a notable proportion on administration (approximately 20% in Nova Scotia and 25% in Ontario). These particular (types of) perceptions are conceptually similar to tech-related burnout; therefore, one would expect robust associations. However, findings were moderate and somewhat unclear, which may indicate that perceptions are mediated or moderated by other yet unstudied attributes. Many open text responses contained some version of “I spend more time with the computer than with my patients” related to documentation and accountability expectations that increase screen-related tasks and dominate the patient encounter. These negative perceptions associated with administrative burdens compete with the perceived benefits to patient care. This aligns with findings elsewhere regarding the contribution of daily frustrations, task, and workload burden to burnout [4,7] and underscores many physicians’ complex relationships with HIT.

HIT is (or should be) a highly modifiable contributor to burnout; improving physicians’ experience of HIT could decrease some burnout overall. Physicians may see the benefits of HIT for patient care in principle; however, a physician’s ability to yield optimal value from HIT depends on usability, ease of switching between platforms and nonstandardized forms and referrals, and lessening administrative burden, all of which influence time spent on patient care. HIT co-designed with physicians [35] could lead to interoperability and usability standards that better address the needs of a variety of different practice types, physician characteristics, and proficiency levels—with positive effects system-wide. A focus on human factors in the design of HIT can offset the usability challenges that may be unintentionally introduced by technology vendors in disregarding the realities of clinical practice [36-38]. An example of co-design supporting interoperability exists in the Belize Health Information System implementation, where Canadian and Belizean health professionals worked with community input for design and deployment [39,40]. The result was a system that was easy to use and reduced administrative tasks, including searches and redundancies—while also generating low development and maintenance costs and improving patient outcomes. Successful implementation of HIT relies on the involvement of end-user health care professionals at the outset to better translate the realities of clinical practice workflow to technological solutions.

The OPOR initiative presents an opportunity for a prestudy and poststudy in the Nova Scotia context, similar to other preimplementation or postimplementation studies in this area—particularly to investigate the impact on dedicated patient care time [41]. In Ontario, government-led initiatives such as the Primary Care Action Plan [29] intend to address several areas of HIT that contribute to administrative burden; plans to standardize implementation of HRM in hospitals will also address a large contributor to unwieldy inboxes [28]. A follow-up questionnaire pursuant to implemented changes could explore whether these have had an impact on administrative burden and burnout. There is also an opportunity to further investigate differences between specialists and primary care physicians. While the Nova Scotia sample had a higher representation of specialists than the Ontario sample, the area of specialty was primarily pediatrics, where physicians often act as de facto family physicians, providing incremental care over years to the same families. As we did find lower levels of self-reported tech burnout among specialists compared to family physicians, capturing the difference between “true” specialists (those who provide secondary rather than primary care) and family physicians may provide insights from the greater contrast in practice style. Although there are similarities across specialists and family physicians in terms of most burdensome tasks, the qualitative analysis in our study suggests that the reality of administrative burden in clinical practice may differ between the 2 groups. In turn, this may be contributing to tensions (eg, primary care physicians express frustration with nonstandardized referral forms and the extra administrative burden these forms generate).

These findings are preliminary and exploratory and provide a foundation on which to further explore tech-related burnout. We recognize as a limitation that our measure is subjective, based on self-report, and not a validated objective measure. These findings are based on perceptions and associations between self-reported items. Because we used multiple avenues to promote the questionnaire, respondents were not prevented from accessing the questionnaire more than once; however, we controlled for duplicate responses via unique identifiers. Because we used various sampling methods and exclusion criteria, it is difficult to discern the sampling frame and calculate response rates. The sample is unlikely to be representative. There are limitations to generalizability associated with the small Nova Scotia sample and lack of variability in characteristics that preclude certain statistical tests, hinder precision, and potentially introduce bias, including overestimation of effect sizes. Certain response biases (eg, respondents with higher levels of burnout may have been more motivated to participate), common in this type of research, may have influenced the analysis. We attempted to address related biases in our analysis and reporting. Because findings primarily reflect physician experiences within 2 Canadian provincial health systems, they may not reflect or represent experiences in other health care settings. Although measured on a scale following best practices and adapted from similar scales noted in the literature, our tech-related burnout scale has not yet been validated and is subject to interpretation. Although yielding a significant predictive model, it is difficult to ascertain what an increase of a burnout level of “5” to “6” entails clinically. Qualitative research may help to describe these types of scales. The inductive qualitative coding approach, in which we created separate codebooks for Ontario and Nova Scotia, posed limitations to the comparability of that coding. Further, thematic coding may be subject to bias due to the small number of coders involved (2 for Ontario, 1 for Nova Scotia). Furthermore, this analysis should be viewed as exploratory as we continue to understand tech-related burnout and further refine our approach. We note evidence of validity via the strong correlation yielded between our burnout scale and “Percent burnout from technology.” Furthermore, some of the regression factors were measured on scales despite evidencing monotonic, near-linear trends from Bayesian generalized nonlinear multilevel modeling tests. Finally, self-report questionnaires are subjective, so we did not secure objective measures of either burnout or time usage. We followed best practice approaches such as item checks of respondents’ attention and piloted ease of questionnaire completion. Our findings are best interpreted as reflecting respondent perceptions and experiences.

Our study provides a broad picture of the physician experience of HIT in the 2 settings we studied. Physician needs may change depending on their location, practice type, specialty, scope of practice, and day-to-day schedule. HIT should have the flexibility to accommodate these needs and be tailored to the people it serves. For this to happen, physicians need to be deeply involved, including in the design of programs to simplify clinical workflows, for example, the use of AI in clinical decision support. Physicians expressed a need to spend more time with patients than with HIT. The current plethora of largely unregulated vendors for an increasingly diverse set of platforms to address the needs of clinical practice, and the lack of full patient-centered integration, may result in high costs for less user-friendly systems. This study reinforces the complexity of HIT systems and their impact on administrative burden. Ongoing involvement of physicians—who actively use these systems—in HIT design and implementation is essential.

This study contributes practical relevance to the growing literature examining the relationship between technology and burnout [7,8,42-45] with analysis and interpretation of physician perspectives on their complex relationships with HIT. HIT may facilitate patient data collection, management, and use, but it can also add cognitive and administrative burdens to physicians’ everyday work life. HIT is not inherently protective or harmful; rather, physician experience depends heavily on usability, interoperability, workflow alignment, and meaningful physician involvement in design and implementation to address their clinical realities.