Clinician Communication and Patient Safety in Pediatrics: A Practical Application of Human-Centered Design for Problem Identification and Analysis

Abstract

Background:

We established a Patient Safety Learning Lab (AHRQ R18HS029473) to examine the sociotechnical system that drives interprofessional communication in pediatric inpatient settings in the context of evolving communication technologies, and to co-create and evaluate solutions with clinician end users. Here, we describe the use of human-centered design and system engineering processes for the Problem Analysis phase of this project.

Methods:

We applied the “Empathize” and “Define” steps of the design thinking process to our Problem Analysis. The goal of the Empathize step is to generate a comprehensive understanding of a problem(s) as experienced by the end user. We conducted interviews and observations with interprofessional clinicians from pediatric inpatient units in a single children’s hospital. We used other operational and clinical data to triangulate findings (clinician secure messaging metadata, survey data, and policies/procedures). In the “Define” step, we iteratively developed user-centered problem statements.

Results:

Data synthesized for the problem analysis included: interviews with 28 clinicians, 32 hours of unit observations; metadata for 433,432 secure messages; 155 free-text clinician survey responses; and 40 communication-related policies/procedures. The Problem Analysis revealed communication challenges in the following domains for clinicians providing frontline care (i.e., bedside nurses, residents, frontline fellows): (1) efficiently locating and contacting other members of the care team; (2) communicating urgency level of information; and (3) managing high volume of minimally informative messages.

Conclusions:

Practical application of human-centered design and systems thinking contributed to a more holistic understanding of communication challenges, and their patient safety implications, from the perspective of multiple end-user groups.

High-quality communication among clinical teams is essential for safe and effective care of hospitalized patients.^1,2 Pediatric patients may be especially vulnerable to errors from health care team miscommunication due to developmental limitations and limited ability to advocate for themselves.³ In a survey of nurses from 330 hospitals providing pediatric care, 60% reported that information was lost in communication.⁴ Parents have also identified clinician-to-clinician communication as a source of error for hospitalized children⁵ and ranked communication as a high-priority area for patient safety research.⁶

Evolving communication modalities create new communication challenges. Increasingly, hospitals are implementing secure messaging platforms for clinician communication,^7–9 augmenting existing systems of voice call, paging, and in-person interactions that already result in frequent and costly miscommunication.^10,11 Over the last several years, multiple studies of secure messaging among clinicians have been conducted, primarily quantifying message frequency.^7,12,13 However, focusing on one communication modality, and at times only one user group, misses the opportunity to understand the full sociotechnical system¹⁴ that produces (mis)communication. Attention to interactions between tools/technology, people, organizational factors, and the physical environment is needed to meaningfully improve interprofessional communication and reduce communication-related patient safety events.

In 2023, we established a Patient Safety Learning Lab funded by Agency for Healthcare Research and Quality (R18HS029473) to examine the sociotechnical system that drives interprofessional clinician-to-clinician communication in pediatric inpatient settings (with particular attention to the role of evolving communication technologies) and to co-create solutions with end-user clinicians. We are conducting work in 3 phases: (1) problem analysis; (2) intervention design and development; and (3) intervention implementation and evaluation. This report focuses on the problem analysis phase, and our application of human-centered design and system engineering approaches.^15,16 A comprehensive problem analysis of the sociotechnical system—often overlooked in quality improvement and safety initiatives¹⁷—is foundational for developing and implementing meaningful solutions to patient safety challenges.

METHODS

Setting and Population

We conducted this study in an urban children’s hospital with over 500 beds. The focus of this work is inpatient clinicians who provide direct care to hospitalized patients [e.g., nurses, physicians, advanced practice providers (APPs)]. In June 2022, the institution adopted clinician smartphones (iPhones) with 2-way text messaging capabilities enterprise-wide (Epic Systems Corporation, Verona, WI) (“secure messaging”), with complete retirement of hospital pagers. Secure messaging is fully integrated with desktop and mobile interfaces, allows clinicians to link medical records directly to messages, and includes group messaging capabilities. Physicians and APPs have individual smartphones, whereas other clinical staff use shared smartphones that function exclusively on hospital premises and are checked in and out at the nurses’ station each shift. Upon introducing the secure messaging platform, guidance for use of secure messaging versus other modes of communication was issued hospital-wide.

Patient Safety Learning Lab (PSLL) Team

Our Patient Safety Learning Lab (PSLL) is led by a nurse scientist principal investigator and co-investigators from the disciplines of pediatric medicine, human factors engineering, informatics, and nursing. Our core PSLL team comprises interprofessional collaborators spanning academic, clinical, and operational roles.

Study Design

This study was approved by the Children’s Hospital of Philadelphia IRB (23-021235). The problem analysis phase consisted of the first 2 steps of the 5-step design thinking process (empathize, define, ideate, prototype, and test).¹⁸ The “empathize” step was conducted in a manner consistent with a convergent parallel mixed methods approach (concurrently using qualitative and quantitative methods), with the goal of more deeply understanding the end users and their problem(s).^19,20 Our core “empathize” data collection methods were interviews and observations with clinicians. We triangulated findings with other operational and clinical data, including clinician secure messaging metadata, survey data, and policies/procedures. The subsequent “define” step of the design thinking process involved the development of user-centered “problem statements”.

Data Collection and Analysis

Our team divided into working groups to collect and analyze data from the following sources: (1) clinician interviews; (2) unit-based observations; (3) clinician messaging data; (4) clinician surveys, and (5) policies and procedures. The methods of each working group are summarized below. Detailed methods and results from individual working groups will be reported elsewhere; in this report we focus on results of data synthesis to generate problem statements.

1. Clinician interviews: We conducted 30 to 60-minute semi-structured interviews with clinicians (i.e., nurses, physicians, APPs, therapists, psychologists) to understand barriers and facilitators of interprofessional communication when caring for hospitalized children. We used convenience sampling, recruiting participants through email. We continued to recruit clinicians until we reached data saturation across ICU and non-ICU areas. Interviews were conducted by videoconferencing software. Participants provided verbal consent and received a $30 gift card for participation. Interviews were audio recorded, professionally transcribed, and reviewed for accuracy. The interview guide can be found in the Supplemental Materials, Supplemental Digital Content 1, http://links.lww.com/JPS/A748. We coded interviews as described in “Data Synthesis” below.

2. Observations: We conducted observations on the clinical units from which we recruited interview participants. We sought to describe interprofessional communication behaviors and identify barriers and facilitators in real time. Trained observers included a nurse (who did not work on the unit) and a graduate student studying integrated product design, to obtain both insider and outsider viewpoints. To observe information transfer between clinicians caring for the same patient, study team members shadowed different members of the clinical team caring for the same patient (e.g., nurse and resident physician). Observations lasted between ~1 and 4 hours each and occurred at different times of day, to ensure the observers were present for interactions such as rounds and sign out. Clinicians were asked to “think aloud” to help the observer understand their thought processes. Observers took field notes, which we coded as described in “Data Synthesis” below. Both interviews and observations were conducted in Winter-Spring 2024.

3. Clinician messaging data: We analyzed clinician secure messaging metadata for the 1-month period of April 2024. We assessed message frequency and response time as an indication of cognitive load. Response time was defined as time between subsequent messages sent between different individuals within a conversation. In 2 clinical units with well-defined clinical teams, we described message frequency and response rate by clinical role and analyzed trends in messaging behaviors by hour of the day. Analyses were performed using R Statistical Software (v 4.0; R Core Team 2021).

4. Clinician survey data: We analyzed results from a survey sent hospital-wide to physician, APPs, and bedside nurses by a hospital operations safety team between March and May 2024. Given its operational goals, the survey was pragmatically delivered through email announcements across hospital divisions and nursing distribution lists. The primary goal of the survey was to understand clinician attitudes toward new communication technologies and identify usability issues that potentially contributed to patient safety. We conducted a secondary analysis of free-text responses to open-ended prompts (e.g., “Other comments about clinical communication:”), given that they likely represent issues most pressing to the respondent. Of the 519 responses to the survey, 155 (30%) included a free-text response. Two members team members reviewed and thematically categorized each of the responses, using an inductive approach.

5. Policies and procedures analysis: We reviewed institutional policies, procedures, and guidelines related to communication of clinical information. Documents were obtained through search of the policy database and from clinical partners. We included documents deemed relevant to the topic of clinician-to-clinician communication of patient information in the inpatient setting. Documents were coded in qualitative software using inductive and deductive methods.

Data Synthesis

Beginning with data collected through direct interaction with clinicians (interviews and observations), we used an empathy mapping process to group data into what participants “did”, “felt”, “thought”, and “said”.²¹ We grouped the results by role (e.g., nurse, resident/APP, fellow, attending) and then further grouped the coded data into themes. We examined similarities and differences across roles and created a communication process map.

We then held a team retreat to triangulate the empathy mapping results with the findings from the other working groups (messaging data, survey data, policy procedure review). We identified common and discrepant findings across the data sources. The retreat led to new avenues of exploration for some working groups. After the retreat, we also used Systems Engineering Initiative for Patient Safety (SEIPS) activities to ground our analysis in systems thinking,¹⁴ including a People, Environments, Tools and Tasks (PETT) scan and development of a configural diagram.^15,22

From our data synthesis results, we developed problem statements using the design thinking formula: “I am (user), I am trying to (goal), but (barrier), because (root cause), which makes me feel (emotion)”. Problem statements were generated by consensus with a small group of PSLL investigators and refined with the rest of the team. We selected 3 of the statements to be the foundation for the next phase of the project, intervention development. We selected statements that were most commonly reflected in the end-user data and that were relevant to patient safety. Statements were reviewed with multiple clinical and operational groups across the organization for face validity.

RESULTS

Data synthesized for the problem analysis included: 28 clinician interviews (7 nurses, 8 residents and APPs; 4 fellows; 2 attending physicians; 4 therapists; 2 clinical nurse specialists/experts; 1 psychologist); 32 hours of unit observations (7 nurses and 5 residents/advanced practice providers); metadata for 433,432 of secure chat messages; 519 survey responses (165 nurses, 142 attendings, 64 APPs, 35 residents, 33 fellows, and 80 other); and 40 communication-related policies/procedures (e.g., iPhone Usage Policy, Communication of Critical Results Policy). Our SEIPS Configural Diagram can be found in the Figure 1 and the PETT scan in the Table 1.

FIGURE 1.

SEIPS configural diagram²² for the work process of health care team communication of an ad hoc patient need/change in the pediatric inpatient setting.

TABLE 1.

PETT (People, Environment, Tools, and Tasks) Scan¹⁵ for Health Care Team Communication in the Pediatric Inpatient Setting

Work system factor	Barriers	Facilitators
People (physical, cognitive, psychosocial characteristics)	Individual preferences about communication mode/frequency/responsiveness contribute to variation in communication processes	In-person or phone communications allow for non-verbal cues, such as tone, to provide context Familiarity of team members can improve communication
Environments (physical, organizational, external)	Limited organizational policies around modes of communication for ad hoc patient communication Policies focused on what to communicate, but not how or who In some units, physicians/advanced practice providers are not physically located on the unit	Work rooms in close proximity facilitate more ad hoc in-person conversations
Tools (objects used to transform input into output and characteristics of the devices, e.g., usability)	Technical challenges with voice calling platform Lower activation energy needed for messaging, more likely to result in incomplete messages, lack of context Nurses experience notifications from messages but also medical device alarms and other notifications that look similar on mobile phone Hesitancy to use or lack of awareness of urgent/important indicators on secure messages	Secure messaging is easy to use Secure messaging can serve as a written task list/reminder of action needed Can add others to secure messaging conversations for awareness
Tasks (activities, their sequence and characteristics—e.g., difficulty/complexity)	Nurses may be engaged in actively intervening with a patient and looking for the fastest way to communicate	Nurse participation in rounds may improve shared awareness and reduce need for messages about plan of care
Interactions (e.g., how well a tool fits a task, how environment affects behavior)	Voice calling takes more steps than messaging, but perceived to be the ‘urgent’ way to communicate Text messaging habits from daily life may impact texting behaviors at work Timing and volume of messages can disrupt workflow Generational differences in preference for calling versus messaging may create variation in interprofessional communications, regardless of tool usability	Secure messaging is easy and less disruptive than other modes of communication Scheduled/anticipated check-ins among clinical team to review non-urgent issues and requests may reduce interruptions

On the basis of iterative analysis of results, we determined that our focus would be communication between 2 frontline clinical roles: staff nurses and ordering clinicians (residents/APPs/fellows in frontline roles), because of the frequency of their communication and direct impact to patient care. We initially generated 7 problem statements and subsequently selected 3 based on our criteria of relevance to patient safety and salience in the data.

Problem Statement 1

I am a staff nurse on an inpatient pediatric unit,

I am trying to efficiently communicate with team members to provide timely patient care,

But I struggle to locate the ordering clinician,

Because I don’t know the best mode of communication to reach them in the moment,

Which makes me feel frustrated.

In our observations, we saw several examples of inefficiencies when nurses were trying to contact the ordering clinician for routine matters. For example, when an ordering clinician did not respond to a secure message, the nurse looked for them in person and called them, and finally received a secure message response that they were busy. In another example, a nurse could not find the ordering clinician in person, so they sent a secure message, to which the ordering clinician responded, “call or come find me.”

Survey data underscored inefficiencies with some modes of communication, for which secure messaging provided a useful alternative. Survey respondents noted delays “…due to trying to find a phone number…” in the current voice calling system. Respondents described that clinicians sometimes “…resort to [secure messaging] though always not the ideal form of communication…” because it involves fewer steps and allows the communication to be initiated so that the receiver can attend to it at their convenience.

Problem Statement 2

I am a frontline ordering clinician on a pediatric inpatient unit,

I am trying to appropriately and safely manage my patients,

But I think I might miss urgent information,

Because senders sometimes use a mode of communication that does not align with my expectations for how urgent information will be communicated,

Which makes me feel anxious.

In interviews and observations, we found general agreement that “urgent” issues warranted a phone call or in-person conversation, but ordering clinicians in particular reported receiving urgent information through secure message. One clinician described the discrepancy as even when information is urgent, “… they’re using the least urgent form of [communication],” referring to secure messaging. Another explained: “…sometimes there’s hidden in [the messages] an important vital sign change that you need to know about, so you have to look through them, even if you’re not responding to them all right away”. These findings were consistent with comments in the survey data.

One core driver was the lack of a shared mental model about urgency and response time. Ordering clinicians generally expressed that issues conveyed through secure messages should be able to wait at least 10 minutes: “if it’s something that you need an answer within, I’d say, the next like 10 to 30 minutes, you should probably call…” Another said: “in theory, we’re supposed to have up to 20 or 30 minutes to answer a [secure message].” In contrast, nurses expressed anticipating faster response times: “Usually, I’ll send the message, and then in my head, I’ll be like, all right. I’m giving them five minutes to respond to this. If they don’t, I’m gonna give them a call.”

The secure messaging metadata revealed very fast response times, averaging about 1 to 2 minutes, which may contribute to expectations about response time. Both nurses and ordering clinicians described “trying” secure messaging first to see if they would get a quick response: “Do people still try to use [secure messaging] first? Yes. And sometimes it works and sometimes it doesn’t, and so you have this intermittent reinforcement of, oh, well, it’s worth a try”.

Problem Statement 3

I am a recipient of messages on a pediatric inpatient unit,

I am trying to attend to clinical work that requires concentration,

But I am frequently interrupted,

Because colleagues send a large volume of messages,

Which makes me feel overwhelmed and frustrated.

Pervasive in interview, observation, and survey data was ordering clinicians’ overwhelm from the number of messages they receive. One clinician wrote in the survey “…[messages are] another version of ‘alarm fatigue’ and it is taking a toll on people’s ability to triage information…” Nurses also expressed this sentiment, but to a lesser extent; for example, one nurse stated: “…you can get a lot of messages in one day, and it can be hard to keep up with sometimes.” Our analysis of the secure messaging data corroborated the interview data. Frontline ordering clinicians received more messages than other clinician roles; for example, in the PICU they received a median of 8 messages per hour, whereas nurses received 2 messages per hour.

Secure messaging created a more frictionlessly way to convey information than was previously available. One ordering clinician observed: “sometimes the threshold to just send a message is so easy that there are a lot of messages that are sent that don’t necessarily need to be messages.” Some nurses explained that secure messaging also makes it easy for them to ‘err on the side of caution’ in terms of sharing information with the team.

Participants noted that lack of context or clear goals in messages resulted in multiple messages back-and-forth. One ordering clinician described that secure messaging “facilitated this enormous burden of micro-messages, as opposed to, here’s a message with a very specific purpose … ” One clinician noted that incomplete sentences contributed to more message burden because they had to ask for clarification “several times”.

Participants cited generational norms for text messaging, such as younger clinicians favoring short messages, as a possible reason for the lack of message clarity. Our data also indicated that perceived clinical hierarchy and professional communication norms played a role; for example, a nurse explained that their approach to messaging was to “recommend it in a way by still asking.” This “hint and hope” approach to clinical communication is widely noted in other studies of clinical communication.^1,23

Despite challenges, clinicians across roles agreed that secure messaging introduced numerous benefits. Messaging allowed clinicians to communicate efficiently, refer back to prior conversations, and maintain a “to do” list of tasks. Clinicians also appreciated the “thumbs up” emoji as a way to “close the loop”. Nurses found that messaging relieved cognitive burden of having to hold questions or requests in their working memory until they could call or find a team member. The messaging platform also facilitated communication peripheral, but essential, to patient care, such as bed and staffing coordination.

DISCUSSION

Using design thinking methods and triangulating findings across multiple data sources, our Patient Safety Learning Lab identified 3 core challenges for nurse-ordering clinician communication in an inpatient pediatric setting. The secure messaging platform presented new workflow challenges, particularly in terms of volume of messages and perception of urgent issues. A more salient challenge for nurses was efficiency, given the variety of communication modes available and the lack of shared mental model for use.

Obvious parallels exist between secure message burden and medical device alarm fatigue.⁹ ‘Alarm fatigue’ due to excessive invalid medical device alarms is a long-standing patient safety concern acknowledged by the Joint Commission.^24,25 Similar patient safety risks could exist with high volumes of secure chat messages. Little progress on alarm fatigue has occurred over the last decade,^26,27 but as technological interruptions increase across clinical roles, we hope to leverage the collective recognition of this problem into meaningful solutions for all. To address alarm fatigue, core strategies have included improving the positive predictive value of alarms (by reducing invalid alarms)²⁸ and using methods such as auditory icons²⁹ to enhance perception and comprehension of alarms. Similar strategies could be applied to manage message burden.

In addition to interruptions from notifications, our results support other predicted patient safety challenges for systems using secure messaging.⁹ We found that lack of clarity in messaging led to the need for more messages. In another study, higher volume of messaging was also associated with higher numbers of phone calls.³⁰ With multiple modes of communication available and variation in how they are used, we also found that communication mode was not an accurate heuristic for triaging urgency. Previous reports have called for more guidelines on use of secure messaging,⁸ but our findings underscore the difficulty of aligning “work as done” with “work as imagined”³¹ (or desired) for secure messaging. Secure messaging may be imagined to be the least urgent mode of communication, but is also often fastest.

Challenges and Limitations

A challenge in applying human-centered design processes to complex systems like health care is the multiple end users with competing needs.^19,32 We are not simply creating an intervention that optimizes for one end user’s experience, but rather one that supports multiple end users and ultimately makes the system safer for patients. As such, we combined traditional design thinking processes with techniques like SEIPS modeling.¹⁵

Each of the methods described have limitations inherent to their design; however, triangulation of qualitative and quantitative analysis of diverse data sources increases our confidence in the validity of our problem statements. Triangulation of data from many sources was an effective approach used in a previous PSLL at our clinical site.³³ We did not conduct observations on nights or weekends, but interview participants worked a variety of shifts. Although we aimed to include clinicians from many roles, we had limited response from some roles; our problem statements were, therefore, generated for roles from which we had sufficient data (nurses, physicians, APPs). Future work including a broader spectrum of clinical roles is necessary. As a single-site study conducted at a large urban teaching children’s hospital, generalizability may be limited, especially to nonteaching institutions or other combinations of communication modalities.

Next steps

We continue to iteratively collect data to understand communication and to help establish outcome measures. For example, we are reviewing over 2000 communication-related events from our institution’s safety event reporting system to categorize based on type of communication failure.³⁴ In subsequent phases of the design thinking process,¹⁸ we will conduct ideation sessions with end users to generate solutions to the problem statements. We have also identified and communicated problem areas that were more appropriate for operational partners to address; for example, technical issues with the voice calling platform. As we move into solution development, our goal is not to remove or restrict use of secure messaging, but rather to proactively design a system that supports safe and effective communication across the multiple modes of communication available. Users of inpatient messaging will likely expand to include patients and families in the future, underscoring the importance of developing interventions to manage and triage message-based communication.

CONCLUSIONS

In summary, our problem analysis led to identification of 3 primary problem areas that will be the focus for the next stages of this work: locating/contacting colleagues, communicating urgency of information, and high volume of minimally informative messages. Practical application of human-centered design processes contributed to a more holistic understanding of challenges, and their patient safety implications, from the perspective of multiple end-user groups.