Technological innovation has expanded the horizons of medicine in recent decades [1], but these advances have been accompanied by increased clerical burden among physicians and other health care providers. Electronic health record (EHR) adoption has improved emphasis on quality monitoring, billing accuracy, and research, but it has often resulted in redundant documentation and inefficient workflows [2-6]. As a result, physicians have greatly increased interactions with a variety of health care information technologies (HITs), with their interactions commonly encompassing a combination of direct clinical and nonclinical goals [7]. For example, outpatient physicians now spend up to twice as much time interacting with EHRs as they do with patients [8,9].

Concurrently, symptoms of burnout, including emotional exhaustion, have risen to epidemic proportions, now affecting over 500,000 US physicians and costing the US health care system $4-5 billion annually [10-12]. Fundamentally, medicine is a human-centered endeavor, and technology can enable or distract from this focus. Unfortunately, almost 50% of physicians believe they spend an excessive amount of time on clerical tasks, and many physicians believe EHRs contribute to burnout [13-15]. High EHR task load, time spent on EHRs, and automated in-basket messages have been associated with greater emotional exhaustion [16-21]. Furthermore, the National Academy of Medicine recognizes the associations of administrative burden, technology usability, and time pressure on burnout (all of which may be attributed as EHR or HIT factors), yet the mechanisms underlying these associations are not well described [22].

One potential mechanism relating to HIT use and burnout may be frustration with technology—an emotional reaction to an obstacle preventing the fulfillment of a perceived need [23]. HIT is at risk for inducing frustration among health care workers, by virtue of its complex interfaces, frequent updates as capabilities improve, and deployment within a high-stakes environment that provides limited opportunity for dedicated training [24-26]. If frustration with technology contributes to emotional exhaustion, this would indicate an opportunity to prioritize reducing frustration through better design, training, and implementation as a mechanism to combat burnout [27].

We sought to quantify health care workers’ frustration with technology and its relationship with emotional exhaustion, after controlling for measures of work-life integration (WLI) that may indicate excessive job demands. We hypothesized that higher frustration with technology corresponds to higher emotional exhaustion.

Of the 23,853 distributed surveys, 16,797 were returned (70.4% response rate). Of these, 915 indicated that technology use was “not applicable” to them and 377 had incomplete responses, resulting in 15,505 complete responses for further analysis. Descriptive statistics are presented in Table 1. The most frequently represented positions among all respondents (N=15,505) were nurses (n=4316, 27.8%), technologists and technicians (n=1890, 12.2%), and administrative support personnel (n=1800, 11.6%). The majority of respondents (n=10,284, 66.3%) reported 5 or more years in their current specialty. Nearly half (n=7286, 47.0%) of the respondents were from units not providing direct patient care, 9.0% (n=1398) were from units providing intensive or emergency care, and 10.0% (n=1559) were from units providing surgical care. Of the 1140 work settings represented, 818 (71.8 %) had 5 or more unique respondents and were included in regression analyses.

WLI scores ranged from 0 to 100 (higher score favorable), with a mean score of 68.4 (SD 23.4) and median score of 71.4 (IQR 52.4-85.7). Separately, frustration with technology was reported as “none/rarely” by 6310 (40.7%) of the 15,505 respondents, “some/a little” by 4130 (26.6%), “occasionally/moderate” by 2815 (18.2%), and “all the time” by 2250 (14.5%) respondents. Frustration with technology scores ranged from 0 to 100 (lower score favorable), with a mean score of 35.0 (SD 35.9), and the score was higher among direct clinical care providers (mean 36.8, SD 36.4) than indirect providers (mean 32.9, SD 35.3; P<.001). The distribution of frustration with technology scores among 818 work settings, with corresponding WLI scores, is shown in Multimedia Appendix 1. The mean frustration with technology scores by job type are presented in Figure 1; the highest scores were reported by physicians, pharmacists, physician assistants, and nurses. Agreement in frustration with technology within work settings was low, with a combined weighted Cohen κ=0.04. Emotional exhaustion scores ranged from 0 to 100 (lower score favorable), with a mean score of 41.6 (SD 30.7).

Work setting correlations among frustration with technology, WLI items, and emotional exhaustion responses are illustrated in Multimedia Appendix 2 and tabulated in Multimedia Appendix 3. Frustration with technology was positively correlated with the full Emotional Exhaustion scale (r=0.35) as well as each individual item on the scale (r=0.29-0.36). The reverse-transposed WLI scale was negatively correlated with the Emotional Exhaustion scale and its individual items (r=–0.55 to –0.63). Each individual WLI item was correlated with the Emotional Exhaustion scale and its individual items, with the smallest correlations for “arrived home late from work” (r=0.32-0.41) and the largest correlations for “had difficulty sleeping” (r=0.57-0.65).

In our primary analysis, frustration with technology and six of the seven WLI items were independently related to emotional exhaustion in multivariable modeling, each associating with a 0.34- to 2.06-point increase in the emotional exhaustion score for each 10-point change (on a 100-point scale), as shown in Table 2. Frustration with technology was associated with a 1.23-point (95% CI 1.07-1.38) increase in the emotional exhaustion score with each 10-point change, and it was second only to difficulty sleeping out of the WLI items most strongly associated with emotional exhaustion. For example, an increase in frustration with technology score from 30 to 40 would correspond to a 1.23-point increase in the emotional exhaustion score, all else being equal. The frequency of sleeping less than 5 hours a night was the only WLI item that was not independently associated with emotional exhaustion. Results were similar when stratified by direct patient care versus indirect patient care. The results of our sensitivity analysis, as shown in Multimedia Appendix 4, are aligned with our primary regression model.

Our secondary analysis relating frustration with technology and emotional exhaustion aggregated within work settings is shown in Figure 2. Work settings with higher mean frustration with technology scores (distributed into quartiles) had higher emotional exhaustion scores. Mean emotional exhaustion scores ranged from 29.0 (SD 18.7, 95% CI 25.6-32.3) in the lowest quartile to 47.3 (SD 19.1, 95% CI 43.9-50.7) in the highest quartile. Results were similar when stratified by direct patient care versus indirect patient care.

This study found that frustration with technology varies with health care worker role and among individuals within work settings. Frustration with technology and 6 of 7 WLI items are independently associated with emotional exhaustion. Although frustration with technology was higher among direct clinical providers, similar relationships with emotional exhaustion were apparent for respondents engaged in direct patient care compared to those engaged in indirect patient care.

These results highlight and build on the evidence relating health care workers’ user experience with well-being outcomes [16,43,44]. HIT differs from consumer technology in that the purchaser (eg, health care administration) often has different priorities than the end users (health care workers), making user experience less incentivized as a driver in technology development. Although the purchaser’s incentives (eg, improving patient care quality and communication, facilitating accurate reimbursement, and developing research and analytics infrastructure) are undeniably important, our findings highlight the need for attention to user experience with HIT [45]. Recently, physician-reported EHR usability was reported to be in the bottom 10th percentile relative to system usability scores in other industries, but physicians reporting better EHR usability had lower odds of burnout [26].

Frustration with technology may arise from discrepancies between expectations and reality [24]. Particularly in settings in which EHRs or other HITs were rapidly implemented, health care workers’ expectations of the benefits of these technologies may not have been accurately set, features may not have been fully explained, or efficient use of the interface may not have been taught [46,47]. The Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009 that incentivized the adoption of HIT was effective at encouraging transition to EHRs, but rapid adoption may have precluded adequate attention to end-user input, development of interoperability standards, setting realistic expectations, or education on the use of new technologies [1,48-50].

Emotional exhaustion, the key construct of burnout evaluated in this study, reflects a depleted state arising from excessive demands, continuous stress, or insufficient resources [34,35,51-54]. If an EHR requires many clicks or otherwise inefficient workflows to accomplish tasks, this would translate to increased job demands. Alternatively, if inadequate training results in a health care worker not knowing how to use the interface effectively, this lack of knowledge translates to insufficient resources to use the available tools. In both cases, frustration with technology may develop as an indicator of imbalanced job demands and resources.

In our primary analysis, frustration with technology was found to be associated with burnout independently of other markers of WLI. This association was not as strong as the one observed for difficulty sleeping, consistent with observations that sleep disturbance may itself be an indicator of psychological distress [55]. However, it was similar in magnitude to the associations for other items reflecting excessive workload or workplace inefficiencies, such as changing personal/family plans because of work, working a shift without any breaks, or arriving home late from work. This pattern of findings suggests that frustration with technology is more closely related to an imbalance between job demands and resources, and it is less similar to psychological distress markers.

At the work-setting level, average frustration with technology was significantly associated with higher emotional exhaustion, which indicates a potential climate-like effect of frustration with technology, similar to that observed for WLI [28]. However, agreement among respondents from the same work setting was low, suggesting that much of frustration with technology is rooted at the individual level rather than the work-setting level (ie, the individual’s experience with technology rather than problems with the technology itself). Even work settings with the highest average frustration with technology scores still contained individuals reporting low frustration, suggesting against the notion that frustration is inevitable or inextricably linked to HIT. Different individuals may have vastly different experiences using the same technology, related to their specific worker tasks, training, personal expectations, comfort with the technology, and acceptance of change [56].

Of note, frustration with technology was only moderately correlated with frustration with one’s job in general, indicating that frustration with technology can occur without generalized job frustration and that individuals may remain motivated to take action to reduce their frustrations. This pattern of findings suggests that ensuring all individuals receive adequate support and training may be effective at reducing frustration with technology, rather than focusing only on making extensive changes to the technology or the user interface itself.

Several practical steps may be taken to reduce the frequency of frustration with technology among health care workers. Although details will require tailoring to specific settings, tasks, and technologies available, our findings suggest an approach of educating and supporting individuals experiencing the most frequent frustrations. Recent studies have described the use of supplemental EHR training to improve comfort with technology and ability to work more efficiently [57,58]. Interventions at the work-setting level may also include measuring and reducing individual workloads, such as employing scribes to assist with clinical documentation, transitioning to team-based documentation and inbox management, or automating data-entry tasks [59-72]. Finally, interventions to reduce workloads placed on the system could have the broadest benefits, but it would likely require changes to current payment structures that promote lengthy documentation and labor-intensive payment authorizations [73].

This study must be interpreted in the context of its design. As a cross-sectional, observational study, it cannot determine causality or directionality of the observed correlations (ie, whether frustration with technology induces burnout, burnout amplifies frustrations, and/or an external factor influences both). Although we were able to adjust for many potential confounders in our regression model, it is possible that residual confounding from unobserved variables such as age or prior experience with technology remains. Emotional exhaustion was also evaluated, but it does not capture other manifestations of burnout such as depersonalization. The 100-point emotional exhaustion scale we used differs from scores generated by other burnout instruments and cannot be used to directly compare effect sizes from studies using different instruments; however, methods to approximate estimates from disparate instruments have been previously described [74]. Although we used the 7-item WLI scale in this study, the 8-item version is commonly used, internally consistent (Cronbach α=.81 for this sample), and appears to add a unique element to the WLI assessment as evidenced by the results presented here.

The response rate of over 70% compares favorably with other studies of this magnitude and exceeds commonly accepted thresholds for survey-based research; however, there may remain some sampling bias, and it is possible that physicians in particular were relatively underrepresented in this sample [75,76]. Similarly, although the survey was confidential, as with any self-reported measures, the responses may be susceptible to recall bias or social desirability bias. The survey respondents reflect a wide variety of health care roles and work settings, with varying interactions with technology. Furthermore, even though many of the direct and indirect patient care roles included in this study heavily feature EHR use, these and other respondents may have referenced other use of technology in their responses, and we did not conduct interviews to further characterize their responses. Prior qualitative research has found that specific drivers of HIT frustration vary among individuals, tasks, and settings [47,77]. Our findings are thus reflective of the overall conceptual relationship between frustration and emotional exhaustion but are unable to provide conclusions regarding any particular piece of technology or source of frustration. Additional research will be necessary to further delineate the specific sources and scope of frustration with technology across health care worker roles, as these data may provide more granular insights of potential interventions to reduce frustrations. Although this survey was administered within a single US state, the Michigan Health and Hospital Association Keystone Center includes all 175 hospitals from 20 health systems within Michigan, making our results likely to be generalizable to community and academic hospitals across the United States.

It remains unknown whether reducing frustrations with technology through improved training, updated interfaces, or redistributed tasks will be effective in reducing burnout. Longitudinal observational studies may enhance our understanding of the directionality of these relationships, but prospective trials will be needed to fully evaluate the effect of interventions to improve health care worker user experience and well-being.

In conclusion, frustration with technology and difficulty sleeping were the biggest WLI factors associated with emotional exhaustion across direct and indirect patient care settings. Interventions designed to reduce health care workers’ frustration with technology and improve other aspects of WLI may be effective strategies to reduce burnout among health care workers.