Unlocking inpatient workload insights with electronic health record event logs

Abstract

High workloads in inpatient settings are associated with clinician burnout and have the potential to negatively impact patient care and the overall performance of organizations.^{1, 2} However, the workload thresholds at which these effects occur are unclear, with a notable lack of evidence-based guidance for optimizing work design in inpatient settings. Conventional measures of workload (i.e., work relative value units and volume of patient encounters) are typically used to capture measures of patient-related workloads but inadequately capture the full work effort or the impact of workloads on key outcomes.³

To address the research to practice re-design gap, an emerging field offers promise: the use of electronic health record (EHR) event logs to understand how work design—which includes strategies around deriving optimal workloads and team structures—influences work patterns and, subsequently, patient outcomes.^{4, 5} EHR event logs represent a wealth of data, recording various aspects of clinical work activities and flow of events based on actions performed within the EHR.⁶ Event log data are unique as they are user-based (as opposed to clinical data that are patient-based) and capture the sequence of clinician care workflows that are typically not part of traditional medical record data elements.⁷ Major EHR vendors have now developed within-EHR platforms (e.g., Epic's Signal) that aggregate event log data into operational measures of clinician work (e.g., time spent on notes) making the data accessible to organizational leaders.

To date, considerable research has used event log data to quantify workload and work patterns in the outpatient setting.^8-10 National organizations, such as the American Medical Association, are encouraging the use of EHR-based measures in practice transformation efforts to improve clinician well-being and reduce burnout.¹¹ Although there has been much excitement in the outpatient space, the use of event logs in inpatient settings has been limited to a handful of studies.^12-15 In this perspective, we identify EHR event log measures relevant to inpatient clinician work, describe challenges with its use, and propose innovative use cases for future work.

In outpatient settings, Sinsky et al. identified seven core measures for assessing practice efficiency through EHR event logs¹⁰; however, inpatient clinician workflows are considerably different from outpatient settings. Recognizing the need for tailored measures to capture the unique demands and dynamics in inpatient settings, our authorship team, consisting of clinician experts in the inpatient setting and outpatient setting and experts in clinical informatics, derived potential measures for use in the field of hospital medicine (i.e., generalist, nonsurgical) given the limited prior applications of such measures in inpatient settings. Our selection process involved considering the work conducted by Sinsky et al.¹⁰ while also considering the distinct differences and challenges faced by inpatient clinicians, such as shift structures, patient encounters that often last for days/weeks (i.e., inpatient stays), communication patterns, structures of teams, interruptions, and the dynamic nature of the work environment. The proposed measures were also reviewed by a group of national experts on clinician workload. Table 1 shows the proposed measures, descriptions, nuances, challenges, and limitations for each measure. This work serves as a starting point for future research and discussion around the most salient measures using event log data and to bring awareness regarding the utility of this data that is available in all modern EHRs. Future efforts should focus on refining, standardizing, and prioritizing these measures.

With potential measures defined, the use cases for event log data in the inpatient setting are immense and include practice management, educational, and quality improvement purposes in addition to research; however, efforts in these areas outside of research settings are sparse. Within the realm of practice management, event log data have the potential to support evidence-based work design practices. As clinicians in the inpatient setting continue to face pressures to increase the volume of patient encounters,^{2, 16} event log-based measures paired with critical workforce, patient, and organizational outcomes may offer pragmatic practice insights. These insights could help organizational leaders understand when work design and workloads may contribute to harm. For example, if certain workload thresholds, as measured by event logs—such as number of patient encounters, secure messages sent or received, or EHR-based interruptions (e.g., alerts, messages)—are found to be associated with burnout or patient harm, then these thresholds could be monitored and proactively addressed. Early research has indicated that conducting ecological assessments through carefully timed clinician surveys can provide valuable insights into how workloads and work environments are perceived. Analyzing these data in conjunction with event log measures may help leaders to understand when workloads and work environments contribute to adverse outcomes or lead to improved outcomes.¹⁷ Additionally, insights into practice patterns such as resource utilization and team member interactions¹⁵ could be incorporated into event log measures.

In education, event log measures can provide valuable insights into trainee work patterns,¹⁸ potentially reducing the need for surveys for auditing work activities and measuring workload. Measures such as total EHR time, time from first sign-in to last log-off, and location of the sign-in could signal shifts in work patterns or identify when trainees exceed work-hour thresholds. Additionally, EHR event logs can be utilized for phenotyping and identifying clinicians who may be struggling with certain activities of the work day such as time spent completing documentation or orders¹⁹ and assessing clinical performance outcomes.²⁰ Understanding the measures that are indicative of struggling learners can also be important. Sebok-Syer et al. have highlighted the importance of understanding how EHR-based measures incorporate the impact of team dynamics on individual performance, as well as the reciprocal influence of the individual on the performance of a team.²¹ Additionally, given that organizations are also increasingly utilizing secure electronic messaging platforms (e.g., Epic Secure Chat), large language models in the future could be utilized to analyze potential knowledge gaps or lapses in professionalism.

In the process improvement domain, event log-based measures may be able to provide valuable insights into care coordination processes, particularly during critical transitions of care. For example, within the context of unexpected transfers to an intensive care unit, clinician leaders could utilize event logs to evaluate care workflows and patterns before the transfer to evaluate chart access, communication patterns, and interactions to assess opportunities to optimize care coordination. There is also an opportunity to explore how workload and work patterns drive cognitive¹⁴ and diagnostic errors.⁵ Innovations in this space include utilizing patterns of orders such as the retract-and-reorder (an indicator of wrong-patient ordering errors),²² which could serve as a measure of cognitive error. Pairing these types of orders with event log measures such as workloads or measures of attention could help operational leaders to understand when work design may be contributing to these types of errors.¹⁴

Although there are many opportunities with event log measures, there are also several challenges. As measures are developed, rigorous validation techniques must be conducted to ensure accuracy and generalizability. As organizational leaders gain access to this data, considerable work will also need to be conducted to build theories and frameworks on how to interpret the data and associated measures. For example, a measure such as work outside of work (WoW) requires contextual interpretation. A high WoW may mean work inefficiencies and/or work overload; in contrast, it could also mean competing work-life demands (e.g., an inpatient clinician may choose to do charting at home to pick children up in a timely fashion, therefore choosing to do WoW) or it could be an adaptive response to fight burnout. Furthermore, understanding the associations with outcomes will be key. Merely defining and assessing these measures will be insufficient and additional work will be needed to understand the significance and the contextual implications of using these measures. If an issue is identified, it will be imperative to collaborate with individuals or groups where the problem is presumed to exist, as there is a risk that this data can be misinterpreted. Additionally, attribution of outcomes to individual clinicians can also be challenging with some work to define best practices already conducted.^{23, 24}

Although all hospitals with EHRs have access to event logs, only 53% of hospitals reported using EHR data to track clinician time.¹⁹ Insights gained (as well as access to event log data) may not trickle down to practice management leaders. Granular event log data also exists (i.e., unaggregated data) and requires considerable data processing and analytics capabilities, which may be challenging for many clinical practices. Although EHR vendors have developed platforms with higher-level reports with summaries of the data, vendors' data aggregation practices have not been standardized, rendering generalizations across organizations with different EHR vendors challenging. Different types of clinical work will influence clinician's EHR use and, thus, the patterns seen in event log data. We have primarily focused on hospitalists' work; however, additional considerations should be pursued to understand how proceduralist or higher acuity services may impact patterns in the data. Thus, measure validation will be an important next step.

Next, inpatient work is also challenging from both a scheduling and attribution perspective, complicating the process of linking event log measures to outcomes. Event logs, in theory, are highly attributable to individual clinicians; however, the data are also impacted by patient complexity, the multidisciplinary nature of inpatient work, and systems-related factors. Next, event log measures present several considerations within the frameworks of Campbell's (i.e., once a measure is used for decision-making, the more it will be subject to corruption pressures, and the processes it is intended to monitor may be undermined)²⁵ and Goodhart's Laws (i.e., once a measure becomes a target it is no longer a good measure).²⁶ Similarly, healthcare leaders with pressures for both financial performance and productivity targets may use these measures to attempt to boost productivity without considering the context for the data (i.e., increasing workloads when electronic measures of workload are perceived to be low). Finally, the concept of surveillance, while commonplace in many occupations (e.g., trucking industry²⁷), is now increasingly prevalent in healthcare workplace settings and must be conducted thoughtfully.

Looking ahead, leveraging EHR event log measures offers the potential for optimizing work design to improve outcomes for clinicians, patients, and organizations. Organizations must develop prerequisites to integrate EHR measures into practice. This involves developing best practices around how event log measures will be used, ensuring data privacy, and emphasizing that interpretation of the data requires some caution and knowledge. Furthermore, fostering a culture of psychological safety will be paramount to ensure individuals feel comfortable sharing insights and concerns arising from this data. Collaboration across disciplines will be needed, particularly when making decisions informed by this data. In summary, EHR event log measures represent a significant opportunity to utilize data collected during routine clinical work to begin to understand how work design impacts clinicians, patient care, and organizational outcomes.

The authors declare no conflict of interest.