Enhancing electronic health records with real-time dashboards to improve patient care and service efficiency: a protocol

Emergency departments (EDs) are critical healthcare settings that often face challenges related to patient overcrowding and sub-optimal resource use, and these issues have been linked to reduced effectiveness of care and increased patient mortality. This protocol describes a study to enhance electronic health records (EHR) with real-time dashboards developed for healthcare staff, administrators, and other stakeholders in order to address these issues. These dashboards will be analytic tools, fed with data directly from the electronic health records, allowing healthcare indicators to be monitored dynamically and interactively. The indicators will provide a comprehensive overview of patient flow, crowding levels, waiting times, and other key operational metrics to support healthcare providers by enabling real-time monitoring and decision-making. Administrators will also benefit from the dashboards by gaining insight into system bottlenecks and overall ED performance, facilitating more effective management of resources. This study is part of the eCREAM (enabling Clinical Research in Emergency and Acute care Medicine through automated data extraction) project, and 32 EDs from six different European countries will be involved in this 36-month study. A panel of experts will test the usefulness of the dashboards, set up for each end-user type, using questionnaires and semi-structured interviews. The EHRs, integrated with these ad-hoc dashboards, will empower clinical teams and policymakers to understand department operations dynamics better, ultimately contributing to improved patient outcomes and reduced crowding in emergency settings. Trial registration: Clinicaltrials.gov, identifier: NCT06372379.

The emergency department (ED) is a healthcare service that provides the first clinical assessment and treatment to patients with various acute conditions. Even though timely, high-quality care is essential to fulfill their mandate, these departments are often overwhelmed by a large volume of patients. This phenomenon, called ED crowding, has become a global concern (1, 2) and has been correlated with reduced timeliness and effectiveness of care and increased patient mortality (3, 4). Crowding occurs when there is an imbalance between the flow of arriving patients (input), the time taken to assess and treat patients already in the ED (throughput), and the flow of patients leaving the ED once their treatment has been completed (output) (5). The main contributor to the input aspect is that most patients self-present to the ED, with only 20%−30% of patients being brought by an ambulance service (6). Most self-presenters have non-urgent conditions (7–9) that would be more appropriately treated by general practitioners or by non-urgent care services (10). They typically wait many hours before being visited, contributing to ED crowding (11).

Regarding throughput, ED crowding is influenced by the time of triage, the registration and care processes, the number of current staff, the accessibility to specialty and diagnostic services, and the informatics systems (12). ED crowding in the output phase is mostly influenced by the availability of hospital beds or post-acute care services related to community mental health and primary and specialty care (12). Since all these three components—input, throughput, and output–are closely linked, an effective program to address ED quality-of-care cannot disregard them and their effect on crowding.

Dashboards in the healthcare setting are analytic tools fed from various information systems that allow healthcare indicators to be monitored dynamically and interactively. By exploiting electronic health records (EHR) data, their application in the ED context could reduce crowding by targeting all three of the aforementioned components. An electronic dashboard tested in the USA, for example, was shown to help overcome hospital capacity inefficiencies, such as those related to the patient handoff process from the ED to the hospital (13). Another dashboard, designed with a different aim (to identify critically ill patients in different participating sites in order to foresee a call with an off-site expert), but which also used EHR data to automatically display the designated information, was also shown to be effective in identifying such patients (14).

Regarding input factors, dashboards could reduce the self-presentation of patients with minor or non-urgent conditions by empowering them through access to relevant information. A real-time dashboard could also help out-of-hospital emergency organizations divert ambulances to the most suitable ED, not simply in terms of the hospital's clinical expertise concerning the patient's conditions (as is usually the case), but also in terms of the average waiting and treatment times at the different EDs.

In terms of throughput and output, dashboards integrating real-time EHR data could allow ED physicians to identify bottlenecks in the flow of the entire patient care process, such as an exceptionally high average waiting time for diagnostic results reporting or specialist consultations, while continuously monitoring the boarding time (i.e., the time between the decision to admit a patient and the time when actually admitted) and the wards' bed availability (7).

The eCREAM project, a European project involving 11 partners across eight countries (see the article “The format of emergency department electronic health records in Europe. The European initiative and the eCREAM proposal” in this issue for details), aims to improve the quality of care in emergency medicine by creating a system that will take advantage of EHR data, which is currently not readily usable for research in the emergency medicine field. Due to the complex context and the scarce resources in emergency medicine, many EHR data are collected in free-text form and in an incomplete manner. eCREAM's system will automatically extract reliable, complete, structured EHR data to enable research and will then validate the system by implementing two use cases using the extracted data. The second use case, Use Case 2 (UC2), concerns the development of real-time dashboards for different end users to be fed with the EHR data and then assessed for usefulness.

This article describes the protocol of the UC2 study, focusing on two interactive dashboards specifically designed to enhance current ED EHR systems. Thanks to the structured data that the eCREAM system will extract, the integrated dashboards will provide healthcare staff, health system administrators, and regional policymakers with indicators revealing a comprehensive, real-time overview of patient flow, crowding levels, waiting times, and other key operational metrics in the ED. This additional ad-hoc information will permit healthcare providers to make appropriate patient and ED management decisions, and this should help reduce waiting times and overcrowding, ultimately increasing the quality of care. The study will also develop a third dashboard, for patients, that will be fed with the ED EHR data. The dashboards' usefulness will then be assessed, their strengths and areas for improvement will be identified, and their effectiveness in supporting decision-making in the ED will be evaluated. UC2 is described in this article.

Study design

The study is an observational, multicenter, retrospective study. Since it aims to develop useful dashboards rather than implement them with real-time data, the dashboards will be fed by data already available in the participating hospitals' databases and run on retrospectively extracted data.

Concerning the citizen dashboard, a preliminary review of existing tools aimed at citizens was conducted through a search of the literature on Medline, and a questionnaire was sent to eCream partners. Findings showed that 32 similar apps or websites on EDs aimed at citizens exist and are variable in terms of information provided (ED crowding levels, overall waiting times, waiting time by level of emergency, etc.) and layout. A comprehensive literature review and a broader survey on the apps available will be carried out and their results will be considered in the development of the dashboard to address citizens' information needs.

Regarding the three different dashboards, three panels of experts representing ED physicians and nurses, healthcare policymakers and service organizers, and patients and citizens will be set up. These expert groups will design their own reference dashboard, based on their needs, via periodic focus groups, and will decide which indicators to include in the dashboards. For the dashboards for professionals, these may include a comprehensive overview of patient flow and congestion indicators, waiting times by triage category, boarding times for admitted patients waiting for inpatient beds, current resource utilization such as available vs. occupied ED beds, but also more specific indicators such as alerts on the presence of patients with time-sensitive conditions such as sepsis and stroke, and high-risk patients (e.g., immunocompromised, or with psychiatric crises). For the dashboards for patients, the expert groups may decide that, in addition to providing information on the EDs and links to health services available in the area, the indicators may include real-time crowding levels and number of patients waiting by priority code.

Technical and functional aspects that will be addressed in designing the dashboards for professionals will include role-based access and different views for clinicians, nurses, administrators, and hospital managers, customisable filters (e.g., specific patient groups, time period), touchscreen support, automated alerts and notifications (e.g., when values exceed accepted ranges).

External experts will revise the work to achieve a version that satisfies as many users as possible. These panels will also be responsible for preparing the questionnaire to assess the usefulness of the dashboards. An example of the dashboard that we envision for the healthcare policymakers in Crete can be seen in Figure 1. Figure 2 provides examples of graphical representations for throughput indicators, which will be considered for inclusion in the dashboard for the ED staff.

Technical details—Integration and interoperability

A local informatics platform, the eCREAM platform, will be installed locally in the participating EDs. The data that will feed the dashboards will be gathered, through the platform, from each ED according to a common dataset specification, which is formalized in a data dictionary. To ensure the integration and interoperability of clinical data from diverse EDs, a data standardization approach will be adopted. Data will initially be gathered locally by each ED and then moved to a central staging database, where standardization will be performed, before being transferred to a target database. All data collected locally at each eCREAM installation will be managed according to the following criteria:

• Data mapping: identifying the data and codes (e.g., diagnostic codes, services, procedures) utilized across different systems and their respective records. Most of the information required for the dashboards is already locally standardized using international benchmarks, such as diagnoses adhering to ICD9-CM or ICD10-CM. However, some information uses regional classifications, such as codes for radiology or laboratory tests.

• Standards adoption: converting all collected data into common formats. This involves using established standards or converting to a chosen reference coding and unit of measure, thereby ensuring that information is comparable and aggregable, at least at the regional level.

While the dashboard software application will be standardized for all eCREAM participants, each region—including ED staff and policymakers—will maintain its own distinct view. Consequently, the consistent use of the same classification system within each geographical area must be ensured.

Technical details—Dashboard development

The three dashboards will be developed separately:

The dashboard for the ED staff will be implemented within the eCREAM EHR. This dashboard will report indicators elaborated based on data gathered from the ED, e.g., the average time needed to complete a radiological test required by the ED clinician, or the patients' boarding times.

The dashboard for the regional policymakers will be implemented as a central web application and will report indicators elaborated on data gathered from the EDs belonging to that region, e.g., the number of influenza cases at the regional level, with the localization of the outbreaks, or the benchmarking of the boarding time between the EDs.

The dashboard for the citizens will report selected indicators of the EDs' workload, e.g., the number of patients currently in each ED.

The dashboards for citizens and for the policymakers will use the same central database, so their information will be coherent. The dashboard for the ED staff will use the database of the EHR application, which is the source of data for the central database. When managing clinical data from different ED electronic health records, it's crucial to align data dictionaries within each region. This ensures that data content is comparable, consistent, and uniformly interpretable. These dictionaries contain standard or proprietary codes for elements such as diagnoses (e.g., ICD-10 and SNOMED CT), service/exam codes, triage classifications, clinical outcomes (e.g., “discharged” or “admitted”), access modes, and lab units. This process enables reliable comparative analyses (e.g., severe access numbers across sites), reduces interpretive errors, allows for unified dashboards, and promotes semantic consistency. The three dashboards will use the same standards regionally.

As for the coherence of indicator values between the dashboards, the coherence between the citizen dashboard and the policymaker dashboard will be guaranteed by design, as both will share the same database and elaboration engine as their source. The coherence with the ED staff dashboard, developed within the EHR, will be monitored by implementing a central web service. This service will make the indicators published on the central dashboard available to the EHR. EHR developers can then decide whether to publish these indicators for the ED directly from the web service or use them to verify consistency with locally calculated values.

Dashboard assessment

Once the dashboards for the physicians and policymakers are developed and integrated into the ED EHR, their suitability will be evaluated by the different end-users. More specifically, a group of emergency physicians and nurses and a group of healthcare policymakers and organization managers of the participating EDs will be asked to complete the prepared questionnaire and participate with in-depth interviews on the usefulness, advantages, and shortcomings of the respective dashboard. A group of citizens and patients will be asked to do the same for the citizen dashboard once it is ready.

These representatives will not be part of the expert panels. A purposive sampling strategy will be employed to select participants for focus groups and semi-structured interviews. Participants will be drawn from the key end-user groups targeted by the dashboards (clinical staff, healthcare decision-makers, and citizens). Selection will aim to ensure a diverse representation in terms of professional roles, experience levels, and institutional contexts (e.g., size and geographic location of EDs).

Selection of subjects

The EDs adhering to the eCREAM project will participate in the study, with the exception of centers in Greece, which are currently not using an EHR. Overall, 24 EDs will be included: 15 from Italy, four from Poland, two each from Slovenia and UK, and one from Switzerland. The centers will not receive any incentive for participation except for expenses incurred.

All patients arriving at participating EDs during the six-month study period in 2023 will be considered. Clinicians, health policymakers, and citizens participating in the dashboard assessment will also provide data for the study through questionnaires and semi-structured interviews.

Study duration

The study will last 36 months: the first six will be used to establish the three panels of experts for the different types of end-users and to develop and refine the dashboards; the following 12 months will be dedicated to installing the local UC2 platform in the participating EDs and retrospectively retrieving the required information, check its quality, and make necessary adjustments; six months will be taken to put the dashboards into operation; the final 12 months will be devoted to assessing the usefulness of the dashboards, perform data analyses, produce research reports, and disseminate the results.

Data collection

The data required to feed the three dashboards will be collected on all patients arriving at the participating EDs between July 1st, 2023, and December 31st, 2023. The retrospective data extraction from the hospital information systems will start as soon as the local UC2 platform is up and functioning. The data will include administrative and clinical information, such as time of arrival at the ED, time of triage and visit, priority code assigned at triage, time of possible hospital admission, and any other data the panels of experts have decided to collect. Information on the EDs' organizational set-up will also be collected (e.g., the number of available workstations and the capacity to perform certain diagnostic tests).

Finally, once the dashboards are in operation, data from the three panels of experts will be collected through a mixed-method approach involving an online survey and semi-structured interviews. A questionnaire and a guide for the interviews will be prepared, covering the usability, usefulness, clarity, and completeness of information, as well as the advantages and shortcomings of each dashboard.

Data processing

Figure 3 represents the data processing that will be carried out in the context of the eCREAM UC2 study. The first data processing step will occur within the participating hospitals' information systems, where the local UC2 platform will be installed. This step consists of extracting all variables needed to feed the dashboards and transmitting them to the platform through an ad hoc interface. Within this study, the factual transmission of the data will be arranged only once. However, the platform will be designed to generate, if necessary, multiple data transmissions. This is important to test the possible infrastructure for future projects if the developed dashboard is considered beneficial. The variables will be retained in the local platform as long as necessary for further processing and subsequent transmission to the central eCREAM server. Data will be physically deleted at the end of the transmission process. The platform will be specifically designed to calculate, from the data received, the indicators that make up the dashboards (second step). Since these indicators are, by definition, aggregate information, such as the time it takes to get the result of a CT scan or the crowding level of the ED, calculating the indicators corresponds to making the data completely anonymous. In the third step, the anonymous data will be transferred through secure protocols to the central eCREAM server in Milan, Italy, for analysis.

Data analysis

Descriptive analyses will be used to summarize the results of the questionnaire. Categorical variables will be described as counts and frequency, continuous variables as mean and standard deviation, or median and interquartile range, as appropriate. Cronbach's α will be used to test survey reliability. The Chi-square test will be used to assess associations between categorical variables and the ANOVA or the Kruskal–Wallis tests will be use to compare continuous variables across multiple groups, as appropriate. A thematic analysis will be used for the interviews, which will be transcribed verbatim. We will use a deductive-inductive analysis starting from the items in the guide and develop additional items coming from the interviewers. The NVivo software will be used to extract themes from textual data and to manage coding.

The present study aims to test a system designed to automatically extract reliable, research-friendly EHR data in a field where carrying out research and identifying areas for improvement is very challenging. Ad-hoc dashboards developed to enhance the EHR and permit healthcare providers and policymakers to make informed decisions based on real-time ED data to address phenomenons contributing to ED crowding will not only test the data extraction system, but will pave the way for improved ED efficiency and patient care.

Achieving situational awareness on the part of clinicians requires acquiring, processing, integrating, and presenting the appropriate information on a dashboard, permitting them to have more control over the ED and to take countermeasures to regain efficiency, reducing disadvantages and possible harm to the patients. These measures can be taken immediately or over the long term, for example if ED managers use the dashboard to study the crowding patterns over time in their ED to improve its efficiency in terms of staffing, accordingly. An appropriate dashboard could also enable policymakers to monitor specific, regional phenomena of interest, such as real-time availability and timeliness of primary and secondary patient transfers and ambulance offload delays, as well as epidemiological phenomena, such as the emergence of certain infectious diseases, in a timely manner.

The citizen dashboard will further contribute to testing the system and could work toward reducing ED crowding. Increasing patient awareness about the EDs' pre-defined characteristics, such as criteria for assigning the priority code, as well as their real-time crowding level and the number of patients waiting by priority code, for example, could prompt patients to consider different, more appropriate, or more efficient healthcare services, thus reducing crowding. Overall, the three dashboards could influence ED efficiency and patient care through different users and, therefore, approaches.

This study has some risks in the implementation of the dashboards and some limits.

Possible, unintended risks from system-level implementation involve the presence of heterogeneous data structures and coding systems across EDs, which could result in inconsistent dashboard outputs or incomplete data interoperability. Early, comprehensive data mapping of local data and coding systems has already begun, as described above, however, in order to avoid this possibility, as well as the conversion of all collected data into common formats using established standards.

The limits of this study are that the dashboards will run on retrospectively extracted data. This is a risk because the dashboards for the ED professionals will be assessed by experts outside of their daily routine and their usefulness will be subjective. The patient dashboard will be assessed by citizens who are not actually searching for real-time data on EDs and cannot test whether the information the dashboard provides helps improve their choices. The study, however, is designed to overcome the limits in using retrospectively extracted data, assisting all three user groups in providing the most realistic assessment of the usefulness, advantages, and shortcomings of their respective dashboards by running up in-depth interviews and providing a specially designed questionnaire.

If considered beneficial by the expert panels, the dashboards and their actual implementation and distribution will be the subject of future projects.