We conducted nine formal observations between August 2022 and October 2022 for a total of 35 hours. Of these, six observations were at the central medical ICU (site A) and three observations were at the other ICU (site B) (see Table 1). The observed 9 care teams (male = 68%, female = 32%) were led by 9 different attending physicians, among which 2 were females and 7 were males.
The journey map (see Fig. 1) depicted three major sequential stages related to clinicians’ respiratory support decision-making processes (i.e., before morning rounds, during morning rounds, and after morning rounds). We identified three overarching themes (see Table 2) that represent clinicians’ use of EHR for their ventilation management decisions: (1) fragmented information and tasks for individual sensemaking before morning rounds; (2) EHR workarounds for collaborative decision-making during morning rounds; and (3) interruptive order entry and order execution after morning rounds.
Table 1
Characteristics of participants and observations sites
| Observation | ICU site | # of patient cases | Care team§ |
| Size | Male | Female |
| n = 9 | A = 6, B = 3 | n = 62 | n = 9 | 54 (70.1%) | 23 (29.9%) |
| 1 | B | 6 | 6 | 4 | 2 |
| 2 | A | 10 | 8 | 5 | 3 |
| 3 | B | 5 | 7 | 6 | 1 |
| 4 | A | 4 | 10 | 7 | 3 |
| 5 | A | 8 | 9 | 6 | 3 |
| 6 | A | 6 | 10 | 6 | 4 |
| 7 | A | 6 | 10 | 7 | 3 |
| 8 | B | 4 | 9 | 7 | 2 |
| 9 | A | 13 | 8 | 6 | 2 |
§: The care teams always had one attending physician and one fellow.
Table 2
Themes and descriptions of EHR use for respiratory support decisions at different stages
| Stages | Themes | Description |
| Before morning rounds | Fragmented information and tasks for individual sensemaking | Information seeking process for sensemaking consisted of fragmented tasks and information: • difficulty retrieving relevant clinical notes; • high variance of EHR use when retrieving relevant structured, quantitative data; • forced screen transitions; • interruptive CDS encounters; • additional communication to validate EHR information |
| During morning rounds | EHR workarounds for collaborative decision-making | EHR workarounds included handwritten rounding sheets, smartphone, portable computers, and tablets. Four occasions where portable computers were used because other workaround methods did not support data input (e.g., order entry or system update) and they also could not satisfactorily display X-ray imaging: • providing missing patient data or correcting incorrect data during resident’s briefing; • resolving conflicting information between EHR and the real world; • accessing and displaying chest imaging; • simple order entry. |
| During or after morning-rounds | Interruptive order entry and order execution | • Clinicians tended to delay entering complicated orders as the order entry was interruptive to their care delivery. • Order entries were sometimes disrupted by system alerts. • Extra team effort was needed to mediate order entries and executions due to lack of EHR system transparency. |
Theme 1: Fragmented information and tasks for individual sensemaking
Before morning rounds, clinicians started their shift by aggregating and making sense of patient data and information in the EHR that will be needed for briefing at morning rounds. This sensemaking process is highly dependent on the individual interaction with the EHR at their workstations. There are several common patterns observed among individual interaction with the EHR: (1) difficulty retrieving relevant clinical notes; (2) high variance of EHR use for retrieving relevant structured, quantitative data; (3) forced screen transitions; (4) infrequent and interruptive alerts or other CDS encounters; and (5) additional communication to validate EHR information. These common patterns embody fragmented tasks related to EHR use that are required for individual sensemaking.
First, it was observed that most clinicians started the sensemaking process by checking notes such as physician progress notes and consultation reports in the documentation section. The major complaint observed during this stage was the difficulty in retrieving needed patient information from notes because pertinent patient information was often buried in repetitive notes in various documents.
Second, there was a high variance in EHR use for retrieving structured data. For example, to find the ventilator setting, some clinicians checked it in the blood gas setting under the results interface, some clinicians found it in the interactive view interface, and some clinicians found the ventilator setting in the assessment section. However, the ventilator setting data was not identically displayed in those three interfaces as one contained more up-to-date ventilator setting data than another. Two fellows postulated that such variance in EHR use in finding the ventilator setting could be attributed to lack of awareness of the nuances in the ventilator data amongst the residents. Our observation also exemplifies that current EHRs lack integration with the same data stored on different interfaces (29).
Third, constant screen transitions were prevalent and did not conform to clinicians’ sensemaking habits. To understand physician progress notes (e.g., why was this patient not intubated?), clinicians had to “pull” related data from multiple sections in EHR, yet the system neither allowed simultaneous display of two or more sections nor provided a semantically integrated view of various data artifacts. Additionally, it was noted that the EHR system only allowed two open patients’ charts on the tab from the patient list, which caused additional navigation when attendings wanted to manage more than two patients. Forced screen transitions were related to fragmented patient data and information across many domains (e.g., results, documentation, interactive view, and medication records), which prompted various workarounds among clinicians. For instance, some clinicians used their smartphones or a tablet as the second screen, whereas some clinicians printed paper copies of notes and compared them against digital/EHR data. These workarounds also extended into supporting later collaborative decision-making elaborated in Theme 2.
Fourth, CDSs were not supporting sensemaking cognitive processes related to respiratory support decisions. Some CDSs appeared in the form of alerts when clinicians were closing patient charts and these notifications (e.g., low glucose alert or difficult airway alert) were interruptive, ill-timed, and often unhelpful. In one observed case, the alert froze the interface requesting data input to proceed yet the data was behind the alert window. Given the lack of built-in CDS, especially for respiratory support management, clinicians had to transition screen out of the EHR interface for external decision support (e.g., logging in to UpToDate).
Last, additional communication was required to validate EHR data and information as it may be out-of-date, inconsistent, and incomplete. For example, as clinicians knew that EHR may not capture up-to-date overnight data, they talked with bedside nurses to make sure data on the ventilator match the EHR. This was due to a lack of interoperability between medical equipment such as ventilators and EHR for automated data collection. Another example is that a clinician detected an anomaly in EHR data and found the patient is actually on mechanical ventilation, but the vent order was not documented. The required walking and talking often happened in the middle of their EHR interaction and were interruptive to their EHR use.
Theme 2: EHR workarounds for collaborative decision-making
While clinicians’ sensemaking was supported by their EHR interaction before morning rounds, the collaborative decision-making during morning rounds outside patient rooms depends on limited EHR use via portable computers and other peripheral devices such as smartphones, tablets, or paper records such as handwritten rounding sheets. The observed morning rounds usually involved attending physicians with their patient list listening to reporting residents briefing on the patient based on their rounding sheets. Then, attending physicians led the discussion of the assessment plan with the team, during which attending physicians requested to see chest imaging on a portable computer and residents were instructed to consult external resources (e.g., search for the five World Health Organization pulmonary hypertension groups on smartphones) or calculate indices (e.g. calculate the ratio of SpO2/FiO2 to respiratory rate (ROX) score using an online calculator) needed for respiratory support decision-making. Before finishing the assessment and plan discussion, some of the care teams entered the patient room to perform bedside evaluations. There was no interaction with EHR observed in the patient room.
The majority of the care teams, except a few attending physicians, depend on handwritten rounding sheets that were derived from sensemaking of EHR data and information before morning rounds for their collaborative decision-making outside patient rooms. Handwritten rounding sheets as a workaround for EHR use were structured differently among individual physicians; however, what they contained were similar – a semantically integrated display of subjective data (e.g., overnight events), objective data (e.g., vitals, labs, physical exam), and assessment and plan.
Two clinicians (one attending and one fellow) were observed using tablets (e.g., Microsoft Surface, and Apple iPad) during morning rounds. However, there are several drawbacks associated with their EHR use on tablets. For example, they used the web-based EHR portal on tablets, which was not completely compatible with tablet platforms (e.g., they could not zoom in or zoom out when looking at imaging on tablets with their fingers). In addition, they had no place to temporarily put away their tablets while interacting with patients because the current ICU work environment does not support the use of tablets.
Although most members of the care teams depended on handwritten rounding sheets, two care team members were typically designated as portable computer users for any interaction with EHR. There were four occasions that workarounds of EHR use could not satisfy clinicians’ needs and portable computers were needed for further EHR interactions to continue supporting collaborative decision-making: (1) providing missing patient data or correcting incorrect data during resident’s briefing; (2) resolving conflicting information between EHR and the real world; (3) accessing and displaying chest imaging; (4) simple order entry.
-
First, residents sometimes missed important clinical data or reported incorrect data in their morning round briefing. In this case, residents with portable computers supplemented missing data and corrected incorrect data. For example, an attending felt some data were off in the resident’s briefing and requested validation from residents using portable computers, who later confirmed that the patient’s blood culture did not increase.
-
Second, as EHR data can be outdated, incomplete, or incorrect (30), residents with portable computers needed to update the system outside the patient room. For example, the attending physician detected inconsistency in the medication summary, went into the patient room, and found that the patient was not on the said medication. Then, the resident with a portable computer updated the system accordingly.
-
Third, portable computers are the only channel for imaging display to the care team during morning rounds. Since chest imaging is an important component of respiratory support decision-making, attending physicians always requested displays of the chest imaging during the care team discussion. Yet, these images can only be satisfactorily viewed on a computer and other workarounds such as tablets are suboptimal.
-
Last, residents or pharmacists with portable computers sometimes enter simple orders outside patient rooms. It allows for the quick order entry that minimizes delays and interruptions to their later workflow.
Theme 3: Interruptive order entry and order execution
Respiratory support decisions finalized through collaborative decision-making outside patient rooms were electronically documented and ensuing actions were entered as orders. The digitization of decisions into EHR was intended for multiple purposes, such as billing, facilitating collaborative teamwork, and streamlining clinical workflow (31); however, the process of digitizing decisions was disruptive to teamwork and clinical workflow. While residents with portable computers could quickly enter simple orders outside patient rooms, some orders were complicated, and clinicians tended to delay those order entries after they finished morning rounds. Nevertheless, clinicians’ care delivery did not stop after morning rounds and the care teams faced a dilemma – should we go see another patient or should we go back to the workroom to enter orders? Hence order entries were seen as interruptions to their care delivery processes. In addition, entering orders into EHR could prompt system alerts that disrupt the order entry. For example, since a Bi-level Positive Airway Pressure (BiPAP) protocol CDS was built into EHR after the COVID-19 pandemic, going through the BiPAP protocol CDS to place an order for BiPAP requires completing a list of procedural actions, such as consulting with a respiratory specialist. However, the BiPAP decision was a result of collaborative decision-making that already involved respiratory specialists. The purpose of the BiPAP protocol CDS was to protect patient safety, but the way it functioned as a post hoc decision check that did not follow clinicians’ workflow and decision-making processes. Last, order execution sometimes required extra teamwork effort to mediate because EHR interfaces were not consistently and uniformly displayed to different care providers. For example, the orders entered by clinicians sometimes were not seen by nurses or respiratory therapists on their EHR interfaces. The clinician had to follow up in person to ensure orders in the EHR were seen and implemented.